1 2018-03-20 Carl Love <cel@us.ibm.com>
4 2017-09-27 Carl Love <cel@us.ibm.com>
6 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
14 instead of __vector bool.
15 (_mm_max_pu8): Likewise.
16 (_mm_min_pi16): Likewise.
18 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
21 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
25 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
27 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
28 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
29 (div_extend): Likewise.
30 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
32 (__builtin_divweuo): Likewise.
33 (__builtin_divdeo): Likewise.
34 (__builtin_divdeuo): Likewise.
36 2018-03-28 Jakub Jelinek <jakub@redhat.com>
39 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
40 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
42 PR tree-optimization/82004
43 * gimple-match-head.c (optimize_pow_to_exp): New function.
44 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
45 Don't fold to exp if optimize_pow_to_exp is false.
47 2018-03-28 Martin Liska <mliska@suse.cz>
50 * calls.c (initialize_argument_information): Fix trailing space.
51 * common.opt: Fix typo and provide better explanation for
52 -fsanitize-coverage option.
53 * config/i386/i386.opt: Fix typo.
55 2018-03-28 Jakub Jelinek <jakub@redhat.com>
56 Martin Liska <mliska@suse.cz>
59 * gimplify.c (asan_poison_variable): Don't do the check for
60 gimplify_omp_ctxp here.
61 (gimplify_decl_expr): Do it here.
62 (gimplify_target_expr): Likewise.
64 2018-03-28 Martin Liska <mliska@suse.cz>
67 * config/i386/i386.c (ix86_function_arg_advance): Do not call
68 chkp_type_bounds_count if MPX is not enabled.
70 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
72 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
74 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
77 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
78 function to create the function decl for complex long double
79 multiply and divide for -mabi=ieeelongdouble.
80 (init_float128_ieee): Call it.
82 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
85 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
86 -fcf-protection=branch -mibt.
87 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
89 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
92 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
94 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
97 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
98 extern array declarations.
100 2018-03-27 Richard Biener <rguenther@suse.de>
103 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
104 explicit single_use checks.
106 2018-03-27 Richard Biener <rguenther@suse.de>
108 PR tree-optimization/85082
109 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
112 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
114 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
115 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
116 Turn on fasynchronous-unwind-tables and funwind-tables.
118 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
121 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
122 (*bmi_blsr_<mode>_ccz): Ditto.
124 2018-03-26 Tom de Vries <tom@codesourcery.com>
126 PR tree-optimization/85063
127 * omp-general.c (offloading_function_p): New function. Factor out
129 * omp-offload.c (pass_omp_target_link::gate): ... here.
130 * omp-general.h (offloading_function_p): Declare.
131 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
132 with attribute omp declare target for offloading functions.
134 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
136 PR tree-optimization/84005
137 * tree-data-ref.h (get_base_for_alignment): Declare.
138 * tree-data-ref.c (get_base_for_alignment_1): New function.
139 (get_base_for_alignment): Likewise.
140 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
141 get_base_for_alignment to find a suitable base object, instead
142 of always using drb->base_address.
144 2018-03-23 Jakub Jelinek <jakub@redhat.com>
147 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
148 known size by default.
150 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
153 * lra-constraints.c (process_alt_operands): Don't match BLKmode
154 and non BLKmode operands.
156 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
159 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
162 2018-03-23 Richard Biener <rguenther@suse.de>
165 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
166 we are going to emit early debug for LTO.
168 2018-03-23 Jakub Jelinek <jakub@redhat.com>
171 * function.c (match_asm_constraints_1): Don't optimize if input
172 doesn't satisfy general_operand predicate for output's mode.
175 * alias.c (write_dependence_p): Don't require for x_canonicalized
176 non-VOIDmode if x has VOIDmode.
179 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
180 just don't try to optimize it rather than assert it never happens.
182 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
184 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
185 macro expansions for definition of ST_INTERNAL_<mode> and
186 LD_INTERNAL_<mode> builtins.
187 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
189 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
191 (altivec_expand_st_builtin): Likewise.
192 (altivec_expand_builtin): Remove calls to deleted functions.
193 (rs6000_address_for_altivec): Delete this function.
194 * config/rs6000/vector.md: Remove expands for
195 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
197 2018-03-22 Sudakshina Das <sudi.das@arm.com>
200 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
201 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
202 re-computing once computed.
203 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
204 (arm_init_machine_status): Initialize
205 machine->static_chain_stack_bytes.
207 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
210 * doc/extend.texi: Add four new prototypes for vec_ld.
211 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
212 definitions for more logical presentation.
213 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
214 entries for V1TI variants of __builtin_altivec_ld builtin.
215 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
216 handling of V1TI variant of LVX icode pattern.
217 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
218 (rs6000_gimple_fold_builtin): Likewise.
219 (altivec_init_builtins): Add code to define
220 __builtin_altivec_lvx_v1ti function.
222 2018-03-22 Jakub Jelinek <jakub@redhat.com>
225 * function.c (match_asm_constraints_1): Don't do the optimization
226 if input isn't a REG, SUBREG, MEM or constant.
228 2018-03-22 Tom de Vries <tom@codesourcery.com>
230 PR tree-optimization/84956
231 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
232 bb_has_abnormal_pred.
234 2018-03-22 Jakub Jelinek <jakub@redhat.com>
237 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
238 DECL_INITIAL (decl) to decl at the end.
239 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
242 2018-03-21 Joseph Myers <joseph@codesourcery.com>
244 * doc/extend.texi (__builtin_tgmath): Document when complex
245 integer types are treated as _Complex _Float64.
247 2018-03-21 Tom de Vries <tom@codesourcery.com>
249 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
251 2018-03-21 Jakub Jelinek <jakub@redhat.com>
253 PR tree-optimization/84960
254 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
255 if it is ENTRY block, move them into single succ of ENTRY in that case.
257 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
259 PR tree-optimization/84811
260 * poly-int.h (poly_span_traits): Remove the T3 parameter and
261 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
262 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
263 (known_subrange_p): Update accordingly. Cast each value involved
264 in the size comparison, rather than casting the result of the
267 2018-03-21 Jakub Jelinek <jakub@redhat.com>
269 PR tree-optimization/84982
270 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
271 by flipping the least significant bit rather than all bits from
272 bitpos to bitpos + bitsize - 1.
274 2018-03-21 Nathan Sidwell <nathan@acm.org>
276 * doc/extend.texi (Deprecated Features): Remove mention of
277 long-deleted deprecations.
279 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
282 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
283 * configure: Regenerate.
285 2018-03-21 Tom de Vries <tom@codesourcery.com>
287 PR tree-optimization/83126
288 * tree-parloops.c (num_phis): New function.
289 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
291 2018-03-21 Nathan Sidwell <nathan@acm.org>
293 * doc/extend.texi (Deprecated Features): Update deprecated flags,
294 mention anon-struct/union members and trailing attributes.
296 2018-03-21 Bin Cheng <bin.cheng@arm.com>
298 PR tree-optimization/84969
299 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
300 builtin memset partitions if they set different rhs values.
302 2018-03-21 Jakub Jelinek <jakub@redhat.com>
304 PR rtl-optimization/84989
305 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
306 VEC_DUPLICATE with scalar result mode.
308 2018-03-21 Martin Liska <mliska@suse.cz>
311 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
312 not intended return statement.
314 2018-03-21 Martin Liska <mliska@suse.cz>
317 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
318 (chkp_find_bound_slots_1): Limit number of iterations.
320 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
323 * Minor grammar fixes for x86 options.
325 2018-03-20 Jakub Jelinek <jakub@redhat.com>
328 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
329 holding REG_CFA_RESTORE notes, instead turn them into a USE.
331 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
334 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
335 (altivec_lvx_<mode>_1op): Likewise.
336 (altivec_stvx_<mode>_2op): Likewise.
337 (altivec_stvx_<mode>_1op): Likewise.
338 (altivec_lvx_<VM2:mode>): New define_expand.
339 (altivec_stvx_<VM2:mode>): Likewise.
340 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
341 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
342 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
343 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
344 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
345 (rs6000_gen_lvx): Likewise.
346 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
347 (altivec_expand_stv_builtin): Likewise.
348 (altivec_expand_builtin): Likewise.
349 * config/rs6000/vector.md: Likewise.
351 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
354 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
357 2018-03-20 Richard Biener <rguenther@suse.de>
360 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
361 sign-conversions as zero, fall back to standard scalar_stmt
364 2018-03-20 Martin Liska <mliska@suse.cz>
367 * predict.c (rebuild_frequencies): Handle case when we have
368 PROFILE_ABSENT, but flag_guess_branch_prob is false.
370 2018-03-20 Jakub Jelinek <jakub@redhat.com>
373 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
374 flag_section_anchors.
375 * varasm.c (use_blocks_for_decl_p): Remove hack for
379 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
381 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
382 be created, use lowpart_subreg of operands[0] rather than operands[0]
384 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
385 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
386 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
387 and n constraint instead of aarch64_shift_imm_di and Usd.
388 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
389 (*aarch64_<optab>_reg_minus<mode>3): ... this.
391 2018-03-20 Sudakshina Das <sudi.das@arm.com>
394 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
395 to favor GPR over NEON registers.
396 (<shift>di3_neon): Likewise.
398 2018-03-20 Tom de Vries <tom@codesourcery.com>
401 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
402 (nvptx_process_pars): Emit bar.sync asap and alap.
404 2018-03-20 Tom de Vries <tom@codesourcery.com>
407 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
408 seen_label if seen_label is already set.
410 2018-03-20 Jakub Jelinek <jakub@redhat.com>
413 * config/i386/i386.c (fold_builtin_cpu): For features above 31
414 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
415 Use 1U instead of 1. Formatting fixes.
418 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
419 instead of TREE_TYPE (s1) for the return value.
421 2018-03-19 Jakub Jelinek <jakub@redhat.com>
423 PR tree-optimization/84946
424 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
425 bitsize + bitsize in poly_uint64 rather than poly_int64.
428 * dwarf2asm.c: Include fold-const.c.
429 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
430 of decl rather than decl itself.
432 PR rtl-optimization/84643
433 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
435 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
438 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
439 calling assemble_variable.
441 2018-03-19 Sudakshina Das <sudi.das@arm.com>
444 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
445 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
447 2018-03-19 Jim Wilson <jimw@sifive.com>
450 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
451 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
452 (riscv_first_stack_step): Likewise.
453 (riscv_option_override): Use STACK_BOUNDARY instead of
455 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
457 (BIGGEST_ALIGNMENT): Set to 128.
458 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
459 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
462 2018-03-19 Richard Biener <rguenther@suse.de>
464 PR tree-optimization/84933
465 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
466 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
468 2018-03-19 Richard Biener <rguenther@suse.de>
470 PR tree-optimization/84859
471 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
472 (cond_if_else_store_replacement): Perform sinking operation on
473 single-store BBs regardless of MAX_STORES_TO_SINK setting.
474 Generalize what a BB with a single eligible store is.
476 2018-03-19 Richard Biener <rguenther@suse.de>
478 PR tree-optimization/84929
479 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
480 chrec_is_positive against non-chrec arg.
482 2018-03-19 Tamar Christina <tamar.christina@arm.com>
485 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
487 2018-03-18 Martin Liska <mliska@suse.cz>
489 PR rtl-optimization/84635
490 * regrename.c (build_def_use): Use matches_mode only when
493 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
495 PR tree-optimization/84913
496 * tree-vect-loop.c (vectorizable_reduction): Don't try to
497 vectorize chains of COND_EXPRs.
499 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
501 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
503 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
505 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
507 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
509 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
511 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
512 Kito Cheng <kito.cheng@gmail.com>
514 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
515 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
516 (nds32_adjust_reg_alloc_order): New function.
517 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
519 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
521 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
522 nds32_print_operand, nds32_print_operand_address): Use
523 HOST_WIDE_INT_PRINT_DEC instead.
525 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
527 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
529 2018-03-17 Jakub Jelinek <jakub@redhat.com>
532 * config/i386/i386.c (initial_ix86_tune_features,
533 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
535 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
536 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
537 rather than 1u << ix86_tune. Formatting fix.
538 (ix86_option_override_internal): Change ix86_arch_mask from
539 unsigned int to unsigned HOST_WIDE_INT, initialize to
540 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
541 (ix86_function_specific_restore): Likewise.
543 2018-03-16 Jakub Jelinek <jakub@redhat.com>
546 * postreload.c (reload_combine_recognize_pattern): Perform
547 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
548 truncate_int_for_mode the result for the destination's mode.
551 * hsa-gen.c (mem_type_for_type): Fix comment typo.
552 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
554 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
557 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
560 * lra-assigns.c (lra_split_hard_reg_for): Don't use
561 regno_allocno_class_array and sorted_pseudos.
562 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
563 insns where regno is used.
565 2018-03-16 Martin Liska <mliska@suse.cz>
568 * multiple_target.c (create_dispatcher_calls): Redirect
569 reference in the symbol table.
571 2018-03-16 Martin Liska <mliska@suse.cz>
574 * multiple_target.c (create_dispatcher_calls): Redirect also
577 2018-03-16 Jakub Jelinek <jakub@redhat.com>
581 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
582 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
583 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
585 2018-03-16 Julia Koval <julia.koval@intel.com>
587 * doc/invoke.texi (Skylake Server): Add CLWB.
588 Cannonlake): Remove CLWB.
590 2018-03-16 Jakub Jelinek <jakub@redhat.com>
592 PR tree-optimization/84841
593 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
595 (FLOAT_ONE_CONST_TYPE): Define.
596 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
597 (sort_by_operand_rank): Put entries with higher constant_type last
598 rather than first to match comments.
600 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
602 * config/nios2/nios2.md (movsi_internal): Fix thinko in
605 2018-03-15 Jakub Jelinek <jakub@redhat.com>
608 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
609 check and use address of target always.
611 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
614 * config/i386/i386.c (indirect_thunk_needed): Update comments.
615 (indirect_thunk_bnd_needed): Likewise.
616 (indirect_thunks_used): Likewise.
617 (indirect_thunks_bnd_used): Likewise.
618 (indirect_return_needed): New.
619 (indirect_return_bnd_needed): Likewise.
620 (output_indirect_thunk_function): Add a bool argument for
622 (output_indirect_thunk_function): Don't generate alias for
623 function return thunk.
624 (ix86_code_end): Call output_indirect_thunk_function to generate
625 function return thunks.
626 (ix86_output_function_return): Set indirect_return_bnd_needed
627 and indirect_return_needed instead of indirect_thunk_bnd_needed
628 and indirect_thunk_needed.
630 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
632 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
633 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
634 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
636 2018-03-15 David Malcolm <dmalcolm@redhat.com>
637 Paul Hua <paul.hua.gm@gmail.com>
640 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
642 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
644 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
645 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
648 2018-03-15 Tamar Christina <tamar.christina@arm.com>
651 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
652 instead of GET_MODE_SIZE when comparing Units.
654 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
657 * varasm.c (hash_section): Return an unchangeble hash value
658 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
659 Return !aarch64_can_use_per_function_literal_pools_p ().
661 2018-03-15 Jakub Jelinek <jakub@redhat.com>
664 * optabs.c (emit_conditional_move): Pass address of cmode's copy
665 rather than address of cmode as last argument to prepare_cmp_insn.
667 2018-03-15 Julia Koval <julia.koval@intel.com>
669 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
670 F_AVX512VNNI, F_AVX512BITALG): New.
672 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
675 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
676 insn for floating-point loads and stores.
678 2018-03-14 Carl Love <cel@us.ibm.com>
680 * config/rs6000/rs6000-c.c: Add macro definitions for
681 ALTIVEC_BUILTIN_VEC_PERMXOR.
682 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
683 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
684 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
685 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
687 * config/doc/extend.texi: Add prototypes for vec_permxor.
689 2018-03-14 David Malcolm <dmalcolm@redhat.com>
692 * diagnostic-show-locus.c (class layout_point): Convert m_line
693 from int to linenum_type.
694 (line_span::comparator): Use linenum "compare" function when
695 comparing line numbers.
696 (test_line_span): New function.
697 (layout_range::contains_point): Convert param "row" from int to
699 (layout_range::intersects_line_p): Likewise.
700 (layout::will_show_line_p): Likewise.
701 (layout::print_source_line): Likewise.
702 (layout::should_print_annotation_line_p): Likewise.
703 (layout::print_annotation_line): Likewise.
704 (layout::print_leading_fixits): Likewise.
705 (layout::annotation_line_showed_range_p): Likewise.
706 (struct line_corrections): Likewise for field m_row.
707 (line_corrections::line_corrections): Likewise for param "row".
708 (layout::print_trailing_fixits): Likewise.
709 (layout::get_state_at_point): Likewise.
710 (layout::get_x_bound_for_row): Likewise.
711 (layout::print_line): Likewise.
712 (diagnostic_show_locus): Likewise for locals "last_line" and
714 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
715 * input.c (selftest::test_linenum_comparisons): New function.
716 (selftest::input_c_tests): Call it.
717 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
718 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
719 * selftest.h (ASSERT_GT): New macro.
720 (ASSERT_GT_AT): New macro.
721 (ASSERT_LT): New macro.
722 (ASSERT_LT_AT): New macro.
724 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
726 PR rtl-optimization/84780
727 * combine.c (distribute_links): Don't make a link based on pc_rtx.
729 2018-03-14 Martin Liska <mliska@suse.cz>
731 * tree.c (record_node_allocation_statistics): Use
733 (get_stats_node_kind): New function extracted from
734 record_node_allocation_statistics.
735 (free_node): Use get_stats_node_kind.
737 2018-03-14 Richard Biener <rguenther@suse.de>
739 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
740 that the value-set of ANTIC_IN doesn't grow.
743 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
745 (BB_VISITED_WITH_VISITED_SUCCS): New define.
746 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
748 2018-03-14 Julia Koval <julia.koval@intel.com>
750 * config.gcc (icelake-client, icelake-server): New.
752 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
753 (initial_ix86_arch_features): Ditto.
754 (PTA_SKYLAKE): Add SGX.
755 (PTA_ICELAKE): Remove.
756 (PTA_ICELAKE_CLIENT): New.
757 (PTA_ICELAKE_SERVER): New.
758 (ix86_option_override_internal): Split up icelake on icelake client and
760 (get_builtin_code_for_version): Ditto.
761 (fold_builtin_cpu): Ditto.
762 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
763 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
764 * config/i386/i386.h (processor_type): Ditto.
765 * doc/invoke.texi: Ditto.
767 2018-03-14 Jakub Jelinek <jakub@redhat.com>
770 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
771 INTEGER_CST offset, add it together with bitpos / 8 and
772 sign extend based on POINTER_SIZE.
776 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
779 * config/i386/constraints.md (Yc): New register constraint.
780 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
781 Use Yc constraint for alternative 2 of operand 0. Remove
782 preferred_for_speed attribute.
784 2018-03-14 Richard Biener <rguenther@suse.de>
786 PR tree-optimization/84830
787 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
788 with the old one to avoid oscillations.
790 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
793 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
795 (assign_by_spills): Return a flag of reload assignment failure.
796 Do not process the reload assignment failures. Do not spill other
797 reload pseudos if they has the same reg class. Update n if
799 (lra_assign): Add a return arg. Set up from the result of
800 assign_by_spills call.
801 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
802 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
803 usage_insns if it is not NULL.
804 (spill_hard_reg_in_range): New function.
805 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
806 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
808 (lra_assign): Change prototype.
809 * lra.c (lra): Add code to deal with fails by splitting hard reg
812 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
814 * config/riscv/riscv.opt (mrelax): New option.
815 * config/riscv/riscv.c (riscv_file_start): Emit ".option
816 "norelax" when riscv_mrelax is disabled.
817 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
819 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
822 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
823 reassociation for int modes.
825 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
827 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
828 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
830 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
831 * config/aarch64/aarch64-sve.md
832 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
833 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
834 (*extend<mode><Vwide>2): Rename to...
835 (aarch64_sve_extend<mode><Vwide>2): ...this.
836 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
837 renaming the old pattern to...
838 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
840 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
841 define_expand, renaming the old pattern to...
842 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
843 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
844 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
845 account when deciding which SVE instruction the optab should use.
846 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
848 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
850 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
851 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
852 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
853 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
854 (tlsdesc_small_<mode>): Turn a define_expand and use
855 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
856 (tlsdesc_small_advsimd_<mode>): ...this.
857 (tlsdesc_small_sve_<mode>): New pattern.
859 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
861 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
862 (UNSPEC_UMUL_HIGHPART): New constants.
863 (MUL_HIGHPART): New int iteraor.
864 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
865 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
867 (*<su>mul<mode>3_highpart): New define_insn.
869 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
872 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
875 2018-03-13 Martin Liska <mliska@suse.cz>
878 * (sem_item_optimizer::sem_item_optimizer): Initialize new
880 (sem_item_optimizer::~sem_item_optimizer): Release it.
881 (sem_item_optimizer::merge_classes): Register variable aliases.
882 (sem_item_optimizer::fixup_pt_set): New function.
883 (sem_item_optimizer::fixup_points_to_sets): Likewise.
884 * ipa-icf.h: Declare new variables and functions.
886 2018-03-13 Jakub Jelinek <jakub@redhat.com>
889 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
890 integer_pow2p@2 and test integer_pow2p in condition.
891 (A < 0 ? C : 0): Similarly for @1.
894 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
895 characters starting at p contain '\0' character, don't look beyond
899 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
900 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
903 * reg-stack.c (change_stack): Change update_end var from int to
904 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
905 also call set_block_for_insn on the newly added insns and rescan.
908 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
912 * tree.c (stabilize_reference_1): Return save_expr (e) for
913 STATEMENT_LIST even if it doesn't have side-effects.
915 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
917 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
919 2018-03-12 Renlin Li <renlin.li@arm.com>
921 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
922 aarch64_output_scalar_simd_mov_immediate.
924 2018-03-12 Martin Sebor <msebor@redhat.com>
926 PR tree-optimization/83456
927 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
928 for perfectly overlapping calls to memcpy.
929 (gimple_fold_builtin_memory_chk): Same.
930 (gimple_fold_builtin_strcpy): Handle no-warning.
931 (gimple_fold_builtin_stxcpy_chk): Same.
932 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
934 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
936 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
937 parameter. Use it for SFmode.
938 (rs6000_function_arg_advance_1): Adjust.
939 (rs6000_function_arg): Adjust.
940 (rs6000_gimplify_va_arg): Pass false for that new parameter.
942 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
944 PR rtl-optimization/84169
945 PR rtl-optimization/84780
946 * combine.c (can_combine_p): Check for a 2-insn combination whether
947 the destination register is used between the two insns, too.
949 2018-03-12 Richard Biener <rguenther@suse.de>
951 PR tree-optimization/84803
952 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
953 for refs DR analysis didn't process.
955 2018-03-12 Richard Biener <rguenther@suse.de>
957 PR tree-optimization/84777
958 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
959 force-vectorize loops ignore whether we are optimizing for size.
961 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
963 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
964 (TARGET_MD_ASM_ADJUST): Define.
966 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
967 Kito Cheng <kito.cheng@gmail.com>
968 Chung-Ju Wu <jasonwucj@gmail.com>
970 * config/nds32/nds32.c (nds32_compute_stack_frame,
971 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
972 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
973 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
974 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
975 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
976 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
977 * config/nds32/nds32.md (prologue, epilogue): Use macro
978 NDS32_V3PUSH_AVAILABLE_P to do checking.
980 2018-03-11 Jakub Jelinek <jakub@redhat.com>
983 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
984 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
985 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
986 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
987 addition of most attributes on !orig_type_die or the attribute not
988 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
990 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
991 Chung-Ju Wu <jasonwucj@gmail.com>
993 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
995 * config/nds32/nds32.opt (mvh): New option.
997 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
998 Chung-Ju Wu <jasonwucj@gmail.com>
1000 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1002 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1003 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1006 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1007 Chung-Ju Wu <jasonwucj@gmail.com>
1009 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1011 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1012 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1014 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1015 Kito Cheng <kito.cheng@gmail.com>
1016 Chung-Ju Wu <jasonwucj@gmail.com>
1018 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1019 UNSPEC_FFMISM and UNSPEC_FLMISM.
1020 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1021 for ffb, ffmism and flmism.
1022 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1023 (unspec_ffmism): Ditto.
1024 (unspec_flmism): Ditto.
1025 (nds32_expand_builtin_impl): Check if string extension is available.
1026 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1027 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1029 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1032 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1035 * lra-assigns.c (assign_by_spills): Return a flag of reload
1036 assignment failure. Do not process the reload assignment
1037 failures. Do not spill other reload pseudos if they has the same
1039 (lra_assign): Add a return arg. Set up from the result of
1040 assign_by_spills call.
1041 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1042 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1043 usage_insns if it is not NULL.
1044 (spill_hard_reg_in_range): New function.
1045 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1046 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1047 function prototypes.
1048 (lra_assign): Change prototype.
1049 * lra.c (lra): Add code to deal with fails by splitting hard reg
1052 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1055 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1057 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1060 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1061 (dw_val_node): Add val_symbolic_view.
1062 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1063 (symview_upper_bound): New.
1064 (new_line_info_table): Initialize symviews_since_reset.
1065 (dwarf2out_source_line): Count symviews_since_reset and set
1066 symview_upper_bound.
1067 (dw_val_equal_p): Handle symview.
1068 (add_AT_symview): New.
1069 (print_dw_val): Handle symview.
1070 (attr_checksum, attr_checksum_ordered): Likewise.
1071 (same_dw_val_p, size_of_die): Likewise.
1072 (value_format, output_die): Likewise.
1073 (add_high_low_attributes): Use add_AT_symview for entry_view.
1074 (dwarf2out_finish): Reset symview_upper_bound, clear
1077 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1080 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1081 Add strict argument and use it.
1082 (rs6000_split_multireg_move): Update for new strict argument.
1083 (mem_operand_gpr): Disallow all non-offsettable addresses.
1084 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1086 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1089 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1090 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1091 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1094 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1095 decl, use remap_type if we want to use the type.
1097 2018-03-09 Martin Sebor <msebor@redhat.com>
1099 PR tree-optimization/84526
1100 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1102 (builtin_access::generic_overlap): Be prepared to handle non-array
1105 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1107 PR rtl-optimization/84682
1108 * lra-constraints.c (process_address_1): Check is_address flag
1109 for address constraints.
1110 (process_alt_operands): Likewise.
1111 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1112 preprocess_constraints.
1113 * recog.h (preprocess_constraints): Add oploc parameter.
