1 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4 * lra-constraints.c (process_alt_operands): Move setting
5 this_alternative_matches below.
7 2018-03-29 Martin Liska <mliska@suse.cz>
10 * doc/invoke.texi: Document how LTO works with debug info.
11 Describe auto-load support of binutils. Mention 'x86-64'
12 as valid option value of -march option.
14 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
19 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
21 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
24 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
27 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
28 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
29 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
30 for RS6000_BTM_POWERPC64.
31 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
32 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
33 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
38 2018-03-28 Carl Love <cel@us.ibm.com>
41 2017-09-27 Carl Love <cel@us.ibm.com>
43 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
44 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
45 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
48 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
50 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
51 instead of __vector bool.
52 (_mm_max_pu8): Likewise.
53 (_mm_min_pi16): Likewise.
55 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
58 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
62 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
64 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
65 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
66 (div_extend): Likewise.
67 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
69 (__builtin_divweuo): Likewise.
70 (__builtin_divdeo): Likewise.
71 (__builtin_divdeuo): Likewise.
73 2018-03-28 Jakub Jelinek <jakub@redhat.com>
76 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
77 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
79 PR tree-optimization/82004
80 * gimple-match-head.c (optimize_pow_to_exp): New function.
81 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
82 Don't fold to exp if optimize_pow_to_exp is false.
84 2018-03-28 Martin Liska <mliska@suse.cz>
87 * calls.c (initialize_argument_information): Fix trailing space.
88 * common.opt: Fix typo and provide better explanation for
89 -fsanitize-coverage option.
90 * config/i386/i386.opt: Fix typo.
92 2018-03-28 Jakub Jelinek <jakub@redhat.com>
93 Martin Liska <mliska@suse.cz>
96 * gimplify.c (asan_poison_variable): Don't do the check for
97 gimplify_omp_ctxp here.
98 (gimplify_decl_expr): Do it here.
99 (gimplify_target_expr): Likewise.
101 2018-03-28 Martin Liska <mliska@suse.cz>
104 * config/i386/i386.c (ix86_function_arg_advance): Do not call
105 chkp_type_bounds_count if MPX is not enabled.
107 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
109 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
111 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
114 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
115 function to create the function decl for complex long double
116 multiply and divide for -mabi=ieeelongdouble.
117 (init_float128_ieee): Call it.
119 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
122 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
123 -fcf-protection=branch -mibt.
124 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
126 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
129 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
131 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
134 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
135 extern array declarations.
137 2018-03-27 Richard Biener <rguenther@suse.de>
140 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
141 explicit single_use checks.
143 2018-03-27 Richard Biener <rguenther@suse.de>
145 PR tree-optimization/85082
146 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
149 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
151 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
152 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
153 Turn on fasynchronous-unwind-tables and funwind-tables.
155 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
158 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
159 (*bmi_blsr_<mode>_ccz): Ditto.
161 2018-03-26 Tom de Vries <tom@codesourcery.com>
163 PR tree-optimization/85063
164 * omp-general.c (offloading_function_p): New function. Factor out
166 * omp-offload.c (pass_omp_target_link::gate): ... here.
167 * omp-general.h (offloading_function_p): Declare.
168 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
169 with attribute omp declare target for offloading functions.
171 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
173 PR tree-optimization/84005
174 * tree-data-ref.h (get_base_for_alignment): Declare.
175 * tree-data-ref.c (get_base_for_alignment_1): New function.
176 (get_base_for_alignment): Likewise.
177 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
178 get_base_for_alignment to find a suitable base object, instead
179 of always using drb->base_address.
181 2018-03-23 Jakub Jelinek <jakub@redhat.com>
184 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
185 known size by default.
187 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
190 * lra-constraints.c (process_alt_operands): Don't match BLKmode
191 and non BLKmode operands.
193 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
196 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
199 2018-03-23 Richard Biener <rguenther@suse.de>
202 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
203 we are going to emit early debug for LTO.
205 2018-03-23 Jakub Jelinek <jakub@redhat.com>
208 * function.c (match_asm_constraints_1): Don't optimize if input
209 doesn't satisfy general_operand predicate for output's mode.
212 * alias.c (write_dependence_p): Don't require for x_canonicalized
213 non-VOIDmode if x has VOIDmode.
216 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
217 just don't try to optimize it rather than assert it never happens.
219 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
221 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
222 macro expansions for definition of ST_INTERNAL_<mode> and
223 LD_INTERNAL_<mode> builtins.
224 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
226 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
228 (altivec_expand_st_builtin): Likewise.
229 (altivec_expand_builtin): Remove calls to deleted functions.
230 (rs6000_address_for_altivec): Delete this function.
231 * config/rs6000/vector.md: Remove expands for
232 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
234 2018-03-22 Sudakshina Das <sudi.das@arm.com>
237 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
238 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
239 re-computing once computed.
240 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
241 (arm_init_machine_status): Initialize
242 machine->static_chain_stack_bytes.
244 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
247 * doc/extend.texi: Add four new prototypes for vec_ld.
248 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
249 definitions for more logical presentation.
250 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
251 entries for V1TI variants of __builtin_altivec_ld builtin.
252 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
253 handling of V1TI variant of LVX icode pattern.
254 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
255 (rs6000_gimple_fold_builtin): Likewise.
256 (altivec_init_builtins): Add code to define
257 __builtin_altivec_lvx_v1ti function.
259 2018-03-22 Jakub Jelinek <jakub@redhat.com>
262 * function.c (match_asm_constraints_1): Don't do the optimization
263 if input isn't a REG, SUBREG, MEM or constant.
265 2018-03-22 Tom de Vries <tom@codesourcery.com>
267 PR tree-optimization/84956
268 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
269 bb_has_abnormal_pred.
271 2018-03-22 Jakub Jelinek <jakub@redhat.com>
274 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
275 DECL_INITIAL (decl) to decl at the end.
276 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
279 2018-03-21 Joseph Myers <joseph@codesourcery.com>
281 * doc/extend.texi (__builtin_tgmath): Document when complex
282 integer types are treated as _Complex _Float64.
284 2018-03-21 Tom de Vries <tom@codesourcery.com>
286 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
288 2018-03-21 Jakub Jelinek <jakub@redhat.com>
290 PR tree-optimization/84960
291 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
292 if it is ENTRY block, move them into single succ of ENTRY in that case.
294 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
296 PR tree-optimization/84811
297 * poly-int.h (poly_span_traits): Remove the T3 parameter and
298 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
299 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
300 (known_subrange_p): Update accordingly. Cast each value involved
301 in the size comparison, rather than casting the result of the
304 2018-03-21 Jakub Jelinek <jakub@redhat.com>
306 PR tree-optimization/84982
307 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
308 by flipping the least significant bit rather than all bits from
309 bitpos to bitpos + bitsize - 1.
311 2018-03-21 Nathan Sidwell <nathan@acm.org>
313 * doc/extend.texi (Deprecated Features): Remove mention of
314 long-deleted deprecations.
316 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
319 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
320 * configure: Regenerate.
322 2018-03-21 Tom de Vries <tom@codesourcery.com>
324 PR tree-optimization/83126
325 * tree-parloops.c (num_phis): New function.
326 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
328 2018-03-21 Nathan Sidwell <nathan@acm.org>
330 * doc/extend.texi (Deprecated Features): Update deprecated flags,
331 mention anon-struct/union members and trailing attributes.
333 2018-03-21 Bin Cheng <bin.cheng@arm.com>
335 PR tree-optimization/84969
336 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
337 builtin memset partitions if they set different rhs values.
339 2018-03-21 Jakub Jelinek <jakub@redhat.com>
341 PR rtl-optimization/84989
342 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
343 VEC_DUPLICATE with scalar result mode.
345 2018-03-21 Martin Liska <mliska@suse.cz>
348 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
349 not intended return statement.
351 2018-03-21 Martin Liska <mliska@suse.cz>
354 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
355 (chkp_find_bound_slots_1): Limit number of iterations.
357 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
360 * Minor grammar fixes for x86 options.
362 2018-03-20 Jakub Jelinek <jakub@redhat.com>
365 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
366 holding REG_CFA_RESTORE notes, instead turn them into a USE.
368 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
371 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
372 (altivec_lvx_<mode>_1op): Likewise.
373 (altivec_stvx_<mode>_2op): Likewise.
374 (altivec_stvx_<mode>_1op): Likewise.
375 (altivec_lvx_<VM2:mode>): New define_expand.
376 (altivec_stvx_<VM2:mode>): Likewise.
377 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
378 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
379 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
380 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
381 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
382 (rs6000_gen_lvx): Likewise.
383 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
384 (altivec_expand_stv_builtin): Likewise.
385 (altivec_expand_builtin): Likewise.
386 * config/rs6000/vector.md: Likewise.
388 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
391 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
394 2018-03-20 Richard Biener <rguenther@suse.de>
397 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
398 sign-conversions as zero, fall back to standard scalar_stmt
401 2018-03-20 Martin Liska <mliska@suse.cz>
404 * predict.c (rebuild_frequencies): Handle case when we have
405 PROFILE_ABSENT, but flag_guess_branch_prob is false.
407 2018-03-20 Jakub Jelinek <jakub@redhat.com>
410 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
411 flag_section_anchors.
412 * varasm.c (use_blocks_for_decl_p): Remove hack for
416 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
418 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
419 be created, use lowpart_subreg of operands[0] rather than operands[0]
421 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
422 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
423 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
424 and n constraint instead of aarch64_shift_imm_di and Usd.
425 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
426 (*aarch64_<optab>_reg_minus<mode>3): ... this.
428 2018-03-20 Sudakshina Das <sudi.das@arm.com>
431 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
432 to favor GPR over NEON registers.
433 (<shift>di3_neon): Likewise.
435 2018-03-20 Tom de Vries <tom@codesourcery.com>
438 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
439 (nvptx_process_pars): Emit bar.sync asap and alap.
441 2018-03-20 Tom de Vries <tom@codesourcery.com>
444 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
445 seen_label if seen_label is already set.
447 2018-03-20 Jakub Jelinek <jakub@redhat.com>
450 * config/i386/i386.c (fold_builtin_cpu): For features above 31
451 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
452 Use 1U instead of 1. Formatting fixes.
455 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
456 instead of TREE_TYPE (s1) for the return value.
458 2018-03-19 Jakub Jelinek <jakub@redhat.com>
460 PR tree-optimization/84946
461 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
462 bitsize + bitsize in poly_uint64 rather than poly_int64.
465 * dwarf2asm.c: Include fold-const.c.
466 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
467 of decl rather than decl itself.
469 PR rtl-optimization/84643
470 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
472 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
475 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
476 calling assemble_variable.
478 2018-03-19 Sudakshina Das <sudi.das@arm.com>
481 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
482 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
484 2018-03-19 Jim Wilson <jimw@sifive.com>
487 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
488 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
489 (riscv_first_stack_step): Likewise.
490 (riscv_option_override): Use STACK_BOUNDARY instead of
492 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
494 (BIGGEST_ALIGNMENT): Set to 128.
495 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
496 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
499 2018-03-19 Richard Biener <rguenther@suse.de>
501 PR tree-optimization/84933
502 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
503 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
505 2018-03-19 Richard Biener <rguenther@suse.de>
507 PR tree-optimization/84859
508 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
509 (cond_if_else_store_replacement): Perform sinking operation on
510 single-store BBs regardless of MAX_STORES_TO_SINK setting.
511 Generalize what a BB with a single eligible store is.
513 2018-03-19 Richard Biener <rguenther@suse.de>
515 PR tree-optimization/84929
516 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
517 chrec_is_positive against non-chrec arg.
519 2018-03-19 Tamar Christina <tamar.christina@arm.com>
522 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
524 2018-03-18 Martin Liska <mliska@suse.cz>
526 PR rtl-optimization/84635
527 * regrename.c (build_def_use): Use matches_mode only when
530 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
532 PR tree-optimization/84913
533 * tree-vect-loop.c (vectorizable_reduction): Don't try to
534 vectorize chains of COND_EXPRs.
536 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
538 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
540 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
542 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
544 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
546 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
548 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
549 Kito Cheng <kito.cheng@gmail.com>
551 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
552 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
553 (nds32_adjust_reg_alloc_order): New function.
554 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
556 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
558 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
559 nds32_print_operand, nds32_print_operand_address): Use
560 HOST_WIDE_INT_PRINT_DEC instead.
562 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
564 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
566 2018-03-17 Jakub Jelinek <jakub@redhat.com>
569 * config/i386/i386.c (initial_ix86_tune_features,
570 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
572 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
573 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
574 rather than 1u << ix86_tune. Formatting fix.
575 (ix86_option_override_internal): Change ix86_arch_mask from
576 unsigned int to unsigned HOST_WIDE_INT, initialize to
577 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
578 (ix86_function_specific_restore): Likewise.
580 2018-03-16 Jakub Jelinek <jakub@redhat.com>
583 * postreload.c (reload_combine_recognize_pattern): Perform
584 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
585 truncate_int_for_mode the result for the destination's mode.
588 * hsa-gen.c (mem_type_for_type): Fix comment typo.
589 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
591 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
594 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
597 * lra-assigns.c (lra_split_hard_reg_for): Don't use
598 regno_allocno_class_array and sorted_pseudos.
599 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
600 insns where regno is used.
602 2018-03-16 Martin Liska <mliska@suse.cz>
605 * multiple_target.c (create_dispatcher_calls): Redirect
606 reference in the symbol table.
608 2018-03-16 Martin Liska <mliska@suse.cz>
611 * multiple_target.c (create_dispatcher_calls): Redirect also
614 2018-03-16 Jakub Jelinek <jakub@redhat.com>
618 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
619 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
620 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
622 2018-03-16 Julia Koval <julia.koval@intel.com>
624 * doc/invoke.texi (Skylake Server): Add CLWB.
625 Cannonlake): Remove CLWB.
627 2018-03-16 Jakub Jelinek <jakub@redhat.com>
629 PR tree-optimization/84841
630 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
632 (FLOAT_ONE_CONST_TYPE): Define.
633 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
634 (sort_by_operand_rank): Put entries with higher constant_type last
635 rather than first to match comments.
637 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
639 * config/nios2/nios2.md (movsi_internal): Fix thinko in
642 2018-03-15 Jakub Jelinek <jakub@redhat.com>
645 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
646 check and use address of target always.
648 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
651 * config/i386/i386.c (indirect_thunk_needed): Update comments.
652 (indirect_thunk_bnd_needed): Likewise.
653 (indirect_thunks_used): Likewise.
654 (indirect_thunks_bnd_used): Likewise.
655 (indirect_return_needed): New.
656 (indirect_return_bnd_needed): Likewise.
657 (output_indirect_thunk_function): Add a bool argument for
659 (output_indirect_thunk_function): Don't generate alias for
660 function return thunk.
661 (ix86_code_end): Call output_indirect_thunk_function to generate
662 function return thunks.
663 (ix86_output_function_return): Set indirect_return_bnd_needed
664 and indirect_return_needed instead of indirect_thunk_bnd_needed
665 and indirect_thunk_needed.
667 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
669 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
670 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
671 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
673 2018-03-15 David Malcolm <dmalcolm@redhat.com>
674 Paul Hua <paul.hua.gm@gmail.com>
677 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
679 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
681 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
682 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
685 2018-03-15 Tamar Christina <tamar.christina@arm.com>
688 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
689 instead of GET_MODE_SIZE when comparing Units.
691 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
694 * varasm.c (hash_section): Return an unchangeble hash value
695 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
696 Return !aarch64_can_use_per_function_literal_pools_p ().
698 2018-03-15 Jakub Jelinek <jakub@redhat.com>
701 * optabs.c (emit_conditional_move): Pass address of cmode's copy
702 rather than address of cmode as last argument to prepare_cmp_insn.
704 2018-03-15 Julia Koval <julia.koval@intel.com>
706 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
707 F_AVX512VNNI, F_AVX512BITALG): New.
709 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
712 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
713 insn for floating-point loads and stores.
715 2018-03-14 Carl Love <cel@us.ibm.com>
717 * config/rs6000/rs6000-c.c: Add macro definitions for
718 ALTIVEC_BUILTIN_VEC_PERMXOR.
719 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
720 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
721 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
722 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
724 * config/doc/extend.texi: Add prototypes for vec_permxor.
726 2018-03-14 David Malcolm <dmalcolm@redhat.com>
729 * diagnostic-show-locus.c (class layout_point): Convert m_line
730 from int to linenum_type.
731 (line_span::comparator): Use linenum "compare" function when
732 comparing line numbers.
733 (test_line_span): New function.
734 (layout_range::contains_point): Convert param "row" from int to
736 (layout_range::intersects_line_p): Likewise.
737 (layout::will_show_line_p): Likewise.
738 (layout::print_source_line): Likewise.
739 (layout::should_print_annotation_line_p): Likewise.
740 (layout::print_annotation_line): Likewise.
741 (layout::print_leading_fixits): Likewise.
742 (layout::annotation_line_showed_range_p): Likewise.
743 (struct line_corrections): Likewise for field m_row.
744 (line_corrections::line_corrections): Likewise for param "row".
745 (layout::print_trailing_fixits): Likewise.
746 (layout::get_state_at_point): Likewise.
747 (layout::get_x_bound_for_row): Likewise.
748 (layout::print_line): Likewise.
749 (diagnostic_show_locus): Likewise for locals "last_line" and
751 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
752 * input.c (selftest::test_linenum_comparisons): New function.
753 (selftest::input_c_tests): Call it.
754 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
755 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
756 * selftest.h (ASSERT_GT): New macro.
757 (ASSERT_GT_AT): New macro.
758 (ASSERT_LT): New macro.
759 (ASSERT_LT_AT): New macro.
761 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
763 PR rtl-optimization/84780
764 * combine.c (distribute_links): Don't make a link based on pc_rtx.
766 2018-03-14 Martin Liska <mliska@suse.cz>
768 * tree.c (record_node_allocation_statistics): Use
770 (get_stats_node_kind): New function extracted from
771 record_node_allocation_statistics.
772 (free_node): Use get_stats_node_kind.
774 2018-03-14 Richard Biener <rguenther@suse.de>
776 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
777 that the value-set of ANTIC_IN doesn't grow.
780 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
782 (BB_VISITED_WITH_VISITED_SUCCS): New define.
783 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
785 2018-03-14 Julia Koval <julia.koval@intel.com>
787 * config.gcc (icelake-client, icelake-server): New.
789 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
790 (initial_ix86_arch_features): Ditto.
791 (PTA_SKYLAKE): Add SGX.
792 (PTA_ICELAKE): Remove.
793 (PTA_ICELAKE_CLIENT): New.
794 (PTA_ICELAKE_SERVER): New.
795 (ix86_option_override_internal): Split up icelake on icelake client and
797 (get_builtin_code_for_version): Ditto.
798 (fold_builtin_cpu): Ditto.
799 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
800 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
801 * config/i386/i386.h (processor_type): Ditto.
802 * doc/invoke.texi: Ditto.
804 2018-03-14 Jakub Jelinek <jakub@redhat.com>
807 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
808 INTEGER_CST offset, add it together with bitpos / 8 and
809 sign extend based on POINTER_SIZE.
813 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
816 * config/i386/constraints.md (Yc): New register constraint.
817 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
818 Use Yc constraint for alternative 2 of operand 0. Remove
819 preferred_for_speed attribute.
821 2018-03-14 Richard Biener <rguenther@suse.de>
823 PR tree-optimization/84830
824 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
825 with the old one to avoid oscillations.
827 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
830 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
832 (assign_by_spills): Return a flag of reload assignment failure.
833 Do not process the reload assignment failures. Do not spill other
834 reload pseudos if they has the same reg class. Update n if
836 (lra_assign): Add a return arg. Set up from the result of
837 assign_by_spills call.
838 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
839 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
840 usage_insns if it is not NULL.
841 (spill_hard_reg_in_range): New function.
842 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
843 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
845 (lra_assign): Change prototype.
846 * lra.c (lra): Add code to deal with fails by splitting hard reg
849 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
851 * config/riscv/riscv.opt (mrelax): New option.
852 * config/riscv/riscv.c (riscv_file_start): Emit ".option
853 "norelax" when riscv_mrelax is disabled.
854 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
856 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
859 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
860 reassociation for int modes.
862 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
864 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
865 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
867 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
868 * config/aarch64/aarch64-sve.md
869 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
870 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
871 (*extend<mode><Vwide>2): Rename to...
872 (aarch64_sve_extend<mode><Vwide>2): ...this.
873 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
874 renaming the old pattern to...
875 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
877 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
878 define_expand, renaming the old pattern to...
879 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
880 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
881 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
882 account when deciding which SVE instruction the optab should use.
883 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
885 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
887 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
888 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
889 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
890 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
891 (tlsdesc_small_<mode>): Turn a define_expand and use
892 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
893 (tlsdesc_small_advsimd_<mode>): ...this.
894 (tlsdesc_small_sve_<mode>): New pattern.
896 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
898 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
899 (UNSPEC_UMUL_HIGHPART): New constants.
900 (MUL_HIGHPART): New int iteraor.
901 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
902 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
904 (*<su>mul<mode>3_highpart): New define_insn.
906 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
909 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
912 2018-03-13 Martin Liska <mliska@suse.cz>
915 * (sem_item_optimizer::sem_item_optimizer): Initialize new
917 (sem_item_optimizer::~sem_item_optimizer): Release it.
918 (sem_item_optimizer::merge_classes): Register variable aliases.
919 (sem_item_optimizer::fixup_pt_set): New function.
920 (sem_item_optimizer::fixup_points_to_sets): Likewise.
921 * ipa-icf.h: Declare new variables and functions.
923 2018-03-13 Jakub Jelinek <jakub@redhat.com>
926 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
927 integer_pow2p@2 and test integer_pow2p in condition.
928 (A < 0 ? C : 0): Similarly for @1.
931 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
932 characters starting at p contain '\0' character, don't look beyond
936 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
937 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
940 * reg-stack.c (change_stack): Change update_end var from int to
941 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
942 also call set_block_for_insn on the newly added insns and rescan.
945 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
949 * tree.c (stabilize_reference_1): Return save_expr (e) for
950 STATEMENT_LIST even if it doesn't have side-effects.
952 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
954 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
956 2018-03-12 Renlin Li <renlin.li@arm.com>
958 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
959 aarch64_output_scalar_simd_mov_immediate.
961 2018-03-12 Martin Sebor <msebor@redhat.com>
963 PR tree-optimization/83456
964 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
965 for perfectly overlapping calls to memcpy.
966 (gimple_fold_builtin_memory_chk): Same.
967 (gimple_fold_builtin_strcpy): Handle no-warning.
968 (gimple_fold_builtin_stxcpy_chk): Same.
969 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
971 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
973 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
974 parameter. Use it for SFmode.
975 (rs6000_function_arg_advance_1): Adjust.
976 (rs6000_function_arg): Adjust.
977 (rs6000_gimplify_va_arg): Pass false for that new parameter.
979 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
981 PR rtl-optimization/84169
982 PR rtl-optimization/84780
983 * combine.c (can_combine_p): Check for a 2-insn combination whether
984 the destination register is used between the two insns, too.
986 2018-03-12 Richard Biener <rguenther@suse.de>
988 PR tree-optimization/84803
989 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
990 for refs DR analysis didn't process.
992 2018-03-12 Richard Biener <rguenther@suse.de>
994 PR tree-optimization/84777
995 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
996 force-vectorize loops ignore whether we are optimizing for size.
998 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1000 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1001 (TARGET_MD_ASM_ADJUST): Define.
1003 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1004 Kito Cheng <kito.cheng@gmail.com>
1005 Chung-Ju Wu <jasonwucj@gmail.com>
1007 * config/nds32/nds32.c (nds32_compute_stack_frame,
1008 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1009 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1010 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1011 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1012 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1013 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1014 * config/nds32/nds32.md (prologue, epilogue): Use macro
1015 NDS32_V3PUSH_AVAILABLE_P to do checking.
1017 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1020 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1021 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1022 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1023 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1024 addition of most attributes on !orig_type_die or the attribute not
1025 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1027 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1028 Chung-Ju Wu <jasonwucj@gmail.com>
1030 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1032 * config/nds32/nds32.opt (mvh): New option.
1034 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1035 Chung-Ju Wu <jasonwucj@gmail.com>
1037 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1039 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1040 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1043 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1044 Chung-Ju Wu <jasonwucj@gmail.com>
1046 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1048 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1049 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1051 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1052 Kito Cheng <kito.cheng@gmail.com>
1053 Chung-Ju Wu <jasonwucj@gmail.com>
1055 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1056 UNSPEC_FFMISM and UNSPEC_FLMISM.
1057 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1058 for ffb, ffmism and flmism.
1059 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1060 (unspec_ffmism): Ditto.
1061 (unspec_flmism): Ditto.
1062 (nds32_expand_builtin_impl): Check if string extension is available.
1063 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1064 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1066 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1069 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1072 * lra-assigns.c (assign_by_spills): Return a flag of reload
1073 assignment failure. Do not process the reload assignment
1074 failures. Do not spill other reload pseudos if they has the same
1076 (lra_assign): Add a return arg. Set up from the result of
1077 assign_by_spills call.
1078 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1079 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1080 usage_insns if it is not NULL.
1081 (spill_hard_reg_in_range): New function.
1082 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1083 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1084 function prototypes.
1085 (lra_assign): Change prototype.
1086 * lra.c (lra): Add code to deal with fails by splitting hard reg
1089 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1092 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1094 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1097 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1098 (dw_val_node): Add val_symbolic_view.
1099 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1100 (symview_upper_bound): New.
1101 (new_line_info_table): Initialize symviews_since_reset.
1102 (dwarf2out_source_line): Count symviews_since_reset and set
1103 symview_upper_bound.
1104 (dw_val_equal_p): Handle symview.
1105 (add_AT_symview): New.
1106 (print_dw_val): Handle symview.
1107 (attr_checksum, attr_checksum_ordered): Likewise.
1108 (same_dw_val_p, size_of_die): Likewise.
1109 (value_format, output_die): Likewise.
1110 (add_high_low_attributes): Use add_AT_symview for entry_view.
1111 (dwarf2out_finish): Reset symview_upper_bound, clear
1114 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1117 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1118 Add strict argument and use it.
1119 (rs6000_split_multireg_move): Update for new strict argument.
1120 (mem_operand_gpr): Disallow all non-offsettable addresses.
1121 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1123 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1126 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1127 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1128 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1131 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1132 decl, use remap_type if we want to use the type.
1134 2018-03-09 Martin Sebor <msebor@redhat.com>
1136 PR tree-optimization/84526
1137 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1139 (builtin_access::generic_overlap): Be prepared to handle non-array
1142 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1144 PR rtl-optimization/84682
1145 * lra-constraints.c (process_address_1): Check is_address flag
1146 for address constraints.
1147 (process_alt_operands): Likewise.
1148 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1149 preprocess_constraints.
1150 * recog.h (preprocess_constraints): Add oploc parameter.
1152 * recog.c (preprocess_constraints): Test address_operand for
1153 CT_ADDRESS constraints.
1155 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1158 * lra-assigns.c (assign_by_spills): Return a flag of reload
1159 assignment failure. Do not process the reload assignment
1160 failures. Do not spill other reload pseudos if they has the same
1162 (lra_assign): Add a return arg. Set up from the result of
1163 assign_by_spills call.
1164 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1165 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1166 usage_insns if it is not NULL.
1167 (spill_hard_reg_in_range): New function.
1168 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1169 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1170 function prototypes.
1171 (lra_assign): Change prototype.
1172 * lra.c (lra): Add code to deal with fails by splitting hard reg
1175 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1178 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1179 Accept complain bool parameter. Only emit errors if it is true.
1180 (arm_parse_cpu_option_name): Likewise.
1181 (arm_target_thumb_only): Adjust callers of the above.
1182 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1183 prototype to take a default true bool parameter.
1184 (arm_parse_arch_option_name): Likewise.
1186 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1187 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1191 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1192 * configure: Regenerate.
1193 * configure.ac ("linker --version-script option"): New.
1194 ("linker soname option"): New.
1196 2018-03-09 Richard Biener <rguenther@suse.de>
1198 PR tree-optimization/84775
1199 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1200 immediate uses of predicate stmts and mark them modified.
1203 PR tree-optimization/84178
1204 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1206 (version_loop_for_if_conversion): Delay update_ssa call.
1207 (tree_if_conversion): Delay update_ssa until after predicate
1210 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1213 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1214 when the function accesses prior frames.
1216 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1219 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1220 gen_llsym, otherwise call maybe_gen_llsym.
1223 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1224 has ',' character inside of it.
1226 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1229 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1230 as clobbering CC_REGNUM.
1232 2018-03-08 Richard Biener <rguenther@suse.de>
1235 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1236 (follow_copies_to_constant): Do not follow SSA names registered
1239 2018-03-08 Richard Biener <rguenther@suse.de>
1241 PR tree-optimization/84178
1242 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1244 (version_loop_for_if_conversion): Delay update_ssa call.
1245 (tree_if_conversion): Delay update_ssa until after predicate
1248 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1250 PR tree-optimization/84178
1251 * tree-if-conv.c (release_bb_predicate): Remove the
1252 the assertion that the stmts have NULL use_ops.
1253 Discard the statements, asserting that they haven't
1254 yet been added to a BB.
1256 2018-03-08 Richard Biener <rguenther@suse.de>
1258 PR tree-optimization/84746
1259 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1260 (phi_translate): Pass in destination ANTIC_OUT set.
1261 (phi_translate_1): Likewise. For a simplified result lookup
1262 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1263 (phi_translate_set): Adjust.
1264 (do_pre_regular_insertion): Likewise.
1265 (do_pre_partial_partial_insertion): Likewise.
1267 2018-03-08 Martin Liska <mliska@suse.cz>
1269 PR gcov-profile/84735
1270 * doc/gcov.texi: Document usage of profile files.
1271 * gcov-io.h: Document changes in the format.
1273 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1277 * dwarf2out.c (struct dw_line_info_table): Update comments for
1279 (FORCE_RESET_NEXT_VIEW): New.
1280 (FORCE_RESETTING_VIEW_P): New.
1281 (RESETTING_VIEW_P): Check for -1 too.
1282 (ZERO_VIEW_P): Likewise.
1283 (new_line_info_table): Force-reset next view.
1284 (dwarf2out_begin_function): Likewise.
1285 (dwarf2out_source_line): Simplify zero_view_p initialization.
1286 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1287 view directly. Omit view when omitting .loc at line 0.
1289 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1291 PR tree-optimization/84740
1292 * tree-switch-conversion.c (process_switch): Call build_constructors
1293 only if info.phi_count is non-zero.
1295 PR tree-optimization/84739
1296 * tree-tailcall.c (find_tail_calls): Check call arguments against
1297 DECL_ARGUMENTS (current_function_decl) rather than
1298 DECL_ARGUMENTS (func) when checking for tail recursion.
1300 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1302 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1303 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1304 Volker Reichelt's entry and add entries for people that perform
1305 GCC fuzzy testing and report numerous bugs.
1307 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1310 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1311 readonly data in sdata, if that is disabled.
1312 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1313 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1314 -mreadonly-in-sdata option.
1316 2018-03-07 Martin Sebor <msebor@redhat.com>
1318 PR tree-optimization/84468
1319 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1320 basic block when looking for nul assignment.
1322 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1325 * except.h (output_function_exception_table): Adjust prototype.
1326 * except.c (output_function_exception_table): Remove FNNAME parameter
1327 and add SECTION parameter. Ouput one part of the table at a time.
1328 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1329 the first part of the exception table and emit unwind directives.
1330 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1331 (i386_pe_seh_cold_init): Likewise.
1332 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1333 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1334 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1335 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1336 * config/i386/winnt.c: Include except.h.
1337 (struct seh_frame_state): Add reg_offset, after_prologue and
1338 in_cold_section fields.
1339 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1340 (i386_pe_seh_cold_init): New function.
1341 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1342 to seh->in_cold_section.
1343 (seh_emit_push): Record the offset of the push.
1344 (seh_emit_save): Record the offet of the save.
1345 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1346 Test seh->after_prologue to disregard the epilogue.
1347 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1348 (i386_pe_end_cold_function): New function.
1350 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1353 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1354 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1357 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1359 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1360 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1363 * multiple_target.c: Include tree-inline.h and intl.h.
1364 (expand_target_clones): Diagnose and fail if node->definition and
1365 !tree_versionable_function_p (node->decl).