1115 * recog.c (preprocess_constraints): Test address_operand for
1116 CT_ADDRESS constraints.
1118 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1121 * lra-assigns.c (assign_by_spills): Return a flag of reload
1122 assignment failure. Do not process the reload assignment
1123 failures. Do not spill other reload pseudos if they has the same
1125 (lra_assign): Add a return arg. Set up from the result of
1126 assign_by_spills call.
1127 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1128 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1129 usage_insns if it is not NULL.
1130 (spill_hard_reg_in_range): New function.
1131 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1132 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1133 function prototypes.
1134 (lra_assign): Change prototype.
1135 * lra.c (lra): Add code to deal with fails by splitting hard reg
1138 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1141 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1142 Accept complain bool parameter. Only emit errors if it is true.
1143 (arm_parse_cpu_option_name): Likewise.
1144 (arm_target_thumb_only): Adjust callers of the above.
1145 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1146 prototype to take a default true bool parameter.
1147 (arm_parse_arch_option_name): Likewise.
1149 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1150 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1154 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1155 * configure: Regenerate.
1156 * configure.ac ("linker --version-script option"): New.
1157 ("linker soname option"): New.
1159 2018-03-09 Richard Biener <rguenther@suse.de>
1161 PR tree-optimization/84775
1162 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1163 immediate uses of predicate stmts and mark them modified.
1166 PR tree-optimization/84178
1167 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1169 (version_loop_for_if_conversion): Delay update_ssa call.
1170 (tree_if_conversion): Delay update_ssa until after predicate
1173 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1176 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1177 when the function accesses prior frames.
1179 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1182 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1183 gen_llsym, otherwise call maybe_gen_llsym.
1186 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1187 has ',' character inside of it.
1189 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1192 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1193 as clobbering CC_REGNUM.
1195 2018-03-08 Richard Biener <rguenther@suse.de>
1198 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1199 (follow_copies_to_constant): Do not follow SSA names registered
1202 2018-03-08 Richard Biener <rguenther@suse.de>
1204 PR tree-optimization/84178
1205 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1207 (version_loop_for_if_conversion): Delay update_ssa call.
1208 (tree_if_conversion): Delay update_ssa until after predicate
1211 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1213 PR tree-optimization/84178
1214 * tree-if-conv.c (release_bb_predicate): Remove the
1215 the assertion that the stmts have NULL use_ops.
1216 Discard the statements, asserting that they haven't
1217 yet been added to a BB.
1219 2018-03-08 Richard Biener <rguenther@suse.de>
1221 PR tree-optimization/84746
1222 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1223 (phi_translate): Pass in destination ANTIC_OUT set.
1224 (phi_translate_1): Likewise. For a simplified result lookup
1225 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1226 (phi_translate_set): Adjust.
1227 (do_pre_regular_insertion): Likewise.
1228 (do_pre_partial_partial_insertion): Likewise.
1230 2018-03-08 Martin Liska <mliska@suse.cz>
1232 PR gcov-profile/84735
1233 * doc/gcov.texi: Document usage of profile files.
1234 * gcov-io.h: Document changes in the format.
1236 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1240 * dwarf2out.c (struct dw_line_info_table): Update comments for
1242 (FORCE_RESET_NEXT_VIEW): New.
1243 (FORCE_RESETTING_VIEW_P): New.
1244 (RESETTING_VIEW_P): Check for -1 too.
1245 (ZERO_VIEW_P): Likewise.
1246 (new_line_info_table): Force-reset next view.
1247 (dwarf2out_begin_function): Likewise.
1248 (dwarf2out_source_line): Simplify zero_view_p initialization.
1249 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1250 view directly. Omit view when omitting .loc at line 0.
1252 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1254 PR tree-optimization/84740
1255 * tree-switch-conversion.c (process_switch): Call build_constructors
1256 only if info.phi_count is non-zero.
1258 PR tree-optimization/84739
1259 * tree-tailcall.c (find_tail_calls): Check call arguments against
1260 DECL_ARGUMENTS (current_function_decl) rather than
1261 DECL_ARGUMENTS (func) when checking for tail recursion.
1263 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1265 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1266 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1267 Volker Reichelt's entry and add entries for people that perform
1268 GCC fuzzy testing and report numerous bugs.
1270 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1273 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1274 readonly data in sdata, if that is disabled.
1275 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1276 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1277 -mreadonly-in-sdata option.
1279 2018-03-07 Martin Sebor <msebor@redhat.com>
1281 PR tree-optimization/84468
1282 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1283 basic block when looking for nul assignment.
1285 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1288 * except.h (output_function_exception_table): Adjust prototype.
1289 * except.c (output_function_exception_table): Remove FNNAME parameter
1290 and add SECTION parameter. Ouput one part of the table at a time.
1291 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1292 the first part of the exception table and emit unwind directives.
1293 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1294 (i386_pe_seh_cold_init): Likewise.
1295 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1296 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1297 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1298 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1299 * config/i386/winnt.c: Include except.h.
1300 (struct seh_frame_state): Add reg_offset, after_prologue and
1301 in_cold_section fields.
1302 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1303 (i386_pe_seh_cold_init): New function.
1304 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1305 to seh->in_cold_section.
1306 (seh_emit_push): Record the offset of the push.
1307 (seh_emit_save): Record the offet of the save.
1308 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1309 Test seh->after_prologue to disregard the epilogue.
1310 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1311 (i386_pe_end_cold_function): New function.
1313 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1316 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1317 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1320 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1322 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1323 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1326 * multiple_target.c: Include tree-inline.h and intl.h.
1327 (expand_target_clones): Diagnose and fail if node->definition and
1328 !tree_versionable_function_p (node->decl).
1330 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1332 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1334 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1335 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1336 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1338 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1341 * combine.c (try_combine): Use reg_or_subregno instead of handling
1342 just paradoxical SUBREGs and REGs.
1344 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1346 * config/arc/arc.c (arc_finalize_pic): Remove function.
1347 (arc_must_save_register): We use single base PIC register, remove
1348 checks to save/restore the PIC register.
1349 (arc_expand_prologue): Likewise.
1350 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1352 (arc_verify_short): Likewise.
1353 (arc_attr_type): Likewise.
1354 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1355 (walk_stores): Likewise.
1356 (arc_address_cost): Make it static.
1357 (arc_verify_short): Likewise.
1358 (branch_dest): Likewise.
1359 (arc_attr_type): Likewise.
1360 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1361 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1362 (arc_final_prescan_insn): Remove inserting the nops due to
1363 hardware hazards. It is done in reorg step.
1364 (insn_length_variant_t): Remove.
1365 (insn_length_parameters_t): Likewise.
1366 (arc_insn_length_parameters): Likewise.
1367 (arc_get_insn_variants): Likewise.
1368 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1370 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1373 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1376 PR tree-optimization/84687
1377 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1379 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1381 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1383 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1384 Rename to ppc_speculation_barrier.
1385 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1386 __builtin_ppc_speculation_barrier.
1388 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1391 * combine.c (combine_simplify_rtx): Don't try to simplify if
1392 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1395 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1397 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1398 to 32 bytes when compiling for POWER9.
1400 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1403 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1404 regparm >= 3 with no arg reg available also for calls with
1405 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1408 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1410 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1412 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1415 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1417 2018-03-05 Richard Biener <rguenther@suse.de>
1419 PR tree-optimization/84486
1420 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1421 When inserting a __builtin_assume_aligned call set the LHS
1422 SSA name alignment info accordingly.
1424 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1426 PR tree-optimization/84114
1427 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1428 Avoid reassociation of FLOAT_MODE addition.
1430 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1432 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1433 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1434 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1435 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1436 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1437 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1438 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1440 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1441 __builtin_ia32_wbinvd): New builtins.
1442 (SPECIAL_ARGS2): New.
1443 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1444 (SPECIAL_ARGS2): New.
1445 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1446 (ix86_valid_target_attribute_inner_p): Ditto.
1447 (ix86_init_mmx_sse_builtins): Add special_args2.
1448 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1449 TARGET_WBNOINVD_P): New.
1450 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1451 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1452 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1453 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1454 * config/i386/pconfigintrin.h: New file.
1455 * config/i386/wbnoinvdintrin.h: Ditto.
1456 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1457 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1459 2018-03-05 Richard Biener <rguenther@suse.de>
1461 PR tree-optimization/84670
1462 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1464 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1465 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1466 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1467 doesn't grow if all successors (recursively) were visited at least
1470 2018-03-05 Richard Biener <rguenther@suse.de>
1472 PR tree-optimization/84650
1473 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1474 if executed in the loop pipeline.
1476 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1478 * doc/configfiles.texi (Configuration Files): Move info about
1479 conditionalizing $target-protos.h to...
1480 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1481 differs from $target-protos.h.
1483 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1484 Chung-Ju Wu <jasonwucj@gmail.com>
1486 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1487 * config/nds32/nds32-multiple.md (setmemsi): Define.
1488 * config/nds32/nds32-memory-manipulation.c
1489 (nds32_gen_dup_4_byte_to_word_value): New.
1490 (emit_setmem_word_loop): New.
1491 (emit_setmem_byte_loop): New.
1492 (nds32_expand_setmem_loop): New.
1493 (nds32_expand_setmem_loop_v3m): New.
1494 (nds32_expand_setmem_unroll): New.
1495 (nds32_expand_setmem): New.
1497 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1498 Chung-Ju Wu <jasonwucj@gmail.com>
1500 * config/nds32/nds32-memory-manipulation.c
1501 (nds32_emit_load_store): New.
1502 (nds32_emit_post_inc_load_store): New.
1503 (nds32_emit_mem_move): New.
1504 (nds32_emit_mem_move_block): New.
1505 (nds32_expand_movmemsi_loop_unknown_size): New.
1506 (nds32_expand_movmemsi_loop_known_size): New.
1507 (nds32_expand_movmemsi_loop): New.
1508 (nds32_expand_movmemsi_unroll): New.
1509 (nds32_expand_movmemqi): Rename ...
1510 (nds32_expand_movmemsi): ... to this.
1511 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1512 (movmemsi): ... to this.
1513 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1514 (nds32_expand_movmemsi): ... to this.
1516 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1517 Monk Chiang <sh.chiang04@gmail.com>
1518 Chung-Ju Wu <jasonwucj@gmail.com>
1520 * config/nds32/nds32-protos.h
1521 (nds32_expand_load_multiple): New arguments.
1522 (nds32_expand_store_multiple): Ditto.
1523 (nds32_valid_multiple_load_store): Rename ...
1524 (nds32_valid_multiple_load_store_p): ... to this.
1525 * config/nds32/nds32-memory-manipulation.c
1526 (nds32_expand_load_multiple): Refine implementation.
1527 (nds32_expand_store_multiple): Ditto.
1528 * config/nds32/nds32-multiple.md
1529 (load_multiple): Update nds32_expand_load_multiple interface.
1530 (store_multiple): Update nds32_expand_store_multiple interface.
1531 * config/nds32/nds32-predicates.c
1532 (nds32_valid_multiple_load_store): Rename ...
1533 (nds32_valid_multiple_load_store_p): ... to this and refine
1535 * config/nds32/predicates.md
1536 (nds32_load_multiple_and_update_address_operation): New predicate.
1537 (nds32_store_multiple_and_update_address_operation): New predicate.
1539 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1540 Chung-Ju Wu <jasonwucj@gmail.com>
1542 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1543 (combo): New attribute.
1544 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1546 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1548 * config/nds32/nds32.opt: Change -mcmodel= default value.
1550 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1551 Monk Chiang <sh.chiang04@gmail.com>
1552 Chung-Ju Wu <jasonwucj@gmail.com>
1554 * config/nds32/constants.md (unspec_element): New enum.
1555 * config/nds32/constraints.md (Umw): New constraint.
1556 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1557 * config/nds32/nds32-intrinsic.md: Likewise.
1558 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1559 (nds32_valid_smw_lwm_base_p): New.
1560 (nds32_output_smw_single_word): New.
1561 (nds32_output_lmw_single_word): New.
1562 (nds32_expand_unaligned_load): New.
1563 (nds32_expand_unaligned_store): New.
1564 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1565 (nds32_output_smw_single_word): Declare.
1566 (nds32_output_lmw_single_word): Declare.
1567 (nds32_expand_unaligned_load): Declare.
1568 (nds32_expand_unaligned_store): Declare.
1569 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1570 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1571 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1572 NDS32_BUILTIN_UASTORE_DW.
1573 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1576 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1577 Kito Cheng <kito.cheng@gmail.com>
1578 Chung-Ju Wu <jasonwucj@gmail.com>
1580 * config/nds32/nds32-intrinsic.c
1581 (nds32_expand_builtin_null_ftype_reg): Delete.
1582 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1583 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1584 (nds32_read_argument): New.
1585 (nds32_legitimize_target): Ditto.
1586 (nds32_legitimize_argument): Ditto.
1587 (nds32_check_constant_argument): Ditto.
1588 (nds32_expand_unop_builtin): Ditto.
1589 (nds32_expand_unopimm_builtin): Ditto.
1590 (nds32_expand_binop_builtin): Ditto.
1591 (nds32_builtin_decl_impl): Ditto.
1592 (builtin_description): Ditto.
1593 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1594 (nds32_init_builtins_impl): Ditto.
1595 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1596 (nds32_builtin_decl): New.
1597 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1598 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1600 2018-03-02 Jeff Law <law@redhat.com>
1602 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1603 (redundant_insn, fill_simple_delay_slots): Likewise.
1604 (fill_slots_from_thread): Likewise.
1605 * resource.c (mark_referenced_resources): Likewise.
1606 (mark_set_resources, find_dead_or_set_registers): Likewise.
1608 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1610 * substring-locations.h (format_warning_va): Formatting fix for
1612 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1614 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1615 * substring-locations.c: Include intl.h.
1616 (format_warning_va): Turned into small wrapper around
1617 format_warning_n_va, renamed to ...
1618 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1619 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1621 (format_warning_at_substring_n): New function.
1622 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1623 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1624 format_warning_at_substring with just a shorter name instead of
1625 const function pointer.
1626 (fmtwarn_n): New function.
1627 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1628 appropriate, get rid of all the fmtstr temporaries, move conditionals
1629 with G_() wrapped string literals directly into fmtwarn arguments,
1630 cast dir.len to (int), formatting fixes.
1632 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1634 * doc/invoke.texi: Remove "Cilk Plus" references.
1636 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1637 Richard Biener <rguenther@suse.de>
1640 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1641 for error or warning attributes if CALL_FROM_THUNK_P is set.
1644 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1647 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1648 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1651 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1652 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1654 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1655 instead of -1U in last predictors element's probability member.
1657 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1660 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1661 arguments if they are comparable.
1663 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1665 PR tree-optimization/84634
1666 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1667 masks and masked_loop_p with a single loop_masks, making sure it's
1668 null for bb vectorization.
1670 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1672 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1673 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1674 loop->force_vectorize to check whether there is no alias.
1676 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1679 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1681 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1683 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1684 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1685 instead of a loop around prev_real_insn.
1686 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1690 * config/i386/i386.c (ix86_print_operand): Use conditional
1691 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1694 2018-03-02 Richard Biener <rguenther@suse.de>
1696 PR tree-optimization/84427
1697 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1698 (bitmap_set_subtract_values): Rewrite to handle multiple
1701 (prune_clobbered_mems): Likewise.
1702 (phi_translate): Take edge instead of pred/phiblock.
1703 (phi_translate_1): Likewise.
1704 (phi_translate_set): Likewise. Insert all translated
1705 exprs for a value into the set, keeping possibly multiple
1706 expressions per value.
1707 (compute_antic_aux): Adjust for phi_translate changes.
1708 When intersecting union the expressions and prune those
1709 not in the final value set, keeping possibly multiple
1710 expressions per value. Do not use value-insertion
1711 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1712 all expressions. Add verification that the value-sets
1713 only shrink during iteration.
1714 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1715 (do_pre_regular_insertion): Likewise.
1716 (do_pre_partial_partial_insertion): Likewise.
1718 2018-03-02 Richard Biener <rguenther@suse.de>
1721 * config/darwin.c (saved_debug_info_level): New static global.
1722 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1723 (darwin_asm_lto_end): Restore debug info generation settings.
1725 2018-03-01 Martin Liska <mliska@suse.cz>
1728 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1731 2018-03-01 Richard Biener <rguenther@suse.de>
1734 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1735 type annotation with LTO when debug was disabled at compile-time.
1737 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1739 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1741 (mips_final_postscan_insn): Likewise.
1743 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1745 PR rtl-optimization/84528
1746 * alias.c (init_alias_target): Add commentary.
1747 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1748 a unique base value if the frame pointer is not eliminated
1749 to the stack pointer.
1751 2018-03-01 Tom de Vries <tom@codesourcery.com>
1753 PR rtl-optimization/83327
1754 * lra-int.h (hard_regs_spilled_into): Declare.
1755 * lra.c (hard_regs_spilled_into): Define.
1756 (init_reg_info): Init hard_regs_spilled_into.
1757 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1758 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1759 (process_bb_lives): Handle hard_regs_spilled_into.
1760 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1761 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1763 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1765 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1766 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1767 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1768 * config/rs6000/aix72.h: New file.
1770 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1772 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1773 instead of warning_at with conditional singular and plural messages
1777 * stor-layout.c (update_alignment_for_field): For
1778 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1779 && !DECL_PACKED (field), do the alignment update, just use
1780 only desired_align instead of MAX (type_align, desired_align)
1782 (place_field): Don't do known_align < desired_align handling
1783 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1784 is non-NULL, instead do it after rli->prev_field handling and
1785 only if not within a bitfield word. For DECL_PACKED (field)
1786 use type_align of BITS_PER_UNIT.
1788 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1790 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1791 superfluous parentheses and trailing spaces.
1793 2018-02-28 Richard Biener <rguenther@suse.de>
1795 PR tree-optimization/84584
1796 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1797 SCoPs with fake exit edge.
1799 2018-02-28 Martin Liska <mliska@suse.cz>
1802 * timevar.c (timer::print): Fix format to properly print 100%
1805 2018-02-28 Richard Biener <rguenther@suse.de>
1808 * genmatch.c (capture_info::walk_match): Do not mark
1809 captured expressions without operands as expr_p given
1810 they act more like predicates and should be subject to
1811 "lost tail" side-effect preserving.
1813 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1815 PR rtl-optimization/81611
1816 * auto-inc-dec.c (attempt_change): Move dead note from
1817 mem_insn if it's the next use of regno
1818 (find_address): Take address use of reg holding
1819 non-incremented value. Add parm to limit search to the named
1821 (merge_in_block): Attempt to use a mem insn that is the next
1822 use of the original regno.
1824 2018-02-27 Martin Sebor <msebor@redhat.com>
1827 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1828 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1830 2018-02-27 Martin Sebor <msebor@redhat.com>
1832 PR translation/84207
1833 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1834 n argument to unsigned HOST_WIDE_INT.
1835 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1836 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1837 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1838 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1840 2018-02-27 Richard Biener <rguenther@suse.de>
1842 PR tree-optimization/84512
1843 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1844 Do not use the estimate returned from record_stmt_cost for
1845 the scalar iteration cost but sum properly using add_stmt_cost.
1847 2018-02-27 Richard Biener <rguenther@suse.de>
1849 PR tree-optimization/84466
1850 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1851 Adjust last change to less strictly validate use operands.
1853 2018-02-27 Martin Liska <mliska@suse.cz>
1855 PR gcov-profile/84548
1856 * gcov.c (process_file): Allow partial overlap and consider it
1857 also as group functions.
1858 (output_lines): Properly calculate range of lines for a group.
1860 2018-02-27 Martin Liska <mliska@suse.cz>
1862 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1863 'ggc' suffixes. Change first column width.
1864 (timer::print): Fix formatting of the column.
1866 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1868 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1869 preserve inline entry blocks for the sake of debug inline
1870 entry point markers alone.
1871 (remove_unused_locals): Suggest in comments a better place to
1872 force the preservation of inline entry blocks that are
1873 otherwise unused, but do not preserve them.
1875 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1877 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1879 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1882 * config/i386/constraints.md (Bs): Replace
1883 ix86_indirect_branch_register with
1884 TARGET_INDIRECT_BRANCH_REGISTER.
1886 * config/i386/i386.md (indirect_jump): Likewise.
1887 (tablejump): Likewise.
1888 (*sibcall_memory): Likewise.
1889 (*sibcall_value_memory): Likewise.
1890 Peepholes of indirect call and jump via memory: Likewise.
1891 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1892 (*sibcall_value_GOT_32): Likewise.
1893 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1894 (GOT_memory_operand): Likewise.
1895 (call_insn_operand): Likewise.
1896 (sibcall_insn_operand): Likewise.
1897 (GOT32_symbol_operand): Likewise.
1898 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1900 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1902 PR rtl-optimization/83496
1903 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1904 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1905 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1906 redundant insn, if any.
1907 (relax_delay_slots): Likewise.
1908 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1910 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1912 PR tree-optimization/83965
1913 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1914 that grouped statements are part of a reduction chain. Return
1915 true if the statement is not marked as a reduction itself but
1917 (vect_recog_dot_prod_pattern): Don't check whether the statement
1918 is part of a group here.
1919 (vect_recog_sad_pattern): Likewise.
1920 (vect_recog_widen_sum_pattern): Likewise.
1922 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1925 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1927 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1930 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1932 (ix86_output_indirect_function_return): New prototype.
1933 (ix86_split_simple_return_pop_internal): Likewise.
1934 * config/i386/i386.c (indirect_return_via_cx): New.
1935 (indirect_return_via_cx_bnd): Likewise.
1936 (indirect_thunk_name): Handle return va CX_REG.
1937 (output_indirect_thunk_function): Create alias for
1938 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1939 (ix86_output_indirect_jmp): Remove the bool argument.
1940 (ix86_output_indirect_function_return): New function.
1941 (ix86_split_simple_return_pop_internal): Likewise.
1942 * config/i386/i386.md (*indirect_jump): Don't pass false
1943 to ix86_output_indirect_jmp.
1944 (*tablejump_1): Likewise.
1945 (simple_return_pop_internal): Change it to define_insn_and_split.
1946 Call ix86_split_simple_return_pop_internal to split it for
1948 (simple_return_indirect_internal): Call
1949 ix86_output_indirect_function_return instead of
1950 ix86_output_indirect_jmp.
1952 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1955 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1956 memset and value initialization afterwards.
1958 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1960 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1962 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1965 * common/config/aarch64/aarch64-common.c
1966 (aarch_option_optimization_table[]): Switch
1967 off fomit-frame-pointer
1969 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1970 Chung-Ju Wu <jasonwucj@gmail.com>
1972 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1974 (store_multiple): Ditto.
1976 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1978 * config.gcc: Add --with-cpu support for nds32 target.
1979 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1980 * config/nds32/nds32.opt: Add -mcpu= option.
1982 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1984 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1985 isel=yes): Warn for these deprecated options.
1987 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1989 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1990 ISA_2_5_MASKS_EMBEDDED.
1992 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1994 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1995 p->max as pointers rather than using iterative_hash_expr.
1997 2018-02-23 Carl Love <cel@us.ibm.com>
1999 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2000 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2002 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2003 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2004 P8V_BUILTIN_VEC_VUNSIGNED2.
2006 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2009 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2010 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2012 * lra-constraints.c (curr_insn_transform): Set up
2013 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2015 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2016 * lra-eliminations.c (spill_pseudos): Ditto.
2017 (process_insn_for_elimination): Ditto.
2018 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2019 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2020 LRA_NON_CLOBBERED_ALT.
2022 2018-02-22 Martin Sebor <msebor@redhat.com>
2024 PR tree-optimization/84480
2025 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2026 to maybe_diag_stxncpy_trunc. Call it.
2027 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2028 from gimple_fold_builtin_strcpy. Print inlining stack.
2029 (handle_builtin_stxncpy): Print inlining stack.
2030 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2032 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2035 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2036 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2037 and -fcheck-pointer-bounds are used together.
2038 (indirect_thunk_prefix): New enum.
2039 (indirect_thunk_need_prefix): New function.
2040 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2041 "_nt" instead of "_bnd" for NOTRACK prefix.
2042 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2043 (output_indirect_thunk_function): Likewise.
2045 (ix86_code_end): Update output_indirect_thunk_function calls.
2046 (ix86_output_indirect_branch_via_reg): Replace
2047 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2048 (ix86_output_indirect_branch_via_push): Likewise.
2049 (ix86_output_function_return): Likewise.
2050 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2051 incompatible with -fcf-protection=branch and
2052 -fcheck-pointer-bounds.
2054 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2057 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2058 Change gcc_assert call to output_operand_lossage.
2060 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2062 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2064 2018-02-22 DJ Delorie <dj@redhat.com>
2065 Sebastian Perta <sebastian.perta@renesas.com>
2066 Oleg Endo <olegendo@gcc.gnu.org>
2068 * config/rx/rx.c (rx_rtx_costs): New function.
2069 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2071 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2073 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2075 2018-02-22 Martin Liska <mliska@suse.cz>
2078 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2079 Add "native" as a possible value.
2081 2018-02-22 Martin Liska <mliska@suse.cz>
2084 * config/i386/i386.c (ix86_option_override_internal):
2085 Add "native" as a possible value for -march and -mtune.
2087 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2090 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2091 to all type variants.
2093 PR tree-optimization/84503
2094 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2095 width as info->bitpos + info->bitsize - start.
2096 (merged_store_group::merge_overlapping): Simplify width computation.
2097 (check_no_overlap): New function.
2098 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2099 start + width and last_order of the group, fail if check_no_overlap
2101 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2102 to group if check_no_overlap fails.
2104 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2106 * config/rs6000/altivec.md: Delete contraint arguments to
2107 define_expand, define_split, and define_peephole2, and in
2108 define_insn_and_split if always unused.
2109 * config/rs6000/darwin.md: Ditto.
2110 * config/rs6000/dfp.md: Ditto.
2111 * config/rs6000/rs6000.md: Ditto.
2112 * config/rs6000/sync.md: Ditto.
2113 * config/rs6000/vector.md: Ditto.
2114 * config/rs6000/vsx.md: Ditto.
2116 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2118 * config/rs6000/altivec.md: Write output control strings as braced
2119 blocks instead of double-quoted strings.
2120 * config/rs6000/darwin.md: Ditto.
2121 * config/rs6000/rs6000.md: Ditto.
2122 * config/rs6000/vector.md: Ditto.
2123 * config/rs6000/vsx.md: Ditto.
2125 2018-02-21 Jason Merrill <jason@redhat.com>
2127 PR c++/84314 - ICE with templates and fastcall attribute.
2128 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2130 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2132 * ipa-cp.c (determine_versionability): Fix comment typos.
2134 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2137 * ipa-cp.c (determine_versionability): Do not version functions caling
2140 2018-02-21 Martin Liska <mliska@suse.cz>
2143 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2144 Add "native" as a possible value.
2145 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2146 the macro when native cpu detection is available.
2148 2018-02-21 Martin Liska <mliska@suse.cz>
2151 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2152 Add "native" as a possible value.
2153 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2154 when native cpu detection is available.
2156 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2157 Martin Sebor <msebor@redhat.com>
2159 PR tree-optimization/84478
2160 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2162 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2163 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2164 support which is conservatively correct, for 2 only stay conservative
2165 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2166 argument to the 2 argument get_range_strlen, adjust 6 arg
2167 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2169 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2170 (gimple_fold_builtin_strlen): Pass true as last argument to
2173 2018-02-20 Martin Sebor <msebor@redhat.com>
2176 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2177 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2178 (builtin_memref::builtin_memref): Factor out parts into
2179 set_base_and_offset and call it.
2181 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2184 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2185 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2186 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2187 search at the associated MODE_INT.
2189 2018-02-20 Jeff Law <law@redhat.com>
2192 PR tree-optimization/81592
2194 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2195 for range data rather than using global data.
2196 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2197 range data rather than using global data.
2198 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2199 pass it to children as needed.
2200 (struct directive::fmtresult): Similarly.
2201 (struct directive::set_width): Similarly.
2202 (struct directive::set_precision): Similarly.
2203 (format_integer, format_directive, parse_directive): Similarly.
2204 (format_none): Accept unnamed vr_values parameter.
2205 (format_percent, format_floating, format_character): Similarly.
2206 (format_string, format_plain): Similarly.
2207 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2208 the EVRP range analyzer for range data rather than using global data.
2209 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2210 gimple-ssa-evrp-analyze.h
2211 (class sprintf_dom_walker): Add after_dom_children member function.
2212 Add evrp_range_analyzer member.
2213 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2214 range analyzer as needed.
2215 (sprintf_dom_walker::after_dom_children): New member function.
2216 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2218 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2219 (evrp_range_analyzer::pop_to_marker): Likewise.
2221 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2223 PR tree-optimization/84419
2224 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2225 with the required type if its current type is compatible but
2228 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2231 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2232 after vectorization.
2234 2018-02-20 Martin Liska <mliska@suse.cz>
2237 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2238 possible values if we don't have a hint.
2240 2018-02-20 Martin Liska <mliska@suse.cz>
2244 * final.c (shorten_branches): Build align_tab array with one
2246 * opts.c (finish_options): Add alignment option limit check.
2247 (MAX_CODE_ALIGN): Likewise.
2248 (MAX_CODE_ALIGN_VALUE): Likewise.
2249 * doc/invoke.texi: Document maximum allowed option value for
2250 all -falign-* options.
2252 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2255 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2256 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2257 * var-tracking.c (emit_note_insn_var_location): Remove all references
2258 to NOTE_INSN_CALL_ARG_LOCATION.
2259 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2260 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2261 Use copy_rtx_if_shared.
2262 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2263 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2264 (dwarf2out_var_location): Remove handling of
2265 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2267 * final.c (final_scan_insn): Remove all references to
2268 NOTE_INSN_CALL_ARG_LOCATION.
2269 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2270 before dumping final insns.
2271 * except.c (emit_note_eh_region_end): Remove all references to
2272 NOTE_INSN_CALL_ARG_LOCATION.
2273 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2274 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2275 * config/arc/arc.c (hwloop_optimize): Likewise.
2276 * config/arm/arm.c (create_fix_barrier): Likewise.
2277 * config/s390/s390.c (s390_chunkify_start): Likewise.
2278 * config/sh/sh.c (find_barrier): Likewise.
2279 * config/i386/i386.c (rest_of_insert_endbranch,
2280 ix86_seh_fixup_eh_fallthru): Likewise.
2281 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2282 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2283 * config/frv/frv.c (frv_function_prologue): Likewise.
2284 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2286 (note_outside_basic_block_p): Remove all references to
2287 NOTE_INSN_CALL_ARG_LOCATION.
2288 * gengtype.c (adjust_field_rtx_def): Likewise.
2289 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2291 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2292 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2295 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2298 PR tree-optimization/84452
2299 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2300 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2303 2018-02-19 Martin Liska <mliska@suse.cz>
2306 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2308 2018-02-19 Martin Liska <mliska@suse.cz>
2309 Richard Sandiford <richard.sandiford@linaro.org>
2311 PR tree-optimization/82491
2312 * gimple-fold.c (get_base_constructor): Make earlier bail out
2315 2018-02-19 Carl Love <cel@us.ibm.com>
2317 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2318 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2320 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2321 P8V_BUILTIN_VEC_NEG.
2323 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2325 * config/rl78/rl78.md (movdf): New define expand.
2327 2018-02-19 Martin Liska <mliska@suse.cz>
2330 * doc/invoke.texi: Fix typo.
2331 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2333 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2335 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2336 handle rs6000_single_float and rs6000_double_float specially for
2339 2018-02-16 Jeff Law <law@redhat.com>
2341 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2344 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2347 * ipa-inline.c (inline_small_functions): Fix a typo.
2349 2018-02-16 Nathan Sidwell <nathan@acm.org>
2351 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2353 2018-02-16 Carl Love <cel@us.ibm.com>
2355 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2356 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2357 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2358 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2359 expansion to P8V_BUILTIN_VEC_FLOAT2.
2361 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2363 PR rtl-optimization/70023
2364 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2365 src_regno into account.
2367 2018-02-16 Carl Love <cel@us.ibm.com>
2369 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2370 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2371 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2372 * config/rs6000/rs6000.c: Remove case statements for
2373 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2374 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2375 and P9V_BUILTIN_VEC_VINSERT4B.
2376 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2377 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2378 * config/rs6000/vsx.md:
2379 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2382 2018-02-16 Carl Love <cel@us.ibm.com>
2384 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2386 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2388 * config/rs6000/rs6000-c.c: Add the definitions for
2389 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2390 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2391 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2392 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2393 definition for insert4b and define insn *insert3b_internal.
2394 * doc/extend.texi: Add documentation for vec_extract4b.
2396 2018-02-16 Nathan Sidwell <nathan@acm.org>
2398 * doc/extend.texi (Backwards Compatibility): Mention friend
2399 injection. Note for-scope is deprecated.
2400 * doc/invoke.texi (-ffriend-injection): Deprecate.
2402 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2404 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2405 that moved to I2, also allow destinations that are a paradoxical
2406 subreg (instead of a normal reg).
2408 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2411 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2414 2018-02-16 Richard Biener <rguenther@suse.de>
2416 PR tree-optimization/84037
2417 PR tree-optimization/84016
2419 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2420 Adjust vec_construct for the fact we need additional higher latency
2421 128bit inserts for AVX256 and AVX512 vector builds.
2422 (ix86_add_stmt_cost): Scale vector construction cost for
2425 2018-02-16 Richard Biener <rguenther@suse.de>
2427 PR tree-optimization/84417
2428 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2429 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2430 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2432 2018-02-16 Martin Liska <mliska@suse.cz>
2435 * internal-fn.def (ASAN_CHECK): Set proper flags.
2436 (ASAN_MARK): Likewise.
2438 2018-02-16 Julia Koval <julia.koval@intel.com>
2440 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2441 from PTA_CANNONLAKE.
2443 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2446 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2447 Use ++iter rather than iter++ for std::list iterators.
2448 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2449 defer deleting them until all nodes in the forest are processed. Do
2450 free even leaf nodes. Change to_process into auto_vec.
2453 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2454 * vec.h (vec_default_construct): Use memset instead of placement new
2455 if BROKEN_VALUE_INITIALIZATION is defined.
2456 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2457 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2460 PR rtl-optimization/83723
2461 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2462 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2463 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2465 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2467 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2469 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2471 PR rtl-optimization/81443
2472 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2473 from inner REGs to paradoxical SUBREGs.
2475 2018-02-16 Richard Biener <rguenther@suse.de>
2477 PR tree-optimization/84399
2478 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2479 For operands we can analyze at their definition make sure we can
2480 analyze them at each use as well.
2482 2018-02-16 Richard Biener <rguenther@suse.de>
2484 PR tree-optimization/84190
2485 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2486 volatile accesses if the decl isn't volatile.
2488 2018-02-15 Jason Merrill <jason@redhat.com>
2490 PR c++/84314 - ICE with templates and fastcall attribute.
2491 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2492 TYPE_CANONICAL on an existing type.
2494 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2496 PR tree-optimization/84383
2497 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2498 dstoff nor call operand_equal_p if dstbase is NULL.
2500 PR tree-optimization/84334
2501 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2502 also a CONSTANT_CLASS_P, punt.
2504 2018-02-14 Jim Wilson <jimw@sifive.com>
2506 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2507 first SMALL_OPERAND check. New local min_second_step. Move assert
2508 to where locals are set. Add TARGET_RVC support.
2509 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2511 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2513 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2515 2018-02-14 Martin Sebor <msebor@redhat.com>
2517 PR tree-optimization/83698
2518 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2519 arrays constrain the offset range to their bounds.
2520 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2521 (builtin_access::overlap): Avoid setting the size of overlap if it's
2523 (maybe_diag_overlap): Also consider arrays when deciding what values
2524 of offsets to include in diagnostics.
2526 2018-02-14 Martin Sebor <msebor@redhat.com>
2529 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2530 that correspond to the kind of a declaration.
2532 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2535 * config/pa/pa.md: Load address of PIC label using the linkage table
2536 if the label is nonlocal.
2538 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2540 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2541 warning message if user requests -maltivec=be.
2542 * doc/invoke.texi: Document deprecation of -maltivec=be.
2544 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2547 * config/rs6000/rs6000-c.c: Update definitions for
2548 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2549 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2551 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2554 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2555 add _get_ssp intrinsics. Remove argument from
2556 __builtin_ia32_rdssp[d|q].
2557 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2558 * config/i386/i386-builtin.def: Remove argument from
2559 __builtin_ia32_rdssp[d|q].
2560 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2561 ix86_expand_special_args_builtin for _rdssp[d|q].
2562 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2563 Clear register before usage.
2564 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2565 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2567 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2569 PR tree-optimization/84357
2570 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2571 operand 1 of an ARRAY_REF too.
2573 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2576 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2577 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2579 (set_of_reg): New struct.
2580 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2581 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2582 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2584 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2585 Split into bitclr, bitset, bitinvert patterns if appropriate.
2586 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2587 use rx_fuse_in_memory_bitop.
2588 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2589 to named insn, correct maximum insn length.
2591 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2594 * machmode.def: Define a complex mode for PARTIAL_INT.
2595 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2597 * doc/rtl.texi: Document CSPImode.
2598 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2600 (msp430_hard_regno_nregs_with_padding): Likewise.
2602 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2605 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2607 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2609 PR rtl-optimization/84169
2610 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2611 we generated a parallel as new i3 and we split that to new i2 and i3
2612 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2613 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2614 those to i2, not i1. Partially rewrite this scan code.
2616 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2619 * stor-layout.c (place_field): For variable length fields, adjust
2620 offset_align afterwards not just based on the field's alignment,
2621 but also on the size.
2624 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2625 of exps and logs in the use_exp2 case.
2627 2018-02-13 Jeff Law <law@redhat.com>
2629 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2632 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2635 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2639 * common.opt (gas-loc-support, gas-locview-support): New.
2640 (ginline-points, ginternal-reset-location-views): New.
2641 * doc/invoke.texi: Document them. Use @itemx where intended.
2642 (gvariable-location-views): Adjust.
2643 * target.def (reset_location_view): New.
2644 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2645 (TARGET_RESET_LOCATION_VIEW): New.
2646 * doc/tm.texi: Rebuilt.
2647 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2648 (dwarf2out_default_as_locview_support): New.
2649 (output_asm_line_debug_info): Use option variables.
2650 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2651 (output_loc_list): Likewise.
2652 (add_high_low_attributes): Check option variables.
2653 Don't output entry view attribute in strict mode.
2654 (gen_inlined_subroutine_die): Check option variables.
2655 (dwarf2out_inline_entry): Likewise.
2656 (init_sections_and_labels): Likewise.
2657 (dwarf2out_early_finish): Likewise.
2658 (maybe_reset_location_view): New, from...
2659 (dwarf2out_var_location): ... here. Call it.
2660 * debug.h (dwarf2out_default_as_loc_support): Declare.
2661 (dwarf2out_default_as_locview_support): Declare.
2662 * hooks.c (hook_int_rtx_insn_0): New.
2663 * hooks.h (hook_int_rtx_insn_0): Declare.
2664 * toplev.c (process_options): Take -gas-loc-support and
2665 -gas-locview-support from dwarf2out. Enable
2666 -gvariable-location-views by default only with locview
2667 assembler support. Enable -ginternal-reset-location-views by
2668 default only if the target defines the corresponding hook.
2669 Enable -ginline-points by default if location views are
2670 enabled; force it disabled if statement frontiers are
2672 * tree-inline.c (expand_call_inline): Check option variables.
2673 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2675 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2677 PR tree-optimization/84321
2678 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2679 handling. Also check whether the anti-range contains any values
2680 that satisfy the mask; switch to a VR_RANGE if not.
2682 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2685 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2687 2018-02-13 Martin Jambor <mjambor@suse.cz>
2690 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2691 of call statements, also set location of a load to a temporary.
2693 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2695 * config/rl78/rl78.c (add_vector_labels): New function.
2696 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2697 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2698 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2699 which checks that no arguments are passed.
2700 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2701 * doc/extend.texi: Documentation for the new attribute.
2703 2018-02-13 Andreas Schwab <schwab@suse.de>
2705 * config/riscv/linux.h (CPP_SPEC): Define.
2707 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2710 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2711 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2712 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2713 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2714 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2715 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2716 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2717 temporarily for AES and PCLMUL builtins.
2719 PR tree-optimization/84339
2720 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2721 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2725 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2726 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2727 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2729 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2731 * omp-simd-clone.h: New file.
2732 * omp-simd-clone.c: Include omp-simd-clone.h.
2733 (expand_simd_clones): No longer static.
2734 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2735 cgraph.h and omp-simd-clone.h.
2736 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2737 (vect_recog_widen_shift_pattern): Formatting fix.
2738 (vect_pattern_recog_1): Don't check optab for calls.
2741 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2742 operands[2] into a REG before using gen_lowpart on it.
2744 2018-02-12 Jeff Law <law@redhat.com>
2747 * config/sh/sh.c (find_barrier): Consider a sibling call
2750 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2751 successfully back substituting a reg.
2753 2018-02-12 Richard Biener <rguenther@suse.de>
2755 PR tree-optimization/84037
2756 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2757 parameter, move visited init to caller.
2758 (vect_slp_analyze_operations): Separate cost from validity
2759 check, initialize visited once for all instances.
2760 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2762 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2764 (vect_model_promotion_demotion_cost): Likewise.
2765 (vectorizable_bswap): Guard cost modeling with !slp_node
2766 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2768 (vectorizable_call): Likewise.
2769 (vectorizable_conversion): Likewise.
2770 (vectorizable_assignment): Likewise.
2771 (vectorizable_shift): Likewise.
2772 (vectorizable_operation): Likewise.
2773 (vectorizable_store): Likewise.
2774 (vectorizable_load): Likewise.
2775 (vectorizable_condition): Likewise.
2776 (vectorizable_comparison): Likewise.
2778 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2781 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2782 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2783 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2785 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2788 * params.def (inline-min-speedup): Increase from 8 to 15.
2789 (max-inline-insns-auto): Decrease from 40 to 30.
2790 * ipa-split.c (consider_split): Add some buffer for function to
2791 be considered inlining candidate.
2792 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2795 2018-02-12 Richard Biener <rguenther@suse.de>
2797 PR tree-optimization/84037
2798 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2799 matched stmts if we cannot swap the non-matched ones.
2801 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2803 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2804 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2805 _mm_maskz_scalef_round_ss): New intrinsics.
2806 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2807 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2808 __builtin_ia32_scalefss_round): Remove.
2809 (__builtin_ia32_scalefsd_mask_round,
2810 __builtin_ia32_scalefss_mask_round): New intrinsics.
2811 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2812 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2813 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2814 "<round_constraint>")): Changed to ...
2815 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2816 "<round_scalar_constraint>")): ... this.
2817 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2818 %0, %1, %2<round_op3>}"): Changed to ...
2819 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2820 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2821 %2<round_scalar_mask_op3>}"): ... this.
2822 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2824 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2826 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2827 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2828 (_mm_maskz_sqrt_round_ss): New intrinsics.
2829 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2830 (__builtin_ia32_sqrtsd_mask_round)
2831 (__builtin_ia32_sqrtss_mask_round): New builtins.
2832 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2833 (__builtin_ia32_sqrtss_round): Remove.
2834 (__builtin_ia32_sqrtsd_mask_round)
2835 (__builtin_ia32_sqrtss_mask_round): New builtins.
2836 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2837 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2838 ((match_operand:VF_128 1 "vector_operand"
2839 "xBm,<round_constraint>")): Changed to ...
2840 ((match_operand:VF_128 1 "vector_operand"
2841 "xBm,<round_scalar_constraint>")): ... this.
2842 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2843 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2844 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2845 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2846 %<iptr>1<round_scalar_mask_op3>}): ... this.
2847 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2848 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2850 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2853 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2854 Cast vec_cmpeq result to correct type.
2855 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2856 Cast vec_cmpgt result to correct type.
2858 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2860 * final.c (final_scan_insn_1): Renamed from...
2861 (final_scan_insn): ... this. New wrapper, to recover
2862 seen from the outermost call in recursive ones.
2863 * config/sparc/sparc.c (output_return): Drop seen from call.
2864 (output_sibcall): Likewise.
2865 * config/visium/visium.c (output_branch): Likewise.
2867 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2869 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2872 2018-02-10 Alan Modra <amodra@gmail.com>
2875 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2876 Specify LR as an input.
2878 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2881 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2882 remove_member_access_dummy_vars): New functions.
2883 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2884 lower_omp_1, execute_lower_omp): Use them.
2886 PR rtl-optimization/84308
2887 * shrink-wrap.c (spread_components): Release todo vector.
2889 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2891 PR rtl-optimization/57193
2892 * ira-color.c (struct allocno_color_data): Add member
2893 conflict_allocno_hard_prefs.
2894 (update_conflict_allocno_hard_prefs): New.
2895 (bucket_allocno_compare_func): Add a preference based on
2896 conflict_allocno_hard_prefs.
2897 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2898 (color_allocnos): Remove a dead code. Initiate
2899 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2901 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2904 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2905 constraint from =wa to wa. Avoid a subreg on the output operand,
2906 instead use a pseudo and subreg it in a move.
2907 (p9_xxbrd_<mode>): Changed to ...
2908 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2909 (p9_xxbrd_v2df): New expander.
2910 (p9_xxbrw_<mode>): Changed to ...
2911 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2912 (p9_xxbrw_v4sf): New expander.
2914 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2916 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2918 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2921 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2922 multiply in 32-bit mode.
2923 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2924 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2927 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2929 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2930 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2931 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2932 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2934 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2937 * dwarf2out.c (is_trivial_indirect_ref): New function.
2938 (dwarf2out_late_global_decl): Do not generate a location
2939 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2940 and that are not defined in the current unit.
2942 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2944 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2945 instead of a libcall for UNORDERED.
2947 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2950 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2951 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2953 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2956 * config/s390/s390.c (s390_set_current_function): Invoke
2957 s390_indirect_branch_settings also if fndecl didn't change.
2959 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2961 * config/rs6000/rs6000.md (blockage): Set length to zero.
2963 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2965 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2967 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2970 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2971 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2975 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2976 PARALLEL incoming that failed vt_get_decl_and_offset check.
2979 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2980 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2982 (get_variable_section): For decls in named .bss* sections pass true as
2983 second argument to bss_initializer_p.
2985 2018-02-09 Marek Polacek <polacek@redhat.com>
2986 Jakub Jelinek <jakub@redhat.com>
2989 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2990 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2991 Sync some changes from cxx_fold_indirect_ref.
2993 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2995 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2997 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2998 (BLOCK_INLINE_ENTRY_LABEL): New.
2999 (dwarf2out_var_location): Disregard inline entry markers.
3000 (inline_entry_data): New struct.
3001 (inline_entry_data_hasher): New hashtable type.
3002 (inline_entry_data_hasher::hash): New.
3003 (inline_entry_data_hasher::equal): New.
3004 (inline_entry_data_table): New variable.
3005 (add_high_low_attributes): Add DW_AT_entry_pc and
3006 DW_AT_GNU_entry_view attributes if a pending entry is found
3007 in inline_entry_data_table. Add old entry_pc attribute only
3008 if debug nonbinding markers are disabled.
3009 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3010 markers are enabled.
3011 (block_within_block_p, dwarf2out_inline_entry): New.
3012 (dwarf2out_finish): Check that no entries remained in
3013 inline_entry_data_table.
3014 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3015 (final_scan_insn, notice_source_line): Likewise.
3016 (rest_of_clean_state): Skip inline entry markers.
3017 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3019 * gimple.c (gimple_build_debug_inline_entry): New.
3020 * gimple.h (enum gimple_debug_subcode): Add
3021 GIMPLE_DEBUG_INLINE_ENTRY.
3022 (gimple_build_debug_inline_entry): Declare.
3023 (gimple_debug_inline_entry_p): New.
3024 (gimple_debug_nonbind_marker_p): Adjust.
3025 * insn-notes.def (INLINE_ENTRY): New.
3026 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3027 inline entry marker notes.
3028 (print_insn): Likewise.
3029 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3030 (INSN_DEBUG_MARKER_KIND): Likewise.
3031 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3032 * tree-inline.c (expand_call_inline): Build and insert
3033 debug_inline_entry stmt.
3034 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3035 inline entry blocks early, if nonbind markers are enabled.
3036 (dump_scope_block): Dump fragment info.
3037 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3038 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3039 (gimple_build_debug_inline_entry): New.
3040 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3041 Enable/disable inline entry points too.
3042 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3043 (DEBUG_INSN): Describe inline entry markers.
3045 * common.opt (gvariable-location-views): New.
3046 (gvariable-location-views=incompat5): New.
3047 * config.in: Rebuilt.
3048 * configure: Rebuilt.
3049 * configure.ac: Test assembler for view support.
3050 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3051 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3052 * dwarf2out.c (var_loc_view): New typedef.
3053 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3054 (dwarf2out_locviews_in_attribute): New.
3055 (dwarf2out_locviews_in_loclist): New.
3056 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3057 (enum dw_line_info_opcode): Add LI_adv_address.
3058 (struct dw_line_info_table): Add view.
3059 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3060 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3061 (zero_view_p): New variable.
3062 (ZERO_VIEW_P): New macro.
3063 (output_asm_line_debug_info): New.
3064 (struct var_loc_node): Add view.
3065 (add_AT_view_list, AT_loc_list): New.
3066 (add_var_loc_to_decl): Add view param. Test it against last.
3067 (new_loc_list): Add view params. Record them.
3068 (AT_loc_list_ptr): Handle loc and view lists.
3069 (view_list_to_loc_list_val_node): New.
3070 (print_dw_val): Handle dw_val_class_view_list.
3071 (size_of_die): Likewise.
3072 (value_format): Likewise.
3073 (loc_list_has_views): New.
3074 (gen_llsym): Set vl_symbol too.
3075 (maybe_gen_llsym, skip_loc_list_entry): New.
3076 (dwarf2out_maybe_output_loclist_view_pair): New.
3077 (output_loc_list): Output view list or entries too.
3078 (output_view_list_offset): New.
3079 (output_die): Handle dw_val_class_view_list.
3080 (output_dwarf_version): New.
3081 (output_compilation_unit_header): Use it.
3082 (output_skeleton_debug_sections): Likewise.
3083 (output_rnglists, output_line_info): Likewise.
3084 (output_pubnames, output_aranges): Update version comments.
3085 (output_one_line_info_table): Output view numbers in asm comments.
3086 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3087 Call maybe_gen_llsym.
3088 (loc_list_from_tree_1): Adjust.
3089 (add_AT_location_description): Create view list attribute if
3090 needed, check it's absent otherwise.
3091 (convert_cfa_to_fb_loc_list): Adjust.
3092 (maybe_emit_file): Call output_asm_line_debug_info for test.
3093 (dwarf2out_var_location): Reset views as needed. Precompute
3094 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3095 attribute. Set view.
3096 (new_line_info_table): Reset next view.
3097 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3098 (dwarf2out_source_line): Likewise. Output view resets and labels to
3099 the assembler, or select appropriate line info opcodes.
3100 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3101 (optimize_string_length): Catch it. Adjust.
3102 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3103 dw_val_class_view_list, and remove it if no longer needed.
3104 (hash_loc_list): Hash view numbers.
3105 (loc_list_hasher::equal): Compare them.
3106 (optimize_location_lists): Check whether a view list symbol is
3107 needed, and whether the locview attribute is present, and
3108 whether they match. Remove the locview attribute if no longer
3110 (index_location_lists): Call skip_loc_list_entry for test.
3111 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3112 Use output_dwarf_version.
3113 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3114 (struct dw_val_node): Add val_view_list.
3115 * final.c (SEEN_NEXT_VIEW): New.
3116 (set_next_view_needed): New.
3117 (clear_next_view_needed): New.
3118 (maybe_output_next_view): New.
3119 (final_start_function): Rename to...
3120 (final_start_function_1): ... this. Take pointer to FIRST,
3121 add SEEN parameter. Emit param bindings in the initial view.
3122 (final_start_function): Reintroduce SEEN-less interface.
3123 (final): Rename to...
3124 (final_1): ... this. Take SEEN parameter. Output final pending
3125 next view at the end.
3126 (final): Reintroduce seen-less interface.
3127 (final_scan_insn): Output pending next view before switching
3128 sections or ending a block. Mark the next view as needed when
3129 outputting variable locations. Notify debug backend of section
3130 changes, and of location view changes.
3131 (rest_of_handle_final): Adjust.
3132 * toplev.c (process_options): Autodetect value for debug variable
3133 location views option. Warn on incompat5 without -gdwarf-5.
3134 * doc/invoke.texi (gvariable-location-views): New.
3135 (gvariable-location-views=incompat5): New.
3136 (gno-variable-location-views): New.
3138 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3140 PR tree-optimization/84136
3141 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3142 that the result of find_edge is non-NULL.
3144 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3147 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3148 storing integer register in SImode. Fix cost of 256 and 512
3149 byte aligned SSE register store.
3151 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3153 * config/i386/i386.c (ix86_multiplication_cost): Fix
3154 multiplication cost for TARGET_AVX512DQ.