1367 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1369 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1371 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1372 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1373 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1375 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1378 * combine.c (try_combine): Use reg_or_subregno instead of handling
1379 just paradoxical SUBREGs and REGs.
1381 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1383 * config/arc/arc.c (arc_finalize_pic): Remove function.
1384 (arc_must_save_register): We use single base PIC register, remove
1385 checks to save/restore the PIC register.
1386 (arc_expand_prologue): Likewise.
1387 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1389 (arc_verify_short): Likewise.
1390 (arc_attr_type): Likewise.
1391 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1392 (walk_stores): Likewise.
1393 (arc_address_cost): Make it static.
1394 (arc_verify_short): Likewise.
1395 (branch_dest): Likewise.
1396 (arc_attr_type): Likewise.
1397 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1398 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1399 (arc_final_prescan_insn): Remove inserting the nops due to
1400 hardware hazards. It is done in reorg step.
1401 (insn_length_variant_t): Remove.
1402 (insn_length_parameters_t): Likewise.
1403 (arc_insn_length_parameters): Likewise.
1404 (arc_get_insn_variants): Likewise.
1405 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1407 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1410 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1413 PR tree-optimization/84687
1414 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1416 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1418 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1420 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1421 Rename to ppc_speculation_barrier.
1422 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1423 __builtin_ppc_speculation_barrier.
1425 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1428 * combine.c (combine_simplify_rtx): Don't try to simplify if
1429 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1432 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1434 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1435 to 32 bytes when compiling for POWER9.
1437 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1440 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1441 regparm >= 3 with no arg reg available also for calls with
1442 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1445 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1447 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1449 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1452 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1454 2018-03-05 Richard Biener <rguenther@suse.de>
1456 PR tree-optimization/84486
1457 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1458 When inserting a __builtin_assume_aligned call set the LHS
1459 SSA name alignment info accordingly.
1461 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1463 PR tree-optimization/84114
1464 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1465 Avoid reassociation of FLOAT_MODE addition.
1467 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1469 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1470 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1471 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1472 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1473 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1474 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1475 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1477 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1478 __builtin_ia32_wbinvd): New builtins.
1479 (SPECIAL_ARGS2): New.
1480 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1481 (SPECIAL_ARGS2): New.
1482 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1483 (ix86_valid_target_attribute_inner_p): Ditto.
1484 (ix86_init_mmx_sse_builtins): Add special_args2.
1485 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1486 TARGET_WBNOINVD_P): New.
1487 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1488 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1489 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1490 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1491 * config/i386/pconfigintrin.h: New file.
1492 * config/i386/wbnoinvdintrin.h: Ditto.
1493 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1494 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1496 2018-03-05 Richard Biener <rguenther@suse.de>
1498 PR tree-optimization/84670
1499 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1501 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1502 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1503 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1504 doesn't grow if all successors (recursively) were visited at least
1507 2018-03-05 Richard Biener <rguenther@suse.de>
1509 PR tree-optimization/84650
1510 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1511 if executed in the loop pipeline.
1513 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1515 * doc/configfiles.texi (Configuration Files): Move info about
1516 conditionalizing $target-protos.h to...
1517 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1518 differs from $target-protos.h.
1520 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1521 Chung-Ju Wu <jasonwucj@gmail.com>
1523 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1524 * config/nds32/nds32-multiple.md (setmemsi): Define.
1525 * config/nds32/nds32-memory-manipulation.c
1526 (nds32_gen_dup_4_byte_to_word_value): New.
1527 (emit_setmem_word_loop): New.
1528 (emit_setmem_byte_loop): New.
1529 (nds32_expand_setmem_loop): New.
1530 (nds32_expand_setmem_loop_v3m): New.
1531 (nds32_expand_setmem_unroll): New.
1532 (nds32_expand_setmem): New.
1534 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1535 Chung-Ju Wu <jasonwucj@gmail.com>
1537 * config/nds32/nds32-memory-manipulation.c
1538 (nds32_emit_load_store): New.
1539 (nds32_emit_post_inc_load_store): New.
1540 (nds32_emit_mem_move): New.
1541 (nds32_emit_mem_move_block): New.
1542 (nds32_expand_movmemsi_loop_unknown_size): New.
1543 (nds32_expand_movmemsi_loop_known_size): New.
1544 (nds32_expand_movmemsi_loop): New.
1545 (nds32_expand_movmemsi_unroll): New.
1546 (nds32_expand_movmemqi): Rename ...
1547 (nds32_expand_movmemsi): ... to this.
1548 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1549 (movmemsi): ... to this.
1550 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1551 (nds32_expand_movmemsi): ... to this.
1553 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1554 Monk Chiang <sh.chiang04@gmail.com>
1555 Chung-Ju Wu <jasonwucj@gmail.com>
1557 * config/nds32/nds32-protos.h
1558 (nds32_expand_load_multiple): New arguments.
1559 (nds32_expand_store_multiple): Ditto.
1560 (nds32_valid_multiple_load_store): Rename ...
1561 (nds32_valid_multiple_load_store_p): ... to this.
1562 * config/nds32/nds32-memory-manipulation.c
1563 (nds32_expand_load_multiple): Refine implementation.
1564 (nds32_expand_store_multiple): Ditto.
1565 * config/nds32/nds32-multiple.md
1566 (load_multiple): Update nds32_expand_load_multiple interface.
1567 (store_multiple): Update nds32_expand_store_multiple interface.
1568 * config/nds32/nds32-predicates.c
1569 (nds32_valid_multiple_load_store): Rename ...
1570 (nds32_valid_multiple_load_store_p): ... to this and refine
1572 * config/nds32/predicates.md
1573 (nds32_load_multiple_and_update_address_operation): New predicate.
1574 (nds32_store_multiple_and_update_address_operation): New predicate.
1576 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1577 Chung-Ju Wu <jasonwucj@gmail.com>
1579 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1580 (combo): New attribute.
1581 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1583 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1585 * config/nds32/nds32.opt: Change -mcmodel= default value.
1587 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1588 Monk Chiang <sh.chiang04@gmail.com>
1589 Chung-Ju Wu <jasonwucj@gmail.com>
1591 * config/nds32/constants.md (unspec_element): New enum.
1592 * config/nds32/constraints.md (Umw): New constraint.
1593 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1594 * config/nds32/nds32-intrinsic.md: Likewise.
1595 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1596 (nds32_valid_smw_lwm_base_p): New.
1597 (nds32_output_smw_single_word): New.
1598 (nds32_output_lmw_single_word): New.
1599 (nds32_expand_unaligned_load): New.
1600 (nds32_expand_unaligned_store): New.
1601 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1602 (nds32_output_smw_single_word): Declare.
1603 (nds32_output_lmw_single_word): Declare.
1604 (nds32_expand_unaligned_load): Declare.
1605 (nds32_expand_unaligned_store): Declare.
1606 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1607 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1608 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1609 NDS32_BUILTIN_UASTORE_DW.
1610 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1613 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1614 Kito Cheng <kito.cheng@gmail.com>
1615 Chung-Ju Wu <jasonwucj@gmail.com>
1617 * config/nds32/nds32-intrinsic.c
1618 (nds32_expand_builtin_null_ftype_reg): Delete.
1619 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1620 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1621 (nds32_read_argument): New.
1622 (nds32_legitimize_target): Ditto.
1623 (nds32_legitimize_argument): Ditto.
1624 (nds32_check_constant_argument): Ditto.
1625 (nds32_expand_unop_builtin): Ditto.
1626 (nds32_expand_unopimm_builtin): Ditto.
1627 (nds32_expand_binop_builtin): Ditto.
1628 (nds32_builtin_decl_impl): Ditto.
1629 (builtin_description): Ditto.
1630 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1631 (nds32_init_builtins_impl): Ditto.
1632 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1633 (nds32_builtin_decl): New.
1634 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1635 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1637 2018-03-02 Jeff Law <law@redhat.com>
1639 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1640 (redundant_insn, fill_simple_delay_slots): Likewise.
1641 (fill_slots_from_thread): Likewise.
1642 * resource.c (mark_referenced_resources): Likewise.
1643 (mark_set_resources, find_dead_or_set_registers): Likewise.
1645 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1647 * substring-locations.h (format_warning_va): Formatting fix for
1649 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1651 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1652 * substring-locations.c: Include intl.h.
1653 (format_warning_va): Turned into small wrapper around
1654 format_warning_n_va, renamed to ...
1655 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1656 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1658 (format_warning_at_substring_n): New function.
1659 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1660 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1661 format_warning_at_substring with just a shorter name instead of
1662 const function pointer.
1663 (fmtwarn_n): New function.
1664 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1665 appropriate, get rid of all the fmtstr temporaries, move conditionals
1666 with G_() wrapped string literals directly into fmtwarn arguments,
1667 cast dir.len to (int), formatting fixes.
1669 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1671 * doc/invoke.texi: Remove "Cilk Plus" references.
1673 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1674 Richard Biener <rguenther@suse.de>
1677 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1678 for error or warning attributes if CALL_FROM_THUNK_P is set.
1681 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1684 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1685 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1688 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1689 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1691 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1692 instead of -1U in last predictors element's probability member.
1694 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1697 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1698 arguments if they are comparable.
1700 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1702 PR tree-optimization/84634
1703 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1704 masks and masked_loop_p with a single loop_masks, making sure it's
1705 null for bb vectorization.
1707 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1709 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1710 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1711 loop->force_vectorize to check whether there is no alias.
1713 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1716 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1718 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1720 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1721 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1722 instead of a loop around prev_real_insn.
1723 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1727 * config/i386/i386.c (ix86_print_operand): Use conditional
1728 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1731 2018-03-02 Richard Biener <rguenther@suse.de>
1733 PR tree-optimization/84427
1734 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1735 (bitmap_set_subtract_values): Rewrite to handle multiple
1738 (prune_clobbered_mems): Likewise.
1739 (phi_translate): Take edge instead of pred/phiblock.
1740 (phi_translate_1): Likewise.
1741 (phi_translate_set): Likewise. Insert all translated
1742 exprs for a value into the set, keeping possibly multiple
1743 expressions per value.
1744 (compute_antic_aux): Adjust for phi_translate changes.
1745 When intersecting union the expressions and prune those
1746 not in the final value set, keeping possibly multiple
1747 expressions per value. Do not use value-insertion
1748 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1749 all expressions. Add verification that the value-sets
1750 only shrink during iteration.
1751 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1752 (do_pre_regular_insertion): Likewise.
1753 (do_pre_partial_partial_insertion): Likewise.
1755 2018-03-02 Richard Biener <rguenther@suse.de>
1758 * config/darwin.c (saved_debug_info_level): New static global.
1759 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1760 (darwin_asm_lto_end): Restore debug info generation settings.
1762 2018-03-01 Martin Liska <mliska@suse.cz>
1765 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1768 2018-03-01 Richard Biener <rguenther@suse.de>
1771 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1772 type annotation with LTO when debug was disabled at compile-time.
1774 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1776 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1778 (mips_final_postscan_insn): Likewise.
1780 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1782 PR rtl-optimization/84528
1783 * alias.c (init_alias_target): Add commentary.
1784 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1785 a unique base value if the frame pointer is not eliminated
1786 to the stack pointer.
1788 2018-03-01 Tom de Vries <tom@codesourcery.com>
1790 PR rtl-optimization/83327
1791 * lra-int.h (hard_regs_spilled_into): Declare.
1792 * lra.c (hard_regs_spilled_into): Define.
1793 (init_reg_info): Init hard_regs_spilled_into.
1794 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1795 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1796 (process_bb_lives): Handle hard_regs_spilled_into.
1797 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1798 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1800 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1802 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1803 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1804 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1805 * config/rs6000/aix72.h: New file.
1807 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1809 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1810 instead of warning_at with conditional singular and plural messages
1814 * stor-layout.c (update_alignment_for_field): For
1815 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1816 && !DECL_PACKED (field), do the alignment update, just use
1817 only desired_align instead of MAX (type_align, desired_align)
1819 (place_field): Don't do known_align < desired_align handling
1820 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1821 is non-NULL, instead do it after rli->prev_field handling and
1822 only if not within a bitfield word. For DECL_PACKED (field)
1823 use type_align of BITS_PER_UNIT.
1825 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1827 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1828 superfluous parentheses and trailing spaces.
1830 2018-02-28 Richard Biener <rguenther@suse.de>
1832 PR tree-optimization/84584
1833 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1834 SCoPs with fake exit edge.
1836 2018-02-28 Martin Liska <mliska@suse.cz>
1839 * timevar.c (timer::print): Fix format to properly print 100%
1842 2018-02-28 Richard Biener <rguenther@suse.de>
1845 * genmatch.c (capture_info::walk_match): Do not mark
1846 captured expressions without operands as expr_p given
1847 they act more like predicates and should be subject to
1848 "lost tail" side-effect preserving.
1850 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1852 PR rtl-optimization/81611
1853 * auto-inc-dec.c (attempt_change): Move dead note from
1854 mem_insn if it's the next use of regno
1855 (find_address): Take address use of reg holding
1856 non-incremented value. Add parm to limit search to the named
1858 (merge_in_block): Attempt to use a mem insn that is the next
1859 use of the original regno.
1861 2018-02-27 Martin Sebor <msebor@redhat.com>
1864 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1865 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1867 2018-02-27 Martin Sebor <msebor@redhat.com>
1869 PR translation/84207
1870 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1871 n argument to unsigned HOST_WIDE_INT.
1872 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1873 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1874 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1875 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1877 2018-02-27 Richard Biener <rguenther@suse.de>
1879 PR tree-optimization/84512
1880 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1881 Do not use the estimate returned from record_stmt_cost for
1882 the scalar iteration cost but sum properly using add_stmt_cost.
1884 2018-02-27 Richard Biener <rguenther@suse.de>
1886 PR tree-optimization/84466
1887 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1888 Adjust last change to less strictly validate use operands.
1890 2018-02-27 Martin Liska <mliska@suse.cz>
1892 PR gcov-profile/84548
1893 * gcov.c (process_file): Allow partial overlap and consider it
1894 also as group functions.
1895 (output_lines): Properly calculate range of lines for a group.
1897 2018-02-27 Martin Liska <mliska@suse.cz>
1899 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1900 'ggc' suffixes. Change first column width.
1901 (timer::print): Fix formatting of the column.
1903 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1905 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1906 preserve inline entry blocks for the sake of debug inline
1907 entry point markers alone.
1908 (remove_unused_locals): Suggest in comments a better place to
1909 force the preservation of inline entry blocks that are
1910 otherwise unused, but do not preserve them.
1912 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1914 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1916 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1919 * config/i386/constraints.md (Bs): Replace
1920 ix86_indirect_branch_register with
1921 TARGET_INDIRECT_BRANCH_REGISTER.
1923 * config/i386/i386.md (indirect_jump): Likewise.
1924 (tablejump): Likewise.
1925 (*sibcall_memory): Likewise.
1926 (*sibcall_value_memory): Likewise.
1927 Peepholes of indirect call and jump via memory: Likewise.
1928 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1929 (*sibcall_value_GOT_32): Likewise.
1930 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1931 (GOT_memory_operand): Likewise.
1932 (call_insn_operand): Likewise.
1933 (sibcall_insn_operand): Likewise.
1934 (GOT32_symbol_operand): Likewise.
1935 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1937 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1939 PR rtl-optimization/83496
1940 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1941 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1942 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1943 redundant insn, if any.
1944 (relax_delay_slots): Likewise.
1945 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1947 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1949 PR tree-optimization/83965
1950 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1951 that grouped statements are part of a reduction chain. Return
1952 true if the statement is not marked as a reduction itself but
1954 (vect_recog_dot_prod_pattern): Don't check whether the statement
1955 is part of a group here.
1956 (vect_recog_sad_pattern): Likewise.
1957 (vect_recog_widen_sum_pattern): Likewise.
1959 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1962 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1964 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1967 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1969 (ix86_output_indirect_function_return): New prototype.
1970 (ix86_split_simple_return_pop_internal): Likewise.
1971 * config/i386/i386.c (indirect_return_via_cx): New.
1972 (indirect_return_via_cx_bnd): Likewise.
1973 (indirect_thunk_name): Handle return va CX_REG.
1974 (output_indirect_thunk_function): Create alias for
1975 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1976 (ix86_output_indirect_jmp): Remove the bool argument.
1977 (ix86_output_indirect_function_return): New function.
1978 (ix86_split_simple_return_pop_internal): Likewise.
1979 * config/i386/i386.md (*indirect_jump): Don't pass false
1980 to ix86_output_indirect_jmp.
1981 (*tablejump_1): Likewise.
1982 (simple_return_pop_internal): Change it to define_insn_and_split.
1983 Call ix86_split_simple_return_pop_internal to split it for
1985 (simple_return_indirect_internal): Call
1986 ix86_output_indirect_function_return instead of
1987 ix86_output_indirect_jmp.
1989 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1992 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1993 memset and value initialization afterwards.
1995 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1997 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1999 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2002 * common/config/aarch64/aarch64-common.c
2003 (aarch_option_optimization_table[]): Switch
2004 off fomit-frame-pointer
2006 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2007 Chung-Ju Wu <jasonwucj@gmail.com>
2009 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2011 (store_multiple): Ditto.
2013 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2015 * config.gcc: Add --with-cpu support for nds32 target.
2016 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2017 * config/nds32/nds32.opt: Add -mcpu= option.
2019 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2021 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2022 isel=yes): Warn for these deprecated options.
2024 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2026 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2027 ISA_2_5_MASKS_EMBEDDED.
2029 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2031 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2032 p->max as pointers rather than using iterative_hash_expr.
2034 2018-02-23 Carl Love <cel@us.ibm.com>
2036 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2037 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2039 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2040 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2041 P8V_BUILTIN_VEC_VUNSIGNED2.
2043 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2046 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2047 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2049 * lra-constraints.c (curr_insn_transform): Set up
2050 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2052 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2053 * lra-eliminations.c (spill_pseudos): Ditto.
2054 (process_insn_for_elimination): Ditto.
2055 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2056 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2057 LRA_NON_CLOBBERED_ALT.
2059 2018-02-22 Martin Sebor <msebor@redhat.com>
2061 PR tree-optimization/84480
2062 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2063 to maybe_diag_stxncpy_trunc. Call it.
2064 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2065 from gimple_fold_builtin_strcpy. Print inlining stack.
2066 (handle_builtin_stxncpy): Print inlining stack.
2067 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2069 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2072 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2073 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2074 and -fcheck-pointer-bounds are used together.
2075 (indirect_thunk_prefix): New enum.
2076 (indirect_thunk_need_prefix): New function.
2077 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2078 "_nt" instead of "_bnd" for NOTRACK prefix.
2079 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2080 (output_indirect_thunk_function): Likewise.
2082 (ix86_code_end): Update output_indirect_thunk_function calls.
2083 (ix86_output_indirect_branch_via_reg): Replace
2084 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2085 (ix86_output_indirect_branch_via_push): Likewise.
2086 (ix86_output_function_return): Likewise.
2087 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2088 incompatible with -fcf-protection=branch and
2089 -fcheck-pointer-bounds.
2091 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2094 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2095 Change gcc_assert call to output_operand_lossage.
2097 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2099 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2101 2018-02-22 DJ Delorie <dj@redhat.com>
2102 Sebastian Perta <sebastian.perta@renesas.com>
2103 Oleg Endo <olegendo@gcc.gnu.org>
2105 * config/rx/rx.c (rx_rtx_costs): New function.
2106 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2108 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2110 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2112 2018-02-22 Martin Liska <mliska@suse.cz>
2115 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2116 Add "native" as a possible value.
2118 2018-02-22 Martin Liska <mliska@suse.cz>
2121 * config/i386/i386.c (ix86_option_override_internal):
2122 Add "native" as a possible value for -march and -mtune.
2124 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2127 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2128 to all type variants.
2130 PR tree-optimization/84503
2131 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2132 width as info->bitpos + info->bitsize - start.
2133 (merged_store_group::merge_overlapping): Simplify width computation.
2134 (check_no_overlap): New function.
2135 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2136 start + width and last_order of the group, fail if check_no_overlap
2138 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2139 to group if check_no_overlap fails.
2141 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2143 * config/rs6000/altivec.md: Delete contraint arguments to
2144 define_expand, define_split, and define_peephole2, and in
2145 define_insn_and_split if always unused.
2146 * config/rs6000/darwin.md: Ditto.
2147 * config/rs6000/dfp.md: Ditto.
2148 * config/rs6000/rs6000.md: Ditto.
2149 * config/rs6000/sync.md: Ditto.
2150 * config/rs6000/vector.md: Ditto.
2151 * config/rs6000/vsx.md: Ditto.
2153 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2155 * config/rs6000/altivec.md: Write output control strings as braced
2156 blocks instead of double-quoted strings.
2157 * config/rs6000/darwin.md: Ditto.
2158 * config/rs6000/rs6000.md: Ditto.
2159 * config/rs6000/vector.md: Ditto.
2160 * config/rs6000/vsx.md: Ditto.
2162 2018-02-21 Jason Merrill <jason@redhat.com>
2164 PR c++/84314 - ICE with templates and fastcall attribute.
2165 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2167 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2169 * ipa-cp.c (determine_versionability): Fix comment typos.
2171 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2174 * ipa-cp.c (determine_versionability): Do not version functions caling
2177 2018-02-21 Martin Liska <mliska@suse.cz>
2180 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2181 Add "native" as a possible value.
2182 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2183 the macro when native cpu detection is available.
2185 2018-02-21 Martin Liska <mliska@suse.cz>
2188 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2189 Add "native" as a possible value.
2190 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2191 when native cpu detection is available.
2193 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2194 Martin Sebor <msebor@redhat.com>
2196 PR tree-optimization/84478
2197 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2199 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2200 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2201 support which is conservatively correct, for 2 only stay conservative
2202 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2203 argument to the 2 argument get_range_strlen, adjust 6 arg
2204 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2206 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2207 (gimple_fold_builtin_strlen): Pass true as last argument to
2210 2018-02-20 Martin Sebor <msebor@redhat.com>
2213 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2214 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2215 (builtin_memref::builtin_memref): Factor out parts into
2216 set_base_and_offset and call it.
2218 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2221 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2222 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2223 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2224 search at the associated MODE_INT.
2226 2018-02-20 Jeff Law <law@redhat.com>
2229 PR tree-optimization/81592
2231 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2232 for range data rather than using global data.
2233 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2234 range data rather than using global data.
2235 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2236 pass it to children as needed.
2237 (struct directive::fmtresult): Similarly.
2238 (struct directive::set_width): Similarly.
2239 (struct directive::set_precision): Similarly.
2240 (format_integer, format_directive, parse_directive): Similarly.
2241 (format_none): Accept unnamed vr_values parameter.
2242 (format_percent, format_floating, format_character): Similarly.
2243 (format_string, format_plain): Similarly.
2244 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2245 the EVRP range analyzer for range data rather than using global data.
2246 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2247 gimple-ssa-evrp-analyze.h
2248 (class sprintf_dom_walker): Add after_dom_children member function.
2249 Add evrp_range_analyzer member.
2250 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2251 range analyzer as needed.
2252 (sprintf_dom_walker::after_dom_children): New member function.
2253 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2255 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2256 (evrp_range_analyzer::pop_to_marker): Likewise.
2258 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2260 PR tree-optimization/84419
2261 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2262 with the required type if its current type is compatible but
2265 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2268 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2269 after vectorization.
2271 2018-02-20 Martin Liska <mliska@suse.cz>
2274 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2275 possible values if we don't have a hint.
2277 2018-02-20 Martin Liska <mliska@suse.cz>
2281 * final.c (shorten_branches): Build align_tab array with one
2283 * opts.c (finish_options): Add alignment option limit check.
2284 (MAX_CODE_ALIGN): Likewise.
2285 (MAX_CODE_ALIGN_VALUE): Likewise.
2286 * doc/invoke.texi: Document maximum allowed option value for
2287 all -falign-* options.
2289 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2292 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2293 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2294 * var-tracking.c (emit_note_insn_var_location): Remove all references
2295 to NOTE_INSN_CALL_ARG_LOCATION.
2296 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2297 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2298 Use copy_rtx_if_shared.
2299 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2300 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2301 (dwarf2out_var_location): Remove handling of
2302 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2304 * final.c (final_scan_insn): Remove all references to
2305 NOTE_INSN_CALL_ARG_LOCATION.
2306 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2307 before dumping final insns.
2308 * except.c (emit_note_eh_region_end): Remove all references to
2309 NOTE_INSN_CALL_ARG_LOCATION.
2310 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2311 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2312 * config/arc/arc.c (hwloop_optimize): Likewise.
2313 * config/arm/arm.c (create_fix_barrier): Likewise.
2314 * config/s390/s390.c (s390_chunkify_start): Likewise.
2315 * config/sh/sh.c (find_barrier): Likewise.
2316 * config/i386/i386.c (rest_of_insert_endbranch,
2317 ix86_seh_fixup_eh_fallthru): Likewise.
2318 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2319 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2320 * config/frv/frv.c (frv_function_prologue): Likewise.
2321 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2323 (note_outside_basic_block_p): Remove all references to
2324 NOTE_INSN_CALL_ARG_LOCATION.
2325 * gengtype.c (adjust_field_rtx_def): Likewise.
2326 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2328 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2329 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2332 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2335 PR tree-optimization/84452
2336 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2337 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2340 2018-02-19 Martin Liska <mliska@suse.cz>
2343 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2345 2018-02-19 Martin Liska <mliska@suse.cz>
2346 Richard Sandiford <richard.sandiford@linaro.org>
2348 PR tree-optimization/82491
2349 * gimple-fold.c (get_base_constructor): Make earlier bail out
2352 2018-02-19 Carl Love <cel@us.ibm.com>
2354 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2355 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2357 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2358 P8V_BUILTIN_VEC_NEG.
2360 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2362 * config/rl78/rl78.md (movdf): New define expand.
2364 2018-02-19 Martin Liska <mliska@suse.cz>
2367 * doc/invoke.texi: Fix typo.
2368 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2370 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2372 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2373 handle rs6000_single_float and rs6000_double_float specially for
2376 2018-02-16 Jeff Law <law@redhat.com>
2378 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2381 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2384 * ipa-inline.c (inline_small_functions): Fix a typo.
2386 2018-02-16 Nathan Sidwell <nathan@acm.org>
2388 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2390 2018-02-16 Carl Love <cel@us.ibm.com>
2392 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2393 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2394 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2395 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2396 expansion to P8V_BUILTIN_VEC_FLOAT2.
2398 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2400 PR rtl-optimization/70023
2401 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2402 src_regno into account.
2404 2018-02-16 Carl Love <cel@us.ibm.com>
2406 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2407 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2408 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2409 * config/rs6000/rs6000.c: Remove case statements for
2410 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2411 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2412 and P9V_BUILTIN_VEC_VINSERT4B.
2413 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2414 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2415 * config/rs6000/vsx.md:
2416 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2419 2018-02-16 Carl Love <cel@us.ibm.com>
2421 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2423 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2425 * config/rs6000/rs6000-c.c: Add the definitions for
2426 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2427 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2428 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2429 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2430 definition for insert4b and define insn *insert3b_internal.
2431 * doc/extend.texi: Add documentation for vec_extract4b.
2433 2018-02-16 Nathan Sidwell <nathan@acm.org>
2435 * doc/extend.texi (Backwards Compatibility): Mention friend
2436 injection. Note for-scope is deprecated.
2437 * doc/invoke.texi (-ffriend-injection): Deprecate.
2439 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2441 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2442 that moved to I2, also allow destinations that are a paradoxical
2443 subreg (instead of a normal reg).
2445 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2448 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2451 2018-02-16 Richard Biener <rguenther@suse.de>
2453 PR tree-optimization/84037
2454 PR tree-optimization/84016
2456 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2457 Adjust vec_construct for the fact we need additional higher latency
2458 128bit inserts for AVX256 and AVX512 vector builds.
2459 (ix86_add_stmt_cost): Scale vector construction cost for
2462 2018-02-16 Richard Biener <rguenther@suse.de>
2464 PR tree-optimization/84417
2465 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2466 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2467 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2469 2018-02-16 Martin Liska <mliska@suse.cz>
2472 * internal-fn.def (ASAN_CHECK): Set proper flags.
2473 (ASAN_MARK): Likewise.
2475 2018-02-16 Julia Koval <julia.koval@intel.com>
2477 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2478 from PTA_CANNONLAKE.
2480 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2483 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2484 Use ++iter rather than iter++ for std::list iterators.
2485 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2486 defer deleting them until all nodes in the forest are processed. Do
2487 free even leaf nodes. Change to_process into auto_vec.
2490 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2491 * vec.h (vec_default_construct): Use memset instead of placement new
2492 if BROKEN_VALUE_INITIALIZATION is defined.
2493 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2494 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2497 PR rtl-optimization/83723
2498 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2499 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2500 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2502 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2504 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2506 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2508 PR rtl-optimization/81443
2509 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2510 from inner REGs to paradoxical SUBREGs.
2512 2018-02-16 Richard Biener <rguenther@suse.de>
2514 PR tree-optimization/84399
2515 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2516 For operands we can analyze at their definition make sure we can
2517 analyze them at each use as well.
2519 2018-02-16 Richard Biener <rguenther@suse.de>
2521 PR tree-optimization/84190
2522 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2523 volatile accesses if the decl isn't volatile.
2525 2018-02-15 Jason Merrill <jason@redhat.com>
2527 PR c++/84314 - ICE with templates and fastcall attribute.
2528 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2529 TYPE_CANONICAL on an existing type.
2531 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2533 PR tree-optimization/84383
2534 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2535 dstoff nor call operand_equal_p if dstbase is NULL.
2537 PR tree-optimization/84334
2538 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2539 also a CONSTANT_CLASS_P, punt.
2541 2018-02-14 Jim Wilson <jimw@sifive.com>
2543 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2544 first SMALL_OPERAND check. New local min_second_step. Move assert
2545 to where locals are set. Add TARGET_RVC support.
2546 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2548 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2550 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2552 2018-02-14 Martin Sebor <msebor@redhat.com>
2554 PR tree-optimization/83698
2555 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2556 arrays constrain the offset range to their bounds.
2557 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2558 (builtin_access::overlap): Avoid setting the size of overlap if it's
2560 (maybe_diag_overlap): Also consider arrays when deciding what values
2561 of offsets to include in diagnostics.
2563 2018-02-14 Martin Sebor <msebor@redhat.com>
2566 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2567 that correspond to the kind of a declaration.
2569 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2572 * config/pa/pa.md: Load address of PIC label using the linkage table
2573 if the label is nonlocal.
2575 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2577 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2578 warning message if user requests -maltivec=be.
2579 * doc/invoke.texi: Document deprecation of -maltivec=be.
2581 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2584 * config/rs6000/rs6000-c.c: Update definitions for
2585 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2586 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2588 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2591 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2592 add _get_ssp intrinsics. Remove argument from
2593 __builtin_ia32_rdssp[d|q].
2594 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2595 * config/i386/i386-builtin.def: Remove argument from
2596 __builtin_ia32_rdssp[d|q].
2597 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2598 ix86_expand_special_args_builtin for _rdssp[d|q].
2599 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2600 Clear register before usage.
2601 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2602 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2604 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2606 PR tree-optimization/84357
2607 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2608 operand 1 of an ARRAY_REF too.
2610 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2613 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2614 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2616 (set_of_reg): New struct.
2617 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2618 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2619 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2621 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2622 Split into bitclr, bitset, bitinvert patterns if appropriate.
2623 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2624 use rx_fuse_in_memory_bitop.
2625 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2626 to named insn, correct maximum insn length.
2628 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2631 * machmode.def: Define a complex mode for PARTIAL_INT.
2632 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2634 * doc/rtl.texi: Document CSPImode.
2635 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2637 (msp430_hard_regno_nregs_with_padding): Likewise.
2639 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2642 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2644 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2646 PR rtl-optimization/84169
2647 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2648 we generated a parallel as new i3 and we split that to new i2 and i3
2649 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2650 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2651 those to i2, not i1. Partially rewrite this scan code.
2653 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2656 * stor-layout.c (place_field): For variable length fields, adjust
2657 offset_align afterwards not just based on the field's alignment,
2658 but also on the size.
2661 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2662 of exps and logs in the use_exp2 case.
2664 2018-02-13 Jeff Law <law@redhat.com>
2666 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2669 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2672 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2676 * common.opt (gas-loc-support, gas-locview-support): New.