3156 2018-02-08 Marek Polacek <polacek@redhat.com>
3158 PR tree-optimization/84238
3159 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3162 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3164 PR tree-optimization/84265
3165 * tree-vect-stmts.c (vectorizable_store): Don't treat
3166 VMAT_CONTIGUOUS accesses as grouped.
3167 (vectorizable_load): Likewise.
3169 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3171 PR tree-optimization/81635
3172 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3173 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3174 (test_round_for_mask): New functions.
3175 (wide_int_cc_tests): Call test_round_for_mask.
3176 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3177 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3178 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3179 range returned by get_range_info.
3181 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3184 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3185 * symtab.c: Include builtins.h
3186 (symtab_node::output_to_lto_symbol_table_p): Move here
3187 from lto-streamer-out.c:output_symbol_p.
3188 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3189 (output_symbol_p): Move all logic to symtab.c
3190 (produce_symtab): Update.
3192 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3194 * config/s390/s390-opts.h (enum indirect_branch): Define.
3195 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3196 (s390_indirect_branch_via_thunk)
3197 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3198 (enum s390_indirect_branch_type): Define.
3199 * config/s390/s390.c (struct s390_frame_layout, struct
3200 machine_function): Remove.
3201 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3202 (indirect_branch_table_label_no, indirect_branch_table_name):
3204 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3205 (enum s390_indirect_branch_option): Define.
3206 (s390_return_addr_from_memory): New function.
3207 (s390_handle_string_attribute): New function.
3208 (s390_attribute_table): Add new attribute handler.
3209 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3210 (s390_indirect_branch_via_thunk): New function.
3211 (s390_indirect_branch_via_inline_thunk): New function.
3212 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3213 sibling call optimization for non z10 compiles.
3214 (s390_emit_call): Force indirect branch target to be a single
3215 register. Add r1 clobber for non-z10 compiles.
3216 (s390_emit_epilogue): Emit return jump via return_use expander.
3217 (s390_reorg): Handle JUMP_INSNs as execute targets.
3218 (s390_option_override_internal): Perform validity checks for the
3219 new command line options.
3220 (s390_indirect_branch_attrvalue): New function.
3221 (s390_indirect_branch_settings): New function.
3222 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3223 (s390_output_indirect_thunk_function): New function.
3224 (s390_code_end): Implement target hook.
3225 (s390_case_values_threshold): Implement target hook.
3226 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3228 * config/s390/s390.h (struct s390_frame_layout)
3229 (struct machine_function): Move here from s390.c.
3230 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3231 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3232 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3233 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3234 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3235 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3236 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3237 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3238 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3239 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3240 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3241 (mnemonic attribute): Add values which aren't recognized
3243 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3244 pattern for branch conversion. Fix mnemonic attribute.
3245 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3246 indirect branch via thunk if requested.
3247 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3248 ("*indirect_jump"): Disable for branch conversion using out of
3250 ("indirect_jump_via_thunk<mode>_z10")
3251 ("indirect_jump_via_thunk<mode>")
3252 ("indirect_jump_via_inlinethunk<mode>_z10")
3253 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3254 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3255 ("casesi_jump_via_inlinethunk<mode>_z10")
3256 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3257 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3258 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3259 ("*indirect2_jump"): Disable for branch conversion.
3260 ("casesi_jump"): Turn into expander and expand patterns for branch
3262 ("return_use"): New expander.
3263 ("*return"): Emit return via thunk and rename it to ...
3264 ("*return<mode>"): ... this one.
3265 * config/s390/s390.opt: Add new options and and enum for the
3268 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3270 * lra-constraints.c (match_reload): Unconditionally use
3271 gen_lowpart_SUBREG, rather than selecting between that
3272 and equivalent gen_rtx_SUBREG code.
3274 2018-02-08 Richard Biener <rguenther@suse.de>
3276 PR tree-optimization/84233
3277 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3278 changed flag instead of boguously re-using phi_inserted.
3280 2018-02-08 Martin Jambor <mjambor@suse.cz>
3282 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3283 static local variables.
3285 2018-02-08 Richard Biener <rguenther@suse.de>
3287 PR tree-optimization/84278
3288 * tree-vect-stmts.c (vectorizable_store): When looking for
3289 smaller vector types to perform grouped strided loads/stores
3290 make sure the mode is supported by the target.
3291 (vectorizable_load): Likewise.
3293 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3295 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3296 Increase LDP/STP opportunities by adding adjacent callee-saves.
3298 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3300 PR rtl-optimization/84068
3301 PR rtl-optimization/83459
3302 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3304 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3306 PR tree-optimization/84224
3307 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3308 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3311 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3314 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3315 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3316 count, remove one USE.
3318 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3320 * doc/install.texi (Configuration): Document the
3321 --with-long-double-format={ibm,ieee} PowerPC configuration
3325 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3326 Convert from define_expand to be define_insn_and_split. Rework
3327 float/double/_Float128 conversions to QI/HI/SImode to work with
3328 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3329 conversions to QI/HImode types did a store and then a load to
3330 truncate the value. For conversions to VSX registers, don't split
3331 the insn, instead emit the code directly. Use the code iterator
3332 any_fix to combine signed and unsigned conversions.
3333 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3334 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3335 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3336 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3337 (fix_<mode>di2_hw): Likewise.
3338 (fixuns_<mode>di2_hw): Likewise.
3339 (fix_<mode>si2_hw): Likewise.
3340 (fixuns_<mode>si2_hw): Likewise.
3341 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3342 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3343 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3344 fix<uns>_trunc<SFDF:mode>si2_p8.
3345 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3347 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3348 (fix<uns>_<mode>_mem): Likewise.
3349 (fctiw<u>z_<mode>_mem): Likewise.
3350 (fix<uns>_<mode>_mem): Likewise.
3351 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3352 the register allocator from doing a direct move to the GPRs to do
3353 a store, and instead use the ISA 3.0 store byte/half-word from
3354 vector register instruction. For IEEE 128-bit floating point,
3355 also optimize stores of 32-bit ints.
3356 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3358 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3360 * genextract.c (push_pathstr_operand): New function to support
3362 (walk_rtx): Call push_pathstr_operand.
3363 (print_path): Support [a-zA-Z].
3365 2018-02-07 Richard Biener <rguenther@suse.de>
3367 PR tree-optimization/84037
3368 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3369 (cse_and_gimplify_to_preheader): Declare.
3370 (vect_get_place_in_interleaving_chain): Likewise.
3371 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3373 (_loop_vec_info::~_loop_vec_info): Delete it.
3374 (cse_and_gimplify_to_preheader): New function.
3375 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3376 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3377 (vectorizable_load): Likewise. For grouped stores always base
3378 the IV on the first element.
3379 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3380 condition before gimplifying.
3382 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3384 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3385 *DIV_EXPR and *MOD_EXPR.
3387 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3390 * config/i386/i386.c (ix86_option_override_internal): Mask out
3391 the CF_SET bit when checking -fcf-protection.
3393 2018-02-07 Tom de Vries <tom@codesourcery.com>
3396 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3399 2018-02-07 Richard Biener <rguenther@suse.de>
3401 PR tree-optimization/84204
3402 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3405 PR tree-optimization/84205
3406 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3407 special-case isl_ast_op_zdiv_r.
3409 PR tree-optimization/84223
3410 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3411 Only add conditions from within the region.
3412 (gather_bbs::after_dom_children): Adjust.
3414 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3417 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3418 * config/avr/avr.md: Only post-reload split REG-REG moves if
3419 either register is GENERAL_REG_P.
3421 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3423 PR tree-optimization/84235
3424 * tree-ssa-scopedtables.c
3425 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3426 if the subtraction is performed in floating point type where NaNs are
3427 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3428 build 1. Formatting fix.
3430 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3433 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3434 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3435 and skip it regardless of bb boundaries. Use CALL_P macro,
3436 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3437 unnecessarily, formatting fix.
3439 2018-02-06 Michael Collison <michael.collison@arm.com>
3441 * config/arm/thumb2.md:
3442 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3443 (*thumb_mov_notscc): Ditto.
3445 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3448 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3449 unsigned_fix, not "s".
3451 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3453 * configure.ac (gcc_fn_eh_frame_ro): New function.
3454 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3455 correct .eh_frame permissions.
3456 * configure: Regenerate.
3458 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3460 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3463 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3465 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3466 Display warning message for -mno-speculate-indirect-jumps.
3468 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3470 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3472 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3474 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3476 PR tree-optimization/84225
3477 * tree-eh.c (find_trapping_overflow): Only call
3478 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3480 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3483 * config/i386/i386.c: Reimplement the check of possible options
3484 -mibt/-mshstk conbination. Change error messages.
3485 * doc/invoke.texi: Fix a typo: remove extra '='.
3487 2018-02-06 Marek Polacek <polacek@redhat.com>
3489 PR tree-optimization/84228
3490 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3492 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3495 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3496 emitted arch directives.
3497 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3498 __ARM_FEATURE_COPROC before changing architectures.
3500 2018-02-06 Richard Biener <rguenther@suse.de>
3502 * config/i386/i386.c (print_reg): Fix typo.
3503 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3505 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3507 * configure: Regenerate.
3509 2018-02-05 Martin Sebor <msebor@redhat.com>
3511 PR tree-optimization/83369
3512 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3515 2018-02-05 Martin Liska <mliska@suse.cz>
3517 * doc/invoke.texi: Cherry-pick upstream r323995.
3519 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3521 * ira.c (ira_init_register_move_cost): Adjust comment.
3523 2018-02-05 Martin Liska <mliska@suse.cz>
3525 PR gcov-profile/84137
3526 * doc/gcov.texi: Fix typo in documentation.
3528 2018-02-05 Martin Liska <mliska@suse.cz>
3530 PR gcov-profile/83879
3531 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3532 using dlopen functionality.
3534 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3536 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3537 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3538 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3539 _mm_maskz_range_round_ss): New intrinsics.
3540 (__builtin_ia32_rangesd128_round)
3541 (__builtin_ia32_rangess128_round): Remove.
3542 (__builtin_ia32_rangesd128_mask_round,
3543 __builtin_ia32_rangess128_mask_round): New builtins.
3544 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3545 __builtin_ia32_rangess128_round): Remove.
3546 (__builtin_ia32_rangesd128_mask_round,
3547 __builtin_ia32_rangess128_mask_round): New builtins.
3548 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3549 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3550 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3551 "<round_saeonly_constraint>")): Changed to ...
3552 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3553 "<round_saeonly_scalar_constraint>")): ... this.
3554 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3555 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3556 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3557 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3558 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3560 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3562 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3564 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3565 Remove all values except native, 8540 and 8548.
3567 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3569 * config/i386/i386.c (ix86_output_function_return): Pass
3570 INVALID_REGNUM, instead of -1, as invalid register number to
3571 indirect_thunk_name and output_indirect_thunk.
3573 2018-02-02 Julia Koval <julia.koval@intel.com>
3575 * config.gcc: Add -march=icelake.
3576 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3577 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3578 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3579 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3580 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3581 (processor_target_table): Add icelake.
3582 (ix86_option_override_internal): Handle new PTAs.
3583 (get_builtin_code_for_version): Handle icelake.
3584 (M_INTEL_COREI7_ICELAKE): New.
3585 (fold_builtin_cpu): Handle icelake.
3586 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3587 * doc/invoke.texi: Add -march=icelake.
3589 2018-02-02 Julia Koval <julia.koval@intel.com>
3591 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3592 to wide_int_bitmask.
3593 * wide-int-bitmask.h: New.
3595 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3598 * config/i386/i386.md: Replace Pmode with word_mode in
3599 builtin_setjmp_setup and builtin_longjmp to support x32.
3601 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3605 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3607 (rs6000_supported_cpu_names): New static variable.
3608 (linux_cpu_translation_table): Likewise.
3609 (elf_platform) <cpu>: Define new static variable and use it.
3610 Translate kernel AT_PLATFORM name to canonical name if needed.
3611 Error if platform name is unknown.
3613 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3616 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3618 2018-02-01 Jeff Law <law@redhat.com>
3621 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3622 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3623 the scratch if RELEASE_VIA_POP is false.
3624 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3625 If we have to save a temporary register, decrement SIZE appropriately.
3626 Pass new arguments to release_scratch_register_on_entry.
3627 (ix86_adjust_stack_and_probe): Likewise.
3628 (ix86_emit_probe_stack_range): Pass new arguments to
3629 release_scratch_register_on_entry.
3631 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3633 PR rtl-optimization/84157
3634 * combine.c (change_zero_ext): Use REG_P predicate in
3635 front of HARD_REGISTER_P predicate.
3637 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3639 * config/avr/avr.c (avr_option_override): Move disabling of
3640 -fdelete-null-pointer-checks to...
3641 * common/config/avr/avr-common.c (avr_option_optimization_table):
3644 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3646 PR tree-optimization/81635
3647 * tree-data-ref.c (split_constant_offset_1): For types that
3648 wrap on overflow, try to use range info to prove that wrapping
3651 2018-02-01 Renlin Li <renlin.li@arm.com>
3654 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3656 (aarch64_register_move_cost): Likewise.
3657 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3659 (REG_CLASS_NAMES): Likewise.
3660 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3661 TAILCALL_ADDR_REGS. Remove IP registers.
3662 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3664 2018-02-01 Richard Biener <rguenther@suse.de>
3666 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3667 for specifying RPO order and allow NULL for that.
3668 * domwalk.c (dom_walker::dom_walker): Likewise.
3669 (dom_walker::walk): Handle NULL RPO order.
3670 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3672 (rewrite_update_dom_walker): Likewise.
3673 (mark_def_dom_walker): Likewise.
3675 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3677 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3678 (aarch64_maybe_expand_sve_subreg_move): Declare.
3679 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3680 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3682 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3683 that are semantically a reverse operation.
3684 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3685 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3686 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3688 (aarch64_can_change_mode_class): For big-endian, forbid changes
3689 between two SVE modes if they have different element sizes.
3691 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3693 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3694 the TImode handling for big-endian targets.
3696 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3698 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3699 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3701 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3702 Remove BSWAP handing for big-endian targets and use the form of
3703 LD1RQ appropariate for the mode.
3705 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3707 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3708 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3711 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3714 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3715 check for operands that need to go through aarch64_sve_reload_be.
3717 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3719 PR tree-optimization/81661
3720 PR tree-optimization/84117
3721 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3722 * tree-eh.c: Include gimplify.h.
3723 (find_trapping_overflow, replace_trapping_overflow,
3724 rewrite_to_non_trapping_overflow): New functions.
3725 * tree-vect-loop.c: Include tree-eh.h.
3726 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3727 * tree-data-ref.c: Include tree-eh.h.
3728 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3730 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3732 PR rtl-optimization/84123
3733 * combine.c (change_zero_ext): Check if hard register satisfies
3734 can_change_dest_mode before calling gen_lowpart_SUBREG.
3736 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3739 * ira.c (ira_init_register_move_cost): Remove assert.
3741 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3743 PR rtl-optimization/84071
3744 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3745 * doc/tm.texi: Regenerate.
3747 2018-01-31 Richard Biener <rguenther@suse.de>
3749 PR tree-optimization/84132
3750 * tree-data-ref.c (analyze_miv_subscript): Properly
3751 check whether evolution_function_is_affine_multivariate_p
3752 before calling gcd_of_steps_may_divide_p.
3754 2018-01-31 Julia Koval <julia.koval@intel.com>
3757 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3758 * config/i386/i386.md (rdpid_rex64) New.
3759 (rdpid): Make 32bit only.
3761 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3764 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3765 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3767 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3770 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3772 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3774 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3776 PR rtl-optimization/84071
3777 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3778 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3780 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3782 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3783 (arc_attribute_table): Add 'aux' attribute.
3784 (arc_in_small_data_p): Consider aux like variables.
3785 (arc_is_aux_reg_p): New function.
3786 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3787 (arc_get_aux_arg): New function.
3788 (prepare_move_operands): Handle aux-register access.
3789 (arc_handle_aux_attribute): New function.
3790 * doc/extend.texi (ARC Variable attributes): Add subsection.
3792 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3794 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3795 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3796 (arc_attribute_table): Add 'uncached' attribute.
3797 (arc_print_operand): Print '.di' flag for uncached memory
3799 (arc_in_small_data_p): Do not consider for small data the uncached
3801 (arc_is_uncached_mem_p): New function.
3802 * config/arc/predicates.md (compact_store_memory_operand): Check
3803 for uncached memory accesses.
3804 (nonvol_nonimm_operand): Likewise.
3805 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3807 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3810 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3811 falign-loops=): Add Optimization flag.
3813 2018-01-30 Jeff Law <law@redhat.com>
3816 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3817 INT_REGISTERS_SAVED. Check it prior to calling
3818 get_scratch_register_on_entry.
3819 (ix86_adjust_stack_and_probe): Similarly.
3820 (ix86_emit_probe_stack_range): Similarly.
3821 (ix86_expand_prologue): Corresponding changes.
3823 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3826 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3827 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3829 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3832 * lra-constraints.c (curr_insn_transform): Process AND in the
3835 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3837 PR rtl-optimization/83986
3838 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3839 dependence against last_pending_memory_flush in addition to
3842 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3844 PR tree-optimization/81611
3845 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3848 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3851 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3854 2018-01-30 Richard Biener <rguenther@suse.de>
3855 Jakub Jelinek <jakub@redhat.com>
3857 PR tree-optimization/84111
3858 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3859 inner loops added during recursion, as they don't have up-to-date
3862 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3865 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3866 (can_inline_edge_by_limits_p): ... here.
3867 (can_early_inline_edge_p, check_callers,
3868 update_caller_keys, update_callee_keys, recursive_inlining,
3869 add_new_edges_to_heap, speculation_useful_p,
3870 inline_small_functions,
3871 inline_small_functions, flatten_function,
3872 inline_to_all_callers_1): Update.
3874 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3876 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3879 2018-01-30 Richard Biener <rguenther@suse.de>
3881 PR tree-optimization/83008
3882 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3883 invariant and constant vector uses in stmts when they need
3886 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3889 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3890 * configure: Regenerate.
3892 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3894 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3896 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3897 Use gen_rtx_REG rather than gen_lowpart.
3899 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3901 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3902 rather than 0 when creating partial subregs.
3904 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3906 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3909 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3912 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3913 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3914 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3915 flags. This restores the settings used before the 2017-07-24.
3916 Turning off pre increment/decrement/modify allows IVOPTS to
3917 optimize DF/SF loops where the index is an int.
3919 2018-01-29 Richard Biener <rguenther@suse.de>
3920 Kelvin Nilsen <kelvin@gcc.gnu.org>
3923 * tree-vect-stmts.c (vectorizable_call): Don't call
3924 targetm.vectorize_builtin_md_vectorized_function if callee is
3927 2018-01-22 Carl Love <cel@us.ibm.com>
3929 * doc/extend.tex: Fix typo in second arg in
3930 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3932 2018-01-29 Richard Biener <rguenther@suse.de>
3934 PR tree-optimization/84086
3935 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3936 (flush_ssaname_freelist): When SSA names were released reset
3937 the SCEV hash table.
3939 2018-01-29 Richard Biener <rguenther@suse.de>
3941 PR tree-optimization/84057
3942 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3943 removed paths when removing edges.
3945 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3947 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3948 -mfunction-return=@var{choice}.
3950 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3953 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3955 (layout::print_source_line): Likewise.
3956 (test_get_line_width_without_trailing_whitespace): Add test cases.
3958 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3961 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3964 2018-01-26 Jim Wilson <jimw@sifive.com>
3966 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3968 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3971 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3973 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3974 and CMP + SUB-immediate -> SUBS.
3976 2018-01-26 Martin Sebor <msebor@redhat.com>
3978 PR tree-optimization/83896
3979 * tree-ssa-strlen.c (get_string_len): Rename...
3980 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3981 Avoid assuming length is constant.
3982 (handle_char_store): Use HOST_WIDE_INT for string length.
3984 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3987 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3988 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3990 2018-01-26 Richard Biener <rguenther@suse.de>
3992 PR rtl-optimization/84003
3993 * dse.c (record_store): Only record redundant stores when
3994 the earlier store aliases at least all accesses the later one does.
3996 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3998 PR rtl-optimization/83985
3999 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4000 REG_CFA_RESTORE insns.
4001 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4002 REG_CFA_RESTORE insns here.
4005 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4006 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4008 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4010 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4011 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4012 (arc_init): Likewise.
4013 (arc_override_options): Likewise.
4014 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4016 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4018 * config/arc/arc.h (TARGET_DBNZ): Define.
4019 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4020 properly set the tune attribute.
4021 (dbnz): Use TARGET_DBNZ guard.
4022 * config/arc/arc.opt (mtune): Add core3 option.
4024 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4026 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4027 recognize new pic like addresses.
4028 (arc_delegitimize_address): Clean up.
4030 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4032 * config/arc/arc-arches.def: Option mrf16 valid for all
4034 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4035 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4036 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4037 * config/arc/arc-tables.opt: Regenerate.
4038 * config/arc/arc.c (arc_conditional_register_usage): Handle
4039 reduced register file case.
4040 (arc_file_start): Set must have build attributes.
4041 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4043 * config/arc/arc.opt (mrf16): Add new option.
4044 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4045 * config/arc/genmultilib.awk: Handle new mrf16 option.
4046 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4047 * config/arc/t-multilib: Regenerate.
4048 * doc/invoke.texi (ARC Options): Document mrf16 option.
4050 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4052 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4053 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4054 (arc_attribute_table): Add 'secure_call' attribute.
4055 (arc_print_operand): Print secure call operand.
4056 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4058 (arc_is_secure_call_p): New function. * config/arc/arc.md
4059 (call_i): Add support for sjli instruction.
4060 (call_value_i): Likewise.
4061 * config/arc/constraints.md (Csc): New constraint.
4063 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4064 John Eric Martin <John.Martin@emmicro-us.com>
4066 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4067 * config/arc/arc.c (_arc_jli_section): New struct.
4068 (arc_jli_section): New type.
4069 (rc_jli_sections): New static variable.
4070 (arc_handle_jli_attribute): New function.
4071 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4072 (arc_file_end): New function.
4073 (TARGET_ASM_FILE_END): Define.
4074 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4075 (arc_add_jli_section): New function.
4076 (jli_call_scan): Likewise.
4077 (arc_reorg): Call jli_call_scan.
4078 (arc_output_addsi): Remove 'S' from printing asm operand.
4079 (arc_is_jli_call_p): New function.
4080 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4082 (movhi_insn): Likewise.
4083 (movsi_insn): Likewise.
4084 (movsi_set_cc_insn): Likewise.
4085 (loadqi_update): Likewise.
4086 (load_zeroextendqisi_update): Likewise.
4087 (load_signextendqisi_update): Likewise.
4088 (loadhi_update): Likewise.
4089 (load_zeroextendhisi_update): Likewise.
4090 (load_signextendhisi_update): Likewise.
4091 (loadsi_update): Likewise.
4092 (loadsf_update): Likewise.
4093 (movsicc_insn): Likewise.
4094 (bset_insn): Likewise.
4095 (bxor_insn): Likewise.
4096 (bclr_insn): Likewise.
4097 (bmsk_insn): Likewise.
4098 (bicsi3_insn): Likewise.
4099 (cmpsi_cc_c_insn): Likewise.
4100 (movsi_ne): Likewise.
4101 (movsi_cond_exec): Likewise.
4102 (clrsbsi2): Likewise.
4113 (call_i): Remove 'S' asm letter, add jli instruction.
4114 (call_value_i): Likewise.
4115 * config/arc/arc.op (mjli-always): New option.
4116 * config/arc/constraints.md (Cji): New constraint.
4117 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4119 (subsf3_fpx): Likewise.
4120 (mulsf3_fpx): Likewise.
4121 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4123 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4125 * doc/invoke.texi (ARC): Document mjli-always option.
4127 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4129 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4130 avoid addition with 0 and use incw and decw where possible.
4132 2018-01-26 Richard Biener <rguenther@suse.de>
4134 PR tree-optimization/81082
4135 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4136 association if it requires casting to unsigned.
4137 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4138 from fold_plusminus_mult_expr to catch important cases late when
4139 range info is available.
4141 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4143 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4144 * configure.ac (hidden_linkonce): New test.
4145 * configure: Regenerate.
4146 * config.in: Regenerate.
4148 2018-01-26 Julia Koval <julia.koval@intel.com>
4150 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4151 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4152 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4153 _mm_mask_bitshuffle_epi64_mask): Fix type.
4154 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4155 USI_FTYPE_V4DI_V4DI_USI): Remove.
4156 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4157 __builtin_ia32_vpshufbitqmb256_mask,
4158 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4159 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4160 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4162 2018-01-26 Alan Modra <amodra@gmail.com>
4165 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4166 UNSPEC_VBPERMQ. Sort other unspecs.
4168 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4170 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4172 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4175 * predict.c (drop_profile): Do not push/pop cfun; update also
4177 (handle_missing_profiles): Fix logic looking for zero profiles.
4179 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4182 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4183 on functions with #pragma omp declare simd or functions with simd
4185 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4186 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4187 Remove trailing \n from warning_at calls.
4189 2018-01-25 Tom de Vries <tom@codesourcery.com>
4192 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4193 for neutered workers.
4195 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4198 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4199 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4201 2018-01-24 Jeff Law <law@redhat.com>
4204 * i386.c (get_probe_interval): Move to earlier point.
4205 (ix86_compute_frame_layout): If -fstack-clash-protection and
4206 the frame is larger than the probe interval, then use pushes
4207 to save registers rather than reg->mem moves.
4208 (ix86_expand_prologue): Remove conditional for int_registers_saved
4211 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4214 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4215 min/max for never referenced object.
4217 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4220 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4222 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4223 attributes from DECL_ATTRIBUTES (decl) without affecting
4224 DECL_ATTRIBUTES (current_function_decl).
4225 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4226 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4228 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4230 PR tree-optimization/83979
4231 * fold-const.c (fold_comparison): Use constant_boolean_node
4232 instead of boolean_{true,false}_node.
4234 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4236 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4239 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4241 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4242 Simplify the clause that sets the length attribute.
4243 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4244 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4245 clause that sets the length attribute.
4246 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4248 2018-01-24 Tom de Vries <tom@codesourcery.com>
4251 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4252 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4253 Add strict parameter.
4254 (prevent_branch_around_nothing): Insert dummy insn between branch to
4255 label and label with no ptx insn inbetween.
4256 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4258 2018-01-24 Tom de Vries <tom@codesourcery.com>
4261 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4262 for neutered threads in warp.
4263 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4265 2018-01-24 Richard Biener <rguenther@suse.de>
4267 PR tree-optimization/83176
4268 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4271 2018-01-24 Richard Biener <rguenther@suse.de>
4273 PR tree-optimization/82819
4274 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4275 code generating pluses that are no-ops in the target precision.