2677 (ginline-points, ginternal-reset-location-views): New.
2678 * doc/invoke.texi: Document them. Use @itemx where intended.
2679 (gvariable-location-views): Adjust.
2680 * target.def (reset_location_view): New.
2681 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2682 (TARGET_RESET_LOCATION_VIEW): New.
2683 * doc/tm.texi: Rebuilt.
2684 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2685 (dwarf2out_default_as_locview_support): New.
2686 (output_asm_line_debug_info): Use option variables.
2687 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2688 (output_loc_list): Likewise.
2689 (add_high_low_attributes): Check option variables.
2690 Don't output entry view attribute in strict mode.
2691 (gen_inlined_subroutine_die): Check option variables.
2692 (dwarf2out_inline_entry): Likewise.
2693 (init_sections_and_labels): Likewise.
2694 (dwarf2out_early_finish): Likewise.
2695 (maybe_reset_location_view): New, from...
2696 (dwarf2out_var_location): ... here. Call it.
2697 * debug.h (dwarf2out_default_as_loc_support): Declare.
2698 (dwarf2out_default_as_locview_support): Declare.
2699 * hooks.c (hook_int_rtx_insn_0): New.
2700 * hooks.h (hook_int_rtx_insn_0): Declare.
2701 * toplev.c (process_options): Take -gas-loc-support and
2702 -gas-locview-support from dwarf2out. Enable
2703 -gvariable-location-views by default only with locview
2704 assembler support. Enable -ginternal-reset-location-views by
2705 default only if the target defines the corresponding hook.
2706 Enable -ginline-points by default if location views are
2707 enabled; force it disabled if statement frontiers are
2709 * tree-inline.c (expand_call_inline): Check option variables.
2710 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2712 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2714 PR tree-optimization/84321
2715 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2716 handling. Also check whether the anti-range contains any values
2717 that satisfy the mask; switch to a VR_RANGE if not.
2719 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2722 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2724 2018-02-13 Martin Jambor <mjambor@suse.cz>
2727 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2728 of call statements, also set location of a load to a temporary.
2730 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2732 * config/rl78/rl78.c (add_vector_labels): New function.
2733 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2734 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2735 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2736 which checks that no arguments are passed.
2737 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2738 * doc/extend.texi: Documentation for the new attribute.
2740 2018-02-13 Andreas Schwab <schwab@suse.de>
2742 * config/riscv/linux.h (CPP_SPEC): Define.
2744 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2747 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2748 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2749 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2750 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2751 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2752 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2753 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2754 temporarily for AES and PCLMUL builtins.
2756 PR tree-optimization/84339
2757 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2758 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2762 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2763 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2764 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2766 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2768 * omp-simd-clone.h: New file.
2769 * omp-simd-clone.c: Include omp-simd-clone.h.
2770 (expand_simd_clones): No longer static.
2771 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2772 cgraph.h and omp-simd-clone.h.
2773 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2774 (vect_recog_widen_shift_pattern): Formatting fix.
2775 (vect_pattern_recog_1): Don't check optab for calls.
2778 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2779 operands[2] into a REG before using gen_lowpart on it.
2781 2018-02-12 Jeff Law <law@redhat.com>
2784 * config/sh/sh.c (find_barrier): Consider a sibling call
2787 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2788 successfully back substituting a reg.
2790 2018-02-12 Richard Biener <rguenther@suse.de>
2792 PR tree-optimization/84037
2793 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2794 parameter, move visited init to caller.
2795 (vect_slp_analyze_operations): Separate cost from validity
2796 check, initialize visited once for all instances.
2797 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2799 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2801 (vect_model_promotion_demotion_cost): Likewise.
2802 (vectorizable_bswap): Guard cost modeling with !slp_node
2803 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2805 (vectorizable_call): Likewise.
2806 (vectorizable_conversion): Likewise.
2807 (vectorizable_assignment): Likewise.
2808 (vectorizable_shift): Likewise.
2809 (vectorizable_operation): Likewise.
2810 (vectorizable_store): Likewise.
2811 (vectorizable_load): Likewise.
2812 (vectorizable_condition): Likewise.
2813 (vectorizable_comparison): Likewise.
2815 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2818 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2819 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2820 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2822 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2825 * params.def (inline-min-speedup): Increase from 8 to 15.
2826 (max-inline-insns-auto): Decrease from 40 to 30.
2827 * ipa-split.c (consider_split): Add some buffer for function to
2828 be considered inlining candidate.
2829 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2832 2018-02-12 Richard Biener <rguenther@suse.de>
2834 PR tree-optimization/84037
2835 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2836 matched stmts if we cannot swap the non-matched ones.
2838 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2840 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2841 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2842 _mm_maskz_scalef_round_ss): New intrinsics.
2843 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2844 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2845 __builtin_ia32_scalefss_round): Remove.
2846 (__builtin_ia32_scalefsd_mask_round,
2847 __builtin_ia32_scalefss_mask_round): New intrinsics.
2848 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2849 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2850 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2851 "<round_constraint>")): Changed to ...
2852 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2853 "<round_scalar_constraint>")): ... this.
2854 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2855 %0, %1, %2<round_op3>}"): Changed to ...
2856 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2857 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2858 %2<round_scalar_mask_op3>}"): ... this.
2859 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2861 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2863 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2864 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2865 (_mm_maskz_sqrt_round_ss): New intrinsics.
2866 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2867 (__builtin_ia32_sqrtsd_mask_round)
2868 (__builtin_ia32_sqrtss_mask_round): New builtins.
2869 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2870 (__builtin_ia32_sqrtss_round): Remove.
2871 (__builtin_ia32_sqrtsd_mask_round)
2872 (__builtin_ia32_sqrtss_mask_round): New builtins.
2873 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2874 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2875 ((match_operand:VF_128 1 "vector_operand"
2876 "xBm,<round_constraint>")): Changed to ...
2877 ((match_operand:VF_128 1 "vector_operand"
2878 "xBm,<round_scalar_constraint>")): ... this.
2879 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2880 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2881 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2882 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2883 %<iptr>1<round_scalar_mask_op3>}): ... this.
2884 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2885 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2887 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2890 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2891 Cast vec_cmpeq result to correct type.
2892 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2893 Cast vec_cmpgt result to correct type.
2895 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2897 * final.c (final_scan_insn_1): Renamed from...
2898 (final_scan_insn): ... this. New wrapper, to recover
2899 seen from the outermost call in recursive ones.
2900 * config/sparc/sparc.c (output_return): Drop seen from call.
2901 (output_sibcall): Likewise.
2902 * config/visium/visium.c (output_branch): Likewise.
2904 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2906 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2909 2018-02-10 Alan Modra <amodra@gmail.com>
2912 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2913 Specify LR as an input.
2915 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2918 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2919 remove_member_access_dummy_vars): New functions.
2920 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2921 lower_omp_1, execute_lower_omp): Use them.
2923 PR rtl-optimization/84308
2924 * shrink-wrap.c (spread_components): Release todo vector.
2926 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2928 PR rtl-optimization/57193
2929 * ira-color.c (struct allocno_color_data): Add member
2930 conflict_allocno_hard_prefs.
2931 (update_conflict_allocno_hard_prefs): New.
2932 (bucket_allocno_compare_func): Add a preference based on
2933 conflict_allocno_hard_prefs.
2934 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2935 (color_allocnos): Remove a dead code. Initiate
2936 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2938 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2941 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2942 constraint from =wa to wa. Avoid a subreg on the output operand,
2943 instead use a pseudo and subreg it in a move.
2944 (p9_xxbrd_<mode>): Changed to ...
2945 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2946 (p9_xxbrd_v2df): New expander.
2947 (p9_xxbrw_<mode>): Changed to ...
2948 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2949 (p9_xxbrw_v4sf): New expander.
2951 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2953 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2955 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2958 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2959 multiply in 32-bit mode.
2960 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2961 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2964 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2966 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2967 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2968 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2969 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2971 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2974 * dwarf2out.c (is_trivial_indirect_ref): New function.
2975 (dwarf2out_late_global_decl): Do not generate a location
2976 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2977 and that are not defined in the current unit.
2979 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2981 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2982 instead of a libcall for UNORDERED.
2984 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2987 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2988 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2990 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2993 * config/s390/s390.c (s390_set_current_function): Invoke
2994 s390_indirect_branch_settings also if fndecl didn't change.
2996 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2998 * config/rs6000/rs6000.md (blockage): Set length to zero.
3000 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3002 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3004 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3007 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3008 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3012 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3013 PARALLEL incoming that failed vt_get_decl_and_offset check.
3016 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3017 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3019 (get_variable_section): For decls in named .bss* sections pass true as
3020 second argument to bss_initializer_p.
3022 2018-02-09 Marek Polacek <polacek@redhat.com>
3023 Jakub Jelinek <jakub@redhat.com>
3026 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3027 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3028 Sync some changes from cxx_fold_indirect_ref.
3030 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3032 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3034 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3035 (BLOCK_INLINE_ENTRY_LABEL): New.
3036 (dwarf2out_var_location): Disregard inline entry markers.
3037 (inline_entry_data): New struct.
3038 (inline_entry_data_hasher): New hashtable type.
3039 (inline_entry_data_hasher::hash): New.
3040 (inline_entry_data_hasher::equal): New.
3041 (inline_entry_data_table): New variable.
3042 (add_high_low_attributes): Add DW_AT_entry_pc and
3043 DW_AT_GNU_entry_view attributes if a pending entry is found
3044 in inline_entry_data_table. Add old entry_pc attribute only
3045 if debug nonbinding markers are disabled.
3046 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3047 markers are enabled.
3048 (block_within_block_p, dwarf2out_inline_entry): New.
3049 (dwarf2out_finish): Check that no entries remained in
3050 inline_entry_data_table.
3051 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3052 (final_scan_insn, notice_source_line): Likewise.
3053 (rest_of_clean_state): Skip inline entry markers.
3054 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3056 * gimple.c (gimple_build_debug_inline_entry): New.
3057 * gimple.h (enum gimple_debug_subcode): Add
3058 GIMPLE_DEBUG_INLINE_ENTRY.
3059 (gimple_build_debug_inline_entry): Declare.
3060 (gimple_debug_inline_entry_p): New.
3061 (gimple_debug_nonbind_marker_p): Adjust.
3062 * insn-notes.def (INLINE_ENTRY): New.
3063 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3064 inline entry marker notes.
3065 (print_insn): Likewise.
3066 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3067 (INSN_DEBUG_MARKER_KIND): Likewise.
3068 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3069 * tree-inline.c (expand_call_inline): Build and insert
3070 debug_inline_entry stmt.
3071 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3072 inline entry blocks early, if nonbind markers are enabled.
3073 (dump_scope_block): Dump fragment info.
3074 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3075 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3076 (gimple_build_debug_inline_entry): New.
3077 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3078 Enable/disable inline entry points too.
3079 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3080 (DEBUG_INSN): Describe inline entry markers.
3082 * common.opt (gvariable-location-views): New.
3083 (gvariable-location-views=incompat5): New.
3084 * config.in: Rebuilt.
3085 * configure: Rebuilt.
3086 * configure.ac: Test assembler for view support.
3087 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3088 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3089 * dwarf2out.c (var_loc_view): New typedef.
3090 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3091 (dwarf2out_locviews_in_attribute): New.
3092 (dwarf2out_locviews_in_loclist): New.
3093 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3094 (enum dw_line_info_opcode): Add LI_adv_address.
3095 (struct dw_line_info_table): Add view.
3096 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3097 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3098 (zero_view_p): New variable.
3099 (ZERO_VIEW_P): New macro.
3100 (output_asm_line_debug_info): New.
3101 (struct var_loc_node): Add view.
3102 (add_AT_view_list, AT_loc_list): New.
3103 (add_var_loc_to_decl): Add view param. Test it against last.
3104 (new_loc_list): Add view params. Record them.
3105 (AT_loc_list_ptr): Handle loc and view lists.
3106 (view_list_to_loc_list_val_node): New.
3107 (print_dw_val): Handle dw_val_class_view_list.
3108 (size_of_die): Likewise.
3109 (value_format): Likewise.
3110 (loc_list_has_views): New.
3111 (gen_llsym): Set vl_symbol too.
3112 (maybe_gen_llsym, skip_loc_list_entry): New.
3113 (dwarf2out_maybe_output_loclist_view_pair): New.
3114 (output_loc_list): Output view list or entries too.
3115 (output_view_list_offset): New.
3116 (output_die): Handle dw_val_class_view_list.
3117 (output_dwarf_version): New.
3118 (output_compilation_unit_header): Use it.
3119 (output_skeleton_debug_sections): Likewise.
3120 (output_rnglists, output_line_info): Likewise.
3121 (output_pubnames, output_aranges): Update version comments.
3122 (output_one_line_info_table): Output view numbers in asm comments.
3123 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3124 Call maybe_gen_llsym.
3125 (loc_list_from_tree_1): Adjust.
3126 (add_AT_location_description): Create view list attribute if
3127 needed, check it's absent otherwise.
3128 (convert_cfa_to_fb_loc_list): Adjust.
3129 (maybe_emit_file): Call output_asm_line_debug_info for test.
3130 (dwarf2out_var_location): Reset views as needed. Precompute
3131 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3132 attribute. Set view.
3133 (new_line_info_table): Reset next view.
3134 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3135 (dwarf2out_source_line): Likewise. Output view resets and labels to
3136 the assembler, or select appropriate line info opcodes.
3137 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3138 (optimize_string_length): Catch it. Adjust.
3139 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3140 dw_val_class_view_list, and remove it if no longer needed.
3141 (hash_loc_list): Hash view numbers.
3142 (loc_list_hasher::equal): Compare them.
3143 (optimize_location_lists): Check whether a view list symbol is
3144 needed, and whether the locview attribute is present, and
3145 whether they match. Remove the locview attribute if no longer
3147 (index_location_lists): Call skip_loc_list_entry for test.
3148 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3149 Use output_dwarf_version.
3150 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3151 (struct dw_val_node): Add val_view_list.
3152 * final.c (SEEN_NEXT_VIEW): New.
3153 (set_next_view_needed): New.
3154 (clear_next_view_needed): New.
3155 (maybe_output_next_view): New.
3156 (final_start_function): Rename to...
3157 (final_start_function_1): ... this. Take pointer to FIRST,
3158 add SEEN parameter. Emit param bindings in the initial view.
3159 (final_start_function): Reintroduce SEEN-less interface.
3160 (final): Rename to...
3161 (final_1): ... this. Take SEEN parameter. Output final pending
3162 next view at the end.
3163 (final): Reintroduce seen-less interface.
3164 (final_scan_insn): Output pending next view before switching
3165 sections or ending a block. Mark the next view as needed when
3166 outputting variable locations. Notify debug backend of section
3167 changes, and of location view changes.
3168 (rest_of_handle_final): Adjust.
3169 * toplev.c (process_options): Autodetect value for debug variable
3170 location views option. Warn on incompat5 without -gdwarf-5.
3171 * doc/invoke.texi (gvariable-location-views): New.
3172 (gvariable-location-views=incompat5): New.
3173 (gno-variable-location-views): New.
3175 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3177 PR tree-optimization/84136
3178 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3179 that the result of find_edge is non-NULL.
3181 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3184 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3185 storing integer register in SImode. Fix cost of 256 and 512
3186 byte aligned SSE register store.
3188 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3190 * config/i386/i386.c (ix86_multiplication_cost): Fix
3191 multiplication cost for TARGET_AVX512DQ.
3193 2018-02-08 Marek Polacek <polacek@redhat.com>
3195 PR tree-optimization/84238
3196 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3199 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3201 PR tree-optimization/84265
3202 * tree-vect-stmts.c (vectorizable_store): Don't treat
3203 VMAT_CONTIGUOUS accesses as grouped.
3204 (vectorizable_load): Likewise.
3206 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3208 PR tree-optimization/81635
3209 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3210 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3211 (test_round_for_mask): New functions.
3212 (wide_int_cc_tests): Call test_round_for_mask.
3213 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3214 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3215 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3216 range returned by get_range_info.
3218 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3221 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3222 * symtab.c: Include builtins.h
3223 (symtab_node::output_to_lto_symbol_table_p): Move here
3224 from lto-streamer-out.c:output_symbol_p.
3225 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3226 (output_symbol_p): Move all logic to symtab.c
3227 (produce_symtab): Update.
3229 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3231 * config/s390/s390-opts.h (enum indirect_branch): Define.
3232 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3233 (s390_indirect_branch_via_thunk)
3234 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3235 (enum s390_indirect_branch_type): Define.
3236 * config/s390/s390.c (struct s390_frame_layout, struct
3237 machine_function): Remove.
3238 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3239 (indirect_branch_table_label_no, indirect_branch_table_name):
3241 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3242 (enum s390_indirect_branch_option): Define.
3243 (s390_return_addr_from_memory): New function.
3244 (s390_handle_string_attribute): New function.
3245 (s390_attribute_table): Add new attribute handler.
3246 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3247 (s390_indirect_branch_via_thunk): New function.
3248 (s390_indirect_branch_via_inline_thunk): New function.
3249 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3250 sibling call optimization for non z10 compiles.
3251 (s390_emit_call): Force indirect branch target to be a single
3252 register. Add r1 clobber for non-z10 compiles.
3253 (s390_emit_epilogue): Emit return jump via return_use expander.
3254 (s390_reorg): Handle JUMP_INSNs as execute targets.
3255 (s390_option_override_internal): Perform validity checks for the
3256 new command line options.
3257 (s390_indirect_branch_attrvalue): New function.
3258 (s390_indirect_branch_settings): New function.
3259 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3260 (s390_output_indirect_thunk_function): New function.
3261 (s390_code_end): Implement target hook.
3262 (s390_case_values_threshold): Implement target hook.
3263 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3265 * config/s390/s390.h (struct s390_frame_layout)
3266 (struct machine_function): Move here from s390.c.
3267 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3268 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3269 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3270 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3271 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3272 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3273 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3274 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3275 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3276 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3277 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3278 (mnemonic attribute): Add values which aren't recognized
3280 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3281 pattern for branch conversion. Fix mnemonic attribute.
3282 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3283 indirect branch via thunk if requested.
3284 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3285 ("*indirect_jump"): Disable for branch conversion using out of
3287 ("indirect_jump_via_thunk<mode>_z10")
3288 ("indirect_jump_via_thunk<mode>")
3289 ("indirect_jump_via_inlinethunk<mode>_z10")
3290 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3291 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3292 ("casesi_jump_via_inlinethunk<mode>_z10")
3293 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3294 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3295 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3296 ("*indirect2_jump"): Disable for branch conversion.
3297 ("casesi_jump"): Turn into expander and expand patterns for branch
3299 ("return_use"): New expander.
3300 ("*return"): Emit return via thunk and rename it to ...
3301 ("*return<mode>"): ... this one.
3302 * config/s390/s390.opt: Add new options and and enum for the
3305 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3307 * lra-constraints.c (match_reload): Unconditionally use
3308 gen_lowpart_SUBREG, rather than selecting between that
3309 and equivalent gen_rtx_SUBREG code.
3311 2018-02-08 Richard Biener <rguenther@suse.de>
3313 PR tree-optimization/84233
3314 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3315 changed flag instead of boguously re-using phi_inserted.
3317 2018-02-08 Martin Jambor <mjambor@suse.cz>
3319 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3320 static local variables.
3322 2018-02-08 Richard Biener <rguenther@suse.de>
3324 PR tree-optimization/84278
3325 * tree-vect-stmts.c (vectorizable_store): When looking for
3326 smaller vector types to perform grouped strided loads/stores
3327 make sure the mode is supported by the target.
3328 (vectorizable_load): Likewise.
3330 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3332 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3333 Increase LDP/STP opportunities by adding adjacent callee-saves.
3335 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3337 PR rtl-optimization/84068
3338 PR rtl-optimization/83459
3339 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3341 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3343 PR tree-optimization/84224
3344 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3345 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3348 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3351 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3352 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3353 count, remove one USE.
3355 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3357 * doc/install.texi (Configuration): Document the
3358 --with-long-double-format={ibm,ieee} PowerPC configuration
3362 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3363 Convert from define_expand to be define_insn_and_split. Rework
3364 float/double/_Float128 conversions to QI/HI/SImode to work with
3365 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3366 conversions to QI/HImode types did a store and then a load to
3367 truncate the value. For conversions to VSX registers, don't split
3368 the insn, instead emit the code directly. Use the code iterator
3369 any_fix to combine signed and unsigned conversions.
3370 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3371 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3372 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3373 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3374 (fix_<mode>di2_hw): Likewise.
3375 (fixuns_<mode>di2_hw): Likewise.
3376 (fix_<mode>si2_hw): Likewise.
3377 (fixuns_<mode>si2_hw): Likewise.
3378 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3379 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3380 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3381 fix<uns>_trunc<SFDF:mode>si2_p8.
3382 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3384 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3385 (fix<uns>_<mode>_mem): Likewise.
3386 (fctiw<u>z_<mode>_mem): Likewise.
3387 (fix<uns>_<mode>_mem): Likewise.
3388 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3389 the register allocator from doing a direct move to the GPRs to do
3390 a store, and instead use the ISA 3.0 store byte/half-word from
3391 vector register instruction. For IEEE 128-bit floating point,
3392 also optimize stores of 32-bit ints.
3393 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3395 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3397 * genextract.c (push_pathstr_operand): New function to support
3399 (walk_rtx): Call push_pathstr_operand.
3400 (print_path): Support [a-zA-Z].
3402 2018-02-07 Richard Biener <rguenther@suse.de>
3404 PR tree-optimization/84037
3405 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3406 (cse_and_gimplify_to_preheader): Declare.
3407 (vect_get_place_in_interleaving_chain): Likewise.
3408 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3410 (_loop_vec_info::~_loop_vec_info): Delete it.
3411 (cse_and_gimplify_to_preheader): New function.
3412 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3413 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3414 (vectorizable_load): Likewise. For grouped stores always base
3415 the IV on the first element.
3416 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3417 condition before gimplifying.
3419 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3421 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3422 *DIV_EXPR and *MOD_EXPR.
3424 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3427 * config/i386/i386.c (ix86_option_override_internal): Mask out
3428 the CF_SET bit when checking -fcf-protection.
3430 2018-02-07 Tom de Vries <tom@codesourcery.com>
3433 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3436 2018-02-07 Richard Biener <rguenther@suse.de>
3438 PR tree-optimization/84204
3439 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3442 PR tree-optimization/84205
3443 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3444 special-case isl_ast_op_zdiv_r.
3446 PR tree-optimization/84223
3447 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3448 Only add conditions from within the region.
3449 (gather_bbs::after_dom_children): Adjust.
3451 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3454 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3455 * config/avr/avr.md: Only post-reload split REG-REG moves if
3456 either register is GENERAL_REG_P.
3458 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3460 PR tree-optimization/84235
3461 * tree-ssa-scopedtables.c
3462 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3463 if the subtraction is performed in floating point type where NaNs are
3464 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3465 build 1. Formatting fix.
3467 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3470 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3471 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3472 and skip it regardless of bb boundaries. Use CALL_P macro,
3473 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3474 unnecessarily, formatting fix.
3476 2018-02-06 Michael Collison <michael.collison@arm.com>
3478 * config/arm/thumb2.md:
3479 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3480 (*thumb_mov_notscc): Ditto.
3482 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3485 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3486 unsigned_fix, not "s".
3488 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3490 * configure.ac (gcc_fn_eh_frame_ro): New function.
3491 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3492 correct .eh_frame permissions.
3493 * configure: Regenerate.
3495 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3497 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3500 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3502 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3503 Display warning message for -mno-speculate-indirect-jumps.
3505 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3507 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3509 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3511 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3513 PR tree-optimization/84225
3514 * tree-eh.c (find_trapping_overflow): Only call
3515 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3517 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3520 * config/i386/i386.c: Reimplement the check of possible options
3521 -mibt/-mshstk conbination. Change error messages.
3522 * doc/invoke.texi: Fix a typo: remove extra '='.
3524 2018-02-06 Marek Polacek <polacek@redhat.com>
3526 PR tree-optimization/84228
3527 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3529 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3532 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3533 emitted arch directives.
3534 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3535 __ARM_FEATURE_COPROC before changing architectures.
3537 2018-02-06 Richard Biener <rguenther@suse.de>
3539 * config/i386/i386.c (print_reg): Fix typo.
3540 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3542 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3544 * configure: Regenerate.
3546 2018-02-05 Martin Sebor <msebor@redhat.com>
3548 PR tree-optimization/83369
3549 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3552 2018-02-05 Martin Liska <mliska@suse.cz>
3554 * doc/invoke.texi: Cherry-pick upstream r323995.
3556 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3558 * ira.c (ira_init_register_move_cost): Adjust comment.
3560 2018-02-05 Martin Liska <mliska@suse.cz>
3562 PR gcov-profile/84137
3563 * doc/gcov.texi: Fix typo in documentation.
3565 2018-02-05 Martin Liska <mliska@suse.cz>
3567 PR gcov-profile/83879
3568 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3569 using dlopen functionality.
3571 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3573 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3574 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3575 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3576 _mm_maskz_range_round_ss): New intrinsics.
3577 (__builtin_ia32_rangesd128_round)
3578 (__builtin_ia32_rangess128_round): Remove.
3579 (__builtin_ia32_rangesd128_mask_round,
3580 __builtin_ia32_rangess128_mask_round): New builtins.
3581 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3582 __builtin_ia32_rangess128_round): Remove.
3583 (__builtin_ia32_rangesd128_mask_round,
3584 __builtin_ia32_rangess128_mask_round): New builtins.
3585 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3586 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3587 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3588 "<round_saeonly_constraint>")): Changed to ...
3589 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3590 "<round_saeonly_scalar_constraint>")): ... this.
3591 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3592 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3593 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3594 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3595 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3597 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3599 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3601 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3602 Remove all values except native, 8540 and 8548.
3604 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3606 * config/i386/i386.c (ix86_output_function_return): Pass
3607 INVALID_REGNUM, instead of -1, as invalid register number to
3608 indirect_thunk_name and output_indirect_thunk.
3610 2018-02-02 Julia Koval <julia.koval@intel.com>
3612 * config.gcc: Add -march=icelake.
3613 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3614 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3615 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3616 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3617 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3618 (processor_target_table): Add icelake.
3619 (ix86_option_override_internal): Handle new PTAs.
3620 (get_builtin_code_for_version): Handle icelake.
3621 (M_INTEL_COREI7_ICELAKE): New.
3622 (fold_builtin_cpu): Handle icelake.
3623 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3624 * doc/invoke.texi: Add -march=icelake.
3626 2018-02-02 Julia Koval <julia.koval@intel.com>
3628 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3629 to wide_int_bitmask.
3630 * wide-int-bitmask.h: New.
3632 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3635 * config/i386/i386.md: Replace Pmode with word_mode in
3636 builtin_setjmp_setup and builtin_longjmp to support x32.
3638 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3642 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3644 (rs6000_supported_cpu_names): New static variable.
3645 (linux_cpu_translation_table): Likewise.
3646 (elf_platform) <cpu>: Define new static variable and use it.
3647 Translate kernel AT_PLATFORM name to canonical name if needed.
3648 Error if platform name is unknown.
3650 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3653 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3655 2018-02-01 Jeff Law <law@redhat.com>
3658 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3659 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3660 the scratch if RELEASE_VIA_POP is false.
3661 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3662 If we have to save a temporary register, decrement SIZE appropriately.
3663 Pass new arguments to release_scratch_register_on_entry.
3664 (ix86_adjust_stack_and_probe): Likewise.
3665 (ix86_emit_probe_stack_range): Pass new arguments to
3666 release_scratch_register_on_entry.
3668 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3670 PR rtl-optimization/84157
3671 * combine.c (change_zero_ext): Use REG_P predicate in
3672 front of HARD_REGISTER_P predicate.
3674 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3676 * config/avr/avr.c (avr_option_override): Move disabling of
3677 -fdelete-null-pointer-checks to...
3678 * common/config/avr/avr-common.c (avr_option_optimization_table):
3681 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3683 PR tree-optimization/81635
3684 * tree-data-ref.c (split_constant_offset_1): For types that
3685 wrap on overflow, try to use range info to prove that wrapping
3688 2018-02-01 Renlin Li <renlin.li@arm.com>
3691 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3693 (aarch64_register_move_cost): Likewise.
3694 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3696 (REG_CLASS_NAMES): Likewise.
3697 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3698 TAILCALL_ADDR_REGS. Remove IP registers.
3699 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3701 2018-02-01 Richard Biener <rguenther@suse.de>
3703 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3704 for specifying RPO order and allow NULL for that.
3705 * domwalk.c (dom_walker::dom_walker): Likewise.
3706 (dom_walker::walk): Handle NULL RPO order.
3707 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3709 (rewrite_update_dom_walker): Likewise.
3710 (mark_def_dom_walker): Likewise.
3712 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3714 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3715 (aarch64_maybe_expand_sve_subreg_move): Declare.
3716 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3717 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3719 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3720 that are semantically a reverse operation.
3721 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3722 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3723 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3725 (aarch64_can_change_mode_class): For big-endian, forbid changes
3726 between two SVE modes if they have different element sizes.
3728 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3730 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3731 the TImode handling for big-endian targets.
3733 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3735 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3736 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3738 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3739 Remove BSWAP handing for big-endian targets and use the form of
3740 LD1RQ appropariate for the mode.
3742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3744 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3745 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3748 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3751 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3752 check for operands that need to go through aarch64_sve_reload_be.
3754 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3756 PR tree-optimization/81661
3757 PR tree-optimization/84117
3758 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3759 * tree-eh.c: Include gimplify.h.
3760 (find_trapping_overflow, replace_trapping_overflow,
3761 rewrite_to_non_trapping_overflow): New functions.
3762 * tree-vect-loop.c: Include tree-eh.h.
3763 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3764 * tree-data-ref.c: Include tree-eh.h.
3765 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3767 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3769 PR rtl-optimization/84123
3770 * combine.c (change_zero_ext): Check if hard register satisfies
3771 can_change_dest_mode before calling gen_lowpart_SUBREG.
3773 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3776 * ira.c (ira_init_register_move_cost): Remove assert.
3778 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3780 PR rtl-optimization/84071
3781 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3782 * doc/tm.texi: Regenerate.
3784 2018-01-31 Richard Biener <rguenther@suse.de>
3786 PR tree-optimization/84132
3787 * tree-data-ref.c (analyze_miv_subscript): Properly
3788 check whether evolution_function_is_affine_multivariate_p
3789 before calling gcd_of_steps_may_divide_p.
3791 2018-01-31 Julia Koval <julia.koval@intel.com>
3794 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3795 * config/i386/i386.md (rdpid_rex64) New.
3796 (rdpid): Make 32bit only.
3798 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3801 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3802 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3804 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3807 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3809 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3811 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3813 PR rtl-optimization/84071
3814 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3815 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3817 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3819 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3820 (arc_attribute_table): Add 'aux' attribute.
3821 (arc_in_small_data_p): Consider aux like variables.
3822 (arc_is_aux_reg_p): New function.
3823 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3824 (arc_get_aux_arg): New function.
3825 (prepare_move_operands): Handle aux-register access.
3826 (arc_handle_aux_attribute): New function.
3827 * doc/extend.texi (ARC Variable attributes): Add subsection.
3829 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3831 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3832 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3833 (arc_attribute_table): Add 'uncached' attribute.
3834 (arc_print_operand): Print '.di' flag for uncached memory
3836 (arc_in_small_data_p): Do not consider for small data the uncached
3838 (arc_is_uncached_mem_p): New function.
3839 * config/arc/predicates.md (compact_store_memory_operand): Check
3840 for uncached memory accesses.
3841 (nonvol_nonimm_operand): Likewise.
3842 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3844 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3847 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3848 falign-loops=): Add Optimization flag.
3850 2018-01-30 Jeff Law <law@redhat.com>
3853 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3854 INT_REGISTERS_SAVED. Check it prior to calling
3855 get_scratch_register_on_entry.
3856 (ix86_adjust_stack_and_probe): Similarly.
3857 (ix86_emit_probe_stack_range): Similarly.
3858 (ix86_expand_prologue): Corresponding changes.