4277 2018-01-24 Richard Biener <rguenther@suse.de>
4280 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4282 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4284 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4285 to merge probabilities.
4286 * predict.c (probably_never_executed): Also mark as cold functions
4287 with global 0 profile and guessed local profile.
4288 * profile-count.c (profile_probability::combine_with_count): New
4290 * profile-count.h (profile_probability::operator*,
4291 profile_probability::operator*=, profile_probability::operator/,
4292 profile_probability::operator/=): Reduce precision to adjusted
4293 and set value to guessed on contradictory divisions.
4294 (profile_probability::combine_with_freq): Remove.
4295 (profile_probability::combine_wiht_count): Declare.
4296 (profile_count::force_nonzero):: Set to adjusted.
4297 (profile_count::probability_in):: Set quality to adjusted.
4298 * tree-ssa-tail-merge.c (replace_block_by): Use
4301 2018-01-23 Andrew Waterman <andrew@sifive.com>
4302 Jim Wilson <jimw@sifive.com>
4304 * config/riscv/riscv.c (riscv_stack_boundary): New.
4305 (riscv_option_override): Set riscv_stack_boundary. Handle
4306 riscv_preferred_stack_boundary_arg.
4307 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4308 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4309 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4310 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4311 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4312 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4314 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4317 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4318 of struct ix86_frame.
4319 (ix86_expand_epilogue): Likewise. Add a local variable for
4320 the reg_save_offset field in struct ix86_frame.
4322 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4324 PR tree-optimization/82604
4325 * tree-loop-distribution.c (enum partition_kind): New enum item
4326 PKIND_PARTIAL_MEMSET.
4327 (partition_builtin_p): Support above new enum item.
4328 (generate_code_for_partition): Ditto.
4329 (compute_access_range): Differentiate cases that equality can be
4330 proven at all loops, the innermost loops or no loops.
4331 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4332 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4333 (finalize_partitions, distribute_loop): Don't fuse partition of
4334 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4335 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4338 2018-01-23 Martin Liska <mliska@suse.cz>
4340 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4341 order to ignore the predictor.
4342 (PRED_POLYMORPHIC_CALL): Likewise.
4343 (PRED_RECURSIVE_CALL): Likewise.
4345 2018-01-23 Martin Liska <mliska@suse.cz>
4347 * tree-profile.c (tree_profiling): Print function header to
4348 aware reader which function we are working on.
4349 * value-prof.c (gimple_find_values_to_profile): Do not print
4350 not interesting value histograms.
4352 2018-01-23 Martin Liska <mliska@suse.cz>
4354 * profile-count.h (enum profile_quality): Add
4355 profile_uninitialized as the first value. Do not number values
4356 as they are zero based.
4357 (profile_count::verify): Update sanity check.
4358 (profile_probability::verify): Likewise.
4360 2018-01-23 Nathan Sidwell <nathan@acm.org>
4362 * doc/invoke.texi (ffor-scope): Deprecate.
4364 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4366 PR tree-optimization/83510
4367 * domwalk.c (set_all_edges_as_executable): New function.
4368 (dom_walker::dom_walker): Convert bool param
4369 "skip_unreachable_blocks" to enum reachability. Move setup of
4370 edge flags to set_all_edges_as_executable and only do it when
4371 reachability is REACHABLE_BLOCKS.
4372 * domwalk.h (enum dom_walker::reachability): New enum.
4373 (dom_walker::dom_walker): Convert bool param
4374 "skip_unreachable_blocks" to enum reachability.
4375 (set_all_edges_as_executable): New decl.
4376 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4377 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4379 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4380 but converting true to REACHABLE_BLOCKS.
4381 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4383 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4384 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4385 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4387 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4388 if check_all_array_refs will be called.
4390 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4392 * tree.c (selftest::test_location_wrappers): Add more test
4395 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4397 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4398 (selftest::test_bit_in_range): Likewise.
4400 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4403 * doc/sourcebuild.texi (vect_float): Say that the selector
4404 only describes the situation when -funsafe-math-optimizations is on.
4405 (vect_float_strict): Document.
4407 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4409 PR tree-optimization/83965
4410 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4411 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4412 instead of checking only for a reduction.
4413 (vect_recog_widen_sum_pattern): Likewise.
4415 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4417 * predict.c (probably_never_executed): Only use precise profile info.
4418 (compute_function_frequency): Skip after inlining hack since we now
4419 have quality checking.
4421 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4423 * profile-count.h (profile_probability::very_unlikely,
4424 profile_probability::unlikely, profile_probability::even): Set
4425 precision to guessed.
4427 2018-01-23 Richard Biener <rguenther@suse.de>
4429 PR tree-optimization/83963
4430 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4431 Properly terminate dominator walk when crossing the exit edge not
4432 when visiting its source block.
4434 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4437 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4438 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4440 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4442 PR tree-optimization/83957
4443 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4444 semicolon after for body surrounded by braces.
4446 PR tree-optimization/83081
4447 * profile-count.h (profile_probability::split): New method.
4448 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4449 Use profile_probability::split.
4450 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4451 when splitting a single conditional jump into 2.
4453 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4455 PR tree-optimization/69452
4456 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4459 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4461 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4462 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4463 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4465 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4467 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4468 * config/rl78/rl78.md (movdi): New define_expand.
4469 * config/rl78/rl78.c (rl78_split_movdi): New function.
4471 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4474 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4476 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4477 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4478 128-bit to produce an UNSPEC move to get the double word with the
4479 signbit and then a shift directly to do signbit.
4480 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4481 implementation with a new version that just does either a direct
4482 move or a regular move. Move memory interface to separate insns.
4483 Move insns so they are next to the expander.
4484 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4485 with signbit move. Split big and little endian case.
4486 (signbit<mode>2_dm_mem_le): Likewise.
4487 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4488 (signbit<mode>2_dm2): Likewise.
4490 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4492 * config/rl78/rl78.md (anddi3): New define_expand.
4494 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4496 * config/rl78/rl78.md (umindi3): New define_expand.
4498 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4500 * config/rl78/rl78.md (smindi3): New define_expand.
4502 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4504 * config/rl78/rl78.md (smaxdi3): New define_expand.
4506 2018-01-22 Carl Love <cel@us.ibm.com>
4508 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4509 LVX_V1TI): Add macro expansion.
4510 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4511 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4512 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4513 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4514 Change check to determine if the instruction is a byte reversing
4515 entry. Fix typo in comment.
4516 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4517 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4518 Add def_builtin calls for new builtins.
4519 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4520 Add define_insn expansion.
4522 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4524 * config/rl78/rl78.md (umaxdi3): New define_expand.
4526 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4528 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4529 for non-QImode registers.
4531 2018-01-22 Richard Biener <rguenther@suse.de>
4533 PR tree-optimization/83963
4534 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4535 including the loop exit block.
4536 (scop_detection::merge_sese): Likewise.
4537 (scop_detection::add_scop): Do it here instead.
4539 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4541 * doc/sourcebuild.texi (arm_softfloat): Document.
4543 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4546 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4547 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4548 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4550 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4551 David Edelsohn <dje.gcc@gmail.com>
4554 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4555 Change "crset eq" to "crset 2".
4556 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4557 (*call_indirect_aix<mode>_nospec): Likewise.
4558 (*call_value_indirect_aix<mode>_nospec): Likewise.
4559 (*call_indirect_elfv2<mode>_nospec): Likewise.
4560 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4561 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4562 change assembly output from . to $.
4563 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4564 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4565 (*tablejump<mode>_internal1_nospec): Likewise.
4567 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4570 * config/sh/sh_optimize_sett_clrt.cc:
4571 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4573 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4575 PR tree-optimization/83940
4576 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4577 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4578 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4579 use it to pass back the definition type.
4580 (vect_check_store_rhs): Likewise rhs_dt_out.
4581 (vect_build_gather_load_calls): Add a mask_dt argument and use
4582 it instead of a call to vect_is_simple_use.
4583 (vectorizable_store): Update calls to vect_check_load_store_mask
4584 and vect_check_store_rhs. Use the dt returned by the latter instead
4585 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4586 instead of calls to vect_is_simple_use. Pass the scalar rather
4587 than the vector operand to vect_is_simple_use when handling
4588 second and subsequent copies of an rhs value.
4589 (vectorizable_load): Update calls to vect_check_load_store_mask
4590 and vect_build_gather_load_calls. Use the cached mask_dt and
4591 gs_info.offset_dt instead of calls to vect_is_simple_use.
4593 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4596 * tree-emutls.c: Include gimplify.h.
4597 (lower_emutls_2): New function.
4598 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4599 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4600 it before further processing.
4603 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4604 UINTVAL (trueop1) instead of INTVAL (op1).
4606 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4610 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4611 INCOMING_FRAME_SP_OFFSET if not defined.
4612 (scan_trace): Add ENTRY argument. If true and
4613 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4614 emit a note to adjust the CFA offset.
4615 (create_cfi_notes): Adjust scan_trace callers.
4616 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4617 INCOMING_FRAME_SP_OFFSET in the CIE.
4618 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4619 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4621 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4622 * doc/tm.texi: Regenerated.
4624 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4626 PR rtl-optimization/83147
4627 * lra-constraints.c (remove_inheritance_pseudos): Use
4628 lra_substitute_pseudo_within_insn.
4630 2018-01-19 Tom de Vries <tom@codesourcery.com>
4631 Cesar Philippidis <cesar@codesourcery.com>
4634 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4636 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4639 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4640 spaces for function labels.
4642 2018-01-19 Martin Liska <mliska@suse.cz>
4644 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4645 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4646 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4647 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4648 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4649 (PRED_CONST_RETURN): Change from 69 to 65.
4650 (PRED_NULL_RETURN): Change from 91 to 71.
4651 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4652 (PRED_LOOP_GUARD): Change from 66 to 73.
4654 2018-01-19 Martin Liska <mliska@suse.cz>
4656 * predict.c (predict_insn_def): Add new assert.
4657 (struct branch_predictor): Change type to signed integer.
4658 (test_prediction_value_range): Amend test to cover
4660 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4661 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4662 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4663 (PRED_LOOP_IV_COMPARE): Likewise.
4664 * predict.h (PROB_UNINITIALIZED): Define new constant.
4666 2018-01-19 Martin Liska <mliska@suse.cz>
4668 * predict.c (dump_prediction): Add new format for
4669 analyze_brprob.py script which is enabled with -details
4671 * profile-count.h (precise_p): New function.
4673 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4675 PR tree-optimization/83922
4676 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4677 there are no statements that need masking.
4678 (vect_active_double_reduction_p): New function.
4679 (vect_analyze_loop_operations): Use it when handling phis that
4680 are not in the loop header.
4682 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4684 PR tree-optimization/83914
4685 * tree-vect-loop.c (vectorizable_induction): Don't convert
4686 init_expr or apply the peeling adjustment for inductions
4687 that are nested within the vectorized loop.
4689 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4691 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4694 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4698 * function.h (gimplify_parameters): Add gimple_seq * argument.
4699 * function.c: Include gimple.h and options.h.
4700 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4701 for the added local temporaries if needed.
4702 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4703 if there are any parameter cleanups, wrap whole body into a
4704 try/finally with the cleanups.
4706 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4709 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4710 Use GET_MODE_CLASS for scalar floating point.
4712 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4716 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4717 Fix call of call_cgraph_insertion_hooks.
4719 2018-01-18 Martin Sebor <msebor@redhat.com>
4721 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4723 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4726 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4729 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4732 * common.opt: (-ffile-prefix-map): New option.
4733 * opts.c (common_handle_option): Defer it.
4734 * opts-global.c (handle_common_deferred_options): Handle it.
4735 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4736 * file-prefix-map.h: New file.
4737 (remap_debug_filename, add_debug_prefix_map): ...here.
4738 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4739 * final.c (debug_prefix_map, add_debug_prefix_map
4740 remap_debug_filename): Move to...
4741 * file-prefix-map.c: New file.
4742 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4743 generalize, get rid of alloca(), use strrchr() instead of strchr().
4744 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4745 Implement in terms of add_prefix_map().
4746 (remap_macro_filename, remap_debug_filename): Implement in term of
4748 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4749 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4750 * dbxout.c: Include file-prefix-map.h.
4751 * varasm.c: Likewise.
4752 * vmsdbgout.c: Likewise.
4753 * xcoffout.c: Likewise.
4754 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4755 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4756 * doc/invoke.texi (-ffile-prefix-map): Document.
4757 (-fdebug-prefix-map): Update description.
4759 2018-01-18 Martin Liska <mliska@suse.cz>
4761 * config/i386/i386.c (indirect_thunk_name): Document that also
4763 (output_indirect_thunk): Document why both instructions
4764 (pause and lfence) are generated.
4766 2018-01-18 Richard Biener <rguenther@suse.de>
4768 PR tree-optimization/83887
4769 * graphite-scop-detection.c
4770 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4771 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4772 (scop_detection::merge_sese): Re-implement with a flood-fill
4773 algorithm that properly finds a SESE region if it exists.
4775 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4778 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4779 pointer_diff optimizations use view_convert instead of convert.
4781 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4783 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4784 Generate different code for -mno-speculate-indirect-jumps.
4785 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4786 (*call_indirect_aix<mode>): Disable for
4787 -mno-speculate-indirect-jumps.
4788 (*call_indirect_aix<mode>_nospec): New define_insn.
4789 (*call_value_indirect_aix<mode>): Disable for
4790 -mno-speculate-indirect-jumps.
4791 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4792 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4793 -mno-speculate-indirect-jumps.
4794 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4796 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4798 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4799 long double type, set the flags for noting the default long double
4800 type, even if we don't pass or return a long double type.
4802 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4805 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4808 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4810 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4811 support for merge[hl].
4812 (fold_mergehl_helper): New helper function.
4813 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4814 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4815 (altivec_vmrglw_direct): Add xxmrglw insn.
4817 2018-01-17 Andrew Waterman <andrew@sifive.com>
4819 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4820 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4822 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4825 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4826 call the lto_location_cache before reading the
4827 DECL_SOURCE_LOCATION of the types.
4829 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4830 Richard Sandiford <richard.sandiford@linaro.org>
4832 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4833 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4834 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4835 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4836 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4838 * config/aarch64/constraints.md (aarch64_movti_operand):
4840 * config/aarch64/predicates.md (Uti): Add new constraint.
4842 2018-01-17 Carl Love <cel@us.ibm.com>
4844 * config/rs6000/vsx.md (define_expand xl_len_r,
4845 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4846 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4848 (define_expand, define_insn): Move the shift left from the
4849 define_insn to the define_expand for lxvl and stxvl instructions.
4850 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4851 and XL_LEN_R definitions to PURE.
4853 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4855 * config/i386/i386.c (indirect_thunk_name): Declare regno
4856 as unsigned int. Compare regno with INVALID_REGNUM.
4857 (output_indirect_thunk): Ditto.
4858 (output_indirect_thunk_function): Ditto.
4859 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4860 in the call to output_indirect_thunk_function.
4862 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4865 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4866 rather than the size of inner_type to determine the stack slot size
4867 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4869 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4872 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4875 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4877 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4878 endian Linux systems to optionally enable multilibs for selecting
4879 the long double type if the user configured an explicit type.
4880 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4881 have no long double multilibs if not defined.
4882 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4883 warn if the user used -mabi={ieee,ibm}longdouble and we built
4884 multilibs for long double.
4885 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4886 appropriate multilib option.
4887 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4889 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4890 for building long double multilibs.
4891 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4893 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4895 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4898 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4900 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4903 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4906 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4909 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4911 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4912 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4914 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4916 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4917 different rtl trees depending on TARGET_64BIT.
4918 (rs6000_gen_lvx): Likewise.
4920 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4922 * config/visium/visium.md (nop): Tweak comment.
4923 (hazard_nop): Likewise.
4925 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4927 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4928 -mspeculate-indirect-jumps.
4929 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4930 for -mno-speculate-indirect-jumps.
4931 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4932 (*call_value_indirect_elfv2<mode>): Disable for
4933 -mno-speculate-indirect-jumps.
4934 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4935 (indirect_jump): Emit different RTL for
4936 -mno-speculate-indirect-jumps.
4937 (*indirect_jump<mode>): Disable for
4938 -mno-speculate-indirect-jumps.
4939 (*indirect_jump<mode>_nospec): New define_insn.
4940 (tablejump): Emit different RTL for
4941 -mno-speculate-indirect-jumps.
4942 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4943 (tablejumpsi_nospec): New define_expand.
4944 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4945 (tablejumpdi_nospec): New define_expand.
4946 (*tablejump<mode>_internal1): Disable for
4947 -mno-speculate-indirect-jumps.
4948 (*tablejump<mode>_internal1_nospec): New define_insn.
4949 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4952 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4954 * caller-save.c (insert_save): Drop unnecessary parameter. All
4957 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4958 Richard Biener <rguenth@suse.de>
4961 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4962 return early, inline manually is_gimple_sizepos. Make sure if we
4963 call gimplify_expr we don't end up with a gimple constant.
4964 * tree.c (variably_modified_type_p): Don't return true for
4965 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4966 * gimplify.h (is_gimple_sizepos): Remove.
4968 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4970 PR tree-optimization/83857
4971 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4972 vectorizable_live_operation for pure SLP statements.
4973 (vectorizable_live_operation): Handle PHIs.
4975 2018-01-16 Richard Biener <rguenther@suse.de>
4977 PR tree-optimization/83867
4978 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4979 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4981 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4984 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4985 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4986 If off is not INTEGER_CST, issue a may not be aligned warning
4987 rather than isn't aligned. Use isn%'t rather than isn't.
4988 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4990 <case MULT_EXPR>: Improve the case when bottom and one of the
4991 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4992 operand, in that case check if the other operand is multiple of
4993 bottom divided by the INTEGER_CST operand.
4995 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4998 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4999 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5000 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5001 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5002 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5003 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5004 (pa_function_arg_size): New function.
5006 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5008 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5009 in a separate statement.
5011 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5013 PR tree-optimization/83847
5014 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5015 group gathers and scatters.
5017 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5019 PR rtl-optimization/86620
5020 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5022 PR rtl-optimization/83213
5023 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5024 to last if both are JUMP_INSNs.
5026 PR tree-optimization/83843
5027 * gimple-ssa-store-merging.c
5028 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5029 store_immediate_info for bswap/nop orig_stores.
5031 2018-01-15 Andrew Waterman <andrew@sifive.com>
5033 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5035 <UDIV>: Increase cost if !TARGET_DIV.
5037 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5039 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5040 (define_attr "cr_logical_3op"): New.
5041 (cceq_ior_compare): Adjust.
5042 (cceq_ior_compare_complement): Adjust.
5043 (*cceq_rev_compare): Adjust.
5044 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5045 (is_cracked_insn): Adjust.
5046 (insn_must_be_first_in_group): Adjust.
5047 * config/rs6000/40x.md: Adjust.
5048 * config/rs6000/440.md: Adjust.
5049 * config/rs6000/476.md: Adjust.
5050 * config/rs6000/601.md: Adjust.
5051 * config/rs6000/603.md: Adjust.
5052 * config/rs6000/6xx.md: Adjust.
5053 * config/rs6000/7450.md: Adjust.
5054 * config/rs6000/7xx.md: Adjust.
5055 * config/rs6000/8540.md: Adjust.
5056 * config/rs6000/cell.md: Adjust.
5057 * config/rs6000/e300c2c3.md: Adjust.
5058 * config/rs6000/e500mc.md: Adjust.
5059 * config/rs6000/e500mc64.md: Adjust.
5060 * config/rs6000/e5500.md: Adjust.
5061 * config/rs6000/e6500.md: Adjust.
5062 * config/rs6000/mpc.md: Adjust.
5063 * config/rs6000/power4.md: Adjust.
5064 * config/rs6000/power5.md: Adjust.
5065 * config/rs6000/power6.md: Adjust.
5066 * config/rs6000/power7.md: Adjust.
5067 * config/rs6000/power8.md: Adjust.
5068 * config/rs6000/power9.md: Adjust.
5069 * config/rs6000/rs64.md: Adjust.
5070 * config/rs6000/titan.md: Adjust.
5072 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5074 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5075 ix86_indirect_branch_register logic.
5077 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5079 * config/i386/constraints.md (Bs): Update
5080 ix86_indirect_branch_register check. Don't check
5081 ix86_indirect_branch_register with GOT_memory_operand.
5083 * config/i386/predicates.md (GOT_memory_operand): Don't check
5084 ix86_indirect_branch_register here.
5085 (GOT32_symbol_operand): Likewise.
5087 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5089 * config/i386/predicates.md (constant_call_address_operand):
5090 Rewrite ix86_indirect_branch_register logic.
5091 (sibcall_insn_operand): Likewise.
5093 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5095 * config/i386/constraints.md (Bs): Replace
5096 ix86_indirect_branch_thunk_register with
5097 ix86_indirect_branch_register.
5099 * config/i386/i386.md (indirect_jump): Likewise.
5100 (tablejump): Likewise.
5101 (*sibcall_memory): Likewise.
5102 (*sibcall_value_memory): Likewise.
5103 Peepholes of indirect call and jump via memory: Likewise.
5104 * config/i386/i386.opt: Likewise.
5105 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5106 (GOT_memory_operand): Likewise.
5107 (call_insn_operand): Likewise.
5108 (sibcall_insn_operand): Likewise.
5109 (GOT32_symbol_operand): Likewise.
5111 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5114 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5115 type rather than type addr's type points to.
5116 (expand_omp_atomic_mutex): Likewise.
5117 (expand_omp_atomic): Likewise.
5119 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5122 * config/i386/i386.c (output_indirect_thunk_function): Use
5123 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5124 for __x86_return_thunk.
5126 2018-01-15 Richard Biener <rguenther@suse.de>
5129 * expmed.c (extract_bit_field_1): Fix typo.
5131 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5134 * config/arm/iterators.md (VF): New mode iterator.
5135 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5136 Remove integer-related logic from pattern.
5137 (neon_vabd<mode>_3): Likewise.
5139 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5142 * common.opt (fstrict-overflow): No longer an alias.
5143 (fwrapv-pointer): New option.
5144 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5145 also for pointer types based on flag_wrapv_pointer.
5146 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5147 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5148 opts->x_flag_wrapv got set.
5149 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5150 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5151 POINTER_TYPE_OVERFLOW_UNDEFINED.
5152 * match.pd: Likewise in address comparison pattern.
5153 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5155 2018-01-15 Richard Biener <rguenther@suse.de>
5158 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5159 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5160 Reset type names to their identifier if their TYPE_DECL doesn't
5161 have linkage (and thus is used for ODR and devirt).
5162 (save_debug_info_for_decl): Remove.
5163 (save_debug_info_for_type): Likewise.
5164 (add_tree_to_fld_list): Adjust.
5165 * tree-pretty-print.c (dump_generic_node): Make dumping of
5166 type names more robust.
5168 2018-01-15 Richard Biener <rguenther@suse.de>
5170 * BASE-VER: Bump to 8.0.1.
5172 2018-01-14 Martin Sebor <msebor@redhat.com>
5175 * builtins.c (check_access): Avoid warning when the no-warning bit
5178 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5180 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5181 * ira-color (allocno_hard_regs_compare): Likewise.
5183 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5186 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5187 Use .pushsection/.popsection.
5189 2018-01-14 Martin Sebor <msebor@redhat.com>
5192 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5194 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5196 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5197 entry from extra_headers.
5198 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5199 extra_headers, make the list bitwise identical to the i?86-*-* one.
5201 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5203 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5204 -mcmodel=large with -mindirect-branch=thunk,
5205 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5206 -mfunction-return=thunk-extern.
5207 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5208 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5209 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5211 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5213 * config/i386/i386.c (print_reg): Print the name of the full
5214 integer register without '%'.
5215 (ix86_print_operand): Handle 'V'.
5216 * doc/extend.texi: Document 'V' modifier.
5218 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5220 * config/i386/constraints.md (Bs): Disallow memory operand for
5221 -mindirect-branch-register.
5223 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5224 (GOT_memory_operand): Likewise.
5225 (call_insn_operand): Likewise.
5226 (sibcall_insn_operand): Likewise.
5227 (GOT32_symbol_operand): Likewise.
5228 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5229 for -mindirect-branch-register.
5230 (tablejump): Likewise.
5231 (*sibcall_memory): Likewise.
5232 (*sibcall_value_memory): Likewise.
5233 Disallow peepholes of indirect call and jump via memory for
5234 -mindirect-branch-register.
5235 (*call_pop): Replace m with Bw.
5236 (*call_value_pop): Likewise.
5237 (*sibcall_pop_memory): Replace m with Bs.
5238 * config/i386/i386.opt (mindirect-branch-register): New option.
5239 * doc/invoke.texi: Document -mindirect-branch-register option.
5241 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5243 * config/i386/i386-protos.h (ix86_output_function_return): New.
5244 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5245 set function_return_type.
5246 (indirect_thunk_name): Add ret_p to indicate thunk for function
5248 (output_indirect_thunk_function): Pass false to
5249 indirect_thunk_name.
5250 (ix86_output_indirect_branch_via_reg): Likewise.
5251 (ix86_output_indirect_branch_via_push): Likewise.
5252 (output_indirect_thunk_function): Create alias for function
5253 return thunk if regno < 0.
5254 (ix86_output_function_return): New function.
5255 (ix86_handle_fndecl_attribute): Handle function_return.
5256 (ix86_attribute_table): Add function_return.
5257 * config/i386/i386.h (machine_function): Add
5258 function_return_type.
5259 * config/i386/i386.md (simple_return_internal): Use
5260 ix86_output_function_return.
5261 (simple_return_internal_long): Likewise.
5262 * config/i386/i386.opt (mfunction-return=): New option.
5263 (indirect_branch): Mention -mfunction-return=.
5264 * doc/extend.texi: Document function_return function attribute.
5265 * doc/invoke.texi: Document -mfunction-return= option.
5267 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5269 * config/i386/i386-opts.h (indirect_branch): New.
5270 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5271 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5272 with local indirect jump when converting indirect call and jump.
5273 (ix86_set_indirect_branch_type): New.
5274 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5275 (indirectlabelno): New.
5276 (indirect_thunk_needed): Likewise.
5277 (indirect_thunk_bnd_needed): Likewise.
5278 (indirect_thunks_used): Likewise.
5279 (indirect_thunks_bnd_used): Likewise.
5280 (INDIRECT_LABEL): Likewise.
5281 (indirect_thunk_name): Likewise.
5282 (output_indirect_thunk): Likewise.
5283 (output_indirect_thunk_function): Likewise.
5284 (ix86_output_indirect_branch_via_reg): Likewise.
5285 (ix86_output_indirect_branch_via_push): Likewise.
5286 (ix86_output_indirect_branch): Likewise.
5287 (ix86_output_indirect_jmp): Likewise.
5288 (ix86_code_end): Call output_indirect_thunk_function if needed.
5289 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5291 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5292 (ix86_attribute_table): Add indirect_branch.
5293 * config/i386/i386.h (machine_function): Add indirect_branch_type
5294 and has_local_indirect_jump.
5295 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5297 (tablejump): Likewise.
5298 (*indirect_jump): Use ix86_output_indirect_jmp.
5299 (*tablejump_1): Likewise.
5300 (simple_return_indirect_internal): Likewise.
5301 * config/i386/i386.opt (mindirect-branch=): New option.
5302 (indirect_branch): New.
5305 (thunk-inline): Likewise.
5306 (thunk-extern): Likewise.
5307 * doc/extend.texi: Document indirect_branch function attribute.
5308 * doc/invoke.texi: Document -mindirect-branch= option.
5310 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5313 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5315 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5317 * ipa-inline.c (want_inline_small_function_p): Return false if
5318 inlining has already failed with CIF_FINAL_ERROR.
5319 (update_caller_keys): Call want_inline_small_function_p before
5321 (update_callee_keys): Likewise.
5323 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5325 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5327 (rs6000_quadword_masked_address_p): Likewise.
5328 (quad_aligned_load_p): Likewise.
5329 (quad_aligned_store_p): Likewise.
5330 (const_load_sequence_p): Add comment to describe the outer-most loop.
5331 (mimic_memory_attributes_and_flags): New function.
5332 (rs6000_gen_stvx): Likewise.
5333 (replace_swapped_aligned_store): Likewise.
5334 (rs6000_gen_lvx): Likewise.
5335 (replace_swapped_aligned_load): Likewise.
5336 (replace_swapped_load_constant): Capitalize argument name in
5337 comment describing this function.
5338 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5339 and stores that access quad-word aligned addresses and replace
5340 with stvx or lvx instructions when appropriate.
5341 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5342 New function prototype.
5343 (rs6000_quadword_masked_address_p): Likewise.
5344 (rs6000_gen_lvx): Likewise.
5345 (rs6000_gen_stvx): Likewise.
5346 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5347 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5348 when memory address is aligned.
5349 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5350 this split to select lvx instruction when memory address is aligned.
5351 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5352 instruction when memory address is aligned.
5353 (*vsx_le_perm_load_v16qi): Likewise.
5354 (four unnamed splitters): Modify to select the stvx instruction
5355 when memory is aligned.
5357 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5359 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5360 which appears in the queue multiple times.
5362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5363 Alan Hayward <alan.hayward@arm.com>
5364 David Sherwood <david.sherwood@arm.com>
5366 * tree-vectorizer.h (vec_lower_bound): New structure.
5367 (_loop_vec_info): Add check_nonzero and lower_bounds.
5368 (LOOP_VINFO_CHECK_NONZERO): New macro.
5369 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5370 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5371 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5372 fields. Make seg_len the distance travelled, not including the
5374 (dr_direction_indicator): Declare.
5375 (dr_zero_step_indicator): Likewise.
5376 (dr_known_forward_stride_p): Likewise.
5377 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5379 (runtime_alias_check_p): Allow runtime alias checks with
5381 (operator ==): Compare access_size and align.
5382 (prune_runtime_alias_test_list): Rework for new distinction between
5383 the access_size and seg_len.
5384 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5386 (get_segment_min_max): New function.
5387 (create_intersect_range_checks): Use it.
5388 (dr_step_indicator): New function.
5389 (dr_direction_indicator): Likewise.
5390 (dr_zero_step_indicator): Likewise.
5391 (dr_known_forward_stride_p): Likewise.
5392 * tree-loop-distribution.c (data_ref_segment_size): Return
5393 DR_STEP * (niters - 1).
5394 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5396 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5397 (vect_preserves_scalar_order_p): New function, split out from...
5398 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5399 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5400 (vect_vfa_access_size): New function.
5401 (vect_vfa_align): Likewise.
5402 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5403 (dump_lower_bound): New function.
5404 (vect_check_lower_bound): Likewise.
5405 (vect_small_gap_p): Likewise.
5406 (vectorizable_with_step_bound_p): Likewise.
5407 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5408 depencies if the vectorization factor is 1. Convert the checks
5409 for nonzero steps into checks on the bounds of DR_STEP. Try using
5410 a bunds check for variable steps if the minimum required step is
5411 relatively small. Update calls to the dr_with_seg_len
5412 constructor and to vect_compile_time_alias.
5413 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5415 (vect_loop_versioning): Call it.
5416 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5418 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5421 Alan Hayward <alan.hayward@arm.com>
5422 David Sherwood <david.sherwood@arm.com>
5424 * doc/sourcebuild.texi (vect_scatter_store): Document.
5425 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5427 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5429 * genopinit.c (main): Add supports_vec_scatter_store and
5430 supports_vec_scatter_store_cached to target_optabs.
5431 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5432 IFN_MASK_SCATTER_STORE.
5433 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5435 * internal-fn.h (internal_store_fn_p): Declare.
5436 (internal_fn_stored_value_index): Likewise.
5437 * internal-fn.c (scatter_store_direct): New macro.
5438 (expand_scatter_store_optab_fn): New function.
5439 (direct_scatter_store_optab_supported_p): New macro.
5440 (internal_store_fn_p): New function.
5441 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5442 IFN_MASK_SCATTER_STORE.
5443 (internal_fn_mask_index): Likewise.
5444 (internal_fn_stored_value_index): New function.
5445 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5447 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5448 * optabs-query.c (supports_vec_scatter_store_p): New function.
5449 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5450 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5451 true for scatter stores.
5452 (vect_gather_scatter_fn_p): Handle scatter stores too.
5453 (vect_check_gather_scatter): Consider using scatter stores if
5454 supports_vec_scatter_store_p.
5455 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5457 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5458 internal_fn_stored_value_index.
5459 (check_load_store_masking): Handle scatter stores too.
5460 (vect_get_store_rhs): Make public.
5461 (vectorizable_call): Use internal_store_fn_p.
5462 (vectorizable_store): Handle scatter store internal functions.
5463 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5464 when deciding whether the end of the group has been reached.
5465 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5466 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5467 (mask_scatter_store<mode>): New insns.
5469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5470 Alan Hayward <alan.hayward@arm.com>
5471 David Sherwood <david.sherwood@arm.com>
5473 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5474 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5475 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5477 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5478 Use vect_truncate_gather_scatter_offset if we can't treat the
5479 operation as a normal gather load or scatter store.
5480 (get_group_load_store_type): Take the gather_scatter_info
5481 as argument. Try using a gather load or scatter store for
5482 single-element groups.
5483 (get_load_store_type): Update calls to get_group_load_store_type
5484 and vect_use_strided_gather_scatters_p.
5486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5487 Alan Hayward <alan.hayward@arm.com>
5488 David Sherwood <david.sherwood@arm.com>
5490 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5491 optional tree argument.
5492 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5494 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5495 but continue to use the current value as a fallback.
5496 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5497 to compare the updates.
5498 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5499 (get_load_store_type): Use it when handling a strided access.
5500 (vect_get_strided_load_store_ops): New function.
5501 (vect_get_data_ptr_increment): Likewise.
5502 (vectorizable_load): Handle strided gather loads. Always pass
5503 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5505 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5506 Alan Hayward <alan.hayward@arm.com>
5507 David Sherwood <david.sherwood@arm.com>
5509 * doc/md.texi (gather_load@var{m}): Document.
5510 (mask_gather_load@var{m}): Likewise.
5511 * genopinit.c (main): Add supports_vec_gather_load and
5512 supports_vec_gather_load_cached to target_optabs.
5513 * optabs-tree.c (init_tree_optimization_optabs): Use
5514 ggc_cleared_alloc to allocate target_optabs.
5515 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5516 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5518 * internal-fn.h (internal_load_fn_p): Declare.
5519 (internal_gather_scatter_fn_p): Likewise.
5520 (internal_fn_mask_index): Likewise.
5521 (internal_gather_scatter_fn_supported_p): Likewise.
5522 * internal-fn.c (gather_load_direct): New macro.
5523 (expand_gather_load_optab_fn): New function.
5524 (direct_gather_load_optab_supported_p): New macro.
5525 (direct_internal_fn_optab): New function.
5526 (internal_load_fn_p): Likewise.
5527 (internal_gather_scatter_fn_p): Likewise.
5528 (internal_fn_mask_index): Likewise.
5529 (internal_gather_scatter_fn_supported_p): Likewise.
5530 * optabs-query.c (supports_at_least_one_mode_p): New function.
5531 (supports_vec_gather_load_p): Likewise.
5532 * optabs-query.h (supports_vec_gather_load_p): Declare.
5533 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5534 and memory_type field.
5535 (NUM_PATTERNS): Bump to 15.
5536 * tree-vect-data-refs.c: Include internal-fn.h.
5537 (vect_gather_scatter_fn_p): New function.
5538 (vect_describe_gather_scatter_call): Likewise.
5539 (vect_check_gather_scatter): Try using internal functions for
5540 gather loads. Recognize existing calls to a gather load function.
5541 (vect_analyze_data_refs): Consider using gather loads if
5542 supports_vec_gather_load_p.
5543 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5544 (vect_get_gather_scatter_offset_type): Likewise.
5545 (vect_convert_mask_for_vectype): Likewise.
5546 (vect_add_conversion_to_patterm): Likewise.
5547 (vect_try_gather_scatter_pattern): Likewise.
5548 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5549 (vect_vect_recog_func_ptrs): Add it.
5550 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5551 internal_fn_mask_index and internal_gather_scatter_fn_p.
5552 (check_load_store_masking): Take the gather_scatter_info as an
5553 argument and handle gather loads.
5554 (vect_get_gather_scatter_ops): New function.
5555 (vectorizable_call): Check internal_load_fn_p.
5556 (vectorizable_load): Likewise. Handle gather load internal
5558 (vectorizable_store): Update call to check_load_store_masking.
5559 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5560 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5561 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5562 (aarch64_gather_scale_operand_d): New predicates.
5563 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5564 (mask_gather_load<mode>): New insns.
5566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5567 Alan Hayward <alan.hayward@arm.com>
5568 David Sherwood <david.sherwood@arm.com>
5570 * optabs.def (fold_left_plus_optab): New optab.
5571 * doc/md.texi (fold_left_plus_@var{m}): Document.
5572 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5573 * internal-fn.c (fold_left_direct): Define.
5574 (expand_fold_left_optab_fn): Likewise.
5575 (direct_fold_left_optab_supported_p): Likewise.
5576 * fold-const-call.c (fold_const_fold_left): New function.
5577 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5578 * tree-parloops.c (valid_reduction_p): New function.
5579 (gather_scalar_reductions): Use it.
5580 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5581 (vect_finish_replace_stmt): Declare.
5582 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5583 (needs_fold_left_reduction_p): New function, split out from...
5584 (vect_is_simple_reduction): ...here. Accept reductions that
5585 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5586 (vect_force_simple_reduction): Also store the reduction type in
5587 the assignment's STMT_VINFO_REDUC_TYPE.
5588 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5589 (merge_with_identity): New function.
5590 (vect_expand_fold_left): Likewise.
5591 (vectorize_fold_left_reduction): Likewise.
5592 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5593 scalar phi in place for it. Check for target support and reject
5594 cases that would reassociate the operation. Defer the transform
5595 phase to vectorize_fold_left_reduction.
5596 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5597 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5598 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5602 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5603 call to ifc_temp_var.
5605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5606 Alan Hayward <alan.hayward@arm.com>
5607 David Sherwood <david.sherwood@arm.com>
5609 * target.def (legitimize_address_displacement): Take the original
5610 offset as a poly_int.
5611 * targhooks.h (default_legitimize_address_displacement): Update
5613 * targhooks.c (default_legitimize_address_displacement): Likewise.
5614 * doc/tm.texi: Regenerate.
5615 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5616 as an argument, moving assert of ad->disp == ad->disp_term to...
5617 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5618 Try calling targetm.legitimize_address_displacement before expanding
5619 the address rather than afterwards, and adjust for the new interface.
5620 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5621 Match the new hook interface. Handle SVE addresses.
5622 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5627 * Makefile.in (OBJS): Add early-remat.o.
5628 * target.def (select_early_remat_modes): New hook.
5629 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5630 * doc/tm.texi: Regenerate.
5631 * targhooks.h (default_select_early_remat_modes): Declare.
5632 * targhooks.c (default_select_early_remat_modes): New function.
5633 * timevar.def (TV_EARLY_REMAT): New timevar.
5634 * passes.def (pass_early_remat): New pass.
5635 * tree-pass.h (make_pass_early_remat): Declare.
5636 * early-remat.c: New file.
5637 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5639 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5642 Alan Hayward <alan.hayward@arm.com>
5643 David Sherwood <david.sherwood@arm.com>
5645 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5646 vfm1 with a bound_epilog parameter.
5647 (vect_do_peeling): Update calls accordingly, and move the prologue
5648 call earlier in the function. Treat the base bound_epilog as 0 for
5649 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5650 this base when peeling for gaps.
5651 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5652 with fully-masked loops.
5653 (vect_estimate_min_profitable_iters): Handle the single peeled
5654 iteration in that case.
5656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5657 Alan Hayward <alan.hayward@arm.com>
5658 David Sherwood <david.sherwood@arm.com>
5660 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5661 single-element interleaving even if the size is not a power of 2.
5662 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5663 accesses for single-element interleaving if the group size is
5666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5667 Alan Hayward <alan.hayward@arm.com>
5668 David Sherwood <david.sherwood@arm.com>
5670 * doc/md.texi (fold_extract_last_@var{m}): Document.
5671 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5672 * optabs.def (fold_extract_last_optab): New optab.
5673 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5674 * internal-fn.c (fold_extract_direct): New macro.
5675 (expand_fold_extract_optab_fn): Likewise.
5676 (direct_fold_extract_optab_supported_p): Likewise.
5677 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5678 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5679 EXTRACT_LAST_REDUCTION.
5680 (get_initial_def_for_reduction): Do not create an initial vector
5681 for EXTRACT_LAST_REDUCTION reductions.
5682 (vectorizable_reduction): Leave the scalar phi in place for
5683 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5684 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5685 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5686 transform phase to vectorizable_condition.
5687 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5689 (vect_finish_stmt_generation): ...here.
5690 (vect_finish_replace_stmt): New function.
5691 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5692 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5694 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5697 Alan Hayward <alan.hayward@arm.com>
5698 David Sherwood <david.sherwood@arm.com>
5700 * doc/md.texi (extract_last_@var{m}): Document.
5701 * optabs.def (extract_last_optab): New optab.
5702 * internal-fn.def (EXTRACT_LAST): New internal function.
5703 * internal-fn.c (cond_unary_direct): New macro.
5704 (expand_cond_unary_optab_fn): Likewise.
5705 (direct_cond_unary_optab_supported_p): Likewise.
5706 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5707 loops using EXTRACT_LAST.
5708 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5709 (extract_last_<mode>): ...this optab.
5710 (vec_extract<mode><Vel>): Update accordingly.
5712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5713 Alan Hayward <alan.hayward@arm.com>
5714 David Sherwood <david.sherwood@arm.com>
5716 * target.def (empty_mask_is_expensive): New hook.
5717 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5718 * doc/tm.texi: Regenerate.
5719 * targhooks.h (default_empty_mask_is_expensive): Declare.
5720 * targhooks.c (default_empty_mask_is_expensive): New function.
5721 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5722 if the target says that empty masks are expensive.
5723 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5725 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5728 Alan Hayward <alan.hayward@arm.com>
5729 David Sherwood <david.sherwood@arm.com>
5731 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5732 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5733 (vect_use_loop_mask_for_alignment_p): New function.
5734 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5735 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5736 niters_skip argument. Make sure that the first niters_skip elements
5737 of the first iteration are inactive.
5738 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5739 Update call to vect_set_loop_masks_directly.
5740 (get_misalign_in_elems): New function, split out from...
5741 (vect_gen_prolog_loop_niters): ...here.
5742 (vect_update_init_of_dr): Take a code argument that specifies whether
5743 the adjustment should be added or subtracted.
5744 (vect_update_init_of_drs): Likewise.
5745 (vect_prepare_for_masked_peels): New function.
5746 (vect_do_peeling): Skip prologue peeling if we're using a mask
5747 instead. Update call to vect_update_inits_of_drs.
5748 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5750 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5751 alignment. Do not include the number of peeled iterations in
5752 the minimum threshold in that case.
5753 (vectorizable_induction): Adjust the start value down by
5754 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5755 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5756 Take the number of skipped iterations into account when calculating
5758 * tree-vect-stmts.c (vect_gen_while_not): New function.
5760 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5761 Alan Hayward <alan.hayward@arm.com>
5762 David Sherwood <david.sherwood@arm.com>
5764 * doc/sourcebuild.texi (vect_fully_masked): Document.
5765 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5767 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5769 (vect_analyze_loop_2): ...here. Don't check the vectorization
5770 factor against the number of loop iterations if the loop is
5773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5774 Alan Hayward <alan.hayward@arm.com>
5775 David Sherwood <david.sherwood@arm.com>
5777 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5778 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5779 (dump_groups): Update accordingly.
5780 (iv_use::mem_type): New member variable.
5781 (address_p): New function.
5782 (record_use): Add a mem_type argument and initialize the new
5784 (record_group_use): Add a mem_type argument. Use address_p.
5785 Remove obsolete null checks of base_object. Update call to record_use.
5786 (find_interesting_uses_op): Update call to record_group_use.
5787 (find_interesting_uses_cond): Likewise.
5788 (find_interesting_uses_address): Likewise.
5789 (get_mem_type_for_internal_fn): New function.
5790 (find_address_like_use): Likewise.
5791 (find_interesting_uses_stmt): Try find_address_like_use before
5792 calling find_interesting_uses_op.
5793 (addr_offset_valid_p): Use the iv mem_type field as the type
5794 of the addressed memory.
5795 (add_autoinc_candidates): Likewise.
5796 (get_address_cost): Likewise.
5797 (split_small_address_groups_p): Use address_p.
5798 (split_address_groups): Likewise.
5799 (add_iv_candidate_for_use): Likewise.
5800 (autoinc_possible_for_pair): Likewise.
5801 (rewrite_groups): Likewise.
5802 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5803 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5804 (get_alias_ptr_type_for_ptr_address): New function.
5805 (rewrite_use_address): Rewrite address uses in calls that were
5806 identified by find_address_like_use.
5808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5809 Alan Hayward <alan.hayward@arm.com>
5810 David Sherwood <david.sherwood@arm.com>
5812 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5814 * gimple-expr.h (is_gimple_addressable: Likewise.
5815 * gimple-expr.c (is_gimple_address): Likewise.
5816 * internal-fn.c (expand_call_mem_ref): New function.
5817 (expand_mask_load_optab_fn): Use it.
5818 (expand_mask_store_optab_fn): Likewise.
5820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5821 Alan Hayward <alan.hayward@arm.com>
5822 David Sherwood <david.sherwood@arm.com>
5824 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5825 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5826 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5827 (cond_umax@var{mode}): Document.
5828 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5829 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5830 (cond_umin_optab, cond_umax_optab): New optabs.
5831 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5832 (COND_IOR, COND_XOR): New internal functions.
5833 * internal-fn.h (get_conditional_internal_fn): Declare.
5834 * internal-fn.c (cond_binary_direct): New macro.
5835 (expand_cond_binary_optab_fn): Likewise.
5836 (direct_cond_binary_optab_supported_p): Likewise.
5837 (get_conditional_internal_fn): New function.
5838 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5839 Cope with reduction statements that are vectorized as calls rather
5841 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5842 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5843 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5844 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5845 (UNSPEC_COND_EOR): New unspecs.
5846 (optab): Add mappings for them.
5847 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5848 (sve_int_op, sve_fp_op): New int attributes.
5850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5851 Alan Hayward <alan.hayward@arm.com>
5852 David Sherwood <david.sherwood@arm.com>
5854 * optabs.def (while_ult_optab): New optab.
5855 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5856 * internal-fn.def (WHILE_ULT): New internal function.
5857 * internal-fn.h (direct_internal_fn_supported_p): New override
5858 that takes two types as argument.
5859 * internal-fn.c (while_direct): New macro.
5860 (expand_while_optab_fn): New function.
5861 (convert_optab_supported_p): Likewise.
5862 (direct_while_optab_supported_p): New macro.
5863 * wide-int.h (wi::udiv_ceil): New function.
5864 * tree-vectorizer.h (rgroup_masks): New structure.
5865 (vec_loop_masks): New typedef.
5866 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5868 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5869 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5870 (vect_max_vf): New function.
5871 (slpeel_make_loop_iterate_ntimes): Delete.
5872 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5873 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5874 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5875 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5876 internal-fn.h, stor-layout.h and optabs-query.h.
5877 (vect_set_loop_mask): New function.
5878 (add_preheader_seq): Likewise.
5879 (add_header_seq): Likewise.
5880 (interleave_supported_p): Likewise.
5881 (vect_maybe_permute_loop_masks): Likewise.
5882 (vect_set_loop_masks_directly): Likewise.
5883 (vect_set_loop_condition_masked): Likewise.
5884 (vect_set_loop_condition_unmasked): New function, split out from
5885 slpeel_make_loop_iterate_ntimes.
5886 (slpeel_make_loop_iterate_ntimes): Rename to..
5887 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5888 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5889 (vect_do_peeling): Update call accordingly.
5890 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5892 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5893 mask_compare_type, can_fully_mask_p and fully_masked_p.
5894 (release_vec_loop_masks): New function.
5895 (_loop_vec_info): Use it to free the loop masks.
5896 (can_produce_all_loop_masks_p): New function.
5897 (vect_get_max_nscalars_per_iter): Likewise.
5898 (vect_verify_full_masking): Likewise.
5899 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5900 retries, and free the mask rgroups before retrying. Check loop-wide
5901 reasons for disallowing fully-masked loops. Make the final decision
5902 about whether use a fully-masked loop or not.
5903 (vect_estimate_min_profitable_iters): Do not assume that peeling
5904 for the number of iterations will be needed for fully-masked loops.
5905 (vectorizable_reduction): Disable fully-masked loops.
5906 (vectorizable_live_operation): Likewise.
5907 (vect_halve_mask_nunits): New function.
5908 (vect_double_mask_nunits): Likewise.
5909 (vect_record_loop_mask): Likewise.
5910 (vect_get_loop_mask): Likewise.
5911 (vect_transform_loop): Handle the case in which the final loop
5912 iteration might handle a partial vector. Call vect_set_loop_condition
5913 instead of slpeel_make_loop_iterate_ntimes.
5914 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5915 (check_load_store_masking): New function.
5916 (prepare_load_store_mask): Likewise.
5917 (vectorizable_store): Handle fully-masked loops.
5918 (vectorizable_load): Likewise.
5919 (supportable_widening_operation): Use vect_halve_mask_nunits for
5921 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5922 (vect_gen_while): New function.
5923 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5924 (aarch64_uqdec<mode>): New insn.
5926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5927 Alan Hayward <alan.hayward@arm.com>
5928 David Sherwood <david.sherwood@arm.com>
5930 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5931 (reduc_xor_scal_optab): New optabs.
5932 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5933 (reduc_xor_scal_@var{m}): Document.
5934 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5935 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5937 * fold-const-call.c (fold_const_call): Handle them.
5938 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5939 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5940 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5941 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5942 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5943 (UNSPEC_XORV): New unspecs.
5944 (optab): Add entries for them.
5945 (BITWISEV): New int iterator.
5946 (bit_reduc_op): New int attributes.
5948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5949 Alan Hayward <alan.hayward@arm.com>
5950 David Sherwood <david.sherwood@arm.com>
5952 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5953 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5954 * optabs.def (vec_shl_insert_optab): New optab.
5955 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5956 (duplicate_and_interleave): Likewise.
5957 * tree-vect-loop.c: Include internal-fn.h.
5958 (neutral_op_for_slp_reduction): New function, split out from
5959 get_initial_defs_for_reduction.
5960 (get_initial_def_for_reduction): Handle option 2 for variable-length
5961 vectors by loading the neutral value into a vector and then shifting
5962 the initial value into element 0.
5963 (get_initial_defs_for_reduction): Replace the code argument with
5964 the neutral value calculated by neutral_op_for_slp_reduction.
5965 Use gimple_build_vector for constant-length vectors.
5966 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5967 but the first group_size elements have a neutral value.
5968 Use duplicate_and_interleave otherwise.
5969 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5970 Update call to get_initial_defs_for_reduction. Handle SLP
5971 reductions for variable-length vectors by creating one vector
5972 result for each scalar result, with the elements associated
5973 with other scalar results stubbed out with the neutral value.
5974 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5975 Require IFN_VEC_SHL_INSERT for double reductions on
5976 variable-length vectors, or SLP reductions that have
5977 a neutral value. Require can_duplicate_and_interleave_p
5978 support for variable-length unchained SLP reductions if there
5979 is no neutral value, such as for MIN/MAX reductions. Also require
5980 the number of vector elements to be a multiple of the number of
5981 SLP statements when doing variable-length unchained SLP reductions.
5982 Update call to vect_create_epilog_for_reduction.
5983 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5984 and remove initial values.
5985 (duplicate_and_interleave): Make public.
5986 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5987 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5990 Alan Hayward <alan.hayward@arm.com>
5991 David Sherwood <david.sherwood@arm.com>
5993 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5994 (can_duplicate_and_interleave_p): New function.
5995 (vect_get_and_check_slp_defs): Take the vector of statements
5996 rather than just the current one. Remove excess parentheses.
5997 Restriction rejectinon of vect_constant_def and vect_external_def
5998 for variable-length vectors to boolean types, or types for which
5999 can_duplicate_and_interleave_p is false.
6000 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6001 (duplicate_and_interleave): New function.
6002 (vect_get_constant_vectors): Use gimple_build_vector for
6003 constant-length vectors and suitable variable-length constant
6004 vectors. Use duplicate_and_interleave for other variable-length
6005 vectors. Don't defer the update when inserting new statements.
6007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6008 Alan Hayward <alan.hayward@arm.com>
6009 David Sherwood <david.sherwood@arm.com>
6011 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6012 min_profitable_iters doesn't go negative.
6014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6015 Alan Hayward <alan.hayward@arm.com>
6016 David Sherwood <david.sherwood@arm.com>
6018 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6019 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6020 * optabs.def (vec_mask_load_lanes_optab): New optab.