3860 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3863 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3864 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3866 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3869 * lra-constraints.c (curr_insn_transform): Process AND in the
3872 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3874 PR rtl-optimization/83986
3875 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3876 dependence against last_pending_memory_flush in addition to
3879 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3881 PR tree-optimization/81611
3882 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3885 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3888 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3891 2018-01-30 Richard Biener <rguenther@suse.de>
3892 Jakub Jelinek <jakub@redhat.com>
3894 PR tree-optimization/84111
3895 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3896 inner loops added during recursion, as they don't have up-to-date
3899 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3902 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3903 (can_inline_edge_by_limits_p): ... here.
3904 (can_early_inline_edge_p, check_callers,
3905 update_caller_keys, update_callee_keys, recursive_inlining,
3906 add_new_edges_to_heap, speculation_useful_p,
3907 inline_small_functions,
3908 inline_small_functions, flatten_function,
3909 inline_to_all_callers_1): Update.
3911 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3913 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3916 2018-01-30 Richard Biener <rguenther@suse.de>
3918 PR tree-optimization/83008
3919 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3920 invariant and constant vector uses in stmts when they need
3923 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3926 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3927 * configure: Regenerate.
3929 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3931 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3933 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3934 Use gen_rtx_REG rather than gen_lowpart.
3936 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3938 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3939 rather than 0 when creating partial subregs.
3941 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3943 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3946 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3949 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3950 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3951 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3952 flags. This restores the settings used before the 2017-07-24.
3953 Turning off pre increment/decrement/modify allows IVOPTS to
3954 optimize DF/SF loops where the index is an int.
3956 2018-01-29 Richard Biener <rguenther@suse.de>
3957 Kelvin Nilsen <kelvin@gcc.gnu.org>
3960 * tree-vect-stmts.c (vectorizable_call): Don't call
3961 targetm.vectorize_builtin_md_vectorized_function if callee is
3964 2018-01-22 Carl Love <cel@us.ibm.com>
3966 * doc/extend.tex: Fix typo in second arg in
3967 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3969 2018-01-29 Richard Biener <rguenther@suse.de>
3971 PR tree-optimization/84086
3972 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3973 (flush_ssaname_freelist): When SSA names were released reset
3974 the SCEV hash table.
3976 2018-01-29 Richard Biener <rguenther@suse.de>
3978 PR tree-optimization/84057
3979 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3980 removed paths when removing edges.
3982 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3984 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3985 -mfunction-return=@var{choice}.
3987 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3990 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3992 (layout::print_source_line): Likewise.
3993 (test_get_line_width_without_trailing_whitespace): Add test cases.
3995 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3998 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4001 2018-01-26 Jim Wilson <jimw@sifive.com>
4003 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4005 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4008 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4010 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4011 and CMP + SUB-immediate -> SUBS.
4013 2018-01-26 Martin Sebor <msebor@redhat.com>
4015 PR tree-optimization/83896
4016 * tree-ssa-strlen.c (get_string_len): Rename...
4017 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4018 Avoid assuming length is constant.
4019 (handle_char_store): Use HOST_WIDE_INT for string length.
4021 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4024 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4025 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4027 2018-01-26 Richard Biener <rguenther@suse.de>
4029 PR rtl-optimization/84003
4030 * dse.c (record_store): Only record redundant stores when
4031 the earlier store aliases at least all accesses the later one does.
4033 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4035 PR rtl-optimization/83985
4036 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4037 REG_CFA_RESTORE insns.
4038 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4039 REG_CFA_RESTORE insns here.
4042 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4043 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4045 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4047 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4048 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4049 (arc_init): Likewise.
4050 (arc_override_options): Likewise.
4051 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4053 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4055 * config/arc/arc.h (TARGET_DBNZ): Define.
4056 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4057 properly set the tune attribute.
4058 (dbnz): Use TARGET_DBNZ guard.
4059 * config/arc/arc.opt (mtune): Add core3 option.
4061 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4063 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4064 recognize new pic like addresses.
4065 (arc_delegitimize_address): Clean up.
4067 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4069 * config/arc/arc-arches.def: Option mrf16 valid for all
4071 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4072 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4073 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4074 * config/arc/arc-tables.opt: Regenerate.
4075 * config/arc/arc.c (arc_conditional_register_usage): Handle
4076 reduced register file case.
4077 (arc_file_start): Set must have build attributes.
4078 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4080 * config/arc/arc.opt (mrf16): Add new option.
4081 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4082 * config/arc/genmultilib.awk: Handle new mrf16 option.
4083 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4084 * config/arc/t-multilib: Regenerate.
4085 * doc/invoke.texi (ARC Options): Document mrf16 option.
4087 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4089 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4090 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4091 (arc_attribute_table): Add 'secure_call' attribute.
4092 (arc_print_operand): Print secure call operand.
4093 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4095 (arc_is_secure_call_p): New function. * config/arc/arc.md
4096 (call_i): Add support for sjli instruction.
4097 (call_value_i): Likewise.
4098 * config/arc/constraints.md (Csc): New constraint.
4100 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4101 John Eric Martin <John.Martin@emmicro-us.com>
4103 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4104 * config/arc/arc.c (_arc_jli_section): New struct.
4105 (arc_jli_section): New type.
4106 (rc_jli_sections): New static variable.
4107 (arc_handle_jli_attribute): New function.
4108 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4109 (arc_file_end): New function.
4110 (TARGET_ASM_FILE_END): Define.
4111 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4112 (arc_add_jli_section): New function.
4113 (jli_call_scan): Likewise.
4114 (arc_reorg): Call jli_call_scan.
4115 (arc_output_addsi): Remove 'S' from printing asm operand.
4116 (arc_is_jli_call_p): New function.
4117 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4119 (movhi_insn): Likewise.
4120 (movsi_insn): Likewise.
4121 (movsi_set_cc_insn): Likewise.
4122 (loadqi_update): Likewise.
4123 (load_zeroextendqisi_update): Likewise.
4124 (load_signextendqisi_update): Likewise.
4125 (loadhi_update): Likewise.
4126 (load_zeroextendhisi_update): Likewise.
4127 (load_signextendhisi_update): Likewise.
4128 (loadsi_update): Likewise.
4129 (loadsf_update): Likewise.
4130 (movsicc_insn): Likewise.
4131 (bset_insn): Likewise.
4132 (bxor_insn): Likewise.
4133 (bclr_insn): Likewise.
4134 (bmsk_insn): Likewise.
4135 (bicsi3_insn): Likewise.
4136 (cmpsi_cc_c_insn): Likewise.
4137 (movsi_ne): Likewise.
4138 (movsi_cond_exec): Likewise.
4139 (clrsbsi2): Likewise.
4150 (call_i): Remove 'S' asm letter, add jli instruction.
4151 (call_value_i): Likewise.
4152 * config/arc/arc.op (mjli-always): New option.
4153 * config/arc/constraints.md (Cji): New constraint.
4154 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4156 (subsf3_fpx): Likewise.
4157 (mulsf3_fpx): Likewise.
4158 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4160 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4162 * doc/invoke.texi (ARC): Document mjli-always option.
4164 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4166 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4167 avoid addition with 0 and use incw and decw where possible.
4169 2018-01-26 Richard Biener <rguenther@suse.de>
4171 PR tree-optimization/81082
4172 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4173 association if it requires casting to unsigned.
4174 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4175 from fold_plusminus_mult_expr to catch important cases late when
4176 range info is available.
4178 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4180 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4181 * configure.ac (hidden_linkonce): New test.
4182 * configure: Regenerate.
4183 * config.in: Regenerate.
4185 2018-01-26 Julia Koval <julia.koval@intel.com>
4187 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4188 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4189 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4190 _mm_mask_bitshuffle_epi64_mask): Fix type.
4191 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4192 USI_FTYPE_V4DI_V4DI_USI): Remove.
4193 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4194 __builtin_ia32_vpshufbitqmb256_mask,
4195 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4196 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4197 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4199 2018-01-26 Alan Modra <amodra@gmail.com>
4202 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4203 UNSPEC_VBPERMQ. Sort other unspecs.
4205 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4207 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4209 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4212 * predict.c (drop_profile): Do not push/pop cfun; update also
4214 (handle_missing_profiles): Fix logic looking for zero profiles.
4216 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4219 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4220 on functions with #pragma omp declare simd or functions with simd
4222 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4223 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4224 Remove trailing \n from warning_at calls.
4226 2018-01-25 Tom de Vries <tom@codesourcery.com>
4229 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4230 for neutered workers.
4232 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4235 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4236 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4238 2018-01-24 Jeff Law <law@redhat.com>
4241 * i386.c (get_probe_interval): Move to earlier point.
4242 (ix86_compute_frame_layout): If -fstack-clash-protection and
4243 the frame is larger than the probe interval, then use pushes
4244 to save registers rather than reg->mem moves.
4245 (ix86_expand_prologue): Remove conditional for int_registers_saved
4248 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4251 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4252 min/max for never referenced object.
4254 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4257 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4259 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4260 attributes from DECL_ATTRIBUTES (decl) without affecting
4261 DECL_ATTRIBUTES (current_function_decl).
4262 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4263 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4265 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4267 PR tree-optimization/83979
4268 * fold-const.c (fold_comparison): Use constant_boolean_node
4269 instead of boolean_{true,false}_node.
4271 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4273 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4276 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4278 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4279 Simplify the clause that sets the length attribute.
4280 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4281 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4282 clause that sets the length attribute.
4283 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4285 2018-01-24 Tom de Vries <tom@codesourcery.com>
4288 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4289 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4290 Add strict parameter.
4291 (prevent_branch_around_nothing): Insert dummy insn between branch to
4292 label and label with no ptx insn inbetween.
4293 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4295 2018-01-24 Tom de Vries <tom@codesourcery.com>
4298 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4299 for neutered threads in warp.
4300 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4302 2018-01-24 Richard Biener <rguenther@suse.de>
4304 PR tree-optimization/83176
4305 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4308 2018-01-24 Richard Biener <rguenther@suse.de>
4310 PR tree-optimization/82819
4311 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4312 code generating pluses that are no-ops in the target precision.
4314 2018-01-24 Richard Biener <rguenther@suse.de>
4317 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4319 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4321 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4322 to merge probabilities.
4323 * predict.c (probably_never_executed): Also mark as cold functions
4324 with global 0 profile and guessed local profile.
4325 * profile-count.c (profile_probability::combine_with_count): New
4327 * profile-count.h (profile_probability::operator*,
4328 profile_probability::operator*=, profile_probability::operator/,
4329 profile_probability::operator/=): Reduce precision to adjusted
4330 and set value to guessed on contradictory divisions.
4331 (profile_probability::combine_with_freq): Remove.
4332 (profile_probability::combine_wiht_count): Declare.
4333 (profile_count::force_nonzero):: Set to adjusted.
4334 (profile_count::probability_in):: Set quality to adjusted.
4335 * tree-ssa-tail-merge.c (replace_block_by): Use
4338 2018-01-23 Andrew Waterman <andrew@sifive.com>
4339 Jim Wilson <jimw@sifive.com>
4341 * config/riscv/riscv.c (riscv_stack_boundary): New.
4342 (riscv_option_override): Set riscv_stack_boundary. Handle
4343 riscv_preferred_stack_boundary_arg.
4344 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4345 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4346 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4347 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4348 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4349 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4351 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4354 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4355 of struct ix86_frame.
4356 (ix86_expand_epilogue): Likewise. Add a local variable for
4357 the reg_save_offset field in struct ix86_frame.
4359 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4361 PR tree-optimization/82604
4362 * tree-loop-distribution.c (enum partition_kind): New enum item
4363 PKIND_PARTIAL_MEMSET.
4364 (partition_builtin_p): Support above new enum item.
4365 (generate_code_for_partition): Ditto.
4366 (compute_access_range): Differentiate cases that equality can be
4367 proven at all loops, the innermost loops or no loops.
4368 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4369 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4370 (finalize_partitions, distribute_loop): Don't fuse partition of
4371 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4372 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4375 2018-01-23 Martin Liska <mliska@suse.cz>
4377 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4378 order to ignore the predictor.
4379 (PRED_POLYMORPHIC_CALL): Likewise.
4380 (PRED_RECURSIVE_CALL): Likewise.
4382 2018-01-23 Martin Liska <mliska@suse.cz>
4384 * tree-profile.c (tree_profiling): Print function header to
4385 aware reader which function we are working on.
4386 * value-prof.c (gimple_find_values_to_profile): Do not print
4387 not interesting value histograms.
4389 2018-01-23 Martin Liska <mliska@suse.cz>
4391 * profile-count.h (enum profile_quality): Add
4392 profile_uninitialized as the first value. Do not number values
4393 as they are zero based.
4394 (profile_count::verify): Update sanity check.
4395 (profile_probability::verify): Likewise.
4397 2018-01-23 Nathan Sidwell <nathan@acm.org>
4399 * doc/invoke.texi (ffor-scope): Deprecate.
4401 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4403 PR tree-optimization/83510
4404 * domwalk.c (set_all_edges_as_executable): New function.
4405 (dom_walker::dom_walker): Convert bool param
4406 "skip_unreachable_blocks" to enum reachability. Move setup of
4407 edge flags to set_all_edges_as_executable and only do it when
4408 reachability is REACHABLE_BLOCKS.
4409 * domwalk.h (enum dom_walker::reachability): New enum.
4410 (dom_walker::dom_walker): Convert bool param
4411 "skip_unreachable_blocks" to enum reachability.
4412 (set_all_edges_as_executable): New decl.
4413 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4414 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4416 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4417 but converting true to REACHABLE_BLOCKS.
4418 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4420 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4421 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4422 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4424 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4425 if check_all_array_refs will be called.
4427 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4429 * tree.c (selftest::test_location_wrappers): Add more test
4432 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4434 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4435 (selftest::test_bit_in_range): Likewise.
4437 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4440 * doc/sourcebuild.texi (vect_float): Say that the selector
4441 only describes the situation when -funsafe-math-optimizations is on.
4442 (vect_float_strict): Document.
4444 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4446 PR tree-optimization/83965
4447 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4448 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4449 instead of checking only for a reduction.
4450 (vect_recog_widen_sum_pattern): Likewise.
4452 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4454 * predict.c (probably_never_executed): Only use precise profile info.
4455 (compute_function_frequency): Skip after inlining hack since we now
4456 have quality checking.
4458 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4460 * profile-count.h (profile_probability::very_unlikely,
4461 profile_probability::unlikely, profile_probability::even): Set
4462 precision to guessed.
4464 2018-01-23 Richard Biener <rguenther@suse.de>
4466 PR tree-optimization/83963
4467 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4468 Properly terminate dominator walk when crossing the exit edge not
4469 when visiting its source block.
4471 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4474 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4475 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4477 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4479 PR tree-optimization/83957
4480 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4481 semicolon after for body surrounded by braces.
4483 PR tree-optimization/83081
4484 * profile-count.h (profile_probability::split): New method.
4485 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4486 Use profile_probability::split.
4487 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4488 when splitting a single conditional jump into 2.
4490 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4492 PR tree-optimization/69452
4493 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4496 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4498 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4499 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4500 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4502 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4504 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4505 * config/rl78/rl78.md (movdi): New define_expand.
4506 * config/rl78/rl78.c (rl78_split_movdi): New function.
4508 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4511 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4513 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4514 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4515 128-bit to produce an UNSPEC move to get the double word with the
4516 signbit and then a shift directly to do signbit.
4517 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4518 implementation with a new version that just does either a direct
4519 move or a regular move. Move memory interface to separate insns.
4520 Move insns so they are next to the expander.
4521 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4522 with signbit move. Split big and little endian case.
4523 (signbit<mode>2_dm_mem_le): Likewise.
4524 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4525 (signbit<mode>2_dm2): Likewise.
4527 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4529 * config/rl78/rl78.md (anddi3): New define_expand.
4531 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4533 * config/rl78/rl78.md (umindi3): New define_expand.
4535 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4537 * config/rl78/rl78.md (smindi3): New define_expand.
4539 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4541 * config/rl78/rl78.md (smaxdi3): New define_expand.
4543 2018-01-22 Carl Love <cel@us.ibm.com>
4545 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4546 LVX_V1TI): Add macro expansion.
4547 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4548 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4549 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4550 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4551 Change check to determine if the instruction is a byte reversing
4552 entry. Fix typo in comment.
4553 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4554 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4555 Add def_builtin calls for new builtins.
4556 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4557 Add define_insn expansion.
4559 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4561 * config/rl78/rl78.md (umaxdi3): New define_expand.
4563 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4565 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4566 for non-QImode registers.
4568 2018-01-22 Richard Biener <rguenther@suse.de>
4570 PR tree-optimization/83963
4571 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4572 including the loop exit block.
4573 (scop_detection::merge_sese): Likewise.
4574 (scop_detection::add_scop): Do it here instead.
4576 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4578 * doc/sourcebuild.texi (arm_softfloat): Document.
4580 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4583 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4584 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4585 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4587 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4588 David Edelsohn <dje.gcc@gmail.com>
4591 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4592 Change "crset eq" to "crset 2".
4593 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4594 (*call_indirect_aix<mode>_nospec): Likewise.
4595 (*call_value_indirect_aix<mode>_nospec): Likewise.
4596 (*call_indirect_elfv2<mode>_nospec): Likewise.
4597 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4598 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4599 change assembly output from . to $.
4600 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4601 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4602 (*tablejump<mode>_internal1_nospec): Likewise.
4604 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4607 * config/sh/sh_optimize_sett_clrt.cc:
4608 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4610 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4612 PR tree-optimization/83940
4613 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4614 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4615 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4616 use it to pass back the definition type.
4617 (vect_check_store_rhs): Likewise rhs_dt_out.
4618 (vect_build_gather_load_calls): Add a mask_dt argument and use
4619 it instead of a call to vect_is_simple_use.
4620 (vectorizable_store): Update calls to vect_check_load_store_mask
4621 and vect_check_store_rhs. Use the dt returned by the latter instead
4622 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4623 instead of calls to vect_is_simple_use. Pass the scalar rather
4624 than the vector operand to vect_is_simple_use when handling
4625 second and subsequent copies of an rhs value.
4626 (vectorizable_load): Update calls to vect_check_load_store_mask
4627 and vect_build_gather_load_calls. Use the cached mask_dt and
4628 gs_info.offset_dt instead of calls to vect_is_simple_use.
4630 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4633 * tree-emutls.c: Include gimplify.h.
4634 (lower_emutls_2): New function.
4635 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4636 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4637 it before further processing.
4640 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4641 UINTVAL (trueop1) instead of INTVAL (op1).
4643 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4647 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4648 INCOMING_FRAME_SP_OFFSET if not defined.
4649 (scan_trace): Add ENTRY argument. If true and
4650 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4651 emit a note to adjust the CFA offset.
4652 (create_cfi_notes): Adjust scan_trace callers.
4653 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4654 INCOMING_FRAME_SP_OFFSET in the CIE.
4655 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4656 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4658 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4659 * doc/tm.texi: Regenerated.
4661 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4663 PR rtl-optimization/83147
4664 * lra-constraints.c (remove_inheritance_pseudos): Use
4665 lra_substitute_pseudo_within_insn.
4667 2018-01-19 Tom de Vries <tom@codesourcery.com>
4668 Cesar Philippidis <cesar@codesourcery.com>
4671 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4673 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4676 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4677 spaces for function labels.
4679 2018-01-19 Martin Liska <mliska@suse.cz>
4681 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4682 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4683 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4684 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4685 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4686 (PRED_CONST_RETURN): Change from 69 to 65.
4687 (PRED_NULL_RETURN): Change from 91 to 71.
4688 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4689 (PRED_LOOP_GUARD): Change from 66 to 73.
4691 2018-01-19 Martin Liska <mliska@suse.cz>
4693 * predict.c (predict_insn_def): Add new assert.
4694 (struct branch_predictor): Change type to signed integer.
4695 (test_prediction_value_range): Amend test to cover
4697 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4698 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4699 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4700 (PRED_LOOP_IV_COMPARE): Likewise.
4701 * predict.h (PROB_UNINITIALIZED): Define new constant.
4703 2018-01-19 Martin Liska <mliska@suse.cz>
4705 * predict.c (dump_prediction): Add new format for
4706 analyze_brprob.py script which is enabled with -details
4708 * profile-count.h (precise_p): New function.
4710 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4712 PR tree-optimization/83922
4713 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4714 there are no statements that need masking.
4715 (vect_active_double_reduction_p): New function.
4716 (vect_analyze_loop_operations): Use it when handling phis that
4717 are not in the loop header.
4719 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4721 PR tree-optimization/83914
4722 * tree-vect-loop.c (vectorizable_induction): Don't convert
4723 init_expr or apply the peeling adjustment for inductions
4724 that are nested within the vectorized loop.
4726 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4728 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4731 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4735 * function.h (gimplify_parameters): Add gimple_seq * argument.
4736 * function.c: Include gimple.h and options.h.
4737 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4738 for the added local temporaries if needed.
4739 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4740 if there are any parameter cleanups, wrap whole body into a
4741 try/finally with the cleanups.
4743 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4746 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4747 Use GET_MODE_CLASS for scalar floating point.
4749 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4753 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4754 Fix call of call_cgraph_insertion_hooks.
4756 2018-01-18 Martin Sebor <msebor@redhat.com>
4758 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4760 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4763 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4766 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4769 * common.opt: (-ffile-prefix-map): New option.
4770 * opts.c (common_handle_option): Defer it.
4771 * opts-global.c (handle_common_deferred_options): Handle it.
4772 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4773 * file-prefix-map.h: New file.
4774 (remap_debug_filename, add_debug_prefix_map): ...here.
4775 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4776 * final.c (debug_prefix_map, add_debug_prefix_map
4777 remap_debug_filename): Move to...
4778 * file-prefix-map.c: New file.
4779 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4780 generalize, get rid of alloca(), use strrchr() instead of strchr().
4781 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4782 Implement in terms of add_prefix_map().
4783 (remap_macro_filename, remap_debug_filename): Implement in term of
4785 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4786 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4787 * dbxout.c: Include file-prefix-map.h.
4788 * varasm.c: Likewise.
4789 * vmsdbgout.c: Likewise.
4790 * xcoffout.c: Likewise.
4791 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4792 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4793 * doc/invoke.texi (-ffile-prefix-map): Document.
4794 (-fdebug-prefix-map): Update description.
4796 2018-01-18 Martin Liska <mliska@suse.cz>
4798 * config/i386/i386.c (indirect_thunk_name): Document that also
4800 (output_indirect_thunk): Document why both instructions
4801 (pause and lfence) are generated.
4803 2018-01-18 Richard Biener <rguenther@suse.de>
4805 PR tree-optimization/83887
4806 * graphite-scop-detection.c
4807 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4808 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4809 (scop_detection::merge_sese): Re-implement with a flood-fill
4810 algorithm that properly finds a SESE region if it exists.
4812 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4815 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4816 pointer_diff optimizations use view_convert instead of convert.
4818 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4820 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4821 Generate different code for -mno-speculate-indirect-jumps.
4822 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4823 (*call_indirect_aix<mode>): Disable for
4824 -mno-speculate-indirect-jumps.
4825 (*call_indirect_aix<mode>_nospec): New define_insn.
4826 (*call_value_indirect_aix<mode>): Disable for
4827 -mno-speculate-indirect-jumps.
4828 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4829 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4830 -mno-speculate-indirect-jumps.
4831 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4833 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4835 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4836 long double type, set the flags for noting the default long double
4837 type, even if we don't pass or return a long double type.
4839 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4842 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4845 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4847 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4848 support for merge[hl].
4849 (fold_mergehl_helper): New helper function.
4850 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4851 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4852 (altivec_vmrglw_direct): Add xxmrglw insn.
4854 2018-01-17 Andrew Waterman <andrew@sifive.com>
4856 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4857 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4859 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4862 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4863 call the lto_location_cache before reading the
4864 DECL_SOURCE_LOCATION of the types.
4866 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4867 Richard Sandiford <richard.sandiford@linaro.org>
4869 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4870 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4871 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4872 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4873 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4875 * config/aarch64/constraints.md (aarch64_movti_operand):
4877 * config/aarch64/predicates.md (Uti): Add new constraint.
4879 2018-01-17 Carl Love <cel@us.ibm.com>
4881 * config/rs6000/vsx.md (define_expand xl_len_r,
4882 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4883 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4885 (define_expand, define_insn): Move the shift left from the
4886 define_insn to the define_expand for lxvl and stxvl instructions.
4887 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4888 and XL_LEN_R definitions to PURE.
4890 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4892 * config/i386/i386.c (indirect_thunk_name): Declare regno
4893 as unsigned int. Compare regno with INVALID_REGNUM.
4894 (output_indirect_thunk): Ditto.
4895 (output_indirect_thunk_function): Ditto.
4896 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4897 in the call to output_indirect_thunk_function.
4899 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4902 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4903 rather than the size of inner_type to determine the stack slot size
4904 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4906 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4909 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4912 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4914 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4915 endian Linux systems to optionally enable multilibs for selecting
4916 the long double type if the user configured an explicit type.
4917 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4918 have no long double multilibs if not defined.
4919 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4920 warn if the user used -mabi={ieee,ibm}longdouble and we built
4921 multilibs for long double.
4922 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4923 appropriate multilib option.
4924 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4926 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4927 for building long double multilibs.
4928 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4930 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4932 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4935 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4937 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4940 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4943 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4946 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4948 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4949 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4951 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4953 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4954 different rtl trees depending on TARGET_64BIT.
4955 (rs6000_gen_lvx): Likewise.
4957 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4959 * config/visium/visium.md (nop): Tweak comment.
4960 (hazard_nop): Likewise.
4962 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4964 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4965 -mspeculate-indirect-jumps.
4966 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4967 for -mno-speculate-indirect-jumps.
4968 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4969 (*call_value_indirect_elfv2<mode>): Disable for
4970 -mno-speculate-indirect-jumps.
4971 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4972 (indirect_jump): Emit different RTL for
4973 -mno-speculate-indirect-jumps.
4974 (*indirect_jump<mode>): Disable for
4975 -mno-speculate-indirect-jumps.
4976 (*indirect_jump<mode>_nospec): New define_insn.
4977 (tablejump): Emit different RTL for
4978 -mno-speculate-indirect-jumps.
4979 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4980 (tablejumpsi_nospec): New define_expand.
4981 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4982 (tablejumpdi_nospec): New define_expand.
4983 (*tablejump<mode>_internal1): Disable for
4984 -mno-speculate-indirect-jumps.
4985 (*tablejump<mode>_internal1_nospec): New define_insn.
4986 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4989 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4991 * caller-save.c (insert_save): Drop unnecessary parameter. All
4994 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4995 Richard Biener <rguenth@suse.de>
4998 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4999 return early, inline manually is_gimple_sizepos. Make sure if we
5000 call gimplify_expr we don't end up with a gimple constant.
5001 * tree.c (variably_modified_type_p): Don't return true for
5002 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5003 * gimplify.h (is_gimple_sizepos): Remove.
5005 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5007 PR tree-optimization/83857
5008 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5009 vectorizable_live_operation for pure SLP statements.
5010 (vectorizable_live_operation): Handle PHIs.
5012 2018-01-16 Richard Biener <rguenther@suse.de>
5014 PR tree-optimization/83867
5015 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5016 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5018 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5021 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5022 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5023 If off is not INTEGER_CST, issue a may not be aligned warning
5024 rather than isn't aligned. Use isn%'t rather than isn't.
5025 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5027 <case MULT_EXPR>: Improve the case when bottom and one of the
5028 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5029 operand, in that case check if the other operand is multiple of
5030 bottom divided by the INTEGER_CST operand.
5032 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5035 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5036 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5037 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5038 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5039 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5040 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5041 (pa_function_arg_size): New function.
5043 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5045 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5046 in a separate statement.
5048 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5050 PR tree-optimization/83847
5051 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5052 group gathers and scatters.
5054 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5056 PR rtl-optimization/86620
5057 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5059 PR rtl-optimization/83213
5060 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5061 to last if both are JUMP_INSNs.
5063 PR tree-optimization/83843
5064 * gimple-ssa-store-merging.c
5065 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5066 store_immediate_info for bswap/nop orig_stores.
5068 2018-01-15 Andrew Waterman <andrew@sifive.com>
5070 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5072 <UDIV>: Increase cost if !TARGET_DIV.
5074 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5076 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5077 (define_attr "cr_logical_3op"): New.
5078 (cceq_ior_compare): Adjust.
5079 (cceq_ior_compare_complement): Adjust.
5080 (*cceq_rev_compare): Adjust.
5081 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5082 (is_cracked_insn): Adjust.
5083 (insn_must_be_first_in_group): Adjust.
5084 * config/rs6000/40x.md: Adjust.
5085 * config/rs6000/440.md: Adjust.
5086 * config/rs6000/476.md: Adjust.
5087 * config/rs6000/601.md: Adjust.
5088 * config/rs6000/603.md: Adjust.
5089 * config/rs6000/6xx.md: Adjust.
5090 * config/rs6000/7450.md: Adjust.
5091 * config/rs6000/7xx.md: Adjust.
5092 * config/rs6000/8540.md: Adjust.
5093 * config/rs6000/cell.md: Adjust.
5094 * config/rs6000/e300c2c3.md: Adjust.
5095 * config/rs6000/e500mc.md: Adjust.
5096 * config/rs6000/e500mc64.md: Adjust.
5097 * config/rs6000/e5500.md: Adjust.
5098 * config/rs6000/e6500.md: Adjust.
5099 * config/rs6000/mpc.md: Adjust.
5100 * config/rs6000/power4.md: Adjust.
5101 * config/rs6000/power5.md: Adjust.
5102 * config/rs6000/power6.md: Adjust.
5103 * config/rs6000/power7.md: Adjust.
5104 * config/rs6000/power8.md: Adjust.
5105 * config/rs6000/power9.md: Adjust.
5106 * config/rs6000/rs64.md: Adjust.
5107 * config/rs6000/titan.md: Adjust.
5109 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5111 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5112 ix86_indirect_branch_register logic.
5114 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5116 * config/i386/constraints.md (Bs): Update
5117 ix86_indirect_branch_register check. Don't check
5118 ix86_indirect_branch_register with GOT_memory_operand.
5120 * config/i386/predicates.md (GOT_memory_operand): Don't check
5121 ix86_indirect_branch_register here.
5122 (GOT32_symbol_operand): Likewise.
5124 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5126 * config/i386/predicates.md (constant_call_address_operand):
5127 Rewrite ix86_indirect_branch_register logic.
5128 (sibcall_insn_operand): Likewise.
5130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5132 * config/i386/constraints.md (Bs): Replace
5133 ix86_indirect_branch_thunk_register with
5134 ix86_indirect_branch_register.
5136 * config/i386/i386.md (indirect_jump): Likewise.
5137 (tablejump): Likewise.
5138 (*sibcall_memory): Likewise.
5139 (*sibcall_value_memory): Likewise.
5140 Peepholes of indirect call and jump via memory: Likewise.
5141 * config/i386/i386.opt: Likewise.
5142 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5143 (GOT_memory_operand): Likewise.
5144 (call_insn_operand): Likewise.
5145 (sibcall_insn_operand): Likewise.
5146 (GOT32_symbol_operand): Likewise.
5148 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5151 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5152 type rather than type addr's type points to.
5153 (expand_omp_atomic_mutex): Likewise.
5154 (expand_omp_atomic): Likewise.
5156 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5159 * config/i386/i386.c (output_indirect_thunk_function): Use
5160 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5161 for __x86_return_thunk.
5163 2018-01-15 Richard Biener <rguenther@suse.de>
5166 * expmed.c (extract_bit_field_1): Fix typo.
5168 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5171 * config/arm/iterators.md (VF): New mode iterator.
5172 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5173 Remove integer-related logic from pattern.
5174 (neon_vabd<mode>_3): Likewise.
5176 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5179 * common.opt (fstrict-overflow): No longer an alias.
5180 (fwrapv-pointer): New option.
5181 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5182 also for pointer types based on flag_wrapv_pointer.
5183 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5184 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5185 opts->x_flag_wrapv got set.
5186 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5187 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5188 POINTER_TYPE_OVERFLOW_UNDEFINED.
5189 * match.pd: Likewise in address comparison pattern.
5190 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5192 2018-01-15 Richard Biener <rguenther@suse.de>
5195 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5196 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5197 Reset type names to their identifier if their TYPE_DECL doesn't
5198 have linkage (and thus is used for ODR and devirt).
5199 (save_debug_info_for_decl): Remove.
5200 (save_debug_info_for_type): Likewise.
5201 (add_tree_to_fld_list): Adjust.
5202 * tree-pretty-print.c (dump_generic_node): Make dumping of
5203 type names more robust.
5205 2018-01-15 Richard Biener <rguenther@suse.de>
5207 * BASE-VER: Bump to 8.0.1.
5209 2018-01-14 Martin Sebor <msebor@redhat.com>
5212 * builtins.c (check_access): Avoid warning when the no-warning bit
5215 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5217 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5218 * ira-color (allocno_hard_regs_compare): Likewise.
5220 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5223 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5224 Use .pushsection/.popsection.
5226 2018-01-14 Martin Sebor <msebor@redhat.com>
5229 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5231 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5233 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5234 entry from extra_headers.
5235 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5236 extra_headers, make the list bitwise identical to the i?86-*-* one.
5238 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5240 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5241 -mcmodel=large with -mindirect-branch=thunk,
5242 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5243 -mfunction-return=thunk-extern.
5244 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5245 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5246 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5248 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5250 * config/i386/i386.c (print_reg): Print the name of the full
5251 integer register without '%'.
5252 (ix86_print_operand): Handle 'V'.
5253 * doc/extend.texi: Document 'V' modifier.
5255 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5257 * config/i386/constraints.md (Bs): Disallow memory operand for
5258 -mindirect-branch-register.
5260 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5261 (GOT_memory_operand): Likewise.
5262 (call_insn_operand): Likewise.
5263 (sibcall_insn_operand): Likewise.
5264 (GOT32_symbol_operand): Likewise.
5265 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5266 for -mindirect-branch-register.
5267 (tablejump): Likewise.
5268 (*sibcall_memory): Likewise.
5269 (*sibcall_value_memory): Likewise.
5270 Disallow peepholes of indirect call and jump via memory for
5271 -mindirect-branch-register.
5272 (*call_pop): Replace m with Bw.
5273 (*call_value_pop): Likewise.
5274 (*sibcall_pop_memory): Replace m with Bs.
5275 * config/i386/i386.opt (mindirect-branch-register): New option.
5276 * doc/invoke.texi: Document -mindirect-branch-register option.
5278 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5280 * config/i386/i386-protos.h (ix86_output_function_return): New.
5281 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5282 set function_return_type.
5283 (indirect_thunk_name): Add ret_p to indicate thunk for function
5285 (output_indirect_thunk_function): Pass false to
5286 indirect_thunk_name.
5287 (ix86_output_indirect_branch_via_reg): Likewise.
5288 (ix86_output_indirect_branch_via_push): Likewise.
5289 (output_indirect_thunk_function): Create alias for function
5290 return thunk if regno < 0.
5291 (ix86_output_function_return): New function.
5292 (ix86_handle_fndecl_attribute): Handle function_return.
5293 (ix86_attribute_table): Add function_return.
5294 * config/i386/i386.h (machine_function): Add
5295 function_return_type.
5296 * config/i386/i386.md (simple_return_internal): Use
5297 ix86_output_function_return.
5298 (simple_return_internal_long): Likewise.
5299 * config/i386/i386.opt (mfunction-return=): New option.
5300 (indirect_branch): Mention -mfunction-return=.
5301 * doc/extend.texi: Document function_return function attribute.
5302 * doc/invoke.texi: Document -mfunction-return= option.
5304 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5306 * config/i386/i386-opts.h (indirect_branch): New.
5307 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5308 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5309 with local indirect jump when converting indirect call and jump.
5310 (ix86_set_indirect_branch_type): New.
5311 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5312 (indirectlabelno): New.
5313 (indirect_thunk_needed): Likewise.
5314 (indirect_thunk_bnd_needed): Likewise.
5315 (indirect_thunks_used): Likewise.
5316 (indirect_thunks_bnd_used): Likewise.
5317 (INDIRECT_LABEL): Likewise.
5318 (indirect_thunk_name): Likewise.
5319 (output_indirect_thunk): Likewise.
5320 (output_indirect_thunk_function): Likewise.
5321 (ix86_output_indirect_branch_via_reg): Likewise.
5322 (ix86_output_indirect_branch_via_push): Likewise.
5323 (ix86_output_indirect_branch): Likewise.
5324 (ix86_output_indirect_jmp): Likewise.
5325 (ix86_code_end): Call output_indirect_thunk_function if needed.
5326 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5328 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5329 (ix86_attribute_table): Add indirect_branch.
5330 * config/i386/i386.h (machine_function): Add indirect_branch_type
5331 and has_local_indirect_jump.
5332 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5334 (tablejump): Likewise.
5335 (*indirect_jump): Use ix86_output_indirect_jmp.
5336 (*tablejump_1): Likewise.
5337 (simple_return_indirect_internal): Likewise.
5338 * config/i386/i386.opt (mindirect-branch=): New option.
5339 (indirect_branch): New.
5342 (thunk-inline): Likewise.
5343 (thunk-extern): Likewise.
5344 * doc/extend.texi: Document indirect_branch function attribute.
5345 * doc/invoke.texi: Document -mindirect-branch= option.
5347 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5350 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5352 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5354 * ipa-inline.c (want_inline_small_function_p): Return false if
5355 inlining has already failed with CIF_FINAL_ERROR.
5356 (update_caller_keys): Call want_inline_small_function_p before
5358 (update_callee_keys): Likewise.
5360 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5362 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5364 (rs6000_quadword_masked_address_p): Likewise.
5365 (quad_aligned_load_p): Likewise.
5366 (quad_aligned_store_p): Likewise.
5367 (const_load_sequence_p): Add comment to describe the outer-most loop.
5368 (mimic_memory_attributes_and_flags): New function.
5369 (rs6000_gen_stvx): Likewise.
5370 (replace_swapped_aligned_store): Likewise.
5371 (rs6000_gen_lvx): Likewise.
5372 (replace_swapped_aligned_load): Likewise.
5373 (replace_swapped_load_constant): Capitalize argument name in
5374 comment describing this function.
5375 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5376 and stores that access quad-word aligned addresses and replace
5377 with stvx or lvx instructions when appropriate.
5378 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5379 New function prototype.
5380 (rs6000_quadword_masked_address_p): Likewise.
5381 (rs6000_gen_lvx): Likewise.
5382 (rs6000_gen_stvx): Likewise.
5383 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5384 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5385 when memory address is aligned.
5386 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5387 this split to select lvx instruction when memory address is aligned.
5388 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5389 instruction when memory address is aligned.
5390 (*vsx_le_perm_load_v16qi): Likewise.
5391 (four unnamed splitters): Modify to select the stvx instruction
5392 when memory is aligned.
5394 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5396 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5397 which appears in the queue multiple times.
5399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5400 Alan Hayward <alan.hayward@arm.com>
5401 David Sherwood <david.sherwood@arm.com>
5403 * tree-vectorizer.h (vec_lower_bound): New structure.
5404 (_loop_vec_info): Add check_nonzero and lower_bounds.
5405 (LOOP_VINFO_CHECK_NONZERO): New macro.
5406 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5407 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5408 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5409 fields. Make seg_len the distance travelled, not including the
5411 (dr_direction_indicator): Declare.
5412 (dr_zero_step_indicator): Likewise.
5413 (dr_known_forward_stride_p): Likewise.
5414 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5416 (runtime_alias_check_p): Allow runtime alias checks with
5418 (operator ==): Compare access_size and align.
5419 (prune_runtime_alias_test_list): Rework for new distinction between
5420 the access_size and seg_len.
5421 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5423 (get_segment_min_max): New function.
5424 (create_intersect_range_checks): Use it.
5425 (dr_step_indicator): New function.
5426 (dr_direction_indicator): Likewise.
5427 (dr_zero_step_indicator): Likewise.
5428 (dr_known_forward_stride_p): Likewise.
5429 * tree-loop-distribution.c (data_ref_segment_size): Return
5430 DR_STEP * (niters - 1).
5431 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5433 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5434 (vect_preserves_scalar_order_p): New function, split out from...
5435 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5436 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5437 (vect_vfa_access_size): New function.
5438 (vect_vfa_align): Likewise.
5439 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5440 (dump_lower_bound): New function.
5441 (vect_check_lower_bound): Likewise.
5442 (vect_small_gap_p): Likewise.
5443 (vectorizable_with_step_bound_p): Likewise.
5444 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5445 depencies if the vectorization factor is 1. Convert the checks
5446 for nonzero steps into checks on the bounds of DR_STEP. Try using
5447 a bunds check for variable steps if the minimum required step is
5448 relatively small. Update calls to the dr_with_seg_len
5449 constructor and to vect_compile_time_alias.
5450 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5452 (vect_loop_versioning): Call it.
5453 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5455 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5458 Alan Hayward <alan.hayward@arm.com>
5459 David Sherwood <david.sherwood@arm.com>
5461 * doc/sourcebuild.texi (vect_scatter_store): Document.
5462 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5464 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5466 * genopinit.c (main): Add supports_vec_scatter_store and
5467 supports_vec_scatter_store_cached to target_optabs.
5468 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5469 IFN_MASK_SCATTER_STORE.
5470 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5472 * internal-fn.h (internal_store_fn_p): Declare.
5473 (internal_fn_stored_value_index): Likewise.
5474 * internal-fn.c (scatter_store_direct): New macro.
5475 (expand_scatter_store_optab_fn): New function.
5476 (direct_scatter_store_optab_supported_p): New macro.
5477 (internal_store_fn_p): New function.
5478 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5479 IFN_MASK_SCATTER_STORE.
5480 (internal_fn_mask_index): Likewise.
5481 (internal_fn_stored_value_index): New function.
5482 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5484 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5485 * optabs-query.c (supports_vec_scatter_store_p): New function.
5486 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5487 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5488 true for scatter stores.
5489 (vect_gather_scatter_fn_p): Handle scatter stores too.
5490 (vect_check_gather_scatter): Consider using scatter stores if
5491 supports_vec_scatter_store_p.
5492 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5494 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5495 internal_fn_stored_value_index.
5496 (check_load_store_masking): Handle scatter stores too.
5497 (vect_get_store_rhs): Make public.
5498 (vectorizable_call): Use internal_store_fn_p.
5499 (vectorizable_store): Handle scatter store internal functions.
5500 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5501 when deciding whether the end of the group has been reached.
5502 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5503 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5504 (mask_scatter_store<mode>): New insns.
5506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5507 Alan Hayward <alan.hayward@arm.com>
5508 David Sherwood <david.sherwood@arm.com>
5510 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5511 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5512 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5514 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5515 Use vect_truncate_gather_scatter_offset if we can't treat the
5516 operation as a normal gather load or scatter store.
5517 (get_group_load_store_type): Take the gather_scatter_info
5518 as argument. Try using a gather load or scatter store for
5519 single-element groups.
5520 (get_load_store_type): Update calls to get_group_load_store_type
5521 and vect_use_strided_gather_scatters_p.
5523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5524 Alan Hayward <alan.hayward@arm.com>
5525 David Sherwood <david.sherwood@arm.com>
5527 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5528 optional tree argument.
5529 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5531 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5532 but continue to use the current value as a fallback.
5533 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5534 to compare the updates.
5535 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5536 (get_load_store_type): Use it when handling a strided access.
5537 (vect_get_strided_load_store_ops): New function.
5538 (vect_get_data_ptr_increment): Likewise.
5539 (vectorizable_load): Handle strided gather loads. Always pass
5540 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5543 Alan Hayward <alan.hayward@arm.com>
5544 David Sherwood <david.sherwood@arm.com>
5546 * doc/md.texi (gather_load@var{m}): Document.
5547 (mask_gather_load@var{m}): Likewise.
5548 * genopinit.c (main): Add supports_vec_gather_load and
5549 supports_vec_gather_load_cached to target_optabs.
5550 * optabs-tree.c (init_tree_optimization_optabs): Use
5551 ggc_cleared_alloc to allocate target_optabs.
5552 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5553 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5555 * internal-fn.h (internal_load_fn_p): Declare.
5556 (internal_gather_scatter_fn_p): Likewise.
5557 (internal_fn_mask_index): Likewise.
5558 (internal_gather_scatter_fn_supported_p): Likewise.
5559 * internal-fn.c (gather_load_direct): New macro.
5560 (expand_gather_load_optab_fn): New function.
5561 (direct_gather_load_optab_supported_p): New macro.
5562 (direct_internal_fn_optab): New function.
5563 (internal_load_fn_p): Likewise.
5564 (internal_gather_scatter_fn_p): Likewise.
5565 (internal_fn_mask_index): Likewise.
5566 (internal_gather_scatter_fn_supported_p): Likewise.
5567 * optabs-query.c (supports_at_least_one_mode_p): New function.
5568 (supports_vec_gather_load_p): Likewise.
5569 * optabs-query.h (supports_vec_gather_load_p): Declare.
5570 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5571 and memory_type field.
5572 (NUM_PATTERNS): Bump to 15.
5573 * tree-vect-data-refs.c: Include internal-fn.h.
5574 (vect_gather_scatter_fn_p): New function.
5575 (vect_describe_gather_scatter_call): Likewise.
5576 (vect_check_gather_scatter): Try using internal functions for
5577 gather loads. Recognize existing calls to a gather load function.
5578 (vect_analyze_data_refs): Consider using gather loads if
5579 supports_vec_gather_load_p.
5580 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5581 (vect_get_gather_scatter_offset_type): Likewise.
5582 (vect_convert_mask_for_vectype): Likewise.
5583 (vect_add_conversion_to_patterm): Likewise.
5584 (vect_try_gather_scatter_pattern): Likewise.
5585 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5586 (vect_vect_recog_func_ptrs): Add it.
5587 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5588 internal_fn_mask_index and internal_gather_scatter_fn_p.
5589 (check_load_store_masking): Take the gather_scatter_info as an
5590 argument and handle gather loads.
5591 (vect_get_gather_scatter_ops): New function.
5592 (vectorizable_call): Check internal_load_fn_p.
5593 (vectorizable_load): Likewise. Handle gather load internal
5595 (vectorizable_store): Update call to check_load_store_masking.
5596 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5597 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5598 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5599 (aarch64_gather_scale_operand_d): New predicates.
5600 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5601 (mask_gather_load<mode>): New insns.
5603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5604 Alan Hayward <alan.hayward@arm.com>
5605 David Sherwood <david.sherwood@arm.com>
5607 * optabs.def (fold_left_plus_optab): New optab.
5608 * doc/md.texi (fold_left_plus_@var{m}): Document.
5609 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5610 * internal-fn.c (fold_left_direct): Define.
5611 (expand_fold_left_optab_fn): Likewise.
5612 (direct_fold_left_optab_supported_p): Likewise.
5613 * fold-const-call.c (fold_const_fold_left): New function.
5614 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5615 * tree-parloops.c (valid_reduction_p): New function.
5616 (gather_scalar_reductions): Use it.
5617 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5618 (vect_finish_replace_stmt): Declare.
5619 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5620 (needs_fold_left_reduction_p): New function, split out from...
5621 (vect_is_simple_reduction): ...here. Accept reductions that
5622 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5623 (vect_force_simple_reduction): Also store the reduction type in
5624 the assignment's STMT_VINFO_REDUC_TYPE.
5625 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5626 (merge_with_identity): New function.
5627 (vect_expand_fold_left): Likewise.
5628 (vectorize_fold_left_reduction): Likewise.
5629 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5630 scalar phi in place for it. Check for target support and reject
5631 cases that would reassociate the operation. Defer the transform
5632 phase to vectorize_fold_left_reduction.
5633 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5634 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5635 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5639 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5640 call to ifc_temp_var.
5642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5643 Alan Hayward <alan.hayward@arm.com>
5644 David Sherwood <david.sherwood@arm.com>
5646 * target.def (legitimize_address_displacement): Take the original
5647 offset as a poly_int.
5648 * targhooks.h (default_legitimize_address_displacement): Update
5650 * targhooks.c (default_legitimize_address_displacement): Likewise.
5651 * doc/tm.texi: Regenerate.
5652 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5653 as an argument, moving assert of ad->disp == ad->disp_term to...
5654 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5655 Try calling targetm.legitimize_address_displacement before expanding
5656 the address rather than afterwards, and adjust for the new interface.
5657 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5658 Match the new hook interface. Handle SVE addresses.
5659 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5664 * Makefile.in (OBJS): Add early-remat.o.
5665 * target.def (select_early_remat_modes): New hook.
5666 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5667 * doc/tm.texi: Regenerate.
5668 * targhooks.h (default_select_early_remat_modes): Declare.
5669 * targhooks.c (default_select_early_remat_modes): New function.
5670 * timevar.def (TV_EARLY_REMAT): New timevar.
5671 * passes.def (pass_early_remat): New pass.
5672 * tree-pass.h (make_pass_early_remat): Declare.
5673 * early-remat.c: New file.
5674 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5676 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5679 Alan Hayward <alan.hayward@arm.com>
5680 David Sherwood <david.sherwood@arm.com>
5682 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5683 vfm1 with a bound_epilog parameter.
5684 (vect_do_peeling): Update calls accordingly, and move the prologue
5685 call earlier in the function. Treat the base bound_epilog as 0 for
5686 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5687 this base when peeling for gaps.
5688 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5689 with fully-masked loops.
5690 (vect_estimate_min_profitable_iters): Handle the single peeled
5691 iteration in that case.
5693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5694 Alan Hayward <alan.hayward@arm.com>
5695 David Sherwood <david.sherwood@arm.com>
5697 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5698 single-element interleaving even if the size is not a power of 2.
5699 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5700 accesses for single-element interleaving if the group size is
5703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5704 Alan Hayward <alan.hayward@arm.com>
5705 David Sherwood <david.sherwood@arm.com>
5707 * doc/md.texi (fold_extract_last_@var{m}): Document.
5708 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5709 * optabs.def (fold_extract_last_optab): New optab.
5710 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5711 * internal-fn.c (fold_extract_direct): New macro.
5712 (expand_fold_extract_optab_fn): Likewise.
5713 (direct_fold_extract_optab_supported_p): Likewise.
5714 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5715 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5716 EXTRACT_LAST_REDUCTION.
5717 (get_initial_def_for_reduction): Do not create an initial vector
5718 for EXTRACT_LAST_REDUCTION reductions.
5719 (vectorizable_reduction): Leave the scalar phi in place for
5720 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5721 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5722 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5723 transform phase to vectorizable_condition.
5724 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5726 (vect_finish_stmt_generation): ...here.
5727 (vect_finish_replace_stmt): New function.
5728 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5729 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5731 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5734 Alan Hayward <alan.hayward@arm.com>
5735 David Sherwood <david.sherwood@arm.com>
5737 * doc/md.texi (extract_last_@var{m}): Document.
5738 * optabs.def (extract_last_optab): New optab.
5739 * internal-fn.def (EXTRACT_LAST): New internal function.
5740 * internal-fn.c (cond_unary_direct): New macro.
5741 (expand_cond_unary_optab_fn): Likewise.
5742 (direct_cond_unary_optab_supported_p): Likewise.
5743 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5744 loops using EXTRACT_LAST.
5745 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5746 (extract_last_<mode>): ...this optab.
5747 (vec_extract<mode><Vel>): Update accordingly.
5749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5750 Alan Hayward <alan.hayward@arm.com>
5751 David Sherwood <david.sherwood@arm.com>
5753 * target.def (empty_mask_is_expensive): New hook.
5754 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5755 * doc/tm.texi: Regenerate.
5756 * targhooks.h (default_empty_mask_is_expensive): Declare.
5757 * targhooks.c (default_empty_mask_is_expensive): New function.
5758 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5759 if the target says that empty masks are expensive.
5760 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5762 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5765 Alan Hayward <alan.hayward@arm.com>
5766 David Sherwood <david.sherwood@arm.com>
5768 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5769 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5770 (vect_use_loop_mask_for_alignment_p): New function.
5771 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5772 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5773 niters_skip argument. Make sure that the first niters_skip elements
5774 of the first iteration are inactive.
5775 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5776 Update call to vect_set_loop_masks_directly.
5777 (get_misalign_in_elems): New function, split out from...
5778 (vect_gen_prolog_loop_niters): ...here.
5779 (vect_update_init_of_dr): Take a code argument that specifies whether
5780 the adjustment should be added or subtracted.
5781 (vect_update_init_of_drs): Likewise.
5782 (vect_prepare_for_masked_peels): New function.
5783 (vect_do_peeling): Skip prologue peeling if we're using a mask
5784 instead. Update call to vect_update_inits_of_drs.
5785 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5787 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5788 alignment. Do not include the number of peeled iterations in
5789 the minimum threshold in that case.
5790 (vectorizable_induction): Adjust the start value down by
5791 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5792 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5793 Take the number of skipped iterations into account when calculating
5795 * tree-vect-stmts.c (vect_gen_while_not): New function.
5797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5798 Alan Hayward <alan.hayward@arm.com>
5799 David Sherwood <david.sherwood@arm.com>
5801 * doc/sourcebuild.texi (vect_fully_masked): Document.
5802 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5804 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5806 (vect_analyze_loop_2): ...here. Don't check the vectorization
5807 factor against the number of loop iterations if the loop is
5810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5811 Alan Hayward <alan.hayward@arm.com>
5812 David Sherwood <david.sherwood@arm.com>
5814 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5815 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5816 (dump_groups): Update accordingly.
5817 (iv_use::mem_type): New member variable.
5818 (address_p): New function.
5819 (record_use): Add a mem_type argument and initialize the new
5821 (record_group_use): Add a mem_type argument. Use address_p.
5822 Remove obsolete null checks of base_object. Update call to record_use.
5823 (find_interesting_uses_op): Update call to record_group_use.
5824 (find_interesting_uses_cond): Likewise.
5825 (find_interesting_uses_address): Likewise.
5826 (get_mem_type_for_internal_fn): New function.
5827 (find_address_like_use): Likewise.
5828 (find_interesting_uses_stmt): Try find_address_like_use before
5829 calling find_interesting_uses_op.
5830 (addr_offset_valid_p): Use the iv mem_type field as the type
5831 of the addressed memory.
5832 (add_autoinc_candidates): Likewise.
5833 (get_address_cost): Likewise.
5834 (split_small_address_groups_p): Use address_p.
5835 (split_address_groups): Likewise.
5836 (add_iv_candidate_for_use): Likewise.
5837 (autoinc_possible_for_pair): Likewise.
5838 (rewrite_groups): Likewise.
5839 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5840 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5841 (get_alias_ptr_type_for_ptr_address): New function.
5842 (rewrite_use_address): Rewrite address uses in calls that were
5843 identified by find_address_like_use.
5845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5846 Alan Hayward <alan.hayward@arm.com>
5847 David Sherwood <david.sherwood@arm.com>
5849 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5851 * gimple-expr.h (is_gimple_addressable: Likewise.
5852 * gimple-expr.c (is_gimple_address): Likewise.
5853 * internal-fn.c (expand_call_mem_ref): New function.
5854 (expand_mask_load_optab_fn): Use it.
5855 (expand_mask_store_optab_fn): Likewise.
5857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5858 Alan Hayward <alan.hayward@arm.com>
5859 David Sherwood <david.sherwood@arm.com>
5861 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5862 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5863 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5864 (cond_umax@var{mode}): Document.
5865 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5866 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5867 (cond_umin_optab, cond_umax_optab): New optabs.
5868 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5869 (COND_IOR, COND_XOR): New internal functions.
5870 * internal-fn.h (get_conditional_internal_fn): Declare.
5871 * internal-fn.c (cond_binary_direct): New macro.
5872 (expand_cond_binary_optab_fn): Likewise.
5873 (direct_cond_binary_optab_supported_p): Likewise.
5874 (get_conditional_internal_fn): New function.
5875 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5876 Cope with reduction statements that are vectorized as calls rather
5878 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5879 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5880 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5881 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5882 (UNSPEC_COND_EOR): New unspecs.
5883 (optab): Add mappings for them.
5884 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5885 (sve_int_op, sve_fp_op): New int attributes.
5887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5888 Alan Hayward <alan.hayward@arm.com>
5889 David Sherwood <david.sherwood@arm.com>
5891 * optabs.def (while_ult_optab): New optab.
5892 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5893 * internal-fn.def (WHILE_ULT): New internal function.
5894 * internal-fn.h (direct_internal_fn_supported_p): New override
5895 that takes two types as argument.
5896 * internal-fn.c (while_direct): New macro.
5897 (expand_while_optab_fn): New function.
5898 (convert_optab_supported_p): Likewise.
5899 (direct_while_optab_supported_p): New macro.
5900 * wide-int.h (wi::udiv_ceil): New function.
5901 * tree-vectorizer.h (rgroup_masks): New structure.
5902 (vec_loop_masks): New typedef.
5903 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5905 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5906 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5907 (vect_max_vf): New function.
5908 (slpeel_make_loop_iterate_ntimes): Delete.
5909 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5910 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5911 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5912 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5913 internal-fn.h, stor-layout.h and optabs-query.h.
5914 (vect_set_loop_mask): New function.
5915 (add_preheader_seq): Likewise.
5916 (add_header_seq): Likewise.
5917 (interleave_supported_p): Likewise.
5918 (vect_maybe_permute_loop_masks): Likewise.
5919 (vect_set_loop_masks_directly): Likewise.
5920 (vect_set_loop_condition_masked): Likewise.
5921 (vect_set_loop_condition_unmasked): New function, split out from
5922 slpeel_make_loop_iterate_ntimes.
5923 (slpeel_make_loop_iterate_ntimes): Rename to..
5924 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5925 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5926 (vect_do_peeling): Update call accordingly.
5927 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5929 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5930 mask_compare_type, can_fully_mask_p and fully_masked_p.
5931 (release_vec_loop_masks): New function.
5932 (_loop_vec_info): Use it to free the loop masks.
5933 (can_produce_all_loop_masks_p): New function.
5934 (vect_get_max_nscalars_per_iter): Likewise.
5935 (vect_verify_full_masking): Likewise.
5936 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5937 retries, and free the mask rgroups before retrying. Check loop-wide
5938 reasons for disallowing fully-masked loops. Make the final decision
5939 about whether use a fully-masked loop or not.
5940 (vect_estimate_min_profitable_iters): Do not assume that peeling
5941 for the number of iterations will be needed for fully-masked loops.
5942 (vectorizable_reduction): Disable fully-masked loops.
5943 (vectorizable_live_operation): Likewise.
5944 (vect_halve_mask_nunits): New function.
5945 (vect_double_mask_nunits): Likewise.
5946 (vect_record_loop_mask): Likewise.
5947 (vect_get_loop_mask): Likewise.
5948 (vect_transform_loop): Handle the case in which the final loop
5949 iteration might handle a partial vector. Call vect_set_loop_condition
5950 instead of slpeel_make_loop_iterate_ntimes.
5951 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5952 (check_load_store_masking): New function.
5953 (prepare_load_store_mask): Likewise.
5954 (vectorizable_store): Handle fully-masked loops.
5955 (vectorizable_load): Likewise.
5956 (supportable_widening_operation): Use vect_halve_mask_nunits for
5958 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5959 (vect_gen_while): New function.
5960 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5961 (aarch64_uqdec<mode>): New insn.
5963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5964 Alan Hayward <alan.hayward@arm.com>
5965 David Sherwood <david.sherwood@arm.com>
5967 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5968 (reduc_xor_scal_optab): New optabs.
5969 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5970 (reduc_xor_scal_@var{m}): Document.
5971 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5972 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5974 * fold-const-call.c (fold_const_call): Handle them.
5975 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5976 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5977 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5978 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5979 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5980 (UNSPEC_XORV): New unspecs.
5981 (optab): Add entries for them.
5982 (BITWISEV): New int iterator.
5983 (bit_reduc_op): New int attributes.
5985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5986 Alan Hayward <alan.hayward@arm.com>
5987 David Sherwood <david.sherwood@arm.com>
5989 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5990 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5991 * optabs.def (vec_shl_insert_optab): New optab.
5992 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5993 (duplicate_and_interleave): Likewise.
5994 * tree-vect-loop.c: Include internal-fn.h.
5995 (neutral_op_for_slp_reduction): New function, split out from
5996 get_initial_defs_for_reduction.
5997 (get_initial_def_for_reduction): Handle option 2 for variable-length
5998 vectors by loading the neutral value into a vector and then shifting
5999 the initial value into element 0.
6000 (get_initial_defs_for_reduction): Replace the code argument with
6001 the neutral value calculated by neutral_op_for_slp_reduction.
6002 Use gimple_build_vector for constant-length vectors.
6003 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6004 but the first group_size elements have a neutral value.
6005 Use duplicate_and_interleave otherwise.
6006 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6007 Update call to get_initial_defs_for_reduction. Handle SLP
6008 reductions for variable-length vectors by creating one vector
6009 result for each scalar result, with the elements associated
6010 with other scalar results stubbed out with the neutral value.
6011 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6012 Require IFN_VEC_SHL_INSERT for double reductions on
6013 variable-length vectors, or SLP reductions that have
6014 a neutral value. Require can_duplicate_and_interleave_p
6015 support for variable-length unchained SLP reductions if there
6016 is no neutral value, such as for MIN/MAX reductions. Also require
6017 the number of vector elements to be a multiple of the number of
6018 SLP statements when doing variable-length unchained SLP reductions.
6019 Update call to vect_create_epilog_for_reduction.
6020 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6021 and remove initial values.
6022 (duplicate_and_interleave): Make public.
6023 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6024 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6027 Alan Hayward <alan.hayward@arm.com>
6028 David Sherwood <david.sherwood@arm.com>
6030 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6031 (can_duplicate_and_interleave_p): New function.
6032 (vect_get_and_check_slp_defs): Take the vector of statements
6033 rather than just the current one. Remove excess parentheses.
6034 Restriction rejectinon of vect_constant_def and vect_external_def
6035 for variable-length vectors to boolean types, or types for which
6036 can_duplicate_and_interleave_p is false.
6037 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6038 (duplicate_and_interleave): New function.
6039 (vect_get_constant_vectors): Use gimple_build_vector for
6040 constant-length vectors and suitable variable-length constant
6041 vectors. Use duplicate_and_interleave for other variable-length
6042 vectors. Don't defer the update when inserting new statements.
6044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6045 Alan Hayward <alan.hayward@arm.com>
6046 David Sherwood <david.sherwood@arm.com>
6048 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6049 min_profitable_iters doesn't go negative.
6051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6052 Alan Hayward <alan.hayward@arm.com>
6053 David Sherwood <david.sherwood@arm.com>
6055 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6056 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6057 * optabs.def (vec_mask_load_lanes_optab): New optab.
6058 (vec_mask_store_lanes_optab): Likewise.
6059 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6060 (MASK_STORE_LANES): Likewise.
6061 * internal-fn.c (mask_load_lanes_direct): New macro.
6062 (mask_store_lanes_direct): Likewise.
6063 (expand_mask_load_optab_fn): Handle masked operations.
6064 (expand_mask_load_lanes_optab_fn): New macro.
6065 (expand_mask_store_optab_fn): Handle masked operations.
6066 (expand_mask_store_lanes_optab_fn): New macro.
6067 (direct_mask_load_lanes_optab_supported_p): Likewise.
6068 (direct_mask_store_lanes_optab_supported_p): Likewise.
6069 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6071 (vect_load_lanes_supported): Likewise.
6072 * tree-vect-data-refs.c (strip_conversion): New function.
6073 (can_group_stmts_p): Likewise.
6074 (vect_analyze_data_ref_accesses): Use it instead of checking
6075 for a pair of assignments.
6076 (vect_store_lanes_supported): Take a masked_p parameter.
6077 (vect_load_lanes_supported): Likewise.
6078 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6079 vect_store_lanes_supported and vect_load_lanes_supported.
6080 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6081 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6082 parameter. Don't allow gaps for masked accesses.
6083 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6084 and vect_load_lanes_supported.
6085 (get_load_store_type): Take a masked_p parameter and update
6086 call to get_group_load_store_type.
6087 (vectorizable_store): Update call to get_load_store_type.
6088 Handle IFN_MASK_STORE_LANES.
6089 (vectorizable_load): Update call to get_load_store_type.
6090 Handle IFN_MASK_LOAD_LANES.
6092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6093 Alan Hayward <alan.hayward@arm.com>
6094 David Sherwood <david.sherwood@arm.com>
6096 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6098 * config/aarch64/aarch64-protos.h
6099 (aarch64_sve_struct_memory_operand_p): Declare.
6100 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6101 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6102 (VPRED, vpred): Handle SVE structure modes.
6103 * config/aarch64/constraints.md (Utx): New constraint.
6104 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6105 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6106 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6107 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6108 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6109 structure modes. Split into pieces after RA.
6110 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6111 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6113 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6114 SVE structure modes.