6021 (vec_mask_store_lanes_optab): Likewise.
6022 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6023 (MASK_STORE_LANES): Likewise.
6024 * internal-fn.c (mask_load_lanes_direct): New macro.
6025 (mask_store_lanes_direct): Likewise.
6026 (expand_mask_load_optab_fn): Handle masked operations.
6027 (expand_mask_load_lanes_optab_fn): New macro.
6028 (expand_mask_store_optab_fn): Handle masked operations.
6029 (expand_mask_store_lanes_optab_fn): New macro.
6030 (direct_mask_load_lanes_optab_supported_p): Likewise.
6031 (direct_mask_store_lanes_optab_supported_p): Likewise.
6032 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6034 (vect_load_lanes_supported): Likewise.
6035 * tree-vect-data-refs.c (strip_conversion): New function.
6036 (can_group_stmts_p): Likewise.
6037 (vect_analyze_data_ref_accesses): Use it instead of checking
6038 for a pair of assignments.
6039 (vect_store_lanes_supported): Take a masked_p parameter.
6040 (vect_load_lanes_supported): Likewise.
6041 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6042 vect_store_lanes_supported and vect_load_lanes_supported.
6043 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6044 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6045 parameter. Don't allow gaps for masked accesses.
6046 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6047 and vect_load_lanes_supported.
6048 (get_load_store_type): Take a masked_p parameter and update
6049 call to get_group_load_store_type.
6050 (vectorizable_store): Update call to get_load_store_type.
6051 Handle IFN_MASK_STORE_LANES.
6052 (vectorizable_load): Update call to get_load_store_type.
6053 Handle IFN_MASK_LOAD_LANES.
6055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6056 Alan Hayward <alan.hayward@arm.com>
6057 David Sherwood <david.sherwood@arm.com>
6059 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6061 * config/aarch64/aarch64-protos.h
6062 (aarch64_sve_struct_memory_operand_p): Declare.
6063 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6064 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6065 (VPRED, vpred): Handle SVE structure modes.
6066 * config/aarch64/constraints.md (Utx): New constraint.
6067 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6068 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6069 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6070 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6071 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6072 structure modes. Split into pieces after RA.
6073 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6074 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6076 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6077 SVE structure modes.
6078 (aarch64_classify_address): Likewise.
6079 (sizetochar): Move earlier in file.
6080 (aarch64_print_operand): Handle SVE register lists.
6081 (aarch64_array_mode): New function.
6082 (aarch64_sve_struct_memory_operand_p): Likewise.
6083 (TARGET_ARRAY_MODE): Redefine.
6085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6086 Alan Hayward <alan.hayward@arm.com>
6087 David Sherwood <david.sherwood@arm.com>
6089 * target.def (array_mode): New target hook.
6090 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6091 * doc/tm.texi: Regenerate.
6092 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6093 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6094 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6096 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6100 Alan Hayward <alan.hayward@arm.com>
6101 David Sherwood <david.sherwood@arm.com>
6103 * fold-const.c (fold_binary_loc): Check the argument types
6104 rather than the result type when testing for a vector operation.
6106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6108 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6109 * doc/tm.texi: Regenerate.
6111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6112 Alan Hayward <alan.hayward@arm.com>
6113 David Sherwood <david.sherwood@arm.com>
6115 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6116 (sve): Document new AArch64 extension.
6117 * doc/md.texi (w): Extend the description of the AArch64
6118 constraint to include SVE vectors.
6119 (Upl, Upa): Document new AArch64 predicate constraints.
6120 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6122 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6123 (msve-vector-bits=): New option.
6124 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6125 SVE when these are disabled.
6126 (sve): New extension.
6127 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6128 modes. Adjust their number of units based on aarch64_sve_vg.
6129 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6130 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6131 aarch64_addr_query_type.
6132 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6133 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6134 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6135 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6136 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6137 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6138 (aarch64_simd_imm_zero_p): Delete.
6139 (aarch64_check_zero_based_sve_index_immediate): Declare.
6140 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6141 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6142 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6143 (aarch64_sve_float_mul_immediate_p): Likewise.
6144 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6146 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6147 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6148 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6149 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6150 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6151 (aarch64_regmode_natural_size): Likewise.
6152 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6153 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6155 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6156 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6157 for VG and the SVE predicate registers.
6158 (V_ALIASES): Add a "z"-prefixed alias.
6159 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6160 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6161 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6162 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6163 (REG_CLASS_NAMES): Add entries for them.
6164 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6165 and the predicate registers.
6166 (aarch64_sve_vg): Declare.
6167 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6168 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6169 (REGMODE_NATURAL_SIZE): Define.
6170 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6172 * config/aarch64/aarch64.c: Include cfgrtl.h.
6173 (simd_immediate_info): Add a constructor for series vectors,
6174 and an associated step field.
6175 (aarch64_sve_vg): New variable.
6176 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6177 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6178 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6179 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6180 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6181 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6182 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6183 (aarch64_get_mask_mode): New functions.
6184 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6185 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6186 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6187 predicate modes and predicate registers. Explicitly restrict
6188 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6189 to store a vector mode if it is recognized by
6190 aarch64_classify_vector_mode.
6191 (aarch64_regmode_natural_size): New function.
6192 (aarch64_hard_regno_caller_save_mode): Return the original mode
6194 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6195 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6196 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6197 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6199 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6200 does not overlap dest if the function is frame-related. Handle
6202 (aarch64_split_add_offset): New function.
6203 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6204 them aarch64_add_offset.
6205 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6206 and update call to aarch64_sub_sp.
6207 (aarch64_add_cfa_expression): New function.
6208 (aarch64_expand_prologue): Pass extra temporary registers to the
6209 functions above. Handle the case in which we need to emit new
6210 DW_CFA_expressions for registers that were originally saved
6211 relative to the stack pointer, but now have to be expressed
6212 relative to the frame pointer.
6213 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6215 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6216 IP0 and IP1 values for SVE frames.
6217 (aarch64_expand_vec_series): New function.
6218 (aarch64_expand_sve_widened_duplicate): Likewise.
6219 (aarch64_expand_sve_const_vector): Likewise.
6220 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6221 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6222 into the register, rather than emitting a SET directly.
6223 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6224 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6225 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6226 (offset_9bit_signed_scaled_p): New functions.
6227 (aarch64_replicate_bitmask_imm): New function.
6228 (aarch64_bitmask_imm): Use it.
6229 (aarch64_cannot_force_const_mem): Reject expressions involving
6230 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6231 (aarch64_classify_index): Handle SVE indices, by requiring
6232 a plain register index with a scale that matches the element size.
6233 (aarch64_classify_address): Handle SVE addresses. Assert that
6234 the mode of the address is VOIDmode or an integer mode.
6235 Update call to aarch64_classify_symbol.
6236 (aarch64_classify_symbolic_expression): Update call to
6237 aarch64_classify_symbol.
6238 (aarch64_const_vec_all_in_range_p): New function.
6239 (aarch64_print_vector_float_operand): Likewise.
6240 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6241 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6242 and the FP immediates 1.0 and 0.5.
6243 (aarch64_print_address_internal): Handle SVE addresses.
6244 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6245 (aarch64_regno_regclass): Handle predicate registers.
6246 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6248 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6249 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6250 (aarch64_convert_sve_vector_bits): New function.
6251 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6252 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6254 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6255 Handle SVE vector and predicate modes. Accept VL-based constants
6256 that need only one temporary register, and VL offsets that require
6257 no temporary registers.
6258 (aarch64_conditional_register_usage): Mark the predicate registers
6259 as fixed if SVE isn't available.
6260 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6261 Return true for SVE vector and predicate modes.
6262 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6263 rather than an unsigned int. Handle SVE modes.
6264 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6266 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6268 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6269 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6270 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6271 (aarch64_sve_float_mul_immediate_p): New functions.
6272 (aarch64_sve_valid_immediate): New function.
6273 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6274 Explicitly reject structure modes. Check for INDEX constants.
6275 Handle PTRUE and PFALSE constants.
6276 (aarch64_check_zero_based_sve_index_immediate): New function.
6277 (aarch64_simd_imm_zero_p): Delete.
6278 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6279 vector modes. Accept constants in the range of CNT[BHWD].
6280 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6281 ask for an Advanced SIMD mode.
6282 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6283 (aarch64_simd_vector_alignment): Handle SVE predicates.
6284 (aarch64_vectorize_preferred_vector_alignment): New function.
6285 (aarch64_simd_vector_alignment_reachable): Use it instead of
6287 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6288 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6290 (MAX_VECT_LEN): Delete.
6291 (expand_vec_perm_d): Add a vec_flags field.
6292 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6293 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6294 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6296 (aarch64_evpc_rev): Rename to...
6297 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6298 (aarch64_evpc_rev_global): New function.
6299 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6300 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6302 (aarch64_evpc_sve_tbl): New function.
6303 (aarch64_expand_vec_perm_const_1): Update after rename of
6304 aarch64_evpc_rev. Handle SVE permutes too, trying
6305 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6306 than aarch64_evpc_tbl.
6307 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6308 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6309 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6310 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6311 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6312 (aarch64_expand_sve_vcond): New functions.
6313 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6314 of aarch64_vector_mode_p.
6315 (aarch64_dwarf_poly_indeterminate_value): New function.
6316 (aarch64_compute_pressure_classes): Likewise.
6317 (aarch64_can_change_mode_class): Likewise.
6318 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6319 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6320 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6321 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6322 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6323 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6324 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6325 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6327 (Dn, Dl, Dr): Accept const as well as const_vector.
6328 (Dz): Likewise. Compare against CONST0_RTX.
6329 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6330 of "vector" where appropriate.
6331 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6332 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6333 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6334 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6335 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6336 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6337 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6338 (v_int_equiv): Extend to SVE modes.
6339 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6341 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6342 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6343 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6344 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6345 (SVE_COND_FP_CMP): New int iterators.
6346 (perm_hilo): Handle the new unpack unspecs.
6347 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6349 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6350 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6351 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6352 (aarch64_equality_operator, aarch64_constant_vector_operand)
6353 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6354 (aarch64_sve_nonimmediate_operand): Likewise.
6355 (aarch64_sve_general_operand): Likewise.
6356 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6357 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6358 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6359 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6360 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6361 (aarch64_sve_float_arith_immediate): Likewise.
6362 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6363 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6364 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6365 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6366 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6367 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6368 (aarch64_sve_float_arith_operand): Likewise.
6369 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6370 (aarch64_sve_float_mul_operand): Likewise.
6371 (aarch64_sve_vec_perm_operand): Likewise.
6372 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6373 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6374 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6375 as well as const_vector.
6376 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6377 in file. Use CONST0_RTX and CONSTM1_RTX.
6378 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6379 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6380 Use aarch64_simd_imm_zero.
6381 * config/aarch64/aarch64-sve.md: New file.
6382 * config/aarch64/aarch64.md: Include it.
6383 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6384 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6385 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6386 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6387 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6388 (sve): New attribute.
6389 (enabled): Disable instructions with the sve attribute unless
6391 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6392 aarch64_expand_mov_immediate.
6393 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6394 CNT[BHSD] immediates.
6395 (movti): Split CONST_POLY_INT moves into two halves.
6396 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6397 Split additions that need a temporary here if the destination
6398 is the stack pointer.
6399 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6400 (*add<mode>3_poly_1): New instruction.
6401 (set_clobber_cc): New expander.
6403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6405 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6406 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6407 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6408 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6409 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6410 Change innermode from fixed_mode_size to machine_mode.
6411 (simplify_subreg): Update call accordingly. Handle a constant-sized
6412 subreg of a variable-length CONST_VECTOR.
6414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6415 Alan Hayward <alan.hayward@arm.com>
6416 David Sherwood <david.sherwood@arm.com>
6418 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6419 (add_offset_to_base): New function, split out from...
6420 (create_mem_ref): ...here. When handling a scale other than 1,
6421 check first whether the address is valid without the offset.
6422 Add it into the base if so, leaving the index and scale as-is.
6424 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6427 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6428 fold_for_warn before checking if arg2 is INTEGER_CST.
6430 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6432 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6433 (store_multiple_operation): Delete.
6434 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6435 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6436 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6437 guarded by TARGET_STRING.
6438 (rs6000_output_load_multiple): Delete.
6439 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6440 OPTION_MASK_STRING / TARGET_STRING handling.
6441 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6442 (const rs6000_opt_masks) <"string">: Change mask to 0.
6443 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6444 (MASK_STRING): Delete.
6445 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6447 (load_multiple): Delete.
6454 (store_multiple): Delete.
6461 (movmemsi_8reg): Delete.
6462 (corresponding unnamed define_insn): Delete.
6463 (movmemsi_6reg): Delete.
6464 (corresponding unnamed define_insn): Delete.
6465 (movmemsi_4reg): Delete.
6466 (corresponding unnamed define_insn): Delete.
6467 (movmemsi_2reg): Delete.
6468 (corresponding unnamed define_insn): Delete.
6469 (movmemsi_1reg): Delete.
6470 (corresponding unnamed define_insn): Delete.
6471 * config/rs6000/rs6000.opt (mno-string): New.
6472 (mstring): Replace by deprecation warning stub.
6473 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6475 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6477 * regrename.c (regrename_do_replace): If replacing the same
6478 reg multiple times, try to reuse last created gen_raw_REG.
6481 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6482 main to workaround a bug in GDB.
6484 2018-01-12 Tom de Vries <tom@codesourcery.com>
6487 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6489 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6491 PR rtl-optimization/80481
6492 * ira-color.c (get_cap_member): New function.
6493 (allocnos_conflict_by_live_ranges_p): Use it.
6494 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6495 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6497 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6500 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6501 (*saddl_se_1): Ditto.
6503 (*ssubl_se_1): Ditto.
6505 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6507 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6508 rather than wi::to_widest for DR_INITs.
6509 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6510 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6511 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6513 (vect_analyze_group_access_1): Note that here.
6515 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6517 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6518 polynomial type sizes.
6520 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6522 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6523 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6524 (gimple_add_tmp_var): Likewise.
6526 2018-01-12 Martin Liska <mliska@suse.cz>
6528 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6529 (gimple_alloc_sizes): Likewise.
6530 (dump_gimple_statistics): Use PRIu64 in printf format.
6531 * gimple.h: Change uint64_t to int.
6533 2018-01-12 Martin Liska <mliska@suse.cz>
6535 * tree-core.h: Use uint64_t instead of int.
6536 * tree.c (tree_node_counts): Likewise.
6537 (tree_node_sizes): Likewise.
6538 (dump_tree_statistics): Use PRIu64 in printf format.
6540 2018-01-12 Martin Liska <mliska@suse.cz>
6542 * Makefile.in: As qsort_chk is implemented in vec.c, add
6543 vec.o to linkage of gencfn-macros.
6544 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6545 passing the info to record_node_allocation_statistics.
6546 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6548 * ggc-common.c (struct ggc_usage): Add operator== and use
6549 it in operator< and compare function.
6550 * mem-stats.h (struct mem_usage): Likewise.
6551 * vec.c (struct vec_usage): Remove operator< and compare
6552 function. Can be simply inherited.
6554 2018-01-12 Martin Jambor <mjambor@suse.cz>
6557 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6558 * tree-ssa-math-opts.c: Include domwalk.h.
6559 (convert_mult_to_fma_1): New function.
6560 (fma_transformation_info): New type.
6561 (fma_deferring_state): Likewise.
6562 (cancel_fma_deferring): New function.
6563 (result_of_phi): Likewise.
6564 (last_fma_candidate_feeds_initial_phi): Likewise.
6565 (convert_mult_to_fma): Added deferring logic, split actual
6566 transformation to convert_mult_to_fma_1.
6567 (math_opts_dom_walker): New type.
6568 (math_opts_dom_walker::after_dom_children): New method, body moved
6569 here from pass_optimize_widening_mul::execute, added deferring logic
6571 (pass_optimize_widening_mul::execute): Moved most of code to
6572 math_opts_dom_walker::after_dom_children.
6573 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6574 * config/i386/i386.c (ix86_option_override_internal): Added
6575 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6577 2018-01-12 Richard Biener <rguenther@suse.de>
6580 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6581 inline instance vars.
6583 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6586 * config/rx/rx.c (rx_is_restricted_memory_address):
6589 2018-01-12 Richard Biener <rguenther@suse.de>
6591 PR tree-optimization/80846
6592 * target.def (split_reduction): New target hook.
6593 * targhooks.c (default_split_reduction): New function.
6594 * targhooks.h (default_split_reduction): Declare.
6595 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6596 target requests first reduce vectors by combining low and high
6598 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6599 (get_vectype_for_scalar_type_and_size): Export.
6600 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6601 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6602 * doc/tm.texi: Regenerate.
6603 * config/i386/i386.c (ix86_split_reduction): Implement
6604 TARGET_VECTORIZE_SPLIT_REDUCTION.
6606 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6609 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6610 in PIC mode except for TARGET_VXWORKS_RTP.
6611 * config/sparc/sparc.c: Include cfgrtl.h.
6612 (TARGET_INIT_PIC_REG): Define.
6613 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6614 (sparc_pic_register_p): New predicate.
6615 (sparc_legitimate_address_p): Use it.
6616 (sparc_legitimize_pic_address): Likewise.
6617 (sparc_delegitimize_address): Likewise.
6618 (sparc_mode_dependent_address_p): Likewise.
6619 (gen_load_pcrel_sym): Remove 4th parameter.
6620 (load_got_register): Adjust call to above. Remove obsolete stuff.
6621 (sparc_expand_prologue): Do not call load_got_register here.
6622 (sparc_flat_expand_prologue): Likewise.
6623 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6624 (sparc_use_pseudo_pic_reg): New function.
6625 (sparc_init_pic_reg): Likewise.
6626 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6627 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6629 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6631 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6632 Add item for branch_cost.
6634 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6636 PR rtl-optimization/83565
6637 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6638 not extend the result to a larger mode for rotate operations.
6639 (num_sign_bit_copies1): Likewise.
6641 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6644 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6646 Use values-Xc.o for -pedantic.
6647 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6649 2018-01-12 Martin Liska <mliska@suse.cz>
6652 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6654 (possible_polymorphic_call_targets): Use it.
6655 (ipa_devirt): Likewise.
6657 2018-01-12 Martin Liska <mliska@suse.cz>
6659 * profile-count.h (enum profile_quality): Use 0 as invalid
6660 enum value of profile_quality.
6662 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6664 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6665 -mext-string options.
6667 2018-01-12 Richard Biener <rguenther@suse.de>
6669 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6670 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6671 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6673 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6675 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6677 * configure.ac (--with-long-double-format): Add support for the
6678 configuration option to change the default long double format on
6680 * config.gcc (powerpc*-linux*-*): Likewise.
6681 * configure: Regenerate.
6682 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6683 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6684 used without modification.
6686 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6688 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6689 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6690 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6691 MISC_BUILTIN_SPEC_BARRIER.
6692 (rs6000_init_builtins): Likewise.
6693 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6695 (speculation_barrier): New define_insn.
6696 * doc/extend.texi: Document __builtin_speculation_barrier.
6698 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6701 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6702 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6703 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6705 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6706 integral modes instead of "ss" and "sd".
6707 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6708 vectors with 32-bit and 64-bit elements.
6709 (vecdupssescalarmodesuffix): New mode attribute.
6710 (vec_dup<mode>): Use it.
6712 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6715 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6716 frame if argument is passed on stack.
6718 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6721 * ree.c (combine_reaching_defs): Optimize also
6722 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6723 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6725 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6728 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6730 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6733 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6734 after they are computed.
6736 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6738 PR tree-optimization/83695
6739 * gimple-loop-linterchange.cc
6740 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6741 reset cached scev information after interchange.
6742 (pass_linterchange::execute): Remove call to scev_reset_htab.
6744 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6746 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6747 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6748 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6749 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6750 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6751 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6752 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6753 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6754 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6755 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6756 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6757 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6758 (V_lane_reg): Likewise.
6759 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6761 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6762 (vfmal_lane_low<mode>_intrinsic,
6763 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6764 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6765 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6766 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6767 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6768 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6770 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6772 * config/arm/arm-cpus.in (fp16fml): New feature.
6773 (ALL_SIMD): Add fp16fml.
6774 (armv8.2-a): Add fp16fml as an option.
6775 (armv8.3-a): Likewise.
6776 (armv8.4-a): Add fp16fml as part of fp16.
6777 * config/arm/arm.h (TARGET_FP16FML): Define.
6778 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6780 * config/arm/arm-modes.def (V2HF): Define.
6781 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6782 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6783 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6784 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6785 vfmsl_low, vfmsl_high): New set of builtins.
6786 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6787 (vfml_op): New code attribute.
6788 (VFMLHALVES): New int iterator.
6789 (VFML, VFMLSEL): New mode attributes.
6790 (V_reg): Define mapping for V2HF.
6791 (V_hi, V_lo): New mode attributes.
6792 (VF_constraint): Likewise.
6793 (vfml_half, vfml_half_selector): New int attributes.
6794 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6796 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6797 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6799 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6800 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6801 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6802 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6804 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6805 Document new effective target and option set.
6807 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6809 * config/arm/arm-cpus.in (armv8_4): New feature.
6810 (ARMv8_4a): New fgroup.
6811 (armv8.4-a): New arch.
6812 * config/arm/arm-tables.opt: Regenerate.
6813 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6814 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6815 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6816 Add matching rules for -march=armv8.4-a and extensions.
6817 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6819 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6822 * config/rx/rx.md (BW): New mode attribute.
6823 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6825 2018-01-11 Richard Biener <rguenther@suse.de>
6827 PR tree-optimization/83435
6828 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6829 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6830 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6832 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6833 Alan Hayward <alan.hayward@arm.com>
6834 David Sherwood <david.sherwood@arm.com>
6836 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6838 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6839 (aarch64_print_address_internal): Use it to check for a zero offset.
6841 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6842 Alan Hayward <alan.hayward@arm.com>
6843 David Sherwood <david.sherwood@arm.com>
6845 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6846 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6847 Return a poly_int64 rather than a HOST_WIDE_INT.
6848 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6849 rather than a HOST_WIDE_INT.
6850 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6851 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6852 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6853 final_offset from HOST_WIDE_INT to poly_int64.
6854 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6855 to_constant when getting the number of units in an Advanced SIMD
6857 (aarch64_builtin_vectorized_function): Check for a constant number
6859 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6861 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6862 attribute instead of GET_MODE_NUNITS.
6863 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6864 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6865 GET_MODE_SIZE for fixed-size registers.
6866 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6867 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6868 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6869 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6870 (aarch64_print_operand, aarch64_print_address_internal)
6871 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6872 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6873 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6874 Handle polynomial GET_MODE_SIZE.
6875 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6876 wider than SImode without modification.
6877 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6878 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6879 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6880 passing and returning SVE modes.
6881 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6882 rather than GEN_INT.
6883 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6884 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6885 (aarch64_allocate_and_probe_stack_space): Likewise.
6886 (aarch64_layout_frame): Cope with polynomial offsets.
6887 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6888 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6890 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6891 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6892 poly_int64 rather than a HOST_WIDE_INT.
6893 (aarch64_get_separate_components, aarch64_process_components)
6894 (aarch64_expand_prologue, aarch64_expand_epilogue)
6895 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6896 (aarch64_anchor_offset): New function, split out from...
6897 (aarch64_legitimize_address): ...here.
6898 (aarch64_builtin_vectorization_cost): Handle polynomial
6899 TYPE_VECTOR_SUBPARTS.
6900 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6902 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6903 number of elements from the PARALLEL rather than the mode.
6904 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6905 rather than GET_MODE_BITSIZE.
6906 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6907 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6908 (aarch64_expand_vec_perm_const_1): Handle polynomial
6909 d->perm.length () and d->perm elements.
6910 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6911 Apply to_constant to d->perm elements.
6912 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6913 polynomial CONST_VECTOR_NUNITS.
6914 (aarch64_move_pointer): Take amount as a poly_int64 rather
6916 (aarch64_progress_pointer): Avoid temporary variable.
6917 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6918 the mode attribute instead of GET_MODE.
6920 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6921 Alan Hayward <alan.hayward@arm.com>
6922 David Sherwood <david.sherwood@arm.com>
6924 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6925 x exists before using it.
6926 (aarch64_add_constant_internal): Rename to...
6927 (aarch64_add_offset_1): ...this. Replace regnum with separate
6928 src and dest rtxes. Handle the case in which they're different,
6929 including when the offset is zero. Replace scratchreg with an rtx.
6930 Use 2 additions if there is no spare register into which we can
6931 move a 16-bit constant.
6932 (aarch64_add_constant): Delete.
6933 (aarch64_add_offset): Replace reg with separate src and dest
6934 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6935 Use aarch64_add_offset_1.
6936 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6937 an rtx rather than an int. Take the delta as a poly_int64
6938 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6939 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6940 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6941 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6942 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6944 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6945 aarch64_add_constant.
6947 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6949 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6950 Use scalar_float_mode.
6952 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6954 * config/aarch64/aarch64-simd.md
6955 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6956 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6957 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6958 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6959 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6960 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6961 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6962 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6963 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6964 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6966 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6969 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6970 targ_options->x_arm_arch_string is non NULL.
6972 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6974 * config/aarch64/aarch64.h
6975 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6977 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6980 * expmed.c (emit_store_flag_force): Swap if const op0
6981 and change VOIDmode to mode of op0.
6983 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6985 PR rtl-optimization/83761
6986 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6987 than bytes to mode_for_size.
6989 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6992 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6993 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6996 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6999 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7000 when in layout mode.
7001 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7002 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7005 2018-01-10 Michael Collison <michael.collison@arm.com>
7007 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7008 * config/aarch64/aarch64-option-extension.def: Add
7009 AARCH64_OPT_EXTENSION of 'fp16fml'.
7010 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7011 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7012 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7013 * config/aarch64/constraints.md (Ui7): New constraint.
7014 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7015 (VFMLA_SEL_W): Ditto.
7018 (VFMLA16_LOW): New int iterator.
7019 (VFMLA16_HIGH): Ditto.
7020 (UNSPEC_FMLAL): New unspec.
7021 (UNSPEC_FMLSL): Ditto.
7022 (UNSPEC_FMLAL2): Ditto.
7023 (UNSPEC_FMLSL2): Ditto.
7024 (f16mac): New code attribute.
7025 * config/aarch64/aarch64-simd-builtins.def
7026 (aarch64_fmlal_lowv2sf): Ditto.
7027 (aarch64_fmlsl_lowv2sf): Ditto.
7028 (aarch64_fmlalq_lowv4sf): Ditto.
7029 (aarch64_fmlslq_lowv4sf): Ditto.
7030 (aarch64_fmlal_highv2sf): Ditto.