6115 (aarch64_classify_address): Likewise.
6116 (sizetochar): Move earlier in file.
6117 (aarch64_print_operand): Handle SVE register lists.
6118 (aarch64_array_mode): New function.
6119 (aarch64_sve_struct_memory_operand_p): Likewise.
6120 (TARGET_ARRAY_MODE): Redefine.
6122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6123 Alan Hayward <alan.hayward@arm.com>
6124 David Sherwood <david.sherwood@arm.com>
6126 * target.def (array_mode): New target hook.
6127 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6128 * doc/tm.texi: Regenerate.
6129 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6130 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6131 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6133 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6137 Alan Hayward <alan.hayward@arm.com>
6138 David Sherwood <david.sherwood@arm.com>
6140 * fold-const.c (fold_binary_loc): Check the argument types
6141 rather than the result type when testing for a vector operation.
6143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6145 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6146 * doc/tm.texi: Regenerate.
6148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6149 Alan Hayward <alan.hayward@arm.com>
6150 David Sherwood <david.sherwood@arm.com>
6152 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6153 (sve): Document new AArch64 extension.
6154 * doc/md.texi (w): Extend the description of the AArch64
6155 constraint to include SVE vectors.
6156 (Upl, Upa): Document new AArch64 predicate constraints.
6157 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6159 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6160 (msve-vector-bits=): New option.
6161 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6162 SVE when these are disabled.
6163 (sve): New extension.
6164 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6165 modes. Adjust their number of units based on aarch64_sve_vg.
6166 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6167 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6168 aarch64_addr_query_type.
6169 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6170 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6171 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6172 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6173 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6174 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6175 (aarch64_simd_imm_zero_p): Delete.
6176 (aarch64_check_zero_based_sve_index_immediate): Declare.
6177 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6178 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6179 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6180 (aarch64_sve_float_mul_immediate_p): Likewise.
6181 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6183 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6184 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6185 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6186 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6187 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6188 (aarch64_regmode_natural_size): Likewise.
6189 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6190 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6192 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6193 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6194 for VG and the SVE predicate registers.
6195 (V_ALIASES): Add a "z"-prefixed alias.
6196 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6197 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6198 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6199 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6200 (REG_CLASS_NAMES): Add entries for them.
6201 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6202 and the predicate registers.
6203 (aarch64_sve_vg): Declare.
6204 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6205 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6206 (REGMODE_NATURAL_SIZE): Define.
6207 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6209 * config/aarch64/aarch64.c: Include cfgrtl.h.
6210 (simd_immediate_info): Add a constructor for series vectors,
6211 and an associated step field.
6212 (aarch64_sve_vg): New variable.
6213 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6214 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6215 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6216 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6217 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6218 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6219 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6220 (aarch64_get_mask_mode): New functions.
6221 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6222 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6223 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6224 predicate modes and predicate registers. Explicitly restrict
6225 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6226 to store a vector mode if it is recognized by
6227 aarch64_classify_vector_mode.
6228 (aarch64_regmode_natural_size): New function.
6229 (aarch64_hard_regno_caller_save_mode): Return the original mode
6231 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6232 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6233 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6234 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6236 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6237 does not overlap dest if the function is frame-related. Handle
6239 (aarch64_split_add_offset): New function.
6240 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6241 them aarch64_add_offset.
6242 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6243 and update call to aarch64_sub_sp.
6244 (aarch64_add_cfa_expression): New function.
6245 (aarch64_expand_prologue): Pass extra temporary registers to the
6246 functions above. Handle the case in which we need to emit new
6247 DW_CFA_expressions for registers that were originally saved
6248 relative to the stack pointer, but now have to be expressed
6249 relative to the frame pointer.
6250 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6252 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6253 IP0 and IP1 values for SVE frames.
6254 (aarch64_expand_vec_series): New function.
6255 (aarch64_expand_sve_widened_duplicate): Likewise.
6256 (aarch64_expand_sve_const_vector): Likewise.
6257 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6258 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6259 into the register, rather than emitting a SET directly.
6260 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6261 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6262 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6263 (offset_9bit_signed_scaled_p): New functions.
6264 (aarch64_replicate_bitmask_imm): New function.
6265 (aarch64_bitmask_imm): Use it.
6266 (aarch64_cannot_force_const_mem): Reject expressions involving
6267 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6268 (aarch64_classify_index): Handle SVE indices, by requiring
6269 a plain register index with a scale that matches the element size.
6270 (aarch64_classify_address): Handle SVE addresses. Assert that
6271 the mode of the address is VOIDmode or an integer mode.
6272 Update call to aarch64_classify_symbol.
6273 (aarch64_classify_symbolic_expression): Update call to
6274 aarch64_classify_symbol.
6275 (aarch64_const_vec_all_in_range_p): New function.
6276 (aarch64_print_vector_float_operand): Likewise.
6277 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6278 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6279 and the FP immediates 1.0 and 0.5.
6280 (aarch64_print_address_internal): Handle SVE addresses.
6281 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6282 (aarch64_regno_regclass): Handle predicate registers.
6283 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6285 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6286 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6287 (aarch64_convert_sve_vector_bits): New function.
6288 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6289 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6291 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6292 Handle SVE vector and predicate modes. Accept VL-based constants
6293 that need only one temporary register, and VL offsets that require
6294 no temporary registers.
6295 (aarch64_conditional_register_usage): Mark the predicate registers
6296 as fixed if SVE isn't available.
6297 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6298 Return true for SVE vector and predicate modes.
6299 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6300 rather than an unsigned int. Handle SVE modes.
6301 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6303 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6305 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6306 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6307 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6308 (aarch64_sve_float_mul_immediate_p): New functions.
6309 (aarch64_sve_valid_immediate): New function.
6310 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6311 Explicitly reject structure modes. Check for INDEX constants.
6312 Handle PTRUE and PFALSE constants.
6313 (aarch64_check_zero_based_sve_index_immediate): New function.
6314 (aarch64_simd_imm_zero_p): Delete.
6315 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6316 vector modes. Accept constants in the range of CNT[BHWD].
6317 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6318 ask for an Advanced SIMD mode.
6319 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6320 (aarch64_simd_vector_alignment): Handle SVE predicates.
6321 (aarch64_vectorize_preferred_vector_alignment): New function.
6322 (aarch64_simd_vector_alignment_reachable): Use it instead of
6324 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6325 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6327 (MAX_VECT_LEN): Delete.
6328 (expand_vec_perm_d): Add a vec_flags field.
6329 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6330 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6331 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6333 (aarch64_evpc_rev): Rename to...
6334 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6335 (aarch64_evpc_rev_global): New function.
6336 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6337 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6339 (aarch64_evpc_sve_tbl): New function.
6340 (aarch64_expand_vec_perm_const_1): Update after rename of
6341 aarch64_evpc_rev. Handle SVE permutes too, trying
6342 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6343 than aarch64_evpc_tbl.
6344 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6345 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6346 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6347 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6348 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6349 (aarch64_expand_sve_vcond): New functions.
6350 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6351 of aarch64_vector_mode_p.
6352 (aarch64_dwarf_poly_indeterminate_value): New function.
6353 (aarch64_compute_pressure_classes): Likewise.
6354 (aarch64_can_change_mode_class): Likewise.
6355 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6356 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6357 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6358 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6359 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6360 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6361 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6362 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6364 (Dn, Dl, Dr): Accept const as well as const_vector.
6365 (Dz): Likewise. Compare against CONST0_RTX.
6366 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6367 of "vector" where appropriate.
6368 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6369 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6370 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6371 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6372 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6373 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6374 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6375 (v_int_equiv): Extend to SVE modes.
6376 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6378 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6379 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6380 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6381 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6382 (SVE_COND_FP_CMP): New int iterators.
6383 (perm_hilo): Handle the new unpack unspecs.
6384 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6386 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6387 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6388 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6389 (aarch64_equality_operator, aarch64_constant_vector_operand)
6390 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6391 (aarch64_sve_nonimmediate_operand): Likewise.
6392 (aarch64_sve_general_operand): Likewise.
6393 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6394 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6395 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6396 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6397 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6398 (aarch64_sve_float_arith_immediate): Likewise.
6399 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6400 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6401 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6402 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6403 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6404 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6405 (aarch64_sve_float_arith_operand): Likewise.
6406 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6407 (aarch64_sve_float_mul_operand): Likewise.
6408 (aarch64_sve_vec_perm_operand): Likewise.
6409 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6410 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6411 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6412 as well as const_vector.
6413 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6414 in file. Use CONST0_RTX and CONSTM1_RTX.
6415 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6416 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6417 Use aarch64_simd_imm_zero.
6418 * config/aarch64/aarch64-sve.md: New file.
6419 * config/aarch64/aarch64.md: Include it.
6420 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6421 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6422 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6423 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6424 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6425 (sve): New attribute.
6426 (enabled): Disable instructions with the sve attribute unless
6428 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6429 aarch64_expand_mov_immediate.
6430 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6431 CNT[BHSD] immediates.
6432 (movti): Split CONST_POLY_INT moves into two halves.
6433 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6434 Split additions that need a temporary here if the destination
6435 is the stack pointer.
6436 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6437 (*add<mode>3_poly_1): New instruction.
6438 (set_clobber_cc): New expander.
6440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6442 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6443 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6444 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6445 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6446 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6447 Change innermode from fixed_mode_size to machine_mode.
6448 (simplify_subreg): Update call accordingly. Handle a constant-sized
6449 subreg of a variable-length CONST_VECTOR.
6451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6452 Alan Hayward <alan.hayward@arm.com>
6453 David Sherwood <david.sherwood@arm.com>
6455 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6456 (add_offset_to_base): New function, split out from...
6457 (create_mem_ref): ...here. When handling a scale other than 1,
6458 check first whether the address is valid without the offset.
6459 Add it into the base if so, leaving the index and scale as-is.
6461 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6464 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6465 fold_for_warn before checking if arg2 is INTEGER_CST.
6467 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6469 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6470 (store_multiple_operation): Delete.
6471 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6472 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6473 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6474 guarded by TARGET_STRING.
6475 (rs6000_output_load_multiple): Delete.
6476 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6477 OPTION_MASK_STRING / TARGET_STRING handling.
6478 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6479 (const rs6000_opt_masks) <"string">: Change mask to 0.
6480 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6481 (MASK_STRING): Delete.
6482 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6484 (load_multiple): Delete.
6491 (store_multiple): Delete.
6498 (movmemsi_8reg): Delete.
6499 (corresponding unnamed define_insn): Delete.
6500 (movmemsi_6reg): Delete.
6501 (corresponding unnamed define_insn): Delete.
6502 (movmemsi_4reg): Delete.
6503 (corresponding unnamed define_insn): Delete.
6504 (movmemsi_2reg): Delete.
6505 (corresponding unnamed define_insn): Delete.
6506 (movmemsi_1reg): Delete.
6507 (corresponding unnamed define_insn): Delete.
6508 * config/rs6000/rs6000.opt (mno-string): New.
6509 (mstring): Replace by deprecation warning stub.
6510 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6512 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6514 * regrename.c (regrename_do_replace): If replacing the same
6515 reg multiple times, try to reuse last created gen_raw_REG.
6518 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6519 main to workaround a bug in GDB.
6521 2018-01-12 Tom de Vries <tom@codesourcery.com>
6524 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6526 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6528 PR rtl-optimization/80481
6529 * ira-color.c (get_cap_member): New function.
6530 (allocnos_conflict_by_live_ranges_p): Use it.
6531 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6532 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6534 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6537 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6538 (*saddl_se_1): Ditto.
6540 (*ssubl_se_1): Ditto.
6542 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6544 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6545 rather than wi::to_widest for DR_INITs.
6546 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6547 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6548 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6550 (vect_analyze_group_access_1): Note that here.
6552 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6554 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6555 polynomial type sizes.
6557 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6559 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6560 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6561 (gimple_add_tmp_var): Likewise.
6563 2018-01-12 Martin Liska <mliska@suse.cz>
6565 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6566 (gimple_alloc_sizes): Likewise.
6567 (dump_gimple_statistics): Use PRIu64 in printf format.
6568 * gimple.h: Change uint64_t to int.
6570 2018-01-12 Martin Liska <mliska@suse.cz>
6572 * tree-core.h: Use uint64_t instead of int.
6573 * tree.c (tree_node_counts): Likewise.
6574 (tree_node_sizes): Likewise.
6575 (dump_tree_statistics): Use PRIu64 in printf format.
6577 2018-01-12 Martin Liska <mliska@suse.cz>
6579 * Makefile.in: As qsort_chk is implemented in vec.c, add
6580 vec.o to linkage of gencfn-macros.
6581 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6582 passing the info to record_node_allocation_statistics.
6583 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6585 * ggc-common.c (struct ggc_usage): Add operator== and use
6586 it in operator< and compare function.
6587 * mem-stats.h (struct mem_usage): Likewise.
6588 * vec.c (struct vec_usage): Remove operator< and compare
6589 function. Can be simply inherited.
6591 2018-01-12 Martin Jambor <mjambor@suse.cz>
6594 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6595 * tree-ssa-math-opts.c: Include domwalk.h.
6596 (convert_mult_to_fma_1): New function.
6597 (fma_transformation_info): New type.
6598 (fma_deferring_state): Likewise.
6599 (cancel_fma_deferring): New function.
6600 (result_of_phi): Likewise.
6601 (last_fma_candidate_feeds_initial_phi): Likewise.
6602 (convert_mult_to_fma): Added deferring logic, split actual
6603 transformation to convert_mult_to_fma_1.
6604 (math_opts_dom_walker): New type.
6605 (math_opts_dom_walker::after_dom_children): New method, body moved
6606 here from pass_optimize_widening_mul::execute, added deferring logic
6608 (pass_optimize_widening_mul::execute): Moved most of code to
6609 math_opts_dom_walker::after_dom_children.
6610 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6611 * config/i386/i386.c (ix86_option_override_internal): Added
6612 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6614 2018-01-12 Richard Biener <rguenther@suse.de>
6617 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6618 inline instance vars.
6620 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6623 * config/rx/rx.c (rx_is_restricted_memory_address):
6626 2018-01-12 Richard Biener <rguenther@suse.de>
6628 PR tree-optimization/80846
6629 * target.def (split_reduction): New target hook.
6630 * targhooks.c (default_split_reduction): New function.
6631 * targhooks.h (default_split_reduction): Declare.
6632 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6633 target requests first reduce vectors by combining low and high
6635 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6636 (get_vectype_for_scalar_type_and_size): Export.
6637 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6638 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6639 * doc/tm.texi: Regenerate.
6640 * config/i386/i386.c (ix86_split_reduction): Implement
6641 TARGET_VECTORIZE_SPLIT_REDUCTION.
6643 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6646 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6647 in PIC mode except for TARGET_VXWORKS_RTP.
6648 * config/sparc/sparc.c: Include cfgrtl.h.
6649 (TARGET_INIT_PIC_REG): Define.
6650 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6651 (sparc_pic_register_p): New predicate.
6652 (sparc_legitimate_address_p): Use it.
6653 (sparc_legitimize_pic_address): Likewise.
6654 (sparc_delegitimize_address): Likewise.
6655 (sparc_mode_dependent_address_p): Likewise.
6656 (gen_load_pcrel_sym): Remove 4th parameter.
6657 (load_got_register): Adjust call to above. Remove obsolete stuff.
6658 (sparc_expand_prologue): Do not call load_got_register here.
6659 (sparc_flat_expand_prologue): Likewise.
6660 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6661 (sparc_use_pseudo_pic_reg): New function.
6662 (sparc_init_pic_reg): Likewise.
6663 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6664 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6666 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6668 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6669 Add item for branch_cost.
6671 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6673 PR rtl-optimization/83565
6674 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6675 not extend the result to a larger mode for rotate operations.
6676 (num_sign_bit_copies1): Likewise.
6678 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6681 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6683 Use values-Xc.o for -pedantic.
6684 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6686 2018-01-12 Martin Liska <mliska@suse.cz>
6689 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6691 (possible_polymorphic_call_targets): Use it.
6692 (ipa_devirt): Likewise.
6694 2018-01-12 Martin Liska <mliska@suse.cz>
6696 * profile-count.h (enum profile_quality): Use 0 as invalid
6697 enum value of profile_quality.
6699 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6701 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6702 -mext-string options.
6704 2018-01-12 Richard Biener <rguenther@suse.de>
6706 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6707 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6708 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6710 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6712 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6714 * configure.ac (--with-long-double-format): Add support for the
6715 configuration option to change the default long double format on
6717 * config.gcc (powerpc*-linux*-*): Likewise.
6718 * configure: Regenerate.
6719 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6720 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6721 used without modification.
6723 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6725 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6726 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6727 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6728 MISC_BUILTIN_SPEC_BARRIER.
6729 (rs6000_init_builtins): Likewise.
6730 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6732 (speculation_barrier): New define_insn.
6733 * doc/extend.texi: Document __builtin_speculation_barrier.
6735 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6738 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6739 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6740 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6742 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6743 integral modes instead of "ss" and "sd".
6744 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6745 vectors with 32-bit and 64-bit elements.
6746 (vecdupssescalarmodesuffix): New mode attribute.
6747 (vec_dup<mode>): Use it.
6749 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6752 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6753 frame if argument is passed on stack.
6755 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6758 * ree.c (combine_reaching_defs): Optimize also
6759 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6760 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6762 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6765 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6767 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6770 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6771 after they are computed.
6773 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6775 PR tree-optimization/83695
6776 * gimple-loop-linterchange.cc
6777 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6778 reset cached scev information after interchange.
6779 (pass_linterchange::execute): Remove call to scev_reset_htab.
6781 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6783 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6784 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6785 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6786 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6787 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6788 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6789 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6790 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6791 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6792 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6793 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6794 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6795 (V_lane_reg): Likewise.
6796 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6798 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6799 (vfmal_lane_low<mode>_intrinsic,
6800 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6801 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6802 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6803 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6804 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6805 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6807 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6809 * config/arm/arm-cpus.in (fp16fml): New feature.
6810 (ALL_SIMD): Add fp16fml.
6811 (armv8.2-a): Add fp16fml as an option.
6812 (armv8.3-a): Likewise.
6813 (armv8.4-a): Add fp16fml as part of fp16.
6814 * config/arm/arm.h (TARGET_FP16FML): Define.
6815 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6817 * config/arm/arm-modes.def (V2HF): Define.
6818 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6819 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6820 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6821 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6822 vfmsl_low, vfmsl_high): New set of builtins.
6823 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6824 (vfml_op): New code attribute.
6825 (VFMLHALVES): New int iterator.
6826 (VFML, VFMLSEL): New mode attributes.
6827 (V_reg): Define mapping for V2HF.
6828 (V_hi, V_lo): New mode attributes.
6829 (VF_constraint): Likewise.
6830 (vfml_half, vfml_half_selector): New int attributes.
6831 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6833 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6834 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6836 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6837 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6838 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6839 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6841 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6842 Document new effective target and option set.
6844 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6846 * config/arm/arm-cpus.in (armv8_4): New feature.
6847 (ARMv8_4a): New fgroup.
6848 (armv8.4-a): New arch.
6849 * config/arm/arm-tables.opt: Regenerate.
6850 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6851 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6852 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6853 Add matching rules for -march=armv8.4-a and extensions.
6854 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6856 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6859 * config/rx/rx.md (BW): New mode attribute.
6860 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6862 2018-01-11 Richard Biener <rguenther@suse.de>
6864 PR tree-optimization/83435
6865 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6866 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6867 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6869 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6870 Alan Hayward <alan.hayward@arm.com>
6871 David Sherwood <david.sherwood@arm.com>
6873 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6875 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6876 (aarch64_print_address_internal): Use it to check for a zero offset.
6878 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6879 Alan Hayward <alan.hayward@arm.com>
6880 David Sherwood <david.sherwood@arm.com>
6882 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6883 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6884 Return a poly_int64 rather than a HOST_WIDE_INT.
6885 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6886 rather than a HOST_WIDE_INT.
6887 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6888 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6889 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6890 final_offset from HOST_WIDE_INT to poly_int64.
6891 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6892 to_constant when getting the number of units in an Advanced SIMD
6894 (aarch64_builtin_vectorized_function): Check for a constant number
6896 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6898 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6899 attribute instead of GET_MODE_NUNITS.
6900 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6901 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6902 GET_MODE_SIZE for fixed-size registers.
6903 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6904 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6905 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6906 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6907 (aarch64_print_operand, aarch64_print_address_internal)
6908 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6909 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6910 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6911 Handle polynomial GET_MODE_SIZE.
6912 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6913 wider than SImode without modification.
6914 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6915 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6916 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6917 passing and returning SVE modes.
6918 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6919 rather than GEN_INT.
6920 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6921 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6922 (aarch64_allocate_and_probe_stack_space): Likewise.
6923 (aarch64_layout_frame): Cope with polynomial offsets.
6924 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6925 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6927 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6928 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6929 poly_int64 rather than a HOST_WIDE_INT.
6930 (aarch64_get_separate_components, aarch64_process_components)
6931 (aarch64_expand_prologue, aarch64_expand_epilogue)
6932 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6933 (aarch64_anchor_offset): New function, split out from...
6934 (aarch64_legitimize_address): ...here.
6935 (aarch64_builtin_vectorization_cost): Handle polynomial
6936 TYPE_VECTOR_SUBPARTS.
6937 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6939 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6940 number of elements from the PARALLEL rather than the mode.
6941 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6942 rather than GET_MODE_BITSIZE.
6943 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6944 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6945 (aarch64_expand_vec_perm_const_1): Handle polynomial
6946 d->perm.length () and d->perm elements.
6947 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6948 Apply to_constant to d->perm elements.
6949 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6950 polynomial CONST_VECTOR_NUNITS.
6951 (aarch64_move_pointer): Take amount as a poly_int64 rather
6953 (aarch64_progress_pointer): Avoid temporary variable.
6954 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6955 the mode attribute instead of GET_MODE.
6957 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6958 Alan Hayward <alan.hayward@arm.com>
6959 David Sherwood <david.sherwood@arm.com>
6961 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6962 x exists before using it.
6963 (aarch64_add_constant_internal): Rename to...
6964 (aarch64_add_offset_1): ...this. Replace regnum with separate
6965 src and dest rtxes. Handle the case in which they're different,
6966 including when the offset is zero. Replace scratchreg with an rtx.
6967 Use 2 additions if there is no spare register into which we can
6968 move a 16-bit constant.
6969 (aarch64_add_constant): Delete.
6970 (aarch64_add_offset): Replace reg with separate src and dest
6971 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6972 Use aarch64_add_offset_1.
6973 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6974 an rtx rather than an int. Take the delta as a poly_int64
6975 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6976 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6977 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6978 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6979 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6981 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6982 aarch64_add_constant.
6984 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6986 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6987 Use scalar_float_mode.
6989 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6991 * config/aarch64/aarch64-simd.md
6992 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6993 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6994 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6995 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6996 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6997 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6998 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6999 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7000 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7001 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7003 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7006 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7007 targ_options->x_arm_arch_string is non NULL.
7009 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7011 * config/aarch64/aarch64.h
7012 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7014 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7017 * expmed.c (emit_store_flag_force): Swap if const op0
7018 and change VOIDmode to mode of op0.
7020 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7022 PR rtl-optimization/83761
7023 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7024 than bytes to mode_for_size.
7026 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7029 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7030 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7033 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7036 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7037 when in layout mode.
7038 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7039 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7042 2018-01-10 Michael Collison <michael.collison@arm.com>
7044 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7045 * config/aarch64/aarch64-option-extension.def: Add
7046 AARCH64_OPT_EXTENSION of 'fp16fml'.
7047 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7048 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7049 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7050 * config/aarch64/constraints.md (Ui7): New constraint.
7051 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7052 (VFMLA_SEL_W): Ditto.
7055 (VFMLA16_LOW): New int iterator.
7056 (VFMLA16_HIGH): Ditto.
7057 (UNSPEC_FMLAL): New unspec.
7058 (UNSPEC_FMLSL): Ditto.
7059 (UNSPEC_FMLAL2): Ditto.
7060 (UNSPEC_FMLSL2): Ditto.
7061 (f16mac): New code attribute.
7062 * config/aarch64/aarch64-simd-builtins.def
7063 (aarch64_fmlal_lowv2sf): Ditto.
7064 (aarch64_fmlsl_lowv2sf): Ditto.
7065 (aarch64_fmlalq_lowv4sf): Ditto.
7066 (aarch64_fmlslq_lowv4sf): Ditto.
7067 (aarch64_fmlal_highv2sf): Ditto.
7068 (aarch64_fmlsl_highv2sf): Ditto.
7069 (aarch64_fmlalq_highv4sf): Ditto.
7070 (aarch64_fmlslq_highv4sf): Ditto.
7071 (aarch64_fmlal_lane_lowv2sf): Ditto.
7072 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7073 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7074 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7075 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7076 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7077 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7078 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7079 (aarch64_fmlal_lane_highv2sf): Ditto.
7080 (aarch64_fmlsl_lane_highv2sf): Ditto.
7081 (aarch64_fmlal_laneq_highv2sf): Ditto.
7082 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7083 (aarch64_fmlalq_lane_highv4sf): Ditto.
7084 (aarch64_fmlsl_lane_highv4sf): Ditto.
7085 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7086 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7087 * config/aarch64/aarch64-simd.md:
7088 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7089 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7090 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7091 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7092 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7093 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7094 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7095 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7096 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7097 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7098 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7099 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7100 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7101 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7102 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7103 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7104 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7105 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7106 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7107 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7108 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7109 (vfmlsl_low_u32): Ditto.
7110 (vfmlalq_low_u32): Ditto.
7111 (vfmlslq_low_u32): Ditto.
7112 (vfmlal_high_u32): Ditto.
7113 (vfmlsl_high_u32): Ditto.
7114 (vfmlalq_high_u32): Ditto.
7115 (vfmlslq_high_u32): Ditto.
7116 (vfmlal_lane_low_u32): Ditto.
7117 (vfmlsl_lane_low_u32): Ditto.
7118 (vfmlal_laneq_low_u32): Ditto.
7119 (vfmlsl_laneq_low_u32): Ditto.
7120 (vfmlalq_lane_low_u32): Ditto.
7121 (vfmlslq_lane_low_u32): Ditto.
7122 (vfmlalq_laneq_low_u32): Ditto.
7123 (vfmlslq_laneq_low_u32): Ditto.
7124 (vfmlal_lane_high_u32): Ditto.
7125 (vfmlsl_lane_high_u32): Ditto.
7126 (vfmlal_laneq_high_u32): Ditto.
7127 (vfmlsl_laneq_high_u32): Ditto.
7128 (vfmlalq_lane_high_u32): Ditto.
7129 (vfmlslq_lane_high_u32): Ditto.
7130 (vfmlalq_laneq_high_u32): Ditto.
7131 (vfmlslq_laneq_high_u32): Ditto.
7132 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7133 (AARCH64_FL_FOR_ARCH8_4): New.
7134 (AARCH64_ISA_F16FML): New ISA flag.
7135 (TARGET_F16FML): New feature flag for fp16fml.
7136 (doc/invoke.texi): Document new fp16fml option.
7138 2018-01-10 Michael Collison <michael.collison@arm.com>
7140 * config/aarch64/aarch64-builtins.c:
7141 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7142 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7143 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7144 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7145 (AARCH64_ISA_SHA3): New ISA flag.
7146 (TARGET_SHA3): New feature flag for sha3.
7147 * config/aarch64/iterators.md (sha512_op): New int attribute.
7148 (CRYPTO_SHA512): New int iterator.
7149 (UNSPEC_SHA512H): New unspec.
7150 (UNSPEC_SHA512H2): Ditto.
7151 (UNSPEC_SHA512SU0): Ditto.
7152 (UNSPEC_SHA512SU1): Ditto.
7153 * config/aarch64/aarch64-simd-builtins.def
7154 (aarch64_crypto_sha512hqv2di): New builtin.
7155 (aarch64_crypto_sha512h2qv2di): Ditto.
7156 (aarch64_crypto_sha512su0qv2di): Ditto.
7157 (aarch64_crypto_sha512su1qv2di): Ditto.
7158 (aarch64_eor3qv8hi): Ditto.
7159 (aarch64_rax1qv2di): Ditto.
7160 (aarch64_xarqv2di): Ditto.
7161 (aarch64_bcaxqv8hi): Ditto.
7162 * config/aarch64/aarch64-simd.md:
7163 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7164 (aarch64_crypto_sha512su0qv2di): Ditto.
7165 (aarch64_crypto_sha512su1qv2di): Ditto.
7166 (aarch64_eor3qv8hi): Ditto.
7167 (aarch64_rax1qv2di): Ditto.
7168 (aarch64_xarqv2di): Ditto.
7169 (aarch64_bcaxqv8hi): Ditto.
7170 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7171 (vsha512h2q_u64): Ditto.
7172 (vsha512su0q_u64): Ditto.
7173 (vsha512su1q_u64): Ditto.
7174 (veor3q_u16): Ditto.
7175 (vrax1q_u64): Ditto.
7177 (vbcaxq_u16): Ditto.
7178 * config/arm/types.md (crypto_sha512): New type attribute.
7179 (crypto_sha3): Ditto.
7180 (doc/invoke.texi): Document new sha3 option.
7182 2018-01-10 Michael Collison <michael.collison@arm.com>
7184 * config/aarch64/aarch64-builtins.c:
7185 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7186 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7187 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7188 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7189 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7190 (AARCH64_ISA_SM4): New ISA flag.
7191 (TARGET_SM4): New feature flag for sm4.
7192 * config/aarch64/aarch64-simd-builtins.def
7193 (aarch64_sm3ss1qv4si): Ditto.
7194 (aarch64_sm3tt1aq4si): Ditto.
7195 (aarch64_sm3tt1bq4si): Ditto.
7196 (aarch64_sm3tt2aq4si): Ditto.
7197 (aarch64_sm3tt2bq4si): Ditto.
7198 (aarch64_sm3partw1qv4si): Ditto.
7199 (aarch64_sm3partw2qv4si): Ditto.
7200 (aarch64_sm4eqv4si): Ditto.
7201 (aarch64_sm4ekeyqv4si): Ditto.
7202 * config/aarch64/aarch64-simd.md:
7203 (aarch64_sm3ss1qv4si): Ditto.
7204 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7205 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7206 (aarch64_sm4eqv4si): Ditto.
7207 (aarch64_sm4ekeyqv4si): Ditto.
7208 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7209 (sm3part_op): Ditto.
7210 (CRYPTO_SM3TT): Ditto.
7211 (CRYPTO_SM3PART): Ditto.
7212 (UNSPEC_SM3SS1): New unspec.
7213 (UNSPEC_SM3TT1A): Ditto.
7214 (UNSPEC_SM3TT1B): Ditto.
7215 (UNSPEC_SM3TT2A): Ditto.
7216 (UNSPEC_SM3TT2B): Ditto.
7217 (UNSPEC_SM3PARTW1): Ditto.
7218 (UNSPEC_SM3PARTW2): Ditto.
7219 (UNSPEC_SM4E): Ditto.
7220 (UNSPEC_SM4EKEY): Ditto.
7221 * config/aarch64/constraints.md (Ui2): New constraint.
7222 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7223 * config/arm/types.md (crypto_sm3): New type attribute.
7224 (crypto_sm4): Ditto.
7225 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7226 (vsm3tt1aq_u32): Ditto.
7227 (vsm3tt1bq_u32): Ditto.
7228 (vsm3tt2aq_u32): Ditto.
7229 (vsm3tt2bq_u32): Ditto.
7230 (vsm3partw1q_u32): Ditto.
7231 (vsm3partw2q_u32): Ditto.
7232 (vsm4eq_u32): Ditto.
7233 (vsm4ekeyq_u32): Ditto.
7234 (doc/invoke.texi): Document new sm4 option.
7236 2018-01-10 Michael Collison <michael.collison@arm.com>
7238 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7239 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7240 (AARCH64_FL_FOR_ARCH8_4): New.
7241 (AARCH64_FL_V8_4): New flag.
7242 (doc/invoke.texi): Document new armv8.4-a option.
7244 2018-01-10 Michael Collison <michael.collison@arm.com>
7246 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7247 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7248 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7249 * config/aarch64/aarch64-option-extension.def: Add
7250 AARCH64_OPT_EXTENSION of 'sha2'.
7251 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7252 (crypto): Disable sha2 and aes if crypto disabled.
7253 (crypto): Enable aes and sha2 if enabled.
7254 (simd): Disable sha2 and aes if simd disabled.
7255 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7257 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7258 (TARGET_SHA2): New feature flag for sha2.
7259 (TARGET_AES): New feature flag for aes.
7260 * config/aarch64/aarch64-simd.md:
7261 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7262 conditional on TARGET_AES.
7263 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7264 (aarch64_crypto_sha1hsi): Make pattern conditional
7266 (aarch64_crypto_sha1hv4si): Ditto.
7267 (aarch64_be_crypto_sha1hv4si): Ditto.
7268 (aarch64_crypto_sha1su1v4si): Ditto.
7269 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7270 (aarch64_crypto_sha1su0v4si): Ditto.
7271 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7272 (aarch64_crypto_sha256su0v4si): Ditto.
7273 (aarch64_crypto_sha256su1v4si): Ditto.
7274 (doc/invoke.texi): Document new aes and sha2 options.
7276 2018-01-10 Martin Sebor <msebor@redhat.com>
7278 PR tree-optimization/83781
7279 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7282 2018-01-11 Martin Sebor <msebor@gmail.com>
7283 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7285 PR tree-optimization/83501
7286 PR tree-optimization/81703
7288 * tree-ssa-strlen.c (get_string_cst): Rename...
7289 (get_string_len): ...to this. Handle global constants.
7290 (handle_char_store): Adjust.
7292 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7293 Jim Wilson <jimw@sifive.com>
7295 * config/riscv/riscv-protos.h (riscv_output_return): New.
7296 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7297 (riscv_attribute_table, riscv_output_return),
7298 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7299 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7300 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7302 (riscv_expand_prologue): Add early return for naked function.
7303 (riscv_expand_epilogue): Likewise.
7304 (riscv_function_ok_for_sibcall): Return false for naked function.
7305 (riscv_set_current_function): New.
7306 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7307 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7308 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7309 * doc/extend.texi (RISC-V Function Attributes): New.
7311 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7313 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7314 check for 128-bit long double before checking TCmode.
7315 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7316 128-bit long doubles before checking TFmode or TCmode.
7317 (FLOAT128_IBM_P): Likewise.
7319 2018-01-10 Martin Sebor <msebor@redhat.com>
7321 PR tree-optimization/83671
7322 * builtins.c (c_strlen): Unconditionally return zero for the empty
7324 Use -Warray-bounds for warnings.
7325 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7326 for non-constant array indices with COMPONENT_REF, arrays of
7327 arrays, and pointers to arrays.
7328 (gimple_fold_builtin_strlen): Determine and set length range for
7329 non-constant character arrays.
7331 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7334 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7337 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7339 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7341 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7344 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7345 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7346 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7347 indexed_or_indirect_operand predicate.
7348 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7349 (*vsx_le_perm_load_v8hi): Likewise.
7350 (*vsx_le_perm_load_v16qi): Likewise.
7351 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7352 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7353 (*vsx_le_perm_store_v8hi): Likewise.
7354 (*vsx_le_perm_store_v16qi): Likewise.
7355 (eight unnamed splitters): Likewise.
7357 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7359 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7360 * config/rs6000/emmintrin.h: Likewise.
7361 * config/rs6000/mmintrin.h: Likewise.
7362 * config/rs6000/xmmintrin.h: Likewise.
7364 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7367 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7369 * tree.c (tree_nop_conversion): Return true for location wrapper
7371 (maybe_wrap_with_location): New function.
7372 (selftest::check_strip_nops): New function.
7373 (selftest::test_location_wrappers): New function.
7374 (selftest::tree_c_tests): Call it.
7375 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7376 (maybe_wrap_with_location): New decl.
7377 (EXPR_LOCATION_WRAPPER_P): New macro.
7378 (location_wrapper_p): New inline function.
7379 (tree_strip_any_location_wrapper): New inline function.
7381 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7384 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7385 stack_realign_offset for the largest alignment of stack slot
7387 (ix86_find_max_used_stack_alignment): New function.
7388 (ix86_finalize_stack_frame_flags): Use it. Set
7389 max_used_stack_alignment if we don't realign stack.
7390 * config/i386/i386.h (machine_function): Add
7391 max_used_stack_alignment.
7393 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7395 * config/arm/arm.opt (-mbranch-cost): New option.
7396 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7399 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7402 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7403 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7405 2018-01-10 Richard Biener <rguenther@suse.de>
7408 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7409 early out so it also covers the case where we have a non-NULL
7412 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7414 PR tree-optimization/83753
7415 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7416 for non-strided grouped accesses if the number of elements is 1.
7418 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7421 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7422 * i386.h (TARGET_USE_GATHER): Define.
7423 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7425 2018-01-10 Martin Liska <mliska@suse.cz>
7428 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7429 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7431 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7432 CLEANUP_NO_PARTITIONING is not set.
7434 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7436 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7437 for vectors, as a partial revert of r254296.
7438 * rtl.h (const_vec_p): Delete.
7439 (const_vec_duplicate_p): Don't test for vector CONSTs.
7440 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7441 * expmed.c (make_tree): Likewise.
7444 * common.md (E, F): Use CONSTANT_P instead of checking for
7446 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7447 checking for CONST_VECTOR.
7449 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7452 * predict.c (force_edge_cold): Handle in more sane way edges
7455 2018-01-09 Carl Love <cel@us.ibm.com>
7457 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7459 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7460 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7461 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7462 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7463 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7464 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7465 * config/rs6000/rs6000-protos.h: Add extern defition for
7466 rs6000_generate_float2_double_code.
7467 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7469 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7470 (float2_v2df): Add define_expand.
7472 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7475 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7476 op_mode in the force_to_mode call.
7478 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7480 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7481 instead of checking each element individually.
7482 (aarch64_evpc_uzp): Likewise.
7483 (aarch64_evpc_zip): Likewise.
7484 (aarch64_evpc_ext): Likewise.
7485 (aarch64_evpc_rev): Likewise.
7486 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7487 instead of checking each element individually. Return true without
7489 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7490 whether all selected elements come from the same input, instead of
7491 checking each element individually. Remove calls to gen_rtx_REG,
7492 start_sequence and end_sequence and instead assert that no rtl is
7495 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7497 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7498 order of HIGH and CONST checks.
7500 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7502 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7503 if the destination isn't an SSA_NAME.
7505 2018-01-09 Richard Biener <rguenther@suse.de>
7507 PR tree-optimization/83668
7508 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7510 (canonicalize_loop_form): ... here, renamed from ...
7511 (canonicalize_loop_closed_ssa_form): ... this and amended to
7512 swap successor edges for loop exit blocks to make us use
7513 the RPO order we need for initial schedule generation.
7515 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7517 PR tree-optimization/64811
7518 * match.pd: When optimizing comparisons with Inf, avoid
7519 introducing or losing exceptions from comparisons with NaN.
7521 2018-01-09 Martin Liska <mliska@suse.cz>
7524 * asan.c (shadow_mem_size): Add gcc_assert.
7526 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7528 Don't save registers in main().
7531 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7532 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7533 * config/avr/avr.c (avr_set_current_function): Don't error if
7534 naked, OS_task or OS_main are specified at the same time.
7535 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7537 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7539 * common/config/avr/avr-common.c (avr_option_optimization_table):
7540 Switch on -mmain-is-OS_task for optimizing compilations.
7542 2018-01-09 Richard Biener <rguenther@suse.de>
7544 PR tree-optimization/83572
7545 * graphite.c: Include cfganal.h.
7546 (graphite_transform_loops): Connect infinite loops to exit
7547 and remove fake edges at the end.
7549 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7551 * ipa-inline.c (edge_badness): Revert accidental checkin.
7553 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7556 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7557 symbols; not inline clones.
7559 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7562 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7563 hard registers. Formatting fixes.
7565 PR preprocessor/83722
7566 * gcc.c (try_generate_repro): Pass
7567 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7568 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7571 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7572 Kito Cheng <kito.cheng@gmail.com>
7574 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7575 (riscv_leaf_function_p): Delete.
7576 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7578 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7580 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7582 (do_ifelse): New function.
7583 (do_isel): New function.
7584 (do_sub3): New function.
7585 (do_add3): New function.
7586 (do_load_mask_compare): New function.
7587 (do_overlap_load_compare): New function.
7588 (expand_compare_loop): New function.
7589 (expand_block_compare): Call expand_compare_loop() when appropriate.
7590 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7592 (-mblock-compare-inline-loop-limit): New option.
7594 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7597 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7598 Reverse order of second and third operands in first alternative.
7599 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7600 of first and second elements in UNSPEC_VPERMR vector.
7601 (altivec_expand_vec_perm_le): Likewise.
7603 2018-01-08 Jeff Law <law@redhat.com>
7605 PR rtl-optimizatin/81308
7606 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7607 (process_switch): If group_case_labels makes a change, then set
7609 (pass_convert_switch::execute): If a switch is converted, then
7610 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7612 PR rtl-optimization/81308
7613 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7616 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7618 PR target/83663 - Revert r255946
7619 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7620 generation for cases where splatting a value is not useful.
7621 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7622 across a vec_duplicate and a paradoxical subreg forming a vector
7623 mode to a vec_concat.
7625 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7627 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7628 -march=armv8.3-a variants.
7629 * config/arm/t-multilib: Likewise.
7630 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7632 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7634 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7636 (cceq_ior_compare_complement): Give it a name so I can use it, and
7637 change boolean_or_operator predicate to boolean_operator so it can
7638 be used to generate a crand.
7639 (eqne): New code iterator.
7640 (bd/bd_neg): New code_attrs.
7641 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7642 a single define_insn.
7643 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7644 decrement (bdnzt/bdnzf/bdzt/bdzf).
7645 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7646 with the new names of the branch decrement patterns, and added the
7647 names of the branch decrement conditional patterns.
7649 2018-01-08 Richard Biener <rguenther@suse.de>
7651 PR tree-optimization/83563
7652 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7655 2018-01-08 Richard Biener <rguenther@suse.de>
7658 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7660 2018-01-08 Richard Biener <rguenther@suse.de>
7662 PR tree-optimization/83685
7663 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7664 references to abnormals.
7666 2018-01-08 Richard Biener <rguenther@suse.de>
7669 * dwarf2out.c (output_indirect_strings): Handle empty
7670 skeleton_debug_str_hash.
7671 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7673 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7675 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7676 (emit_store_direct): Likewise.
7677 (arc_trampoline_adjust_address): Likewise.
7678 (arc_asm_trampoline_template): New function.
7679 (arc_initialize_trampoline): Use asm_trampoline_template.
7680 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7681 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7682 * config/arc/arc.md (flush_icache): Delete pattern.
7684 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7686 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7687 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7690 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7693 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7694 by not USED_FOR_TARGET.
7695 (make_pass_resolve_sw_modes): Likewise.
7697 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7699 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7702 2018-01-08 Richard Biener <rguenther@suse.de>
7705 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7707 2018-01-08 Richard Biener <rguenther@suse.de>
7710 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7712 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7715 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7716 basic blocks with a small number of successors.
7717 (convert_control_dep_chain_into_preds): Improve handling of
7719 (dump_predicates): Split apart into...
7720 (dump_pred_chain): ...here...
7721 (dump_pred_info): ...and here.
7722 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7723 (can_chain_union_be_invalidated_p): Improve check for invalidation
7725 (uninit_uses_cannot_happen): Avoid unnecessary if
7726 convert_control_dep_chain_into_preds yielded nothing.
7728 2018-01-06 Martin Sebor <msebor@redhat.com>
7730 PR tree-optimization/83640
7731 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7732 subtracting negative offset from size.
7733 (builtin_access::overlap): Adjust offset bounds of the access to fall
7734 within the size of the object if possible.
7736 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7738 PR rtl-optimization/83699
7739 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7740 extract_bit_field_as_subreg to cases in which the extracted
7741 value is also a vector.
7743 * lra-constraints.c (process_alt_operands): Test for the equivalence
7744 substitutions when detecting a possible reload cycle.
7746 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7749 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7750 by default if flag_selective_schedling{,2}. Formatting fixes.
7752 PR rtl-optimization/83682
7753 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7754 if it has non-VECTOR_MODE element mode.
7755 (vec_duplicate_p): Likewise.
7758 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7759 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7761 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7764 * config/i386/i386-builtin.def
7765 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7766 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7767 Require also OPTION_MASK_ISA_AVX512F in addition to
7768 OPTION_MASK_ISA_GFNI.
7769 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7770 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7771 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7772 to OPTION_MASK_ISA_GFNI.
7773 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7774 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7775 OPTION_MASK_ISA_AVX512BW.
7776 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7777 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7778 addition to OPTION_MASK_ISA_GFNI.
7779 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7780 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7781 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7782 to OPTION_MASK_ISA_GFNI.
7783 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7784 a requirement for all ISAs rather than any of them with a few
7786 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7788 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7789 bitmasks to be enabled with 3 exceptions, instead of requiring any
7790 enabled ISA with lots of exceptions.
7791 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7792 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7793 Change avx512bw in isa attribute to avx512f.
7794 * config/i386/sgxintrin.h: Add license boilerplate.
7795 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7796 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7797 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7798 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7800 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7801 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7802 temporarily sse2 rather than sse if not enabled already.
7805 * config/i386/sse.md (VI248_VLBW): Rename to ...
7806 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7807 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7808 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7809 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7810 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7811 mode iterator instead of VI248_VLBW.
7813 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7815 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7816 (record_modified): Skip clobbers; add debug output.
7817 (param_change_prob): Use sreal frequencies.
7819 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7821 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7822 punt for user-aligned variables.
7824 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7826 * tree-chrec.c (chrec_contains_symbols): Return true for
7829 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7832 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7833 of (x|y) == x for BICS pattern.
7835 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7837 PR tree-optimization/83605
7838 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7839 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7842 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7844 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7845 * config/epiphany/rtems.h: New file.
7847 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7848 Uros Bizjak <ubizjak@gmail.com>
7851 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7852 QIreg_operand instead of register_operand predicate.
7853 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7854 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7855 comments instead of -fmitigate[-_]rop.
7857 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7860 * cgraphunit.c (symbol_table::compile): Switch to text_section
7861 before calling assembly_start debug hook.
7862 * run-rtl-passes.c (run_rtl_passes): Likewise.
7865 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7867 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7868 range_int_cst_p rather than !symbolic_range_p before calling
7869 extract_range_from_multiplicative_op_1.
7871 2018-01-04 Jeff Law <law@redhat.com>
7873 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7874 redundant test in assertion.
7876 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7878 * doc/rtl.texi: Document machine_mode wrapper classes.
7880 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7882 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7885 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7887 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7888 the VEC_PERM_EXPR fold to fail.
7890 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7893 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7894 to switched_sections.
7896 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7899 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7902 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7905 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7906 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7908 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7911 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7912 is BLKmode and bitpos not zero or mode change is needed.
7914 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7917 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7920 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7923 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7924 instead of MULT rtx. Update all corresponding splitters.
7926 (*ssub<modesuffix>): Ditto.
7928 (*cmp_sadd_di): Update split patterns.
7929 (*cmp_sadd_si): Ditto.
7930 (*cmp_sadd_sidi): Ditto.
7931 (*cmp_ssub_di): Ditto.
7932 (*cmp_ssub_si): Ditto.
7933 (*cmp_ssub_sidi): Ditto.
7934 * config/alpha/predicates.md (const23_operand): New predicate.
7935 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7936 Look for ASHIFT, not MULT inner operand.
7937 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7939 2018-01-04 Martin Liska <mliska@suse.cz>
7941 PR gcov-profile/83669
7942 * gcov.c (output_intermediate_file): Add version to intermediate
7944 * doc/gcov.texi: Document new field 'version' in intermediate
7945 file format. Fix location of '-k' option of gcov command.
7947 2018-01-04 Martin Liska <mliska@suse.cz>
7950 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7952 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7954 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7956 2018-01-03 Martin Sebor <msebor@redhat.com>
7958 PR tree-optimization/83655
7959 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7960 checking calls with invalid arguments.
7962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7964 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7965 (vectorizable_mask_load_store): Delete.
7966 (vectorizable_call): Return false for masked loads and stores.
7967 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7968 instead of gimple_assign_rhs1.
7969 (vectorizable_load): Handle IFN_MASK_LOAD.
7970 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7974 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7976 (vectorizable_mask_load_store): ...here.
7977 (vectorizable_load): ...and here.
7979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7981 * tree-vect-stmts.c (vect_build_all_ones_mask)
7982 (vect_build_zero_merge_argument): New functions, split out from...
7983 (vectorizable_load): ...here.
7985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7989 (vectorizable_mask_load_store): ...here.
7990 (vectorizable_store): ...and here.
7992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7994 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7996 (vectorizable_mask_load_store): ...here.
7998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8000 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8001 (vect_model_store_cost): Take a vec_load_store_type instead of a
8003 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8004 (vect_model_store_cost): Take a vec_load_store_type instead of a
8006 (vectorizable_mask_load_store): Update accordingly.
8007 (vectorizable_store): Likewise.
8008 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8012 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8013 IFN_MASK_LOAD calls here rather than...
8014 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8017 Alan Hayward <alan.hayward@arm.com>
8018 David Sherwood <david.sherwood@arm.com>
8020 * expmed.c (extract_bit_field_1): For vector extracts,
8021 fall back to extract_bit_field_as_subreg if vec_extract
8024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8025 Alan Hayward <alan.hayward@arm.com>
8026 David Sherwood <david.sherwood@arm.com>
8028 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8029 they are variable or constant sized.
8030 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8031 slots for constant-sized data.
8033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8034 Alan Hayward <alan.hayward@arm.com>
8035 David Sherwood <david.sherwood@arm.com>
8037 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8038 handling COND_EXPRs with boolean comparisons, try to find a better
8039 basis for the mask type than the boolean itself.
8041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8043 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8044 is calculated and how it can be overridden.
8045 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8046 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8048 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8052 Alan Hayward <alan.hayward@arm.com>
8053 David Sherwood <david.sherwood@arm.com>
8055 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8056 Remove the mode argument.
8057 (aarch64_simd_valid_immediate): Remove the mode and inverse
8059 * config/aarch64/iterators.md (bitsize): New iterator.
8060 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8061 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8062 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8063 aarch64_simd_valid_immediate.
8064 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8065 (aarch64_reg_or_bic_imm): Likewise.
8066 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8067 with an insn_type enum and msl with a modifier_type enum.
8068 Replace element_width with a scalar_mode. Change the shift
8069 to unsigned int. Add constructors for scalar_float_mode and
8070 scalar_int_mode elements.
8071 (aarch64_vect_float_const_representable_p): Delete.
8072 (aarch64_can_const_movi_rtx_p)
8073 (aarch64_simd_scalar_immediate_valid_for_move)
8074 (aarch64_simd_make_constant): Update call to
8075 aarch64_simd_valid_immediate.
8076 (aarch64_advsimd_valid_immediate_hs): New function.
8077 (aarch64_advsimd_valid_immediate): Likewise.
8078 (aarch64_simd_valid_immediate): Remove mode and inverse
8079 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8080 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8081 and aarch64_float_const_representable_p on the result.
8082 (aarch64_output_simd_mov_immediate): Remove mode argument.
8083 Update call to aarch64_simd_valid_immediate and use of
8084 simd_immediate_info.
8085 (aarch64_output_scalar_simd_mov_immediate): Update call
8088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8089 Alan Hayward <alan.hayward@arm.com>
8090 David Sherwood <david.sherwood@arm.com>
8092 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8093 (mode_nunits): Likewise CONST_MODE_NUNITS.
8094 * machmode.def (ADJUST_NUNITS): Document.
8095 * genmodes.c (mode_data::need_nunits_adj): New field.
8096 (blank_mode): Update accordingly.
8097 (adj_nunits): New variable.
8098 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8100 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8101 listed in adj_nunits.
8102 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8103 listed in adj_nunits. Don't emit case statements for such modes.
8104 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8105 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8106 nothing if adj_nunits is nonnull.
8107 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8108 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8109 (emit_mode_fbit): Update use of print_maybe_const_decl.
8110 (emit_move_size): Likewise. Treat the array as non-const
8112 (emit_mode_adjustments): Handle adj_nunits.
8114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8116 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8117 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8118 (VECTOR_MODES): Use it.
8119 (make_vector_modes): Take the prefix as an argument.
8121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8122 Alan Hayward <alan.hayward@arm.com>
8123 David Sherwood <david.sherwood@arm.com>
8125 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8126 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8127 for MODE_VECTOR_BOOL.
8128 * machmode.def (VECTOR_BOOL_MODE): Document.
8129 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8130 (make_vector_bool_mode): New function.
8131 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8133 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8134 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8136 * stor-layout.c (int_mode_for_mode): Likewise.
8137 * tree.c (build_vector_type_for_mode): Likewise.
8138 * varasm.c (output_constant_pool_2): Likewise.
8139 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8140 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8141 for MODE_VECTOR_BOOL.
8142 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8143 of mode class checks.
8144 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8145 instead of a list of mode class checks.
8146 (expand_vector_scalar_condition): Likewise.
8147 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8150 Alan Hayward <alan.hayward@arm.com>
8151 David Sherwood <david.sherwood@arm.com>
8153 * machmode.h (mode_size): Change from unsigned short to
8155 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8156 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8157 or if measurement_type is not polynomial.
8158 (fixed_size_mode::includes_p): Check for constant-sized modes.
8159 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8160 return a poly_uint16 rather than an unsigned short.
8161 (emit_mode_size): Change the type of mode_size from unsigned short
8162 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8163 (emit_mode_adjustments): Cope with polynomial vector sizes.
8164 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8166 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8168 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8169 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8170 * caller-save.c (setup_save_areas): Likewise.
8171 (replace_reg_with_saved_mem): Likewise.
8172 * calls.c (emit_library_call_value_1): Likewise.
8173 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8174 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8175 (gen_lowpart_for_combine): Likewise.
8176 * convert.c (convert_to_integer_1): Likewise.
8177 * cse.c (equiv_constant, cse_insn): Likewise.
8178 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8179 (cselib_subst_to_values): Likewise.
8180 * dce.c (word_dce_process_block): Likewise.
8181 * df-problems.c (df_word_lr_mark_ref): Likewise.
8182 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8183 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8184 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8185 (rtl_for_decl_location): Likewise.
8186 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8187 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8188 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8189 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8190 (expand_expr_real_1): Likewise.
8191 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8192 (pad_below): Likewise.
8193 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8194 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8195 * ira.c (get_subreg_tracking_sizes): Likewise.
8196 * ira-build.c (ira_create_allocno_objects): Likewise.
8197 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8198 (ira_sort_regnos_for_alter_reg): Likewise.
8199 * ira-costs.c (record_operand_costs): Likewise.
8200 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8201 (resolve_simple_move): Likewise.
8202 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8203 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8204 (lra_constraints): Likewise.
8205 (CONST_POOL_OK_P): Reject variable-sized modes.
8206 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8207 (add_pseudo_to_slot, lra_spill): Likewise.
8208 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8209 * optabs-query.c (get_best_extraction_insn): Likewise.
8210 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8211 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8212 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8213 * recog.c (offsettable_address_addr_space_p): Likewise.
8214 * regcprop.c (maybe_mode_change): Likewise.
8215 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8216 * regrename.c (build_def_use): Likewise.
8217 * regstat.c (dump_reg_info): Likewise.
8218 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8219 (find_reloads, find_reloads_subreg_address): Likewise.
8220 * reload1.c (eliminate_regs_1): Likewise.
8221 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8222 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8223 (simplify_binary_operation_1, simplify_subreg): Likewise.
8224 * targhooks.c (default_function_arg_padding): Likewise.
8225 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8226 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8227 (verify_gimple_assign_ternary): Likewise.
8228 * tree-inline.c (estimate_move_cost): Likewise.
8229 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8230 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8231 (get_address_cost_ainc): Likewise.
8232 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8233 (vect_supportable_dr_alignment): Likewise.
8234 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8235 (vectorizable_reduction): Likewise.
8236 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8237 (vectorizable_operation, vectorizable_load): Likewise.
8238 * tree.c (build_same_sized_truth_vector_type): Likewise.
8239 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8240 * var-tracking.c (emit_note_insn_var_location): Likewise.
8241 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8242 (ADDR_VEC_ALIGN): Likewise.
8244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8245 Alan Hayward <alan.hayward@arm.com>
8246 David Sherwood <david.sherwood@arm.com>
8248 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8250 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8251 or if measurement_type is polynomial.
8252 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8253 * combine.c (make_extraction): Likewise.
8254 * dse.c (find_shift_sequence): Likewise.
8255 * dwarf2out.c (mem_loc_descriptor): Likewise.
8256 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8257 (extract_bit_field, extract_low_bits): Likewise.
8258 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8259 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8260 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8261 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8262 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8263 * reload.c (find_reloads): Likewise.
8264 * reload1.c (alter_reg): Likewise.
8265 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8266 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8267 * tree-if-conv.c (predicate_mem_writes): Likewise.
8268 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8269 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8270 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8271 * valtrack.c (dead_debug_insert_temp): Likewise.
8272 * varasm.c (mergeable_constant_section): Likewise.
8273 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8276 Alan Hayward <alan.hayward@arm.com>
8277 David Sherwood <david.sherwood@arm.com>
8279 * expr.c (expand_assignment): Cope with polynomial mode sizes
8280 when assigning to a CONCAT.
8282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8283 Alan Hayward <alan.hayward@arm.com>
8284 David Sherwood <david.sherwood@arm.com>
8286 * machmode.h (mode_precision): Change from unsigned short to
8288 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8290 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8291 or if measurement_type is not polynomial.
8292 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8293 in which the mode is already known to be a scalar_int_mode.
8294 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8295 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8297 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8298 for GET_MODE_PRECISION.
8299 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8300 for GET_MODE_PRECISION.
8301 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8303 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8304 (expand_field_assignment, make_extraction): Likewise.
8305 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8306 (get_last_value): Likewise.
8307 * convert.c (convert_to_integer_1): Likewise.
8308 * cse.c (cse_insn): Likewise.
8309 * expr.c (expand_expr_real_1): Likewise.
8310 * lra-constraints.c (simplify_operand_subreg): Likewise.
8311 * optabs-query.c (can_atomic_load_p): Likewise.
8312 * optabs.c (expand_atomic_load): Likewise.
8313 (expand_atomic_store): Likewise.
8314 * ree.c (combine_reaching_defs): Likewise.
8315 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8316 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8317 * tree.h (type_has_mode_precision_p): Likewise.
8318 * ubsan.c (instrument_si_overflow): Likewise.
8320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8321 Alan Hayward <alan.hayward@arm.com>
8322 David Sherwood <david.sherwood@arm.com>
8324 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8325 polynomial numbers of units.
8326 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8327 (valid_vector_subparts_p): New function.
8328 (build_vector_type): Remove temporary shim and take the number
8329 of units as a poly_uint64 rather than an int.
8330 (build_opaque_vector_type): Take the number of units as a
8331 poly_uint64 rather than an int.
8332 * tree.c (build_vector_from_ctor): Handle polynomial
8333 TYPE_VECTOR_SUBPARTS.
8334 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8335 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8336 (build_vector_from_val): If the number of units is variable,
8337 use build_vec_duplicate_cst for constant operands and
8338 VEC_DUPLICATE_EXPR otherwise.
8339 (make_vector_type): Remove temporary is_constant ().
8340 (build_vector_type, build_opaque_vector_type): Take the number of
8341 units as a poly_uint64 rather than an int.
8342 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8344 * cfgexpand.c (expand_debug_expr): Likewise.
8345 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8346 (store_constructor, expand_expr_real_1): Likewise.
8347 (const_scalar_mask_from_tree): Likewise.
8348 * fold-const-call.c (fold_const_reduction): Likewise.
8349 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8350 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8351 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8352 (fold_relational_const): Likewise.
8353 (native_interpret_vector): Likewise. Change the size from an
8354 int to an unsigned int.
8355 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8356 TYPE_VECTOR_SUBPARTS.
8357 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8358 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8359 duplicating a non-constant operand into a variable-length vector.
8360 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8361 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8362 * ipa-icf.c (sem_variable::equals): Likewise.
8363 * match.pd: Likewise.
8364 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8365 * print-tree.c (print_node): Likewise.
8366 * stor-layout.c (layout_type): Likewise.
8367 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8368 * tree-cfg.c (verify_gimple_comparison): Likewise.
8369 (verify_gimple_assign_binary): Likewise.
8370 (verify_gimple_assign_ternary): Likewise.
8371 (verify_gimple_assign_single): Likewise.
8372 * tree-pretty-print.c (dump_generic_node): Likewise.
8373 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8374 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8375 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8376 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8377 (vect_shift_permute_load_chain): Likewise.
8378 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8379 (expand_vector_condition, optimize_vector_constructor): Likewise.
8380 (lower_vec_perm, get_compute_type): Likewise.
8381 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8382 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8383 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8384 (vect_recog_mask_conversion_pattern): Likewise.
8385 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8386 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8387 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8388 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8389 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8390 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8391 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8392 (supportable_widening_operation): Likewise.
8393 (supportable_narrowing_operation): Likewise.
8394 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8396 * varasm.c (output_constant): Likewise.
8398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8399 Alan Hayward <alan.hayward@arm.com>
8400 David Sherwood <david.sherwood@arm.com>
8402 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8403 so that both the length == 3 and length != 3 cases set up their
8404 own permute vectors. Add comments explaining why we know the
8405 number of elements is constant.
8406 (vect_permute_load_chain): Likewise.
8408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8409 Alan Hayward <alan.hayward@arm.com>
8410 David Sherwood <david.sherwood@arm.com>
8412 * machmode.h (mode_nunits): Change from unsigned char to
8414 (ONLY_FIXED_SIZE_MODES): New macro.
8415 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8416 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8417 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8419 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8420 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8421 or if measurement_type is not polynomial.
8422 * genmodes.c (ZERO_COEFFS): New macro.
8423 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8425 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8426 Use ZERO_COEFFS when emitting initializers.
8427 * data-streamer.h (bp_pack_poly_value): New function.
8428 (bp_unpack_poly_value): Likewise.
8429 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8430 for GET_MODE_NUNITS.
8431 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8432 for GET_MODE_NUNITS.
8433 * tree.c (make_vector_type): Remove temporary shim and make
8434 the real function take the number of units as a poly_uint64
8436 (build_vector_type_for_mode): Handle polynomial nunits.
8437 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8438 * emit-rtl.c (const_vec_series_p_1): Likewise.
8439 (gen_rtx_CONST_VECTOR): Likewise.
8440 * fold-const.c (test_vec_duplicate_folding): Likewise.
8441 * genrecog.c (validate_pattern): Likewise.
8442 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8443 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8444 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8445 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8446 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8447 * rtlanal.c (subreg_get_info): Likewise.
8448 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8449 (vect_grouped_load_supported): Likewise.
8450 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8451 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8452 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8453 (simplify_const_unary_operation, simplify_binary_operation_1)
8454 (simplify_const_binary_operation, simplify_ternary_operation)
8455 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8456 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8457 instead of CONST_VECTOR_NUNITS.
8458 * varasm.c (output_constant_pool_2): Likewise.
8459 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8460 explicit-encoded elements in the XVEC for variable-length vectors.
8462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8464 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8467 Alan Hayward <alan.hayward@arm.com>
8468 David Sherwood <david.sherwood@arm.com>
8470 * coretypes.h (fixed_size_mode): Declare.
8471 (fixed_size_mode_pod): New typedef.
8472 * builtins.h (target_builtins::x_apply_args_mode)
8473 (target_builtins::x_apply_result_mode): Change type to
8474 fixed_size_mode_pod.
8475 * builtins.c (apply_args_size, apply_result_size, result_vector)
8476 (expand_builtin_apply_args_1, expand_builtin_apply)
8477 (expand_builtin_return): Update accordingly.
8479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8481 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8482 * cselib.c (cselib_hash_rtx): Likewise.
8483 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8484 CONST_VECTOR encoding.
8486 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8487 Jeff Law <law@redhat.com>
8490 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8491 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8492 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8493 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8496 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8497 explicitly probe *sp in a noreturn function if there were any callee
8498 register saves or frame pointer is needed.
8500 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8503 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8504 BLKmode for ternary, binary or unary expressions.
8507 * var-tracking.c (delete_vta_debug_insn): New inline function.
8508 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8509 insns from get_insns () to NULL instead of each bb separately.
8510 Use delete_vta_debug_insn. No longer static.
8511 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8512 delete_vta_debug_insns callers.
8513 * rtl.h (delete_vta_debug_insns): Declare.
8514 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8515 instead of variable_tracking_main.
8517 2018-01-03 Martin Sebor <msebor@redhat.com>
8519 PR tree-optimization/83603
8520 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8521 arguments past the endof the argument list in functions declared
8522 without a prototype.
8523 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8524 Avoid checking when arguments are null.
8526 2018-01-03 Martin Sebor <msebor@redhat.com>
8529 * doc/extend.texi (attribute const): Fix a typo.
8530 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8531 issuing -Wsuggest-attribute for void functions.
8533 2018-01-03 Martin Sebor <msebor@redhat.com>
8535 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8536 offset_int::from instead of wide_int::to_shwi.
8537 (maybe_diag_overlap): Remove assertion.
8538 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8539 * gimple-ssa-sprintf.c (format_directive): Same.
8540 (parse_directive): Same.
8541 (sprintf_dom_walker::compute_format_length): Same.
8542 (try_substitute_return_value): Same.
8544 2018-01-03 Jeff Law <law@redhat.com>
8547 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8548 non-constant residual for zero at runtime and avoid probing in
8549 that case. Reorganize code for trailing problem to mirror handling
8552 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8554 PR tree-optimization/83501
8555 * tree-ssa-strlen.c (get_string_cst): New.
8556 (handle_char_store): Call get_string_cst.
8558 2018-01-03 Martin Liska <mliska@suse.cz>
8560 PR tree-optimization/83593
8561 * tree-ssa-strlen.c: Include tree-cfg.h.
8562 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8563 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8564 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8566 (strlen_dom_walker::before_dom_children): Call
8567 gimple_purge_dead_eh_edges. Dump tranformation with details
8569 (strlen_dom_walker::before_dom_children): Update call by adding
8570 new argument cleanup_eh.
8571 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8573 2018-01-03 Martin Liska <mliska@suse.cz>
8576 * cif-code.def (VARIADIC_THUNK): New enum value.
8577 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8580 2018-01-03 Jan Beulich <jbeulich@suse.com>
8582 * sse.md (mov<mode>_internal): Tighten condition for when to use
8583 vmovdqu<ssescalarsize> for TI and OI modes.
8585 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8587 Update copyright years.
8589 2018-01-03 Martin Liska <mliska@suse.cz>
8592 * ipa-visibility.c (function_and_variable_visibility): Skip
8593 functions with noipa attribure.
8595 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8597 * gcc.c (process_command): Update copyright notice dates.
8598 * gcov-dump.c (print_version): Ditto.
8599 * gcov.c (print_version): Ditto.
8600 * gcov-tool.c (print_version): Ditto.
8601 * gengtype.c (create_file): Ditto.
8602 * doc/cpp.texi: Bump @copying's copyright year.
8603 * doc/cppinternals.texi: Ditto.
8604 * doc/gcc.texi: Ditto.
8605 * doc/gccint.texi: Ditto.
8606 * doc/gcov.texi: Ditto.
8607 * doc/install.texi: Ditto.
8608 * doc/invoke.texi: Ditto.
8610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8612 * vector-builder.h (vector_builder::m_full_nelts): Change from
8613 unsigned int to poly_uint64.
8614 (vector_builder::full_nelts): Update prototype accordingly.
8615 (vector_builder::new_vector): Likewise.
8616 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8617 (vector_builder::operator ==): Likewise.
8618 (vector_builder::finalize): Likewise.
8619 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8620 Take the number of elements as a poly_uint64 rather than an
8622 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8623 from unsigned int to poly_uint64.
8624 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8625 (vec_perm_indices::new_vector): Likewise.
8626 (vec_perm_indices::length): Likewise.
8627 (vec_perm_indices::nelts_per_input): Likewise.
8628 (vec_perm_indices::input_nelts): Likewise.
8629 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8630 number of elements per input as a poly_uint64 rather than an
8631 unsigned int. Use the original encoding for variable-length
8632 vectors, rather than clamping each individual element.
8633 For the second and subsequent elements in each pattern,
8634 clamp the step and base before clamping their sum.
8635 (vec_perm_indices::series_p): Handle polynomial element counts.
8636 (vec_perm_indices::all_in_range_p): Likewise.
8637 (vec_perm_indices_to_tree): Likewise.
8638 (vec_perm_indices_to_rtx): Likewise.
8639 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8640 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8641 (tree_vector_builder::new_binary_operation): Handle polynomial
8642 element counts. Return false if we need to know the number
8643 of elements at compile time.
8644 * fold-const.c (fold_vec_perm): Punt if the number of elements
8645 isn't known at compile time.
8647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8649 * vec-perm-indices.h (vec_perm_builder): Change element type
8650 from HOST_WIDE_INT to poly_int64.
8651 (vec_perm_indices::element_type): Update accordingly.
8652 (vec_perm_indices::clamp): Handle polynomial element_types.
8653 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8654 (vec_perm_indices::all_in_range_p): Likewise.
8655 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8657 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8658 polynomial vec_perm_indices element types.
8659 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8660 * fold-const.c (fold_vec_perm): Likewise.
8661 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8662 * tree-vect-generic.c (lower_vec_perm): Likewise.
8663 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8664 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8665 element type to HOST_WIDE_INT.
8667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8668 Alan Hayward <alan.hayward@arm.com>
8669 David Sherwood <david.sherwood@arm.com>
8671 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8672 rather than an int. Use plus_constant.
8673 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8674 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8677 Alan Hayward <alan.hayward@arm.com>
8678 David Sherwood <david.sherwood@arm.com>
8680 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8681 a HOST_WIDE_INT to a poly_int64.
8683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8684 Alan Hayward <alan.hayward@arm.com>
8685 David Sherwood <david.sherwood@arm.com>
8687 * calls.c (load_register_parameters): Cope with polynomial
8688 mode sizes. Require a constant size for BLKmode parameters
8689 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8690 forces a parameter to be padded at the lsb end in order to
8691 fill a complete number of words, require the parameter size
8692 to be ordered wrt UNITS_PER_WORD.
8694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8695 Alan Hayward <alan.hayward@arm.com>
8696 David Sherwood <david.sherwood@arm.com>
8698 * reload1.c (spill_stack_slot_width): Change element type
8699 from unsigned int to poly_uint64_pod.
8700 (alter_reg): Treat mode sizes as polynomial.
8702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8703 Alan Hayward <alan.hayward@arm.com>
8704 David Sherwood <david.sherwood@arm.com>
8706 * reload.c (complex_word_subreg_p): New function.
8707 (reload_inner_reg_of_subreg, push_reload): Use it.
8709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8710 Alan Hayward <alan.hayward@arm.com>
8711 David Sherwood <david.sherwood@arm.com>
8713 * lra-constraints.c (process_alt_operands): Reject matched
8714 operands whose sizes aren't ordered.
8715 (match_reload): Refer to this check here.
8717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8718 Alan Hayward <alan.hayward@arm.com>
8719 David Sherwood <david.sherwood@arm.com>
8721 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8722 that the mode size is in the set {1, 2, 4, 8, 16}.
8724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8725 Alan Hayward <alan.hayward@arm.com>
8726 David Sherwood <david.sherwood@arm.com>
8728 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8729 Use plus_constant instead of gen_rtx_PLUS.
8731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8732 Alan Hayward <alan.hayward@arm.com>
8733 David Sherwood <david.sherwood@arm.com>
8735 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8736 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8737 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8738 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8739 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8740 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8741 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8742 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8743 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8744 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8746 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8747 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8748 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8749 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8750 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8751 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8752 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8753 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8754 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8755 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8757 * expr.c (emit_move_resolve_push): Treat the input and result
8758 of PUSH_ROUNDING as a poly_int64.
8759 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8760 (emit_push_insn): Likewise.
8761 * lra-eliminations.c (mark_not_eliminable): Likewise.
8762 * recog.c (push_operand): Likewise.
8763 * reload1.c (elimination_effects): Likewise.
8764 * rtlanal.c (nonzero_bits1): Likewise.
8765 * calls.c (store_one_arg): Likewise. Require the padding to be
8766 known at compile time.
8768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8769 Alan Hayward <alan.hayward@arm.com>
8770 David Sherwood <david.sherwood@arm.com>
8772 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8773 Use plus_constant instead of gen_rtx_PLUS.
8775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8776 Alan Hayward <alan.hayward@arm.com>
8777 David Sherwood <david.sherwood@arm.com>
8779 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8783 Alan Hayward <alan.hayward@arm.com>
8784 David Sherwood <david.sherwood@arm.com>
8786 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8787 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8788 via stack temporaries. Treat the mode size as polynomial too.
8790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8791 Alan Hayward <alan.hayward@arm.com>
8792 David Sherwood <david.sherwood@arm.com>
8794 * expr.c (expand_expr_real_2): When handling conversions involving
8795 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8796 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8797 as a poly_uint64 too.
8799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8800 Alan Hayward <alan.hayward@arm.com>
8801 David Sherwood <david.sherwood@arm.com>
8803 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8806 Alan Hayward <alan.hayward@arm.com>
8807 David Sherwood <david.sherwood@arm.com>
8809 * combine.c (can_change_dest_mode): Handle polynomial
8810 REGMODE_NATURAL_SIZE.
8811 * expmed.c (store_bit_field_1): Likewise.
8812 * expr.c (store_constructor): Likewise.
8813 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8814 and polynomial REGMODE_NATURAL_SIZE.
8815 (gen_lowpart_common): Likewise.
8816 * reginfo.c (record_subregs_of_mode): Likewise.
8817 * rtlanal.c (read_modify_subreg_p): Likewise.
8819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8820 Alan Hayward <alan.hayward@arm.com>
8821 David Sherwood <david.sherwood@arm.com>
8823 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8824 numbers of elements.
8826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8827 Alan Hayward <alan.hayward@arm.com>
8828 David Sherwood <david.sherwood@arm.com>
8830 * match.pd: Cope with polynomial numbers of vector elements.
8832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8833 Alan Hayward <alan.hayward@arm.com>
8834 David Sherwood <david.sherwood@arm.com>
8836 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8837 in a POINTER_PLUS_EXPR.
8839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8840 Alan Hayward <alan.hayward@arm.com>
8841 David Sherwood <david.sherwood@arm.com>
8843 * omp-simd-clone.c (simd_clone_subparts): New function.
8844 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8845 (ipa_simd_modify_function_body): Likewise.
8847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8848 Alan Hayward <alan.hayward@arm.com>
8849 David Sherwood <david.sherwood@arm.com>
8851 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8852 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8853 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8854 (expand_vector_condition, vector_element): Likewise.
8855 (subparts_gt): New function.
8856 (get_compute_type): Use subparts_gt.
8857 (count_type_subparts): Delete.
8858 (expand_vector_operations_1): Use subparts_gt instead of
8859 count_type_subparts.
8861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8862 Alan Hayward <alan.hayward@arm.com>
8863 David Sherwood <david.sherwood@arm.com>
8865 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8866 (vect_compile_time_alias): ...this new function. Do the calculation
8867 on poly_ints rather than trees.
8868 (vect_prune_runtime_alias_test_list): Update call accordingly.
8870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8871 Alan Hayward <alan.hayward@arm.com>
8872 David Sherwood <david.sherwood@arm.com>
8874 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8876 (vect_schedule_slp_instance): Likewise.
8878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8879 Alan Hayward <alan.hayward@arm.com>
8880 David Sherwood <david.sherwood@arm.com>
8882 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8883 constant and extern definitions for variable-length vectors.
8884 (vect_get_constant_vectors): Note that the number of units
8885 is known to be constant.
8887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8888 Alan Hayward <alan.hayward@arm.com>
8889 David Sherwood <david.sherwood@arm.com>
8891 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8892 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
8900 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8907 vectors as polynomial. Use build_index_vector for
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 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8915 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8916 for variable-length vectors.
8917 (vectorizable_mask_load_store): Treat the number of units as
8918 polynomial, asserting that it is constant if the condition has
8919 already been enforced.
8920 (vectorizable_store, vectorizable_load): Likewise.
8922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8923 Alan Hayward <alan.hayward@arm.com>
8924 David Sherwood <david.sherwood@arm.com>
8926 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8927 of units as polynomial. Punt if we can't tell at compile time
8928 which vector contains the final result.
8930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8931 Alan Hayward <alan.hayward@arm.com>
8932 David Sherwood <david.sherwood@arm.com>
8934 * tree-vect-loop.c (vectorizable_induction): Treat the number
8935 of units as polynomial. Punt on SLP inductions. Use an integer
8936 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8937 cast of such a series for variable-length floating-point
8940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8941 Alan Hayward <alan.hayward@arm.com>
8942 David Sherwood <david.sherwood@arm.com>
8944 * tree.h (build_index_vector): Declare.
8945 * tree.c (build_index_vector): New function.
8946 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8947 of units as polynomial, forcibly converting it to a constant if
8948 vectorizable_reduction has already enforced the condition.
8949 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8950 to create a {1,2,3,...} vector.
8951 (vectorizable_reduction): Treat the number of units as polynomial.
8952 Choose vectype_in based on the largest scalar element size rather
8953 than the smallest number of units. Enforce the restrictions
8956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8957 Alan Hayward <alan.hayward@arm.com>
8958 David Sherwood <david.sherwood@arm.com>
8960 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8961 number of units as polynomial.
8963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8964 Alan Hayward <alan.hayward@arm.com>
8965 David Sherwood <david.sherwood@arm.com>
8967 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8968 * target.def (autovectorize_vector_sizes): Return the vector sizes
8969 by pointer, using vector_sizes rather than a bitmask.
8970 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8971 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8972 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8974 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8975 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8976 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8977 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8978 * omp-general.c (omp_max_vf): Likewise.
8979 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8980 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8981 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8982 * tree-vect-slp.c (vect_slp_bb): Likewise.
8983 * doc/tm.texi: Regenerate.
8984 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8986 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8987 the vector size as a poly_uint64 rather than an unsigned int.
8988 (current_vector_size): Change from an unsigned int to a poly_uint64.
8989 (get_vectype_for_scalar_type): Update accordingly.
8990 * tree.h (build_truth_vector_type): Take the size and number of
8991 units as a poly_uint64 rather than an unsigned int.
8992 (build_vector_type): Add a temporary overload that takes
8993 the number of units as a poly_uint64 rather than an unsigned int.
8994 * tree.c (make_vector_type): Likewise.
8995 (build_truth_vector_type): Take the number of units as a poly_uint64
8996 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
9003 as poly_uint64s rather than unsigned ints.
9004 * targhooks.h (default_get_mask_mode): Update accordingly.
9005 * targhooks.c (default_get_mask_mode): Likewise.
9006 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9007 * doc/tm.texi: Regenerate.
9009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9014 * omp-general.c (omp_max_vf): Likewise.
9015 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9016 (expand_omp_simd): Handle polynomial safelen.
9017 * omp-low.c (omplow_simd_context): Add a default constructor.
9018 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9019 (lower_rec_simd_input_clauses): Update accordingly.
9020 (lower_rec_input_clauses): Likewise.
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-vectorizer.h (vect_nunits_for_cost): New function.
9027 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9028 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9029 (vect_analyze_slp_cost): Likewise.
9030 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9031 (vect_model_load_cost): Likewise.
9033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9034 Alan Hayward <alan.hayward@arm.com>
9035 David Sherwood <david.sherwood@arm.com>
9037 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9038 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9039 from an unsigned int * to a poly_uint64_pod *.
9040 (calculate_unrolling_factor): New function.
9041 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9044 Alan Hayward <alan.hayward@arm.com>
9045 David Sherwood <david.sherwood@arm.com>
9047 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9048 from an unsigned int to a poly_uint64.
9049 (_loop_vec_info::slp_unrolling_factor): Likewise.
9050 (_loop_vec_info::vectorization_factor): Change from an int
9052 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9053 (vect_get_num_vectors): New function.
9054 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9055 (vect_get_num_copies): Use vect_get_num_vectors.
9056 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9057 to an unsigned int *.
9058 (vect_analyze_data_refs): Change min_vf from an int * to a
9060 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9061 than an unsigned HOST_WIDE_INT.
9062 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9063 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9064 to an unsigned int *.
9065 (vect_analyze_data_ref_dependences): Likewise.
9066 (vect_compute_data_ref_alignment): Handle polynomial vf.
9067 (vect_enhance_data_refs_alignment): Likewise.
9068 (vect_prune_runtime_alias_test_list): Likewise.
9069 (vect_shift_permute_load_chain): Likewise.
9070 (vect_supportable_dr_alignment): Likewise.
9071 (dependence_distance_ge_vf): Take the vectorization factor as a
9072 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9073 (vect_analyze_data_refs): Change min_vf from an int * to a
9075 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9076 vfm1 as a poly_uint64 rather than an int. Make the same change
9077 for the returned bound_scalar.
9078 (vect_gen_vector_loop_niters): Handle polynomial vf.
9079 (vect_do_peeling): Likewise. Update call to
9080 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9081 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9083 * tree-vect-loop.c (vect_determine_vectorization_factor)
9084 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9085 (vect_get_known_peeling_cost): Likewise.
9086 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9087 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9088 (vect_transform_loop): Likewise. Use the lowest possible VF when
9089 updating the upper bounds of the loop.
9090 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9092 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9093 polynomial unroll factors.
9094 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9095 (vect_make_slp_decision): Likewise.
9096 (vect_supported_load_permutation_p): Likewise, and polynomial
9098 (vect_analyze_slp_cost): Handle polynomial vf.
9099 (vect_slp_analyze_node_operations): Likewise.
9100 (vect_slp_analyze_bb_1): Likewise.
9101 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9102 than an unsigned HOST_WIDE_INT.
9103 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9104 (vectorizable_load): Handle polynomial vf.
9105 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9107 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9110 Alan Hayward <alan.hayward@arm.com>
9111 David Sherwood <david.sherwood@arm.com>
9113 * match.pd: Handle bit operations involving three constants
9114 and try to fold one pair.
9116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9118 * tree-vect-loop-manip.c: Include gimple-fold.h.
9119 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9120 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9121 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9122 Add a path that uses a step of VF instead of 1, but disable it
9124 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9125 and niters_no_overflow parameters. Update calls to
9126 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9127 Create a new SSA name if the latter choses to use a ste other
9128 than zero, and return it via niters_vector_mult_vf_var.
9129 * tree-vect-loop.c (vect_transform_loop): Update calls to
9130 vect_do_peeling, vect_gen_vector_loop_niters and
9131 slpeel_make_loop_iterate_ntimes.
9132 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9133 (vect_gen_vector_loop_niters): Update declarations after above changes.
9135 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9137 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9138 128-bit round to integer instructions.
9139 (ceil<mode>2): Likewise.
9140 (btrunc<mode>2): Likewise.
9141 (round<mode>2): Likewise.
9143 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9145 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9146 unaligned VSX load/store on P8/P9.
9147 (expand_block_clear): Allow the use of unaligned VSX
9148 load/store on P8/P9.
9150 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9152 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9154 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9155 swap associated with both a load and a store.
9157 2018-01-02 Andrew Waterman <andrew@sifive.com>
9159 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9160 * config/riscv/riscv.md (clear_cache): Use it.
9162 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9164 * web.c: Remove out-of-date comment.
9166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9168 * expr.c (fixup_args_size_notes): Check that any existing
9169 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9170 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9171 (emit_single_push_insn): ...here.
9173 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9175 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9176 (const_vector_encoded_nelts): New function.
9177 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9178 (const_vector_int_elt, const_vector_elt): Declare.
9179 * emit-rtl.c (const_vector_int_elt_1): New function.
9180 (const_vector_elt): Likewise.
9181 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9182 of CONST_VECTOR_ELT.
9184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9186 * expr.c: Include rtx-vector-builder.h.
9187 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9188 directly on the tree encoding.
9189 (const_vector_from_tree): Likewise.
9190 * optabs.c: Include rtx-vector-builder.h.
9191 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9192 sequence of "u" values.
9193 * vec-perm-indices.c: Include rtx-vector-builder.h.
9194 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9195 directly on the vec_perm_indices encoding.
9197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9199 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9200 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9201 * rtx-vector-builder.h: New file.
9202 * rtx-vector-builder.c: Likewise.
9203 * rtl.h (rtx_def::u2): Add a const_vector field.
9204 (CONST_VECTOR_NPATTERNS): New macro.
9205 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9206 (CONST_VECTOR_DUPLICATE_P): Likewise.
9207 (CONST_VECTOR_STEPPED_P): Likewise.
9208 (CONST_VECTOR_ENCODED_ELT): Likewise.
9209 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9210 (unwrap_const_vec_duplicate): Likewise.
9211 (const_vec_series_p): Check for a non-duplicated vector encoding.
9212 Say that the function only returns true for integer vectors.
9213 * emit-rtl.c: Include rtx-vector-builder.h.
9214 (gen_const_vec_duplicate_1): Delete.
9215 (gen_const_vector): Call gen_const_vec_duplicate instead of
9216 gen_const_vec_duplicate_1.
9217 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9218 (gen_const_vec_duplicate): Use rtx_vector_builder.
9219 (gen_const_vec_series): Likewise.
9220 (gen_rtx_CONST_VECTOR): Likewise.
9221 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9222 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9223 Build a new vector rather than modifying a CONST_VECTOR in-place.
9224 (handle_special_swappables): Update call accordingly.
9225 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9226 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9227 Build a new vector rather than modifying a CONST_VECTOR in-place.
9228 (handle_special_swappables): Update call accordingly.
9230 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9232 * simplify-rtx.c (simplify_const_binary_operation): Use
9233 CONST_VECTOR_ELT instead of XVECEXP.
9235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9237 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9238 the selector elements to be different from the data elements
9239 if the selector is a VECTOR_CST.
9240 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9241 ssizetype for the selector.
9243 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9245 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9246 before testing each element individually.
9247 * tree-vect-generic.c (lower_vec_perm): Likewise.
9249 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9251 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9252 * selftest-run-tests.c (selftest::run_tests): Call it.
9253 * vector-builder.h (vector_builder::operator ==): New function.
9254 (vector_builder::operator !=): Likewise.
9255 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9256 (vec_perm_indices::all_from_input_p): New function.
9257 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9258 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9259 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9260 instead of reading the VECTOR_CST directly. Detect whether both
9261 vector inputs are the same before constructing the vec_perm_indices,
9262 and update the number of inputs argument accordingly. Use the
9263 utility functions added above. Only construct sel2 if we need to.
9265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9267 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9268 the broadcast of the low byte.
9269 (expand_mult_highpart): Use an explicit encoding for the permutes.
9270 * optabs-query.c (can_mult_highpart_p): Likewise.
9271 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9272 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9273 (vectorizable_bswap): Likewise.
9274 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9275 explicit encoding for the power-of-2 permutes.
9276 (vect_permute_store_chain): Likewise.
9277 (vect_grouped_load_supported): Likewise.
9278 (vect_permute_load_chain): Likewise.
9280 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9282 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9283 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9284 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9285 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9286 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9287 (vect_gen_perm_mask_any): Likewise.
9289 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9291 * int-vector-builder.h: New file.
9292 * vec-perm-indices.h: Include int-vector-builder.h.
9293 (vec_perm_indices): Redefine as an int_vector_builder.
9294 (auto_vec_perm_indices): Delete.
9295 (vec_perm_builder): Redefine as a stand-alone class.
9296 (vec_perm_indices::vec_perm_indices): New function.
9297 (vec_perm_indices::clamp): Likewise.
9298 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9299 (vec_perm_indices::new_vector): New function.
9300 (vec_perm_indices::new_expanded_vector): Update for new
9301 vec_perm_indices class.
9302 (vec_perm_indices::rotate_inputs): New function.
9303 (vec_perm_indices::all_in_range_p): Operate directly on the
9304 encoded form, without computing elided elements.
9305 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9306 encoding. Update for new vec_perm_indices class.
9307 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9308 the given vec_perm_builder.
9309 (expand_vec_perm_var): Update vec_perm_builder constructor.
9310 (expand_mult_highpart): Use vec_perm_builder instead of
9311 auto_vec_perm_indices.
9312 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9313 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9314 or double series encoding as appropriate.
9315 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9316 vec_perm_indices instead of auto_vec_perm_indices.
9317 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9318 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9319 (vect_permute_store_chain): Likewise.
9320 (vect_grouped_load_supported): Likewise.
9321 (vect_permute_load_chain): Likewise.
9322 (vect_shift_permute_load_chain): Likewise.
9323 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9324 (vect_transform_slp_perm_load): Likewise.
9325 (vect_schedule_slp_instance): Likewise.
9326 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9327 (vectorizable_mask_load_store): Likewise.
9328 (vectorizable_bswap): Likewise.
9329 (vectorizable_store): Likewise.
9330 (vectorizable_load): Likewise.
9331 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9332 vec_perm_indices instead of auto_vec_perm_indices. Use
9333 tree_to_vec_perm_builder to read the vector from a tree.
9334 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9335 vec_perm_builder instead of a vec_perm_indices.
9336 (have_whole_vector_shift): Use vec_perm_builder and
9337 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9338 truncation to calc_vec_perm_mask_for_shift.
9339 (vect_create_epilog_for_reduction): Likewise.
9340 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9341 from auto_vec_perm_indices to vec_perm_indices.
9342 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9343 instead of changing individual elements.
9344 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9345 the vector in d.perm.
9346 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9347 from auto_vec_perm_indices to vec_perm_indices.
9348 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9349 instead of changing individual elements.
9350 (arm_vectorize_vec_perm_const): Use new_vector to install
9351 the vector in d.perm.
9352 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9353 Update vec_perm_builder constructor.
9354 (rs6000_expand_interleave): Likewise.
9355 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9356 (rs6000_expand_interleave): Likewise.
9358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9360 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9361 to qimode could truncate the indices.
9362 * optabs.c (expand_vec_perm_var): Likewise.
9364 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9366 * Makefile.in (OBJS): Add vec-perm-indices.o.
9367 * vec-perm-indices.h: New file.
9368 * vec-perm-indices.c: Likewise.
9369 * target.h (vec_perm_indices): Replace with a forward class
9371 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9372 * optabs.h: Include vec-perm-indices.h.
9373 (expand_vec_perm): Delete.
9374 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9375 (expand_vec_perm_const): Declare.
9376 * target.def (vec_perm_const_ok): Replace with...
9377 (vec_perm_const): ...this new hook.
9378 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9379 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9380 * doc/tm.texi: Regenerate.
9381 * optabs.def (vec_perm_const): Delete.
9382 * doc/md.texi (vec_perm_const): Likewise.
9383 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9384 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9385 expand_vec_perm for constant permutation vectors. Assert that
9386 the mode of variable permutation vectors is the integer equivalent
9387 of the mode that is being permuted.
9388 * optabs-query.h (selector_fits_mode_p): Declare.
9389 * optabs-query.c: Include vec-perm-indices.h.
9390 (selector_fits_mode_p): New function.
9391 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9392 is defined, instead of checking whether the vec_perm_const_optab
9393 exists. Use targetm.vectorize.vec_perm_const instead of
9394 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9395 fit in the vector mode before using a variable permute.
9396 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9397 vec_perm_indices instead of an rtx.
9398 (expand_vec_perm): Replace with...
9399 (expand_vec_perm_const): ...this new function. Take the selector
9400 as a vec_perm_indices rather than an rtx. Also take the mode of
9401 the selector. Update call to shift_amt_for_vec_perm_mask.
9402 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9403 Use vec_perm_indices::new_expanded_vector to expand the original
9404 selector into bytes. Check whether the indices fit in the vector
9405 mode before using a variable permute.
9406 (expand_vec_perm_var): Make global.
9407 (expand_mult_highpart): Use expand_vec_perm_const.
9408 * fold-const.c: Includes vec-perm-indices.h.
9409 * tree-ssa-forwprop.c: Likewise.
9410 * tree-vect-data-refs.c: Likewise.
9411 * tree-vect-generic.c: Likewise.
9412 * tree-vect-loop.c: Likewise.
9413 * tree-vect-slp.c: Likewise.
9414 * tree-vect-stmts.c: Likewise.
9415 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9417 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9418 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9419 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9420 (aarch64_vectorize_vec_perm_const): ...this new function.
9421 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9422 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9423 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9424 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9425 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9426 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9427 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9429 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9430 check for NEON modes.
9431 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9432 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9433 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9434 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9436 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9437 the old VEC_PERM_CONST conditions.
9438 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9439 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9440 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9441 (ia64_vectorize_vec_perm_const_ok): Merge into...
9442 (ia64_vectorize_vec_perm_const): ...this new function.
9443 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9444 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9445 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9446 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9447 * config/mips/mips.c (mips_expand_vec_perm_const)
9448 (mips_vectorize_vec_perm_const_ok): Merge into...
9449 (mips_vectorize_vec_perm_const): ...this new function.
9450 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9451 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9452 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9453 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9454 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9455 (rs6000_expand_vec_perm_const): Delete.
9456 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9458 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9459 (altivec_expand_vec_perm_const_le): Take each operand individually.
9460 Operate on constant selectors rather than rtxes.
9461 (altivec_expand_vec_perm_const): Likewise. Update call to
9462 altivec_expand_vec_perm_const_le.
9463 (rs6000_expand_vec_perm_const): Delete.
9464 (rs6000_vectorize_vec_perm_const_ok): Delete.
9465 (rs6000_vectorize_vec_perm_const): New function.
9466 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9467 an element count and rtx array.
9468 (rs6000_expand_extract_even): Update call accordingly.
9469 (rs6000_expand_interleave): Likewise.
9470 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9471 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9472 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9473 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9474 (rs6000_expand_vec_perm_const): Delete.
9475 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9476 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9477 (altivec_expand_vec_perm_const_le): Take each operand individually.
9478 Operate on constant selectors rather than rtxes.
9479 (altivec_expand_vec_perm_const): Likewise. Update call to
9480 altivec_expand_vec_perm_const_le.
9481 (rs6000_expand_vec_perm_const): Delete.
9482 (rs6000_vectorize_vec_perm_const_ok): Delete.
9483 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9484 reference to the SPE evmerge intructions.
9485 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9486 an element count and rtx array.
9487 (rs6000_expand_extract_even): Update call accordingly.
9488 (rs6000_expand_interleave): Likewise.
9489 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9490 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9492 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9496 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9497 vector mode and that that mode matches the mode of the data
9499 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9500 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9501 directly using expand_vec_perm_1 when forcing selectors into
9503 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9507 * optabs-query.h (can_vec_perm_p): Delete.
9508 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9509 * optabs-query.c (can_vec_perm_p): Split into...
9510 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9511 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9512 particular selector is valid.
9513 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9514 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9515 (vect_grouped_load_supported): Likewise.
9516 (vect_shift_permute_load_chain): Likewise.
9517 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9518 (vect_transform_slp_perm_load): Likewise.
9519 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9520 (vectorizable_bswap): Likewise.
9521 (vect_gen_perm_mask_checked): Likewise.
9522 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9523 implementations of variable permutation vectors into account
9524 when deciding which selector to use.
9525 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9526 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9527 with a false third argument.
9528 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9529 to test whether the constant selector is valid and can_vec_perm_var_p
9530 to test whether a variable selector is valid.
9532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9534 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9535 * optabs-query.c (can_vec_perm_p): Likewise.
9536 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9537 instead of vec_perm_indices.
9538 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9539 (vect_gen_perm_mask_checked): Likewise,
9540 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9541 (vect_gen_perm_mask_checked): Likewise,
9543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9545 * optabs-query.h (qimode_for_vec_perm): Declare.
9546 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9547 (qimode_for_vec_perm): ...this new function.
9548 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9550 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9552 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9553 does not have a conditional at the top.
9555 2018-01-02 Richard Biener <rguenther@suse.de>
9557 * ipa-inline.c (big_speedup_p): Fix expression.
9559 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9562 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9565 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9569 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9570 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9571 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9572 cond_taken_branch_cost 3->4.
9574 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9576 PR tree-optimization/83581
9577 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9578 TODO_cleanup_cfg if any changes have been made.
9581 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9582 convert_modes if target mode has the right side, but different mode
9586 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9587 last argument when extracting from CONCAT. If either from_real or
9588 from_imag is NULL, use expansion through memory. If result is not
9589 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9590 the parts directly to inner mode, if even that fails, use expansion
9594 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9595 check for bswap in mode rather than HImode and use that in expand_unop
9598 Copyright (C) 2018 Free Software Foundation, Inc.
9600 Copying and distribu