7031 (aarch64_fmlsl_highv2sf): Ditto.
7032 (aarch64_fmlalq_highv4sf): Ditto.
7033 (aarch64_fmlslq_highv4sf): Ditto.
7034 (aarch64_fmlal_lane_lowv2sf): Ditto.
7035 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7036 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7037 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7038 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7039 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7040 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7041 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7042 (aarch64_fmlal_lane_highv2sf): Ditto.
7043 (aarch64_fmlsl_lane_highv2sf): Ditto.
7044 (aarch64_fmlal_laneq_highv2sf): Ditto.
7045 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7046 (aarch64_fmlalq_lane_highv4sf): Ditto.
7047 (aarch64_fmlsl_lane_highv4sf): Ditto.
7048 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7049 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7050 * config/aarch64/aarch64-simd.md:
7051 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7052 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7053 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7054 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7055 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7056 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7057 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7058 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7059 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7060 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7061 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7062 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7063 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7064 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7065 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7066 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7067 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7068 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7069 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7070 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7071 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7072 (vfmlsl_low_u32): Ditto.
7073 (vfmlalq_low_u32): Ditto.
7074 (vfmlslq_low_u32): Ditto.
7075 (vfmlal_high_u32): Ditto.
7076 (vfmlsl_high_u32): Ditto.
7077 (vfmlalq_high_u32): Ditto.
7078 (vfmlslq_high_u32): Ditto.
7079 (vfmlal_lane_low_u32): Ditto.
7080 (vfmlsl_lane_low_u32): Ditto.
7081 (vfmlal_laneq_low_u32): Ditto.
7082 (vfmlsl_laneq_low_u32): Ditto.
7083 (vfmlalq_lane_low_u32): Ditto.
7084 (vfmlslq_lane_low_u32): Ditto.
7085 (vfmlalq_laneq_low_u32): Ditto.
7086 (vfmlslq_laneq_low_u32): Ditto.
7087 (vfmlal_lane_high_u32): Ditto.
7088 (vfmlsl_lane_high_u32): Ditto.
7089 (vfmlal_laneq_high_u32): Ditto.
7090 (vfmlsl_laneq_high_u32): Ditto.
7091 (vfmlalq_lane_high_u32): Ditto.
7092 (vfmlslq_lane_high_u32): Ditto.
7093 (vfmlalq_laneq_high_u32): Ditto.
7094 (vfmlslq_laneq_high_u32): Ditto.
7095 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7096 (AARCH64_FL_FOR_ARCH8_4): New.
7097 (AARCH64_ISA_F16FML): New ISA flag.
7098 (TARGET_F16FML): New feature flag for fp16fml.
7099 (doc/invoke.texi): Document new fp16fml option.
7101 2018-01-10 Michael Collison <michael.collison@arm.com>
7103 * config/aarch64/aarch64-builtins.c:
7104 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7105 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7106 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7107 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7108 (AARCH64_ISA_SHA3): New ISA flag.
7109 (TARGET_SHA3): New feature flag for sha3.
7110 * config/aarch64/iterators.md (sha512_op): New int attribute.
7111 (CRYPTO_SHA512): New int iterator.
7112 (UNSPEC_SHA512H): New unspec.
7113 (UNSPEC_SHA512H2): Ditto.
7114 (UNSPEC_SHA512SU0): Ditto.
7115 (UNSPEC_SHA512SU1): Ditto.
7116 * config/aarch64/aarch64-simd-builtins.def
7117 (aarch64_crypto_sha512hqv2di): New builtin.
7118 (aarch64_crypto_sha512h2qv2di): Ditto.
7119 (aarch64_crypto_sha512su0qv2di): Ditto.
7120 (aarch64_crypto_sha512su1qv2di): Ditto.
7121 (aarch64_eor3qv8hi): Ditto.
7122 (aarch64_rax1qv2di): Ditto.
7123 (aarch64_xarqv2di): Ditto.
7124 (aarch64_bcaxqv8hi): Ditto.
7125 * config/aarch64/aarch64-simd.md:
7126 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7127 (aarch64_crypto_sha512su0qv2di): Ditto.
7128 (aarch64_crypto_sha512su1qv2di): Ditto.
7129 (aarch64_eor3qv8hi): Ditto.
7130 (aarch64_rax1qv2di): Ditto.
7131 (aarch64_xarqv2di): Ditto.
7132 (aarch64_bcaxqv8hi): Ditto.
7133 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7134 (vsha512h2q_u64): Ditto.
7135 (vsha512su0q_u64): Ditto.
7136 (vsha512su1q_u64): Ditto.
7137 (veor3q_u16): Ditto.
7138 (vrax1q_u64): Ditto.
7140 (vbcaxq_u16): Ditto.
7141 * config/arm/types.md (crypto_sha512): New type attribute.
7142 (crypto_sha3): Ditto.
7143 (doc/invoke.texi): Document new sha3 option.
7145 2018-01-10 Michael Collison <michael.collison@arm.com>
7147 * config/aarch64/aarch64-builtins.c:
7148 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7149 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7150 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7151 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7152 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7153 (AARCH64_ISA_SM4): New ISA flag.
7154 (TARGET_SM4): New feature flag for sm4.
7155 * config/aarch64/aarch64-simd-builtins.def
7156 (aarch64_sm3ss1qv4si): Ditto.
7157 (aarch64_sm3tt1aq4si): Ditto.
7158 (aarch64_sm3tt1bq4si): Ditto.
7159 (aarch64_sm3tt2aq4si): Ditto.
7160 (aarch64_sm3tt2bq4si): Ditto.
7161 (aarch64_sm3partw1qv4si): Ditto.
7162 (aarch64_sm3partw2qv4si): Ditto.
7163 (aarch64_sm4eqv4si): Ditto.
7164 (aarch64_sm4ekeyqv4si): Ditto.
7165 * config/aarch64/aarch64-simd.md:
7166 (aarch64_sm3ss1qv4si): Ditto.
7167 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7168 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7169 (aarch64_sm4eqv4si): Ditto.
7170 (aarch64_sm4ekeyqv4si): Ditto.
7171 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7172 (sm3part_op): Ditto.
7173 (CRYPTO_SM3TT): Ditto.
7174 (CRYPTO_SM3PART): Ditto.
7175 (UNSPEC_SM3SS1): New unspec.
7176 (UNSPEC_SM3TT1A): Ditto.
7177 (UNSPEC_SM3TT1B): Ditto.
7178 (UNSPEC_SM3TT2A): Ditto.
7179 (UNSPEC_SM3TT2B): Ditto.
7180 (UNSPEC_SM3PARTW1): Ditto.
7181 (UNSPEC_SM3PARTW2): Ditto.
7182 (UNSPEC_SM4E): Ditto.
7183 (UNSPEC_SM4EKEY): Ditto.
7184 * config/aarch64/constraints.md (Ui2): New constraint.
7185 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7186 * config/arm/types.md (crypto_sm3): New type attribute.
7187 (crypto_sm4): Ditto.
7188 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7189 (vsm3tt1aq_u32): Ditto.
7190 (vsm3tt1bq_u32): Ditto.
7191 (vsm3tt2aq_u32): Ditto.
7192 (vsm3tt2bq_u32): Ditto.
7193 (vsm3partw1q_u32): Ditto.
7194 (vsm3partw2q_u32): Ditto.
7195 (vsm4eq_u32): Ditto.
7196 (vsm4ekeyq_u32): Ditto.
7197 (doc/invoke.texi): Document new sm4 option.
7199 2018-01-10 Michael Collison <michael.collison@arm.com>
7201 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7202 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7203 (AARCH64_FL_FOR_ARCH8_4): New.
7204 (AARCH64_FL_V8_4): New flag.
7205 (doc/invoke.texi): Document new armv8.4-a option.
7207 2018-01-10 Michael Collison <michael.collison@arm.com>
7209 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7210 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7211 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7212 * config/aarch64/aarch64-option-extension.def: Add
7213 AARCH64_OPT_EXTENSION of 'sha2'.
7214 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7215 (crypto): Disable sha2 and aes if crypto disabled.
7216 (crypto): Enable aes and sha2 if enabled.
7217 (simd): Disable sha2 and aes if simd disabled.
7218 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7220 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7221 (TARGET_SHA2): New feature flag for sha2.
7222 (TARGET_AES): New feature flag for aes.
7223 * config/aarch64/aarch64-simd.md:
7224 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7225 conditional on TARGET_AES.
7226 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7227 (aarch64_crypto_sha1hsi): Make pattern conditional
7229 (aarch64_crypto_sha1hv4si): Ditto.
7230 (aarch64_be_crypto_sha1hv4si): Ditto.
7231 (aarch64_crypto_sha1su1v4si): Ditto.
7232 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7233 (aarch64_crypto_sha1su0v4si): Ditto.
7234 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7235 (aarch64_crypto_sha256su0v4si): Ditto.
7236 (aarch64_crypto_sha256su1v4si): Ditto.
7237 (doc/invoke.texi): Document new aes and sha2 options.
7239 2018-01-10 Martin Sebor <msebor@redhat.com>
7241 PR tree-optimization/83781
7242 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7245 2018-01-11 Martin Sebor <msebor@gmail.com>
7246 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7248 PR tree-optimization/83501
7249 PR tree-optimization/81703
7251 * tree-ssa-strlen.c (get_string_cst): Rename...
7252 (get_string_len): ...to this. Handle global constants.
7253 (handle_char_store): Adjust.
7255 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7256 Jim Wilson <jimw@sifive.com>
7258 * config/riscv/riscv-protos.h (riscv_output_return): New.
7259 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7260 (riscv_attribute_table, riscv_output_return),
7261 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7262 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7263 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7265 (riscv_expand_prologue): Add early return for naked function.
7266 (riscv_expand_epilogue): Likewise.
7267 (riscv_function_ok_for_sibcall): Return false for naked function.
7268 (riscv_set_current_function): New.
7269 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7270 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7271 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7272 * doc/extend.texi (RISC-V Function Attributes): New.
7274 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7276 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7277 check for 128-bit long double before checking TCmode.
7278 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7279 128-bit long doubles before checking TFmode or TCmode.
7280 (FLOAT128_IBM_P): Likewise.
7282 2018-01-10 Martin Sebor <msebor@redhat.com>
7284 PR tree-optimization/83671
7285 * builtins.c (c_strlen): Unconditionally return zero for the empty
7287 Use -Warray-bounds for warnings.
7288 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7289 for non-constant array indices with COMPONENT_REF, arrays of
7290 arrays, and pointers to arrays.
7291 (gimple_fold_builtin_strlen): Determine and set length range for
7292 non-constant character arrays.
7294 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7297 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7300 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7302 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7304 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7307 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7308 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7309 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7310 indexed_or_indirect_operand predicate.
7311 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7312 (*vsx_le_perm_load_v8hi): Likewise.
7313 (*vsx_le_perm_load_v16qi): Likewise.
7314 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7315 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7316 (*vsx_le_perm_store_v8hi): Likewise.
7317 (*vsx_le_perm_store_v16qi): Likewise.
7318 (eight unnamed splitters): Likewise.
7320 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7322 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7323 * config/rs6000/emmintrin.h: Likewise.
7324 * config/rs6000/mmintrin.h: Likewise.
7325 * config/rs6000/xmmintrin.h: Likewise.
7327 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7330 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7332 * tree.c (tree_nop_conversion): Return true for location wrapper
7334 (maybe_wrap_with_location): New function.
7335 (selftest::check_strip_nops): New function.
7336 (selftest::test_location_wrappers): New function.
7337 (selftest::tree_c_tests): Call it.
7338 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7339 (maybe_wrap_with_location): New decl.
7340 (EXPR_LOCATION_WRAPPER_P): New macro.
7341 (location_wrapper_p): New inline function.
7342 (tree_strip_any_location_wrapper): New inline function.
7344 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7347 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7348 stack_realign_offset for the largest alignment of stack slot
7350 (ix86_find_max_used_stack_alignment): New function.
7351 (ix86_finalize_stack_frame_flags): Use it. Set
7352 max_used_stack_alignment if we don't realign stack.
7353 * config/i386/i386.h (machine_function): Add
7354 max_used_stack_alignment.
7356 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7358 * config/arm/arm.opt (-mbranch-cost): New option.
7359 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7362 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7365 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7366 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7368 2018-01-10 Richard Biener <rguenther@suse.de>
7371 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7372 early out so it also covers the case where we have a non-NULL
7375 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7377 PR tree-optimization/83753
7378 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7379 for non-strided grouped accesses if the number of elements is 1.
7381 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7384 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7385 * i386.h (TARGET_USE_GATHER): Define.
7386 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7388 2018-01-10 Martin Liska <mliska@suse.cz>
7391 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7392 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7394 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7395 CLEANUP_NO_PARTITIONING is not set.
7397 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7399 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7400 for vectors, as a partial revert of r254296.
7401 * rtl.h (const_vec_p): Delete.
7402 (const_vec_duplicate_p): Don't test for vector CONSTs.
7403 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7404 * expmed.c (make_tree): Likewise.
7407 * common.md (E, F): Use CONSTANT_P instead of checking for
7409 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7410 checking for CONST_VECTOR.
7412 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7415 * predict.c (force_edge_cold): Handle in more sane way edges
7418 2018-01-09 Carl Love <cel@us.ibm.com>
7420 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7422 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7423 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7424 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7425 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7426 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7427 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7428 * config/rs6000/rs6000-protos.h: Add extern defition for
7429 rs6000_generate_float2_double_code.
7430 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7432 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7433 (float2_v2df): Add define_expand.
7435 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7438 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7439 op_mode in the force_to_mode call.
7441 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7443 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7444 instead of checking each element individually.
7445 (aarch64_evpc_uzp): Likewise.
7446 (aarch64_evpc_zip): Likewise.
7447 (aarch64_evpc_ext): Likewise.
7448 (aarch64_evpc_rev): Likewise.
7449 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7450 instead of checking each element individually. Return true without
7452 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7453 whether all selected elements come from the same input, instead of
7454 checking each element individually. Remove calls to gen_rtx_REG,
7455 start_sequence and end_sequence and instead assert that no rtl is
7458 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7460 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7461 order of HIGH and CONST checks.
7463 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7465 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7466 if the destination isn't an SSA_NAME.
7468 2018-01-09 Richard Biener <rguenther@suse.de>
7470 PR tree-optimization/83668
7471 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7473 (canonicalize_loop_form): ... here, renamed from ...
7474 (canonicalize_loop_closed_ssa_form): ... this and amended to
7475 swap successor edges for loop exit blocks to make us use
7476 the RPO order we need for initial schedule generation.
7478 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7480 PR tree-optimization/64811
7481 * match.pd: When optimizing comparisons with Inf, avoid
7482 introducing or losing exceptions from comparisons with NaN.
7484 2018-01-09 Martin Liska <mliska@suse.cz>
7487 * asan.c (shadow_mem_size): Add gcc_assert.
7489 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7491 Don't save registers in main().
7494 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7495 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7496 * config/avr/avr.c (avr_set_current_function): Don't error if
7497 naked, OS_task or OS_main are specified at the same time.
7498 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7500 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7502 * common/config/avr/avr-common.c (avr_option_optimization_table):
7503 Switch on -mmain-is-OS_task for optimizing compilations.
7505 2018-01-09 Richard Biener <rguenther@suse.de>
7507 PR tree-optimization/83572
7508 * graphite.c: Include cfganal.h.
7509 (graphite_transform_loops): Connect infinite loops to exit
7510 and remove fake edges at the end.
7512 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7514 * ipa-inline.c (edge_badness): Revert accidental checkin.
7516 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7519 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7520 symbols; not inline clones.
7522 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7525 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7526 hard registers. Formatting fixes.
7528 PR preprocessor/83722
7529 * gcc.c (try_generate_repro): Pass
7530 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7531 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7534 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7535 Kito Cheng <kito.cheng@gmail.com>
7537 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7538 (riscv_leaf_function_p): Delete.
7539 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7541 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7543 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7545 (do_ifelse): New function.
7546 (do_isel): New function.
7547 (do_sub3): New function.
7548 (do_add3): New function.
7549 (do_load_mask_compare): New function.
7550 (do_overlap_load_compare): New function.
7551 (expand_compare_loop): New function.
7552 (expand_block_compare): Call expand_compare_loop() when appropriate.
7553 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7555 (-mblock-compare-inline-loop-limit): New option.
7557 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7560 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7561 Reverse order of second and third operands in first alternative.
7562 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7563 of first and second elements in UNSPEC_VPERMR vector.
7564 (altivec_expand_vec_perm_le): Likewise.
7566 2018-01-08 Jeff Law <law@redhat.com>
7568 PR rtl-optimizatin/81308
7569 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7570 (process_switch): If group_case_labels makes a change, then set
7572 (pass_convert_switch::execute): If a switch is converted, then
7573 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7575 PR rtl-optimization/81308
7576 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7579 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7581 PR target/83663 - Revert r255946
7582 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7583 generation for cases where splatting a value is not useful.
7584 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7585 across a vec_duplicate and a paradoxical subreg forming a vector
7586 mode to a vec_concat.
7588 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7590 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7591 -march=armv8.3-a variants.
7592 * config/arm/t-multilib: Likewise.
7593 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7595 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7597 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7599 (cceq_ior_compare_complement): Give it a name so I can use it, and
7600 change boolean_or_operator predicate to boolean_operator so it can
7601 be used to generate a crand.
7602 (eqne): New code iterator.
7603 (bd/bd_neg): New code_attrs.
7604 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7605 a single define_insn.
7606 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7607 decrement (bdnzt/bdnzf/bdzt/bdzf).
7608 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7609 with the new names of the branch decrement patterns, and added the
7610 names of the branch decrement conditional patterns.
7612 2018-01-08 Richard Biener <rguenther@suse.de>
7614 PR tree-optimization/83563
7615 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7618 2018-01-08 Richard Biener <rguenther@suse.de>
7621 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7623 2018-01-08 Richard Biener <rguenther@suse.de>
7625 PR tree-optimization/83685
7626 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7627 references to abnormals.
7629 2018-01-08 Richard Biener <rguenther@suse.de>
7632 * dwarf2out.c (output_indirect_strings): Handle empty
7633 skeleton_debug_str_hash.
7634 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7636 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7638 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7639 (emit_store_direct): Likewise.
7640 (arc_trampoline_adjust_address): Likewise.
7641 (arc_asm_trampoline_template): New function.
7642 (arc_initialize_trampoline): Use asm_trampoline_template.
7643 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7644 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7645 * config/arc/arc.md (flush_icache): Delete pattern.
7647 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7649 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7650 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7653 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7656 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7657 by not USED_FOR_TARGET.
7658 (make_pass_resolve_sw_modes): Likewise.
7660 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7662 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7665 2018-01-08 Richard Biener <rguenther@suse.de>
7668 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7670 2018-01-08 Richard Biener <rguenther@suse.de>
7673 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7675 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7678 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7679 basic blocks with a small number of successors.
7680 (convert_control_dep_chain_into_preds): Improve handling of
7682 (dump_predicates): Split apart into...
7683 (dump_pred_chain): ...here...
7684 (dump_pred_info): ...and here.
7685 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7686 (can_chain_union_be_invalidated_p): Improve check for invalidation
7688 (uninit_uses_cannot_happen): Avoid unnecessary if
7689 convert_control_dep_chain_into_preds yielded nothing.
7691 2018-01-06 Martin Sebor <msebor@redhat.com>
7693 PR tree-optimization/83640
7694 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7695 subtracting negative offset from size.
7696 (builtin_access::overlap): Adjust offset bounds of the access to fall
7697 within the size of the object if possible.
7699 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7701 PR rtl-optimization/83699
7702 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7703 extract_bit_field_as_subreg to cases in which the extracted
7704 value is also a vector.
7706 * lra-constraints.c (process_alt_operands): Test for the equivalence
7707 substitutions when detecting a possible reload cycle.
7709 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7712 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7713 by default if flag_selective_schedling{,2}. Formatting fixes.
7715 PR rtl-optimization/83682
7716 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7717 if it has non-VECTOR_MODE element mode.
7718 (vec_duplicate_p): Likewise.
7721 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7722 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7724 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7727 * config/i386/i386-builtin.def
7728 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7729 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7730 Require also OPTION_MASK_ISA_AVX512F in addition to
7731 OPTION_MASK_ISA_GFNI.
7732 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7733 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7734 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7735 to OPTION_MASK_ISA_GFNI.
7736 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7737 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7738 OPTION_MASK_ISA_AVX512BW.
7739 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7740 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7741 addition to OPTION_MASK_ISA_GFNI.
7742 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7743 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7744 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7745 to OPTION_MASK_ISA_GFNI.
7746 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7747 a requirement for all ISAs rather than any of them with a few
7749 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7751 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7752 bitmasks to be enabled with 3 exceptions, instead of requiring any
7753 enabled ISA with lots of exceptions.
7754 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7755 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7756 Change avx512bw in isa attribute to avx512f.
7757 * config/i386/sgxintrin.h: Add license boilerplate.
7758 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7759 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7760 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7761 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7763 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7764 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7765 temporarily sse2 rather than sse if not enabled already.
7768 * config/i386/sse.md (VI248_VLBW): Rename to ...
7769 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7770 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7771 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7772 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7773 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7774 mode iterator instead of VI248_VLBW.
7776 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7778 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7779 (record_modified): Skip clobbers; add debug output.
7780 (param_change_prob): Use sreal frequencies.
7782 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7784 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7785 punt for user-aligned variables.
7787 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7789 * tree-chrec.c (chrec_contains_symbols): Return true for
7792 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7795 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7796 of (x|y) == x for BICS pattern.
7798 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7800 PR tree-optimization/83605
7801 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7802 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7805 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7807 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7808 * config/epiphany/rtems.h: New file.
7810 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7811 Uros Bizjak <ubizjak@gmail.com>
7814 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7815 QIreg_operand instead of register_operand predicate.
7816 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7817 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7818 comments instead of -fmitigate[-_]rop.
7820 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7823 * cgraphunit.c (symbol_table::compile): Switch to text_section
7824 before calling assembly_start debug hook.
7825 * run-rtl-passes.c (run_rtl_passes): Likewise.
7828 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7830 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7831 range_int_cst_p rather than !symbolic_range_p before calling
7832 extract_range_from_multiplicative_op_1.
7834 2018-01-04 Jeff Law <law@redhat.com>
7836 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7837 redundant test in assertion.
7839 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7841 * doc/rtl.texi: Document machine_mode wrapper classes.
7843 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7845 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7848 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7850 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7851 the VEC_PERM_EXPR fold to fail.
7853 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7856 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7857 to switched_sections.
7859 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7862 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7865 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7868 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7869 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7871 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7874 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7875 is BLKmode and bitpos not zero or mode change is needed.
7877 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7880 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7883 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7886 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7887 instead of MULT rtx. Update all corresponding splitters.
7889 (*ssub<modesuffix>): Ditto.
7891 (*cmp_sadd_di): Update split patterns.
7892 (*cmp_sadd_si): Ditto.
7893 (*cmp_sadd_sidi): Ditto.
7894 (*cmp_ssub_di): Ditto.
7895 (*cmp_ssub_si): Ditto.
7896 (*cmp_ssub_sidi): Ditto.
7897 * config/alpha/predicates.md (const23_operand): New predicate.
7898 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7899 Look for ASHIFT, not MULT inner operand.
7900 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7902 2018-01-04 Martin Liska <mliska@suse.cz>
7904 PR gcov-profile/83669
7905 * gcov.c (output_intermediate_file): Add version to intermediate
7907 * doc/gcov.texi: Document new field 'version' in intermediate
7908 file format. Fix location of '-k' option of gcov command.
7910 2018-01-04 Martin Liska <mliska@suse.cz>
7913 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7915 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7917 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7919 2018-01-03 Martin Sebor <msebor@redhat.com>
7921 PR tree-optimization/83655
7922 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7923 checking calls with invalid arguments.
7925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7927 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7928 (vectorizable_mask_load_store): Delete.
7929 (vectorizable_call): Return false for masked loads and stores.
7930 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7931 instead of gimple_assign_rhs1.
7932 (vectorizable_load): Handle IFN_MASK_LOAD.
7933 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7937 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7939 (vectorizable_mask_load_store): ...here.
7940 (vectorizable_load): ...and here.
7942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7944 * tree-vect-stmts.c (vect_build_all_ones_mask)
7945 (vect_build_zero_merge_argument): New functions, split out from...
7946 (vectorizable_load): ...here.
7948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7950 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7952 (vectorizable_mask_load_store): ...here.
7953 (vectorizable_store): ...and here.
7955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7957 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7959 (vectorizable_mask_load_store): ...here.
7961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7963 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7964 (vect_model_store_cost): Take a vec_load_store_type instead of a
7966 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7967 (vect_model_store_cost): Take a vec_load_store_type instead of a
7969 (vectorizable_mask_load_store): Update accordingly.
7970 (vectorizable_store): Likewise.
7971 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7975 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7976 IFN_MASK_LOAD calls here rather than...
7977 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7980 Alan Hayward <alan.hayward@arm.com>
7981 David Sherwood <david.sherwood@arm.com>
7983 * expmed.c (extract_bit_field_1): For vector extracts,
7984 fall back to extract_bit_field_as_subreg if vec_extract
7987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7988 Alan Hayward <alan.hayward@arm.com>
7989 David Sherwood <david.sherwood@arm.com>
7991 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7992 they are variable or constant sized.
7993 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7994 slots for constant-sized data.
7996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7997 Alan Hayward <alan.hayward@arm.com>
7998 David Sherwood <david.sherwood@arm.com>
8000 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8001 handling COND_EXPRs with boolean comparisons, try to find a better
8002 basis for the mask type than the boolean itself.
8004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8006 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8007 is calculated and how it can be overridden.
8008 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8009 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8011 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8015 Alan Hayward <alan.hayward@arm.com>
8016 David Sherwood <david.sherwood@arm.com>
8018 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8019 Remove the mode argument.
8020 (aarch64_simd_valid_immediate): Remove the mode and inverse
8022 * config/aarch64/iterators.md (bitsize): New iterator.
8023 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8024 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8025 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8026 aarch64_simd_valid_immediate.
8027 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8028 (aarch64_reg_or_bic_imm): Likewise.
8029 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8030 with an insn_type enum and msl with a modifier_type enum.
8031 Replace element_width with a scalar_mode. Change the shift
8032 to unsigned int. Add constructors for scalar_float_mode and
8033 scalar_int_mode elements.
8034 (aarch64_vect_float_const_representable_p): Delete.
8035 (aarch64_can_const_movi_rtx_p)
8036 (aarch64_simd_scalar_immediate_valid_for_move)
8037 (aarch64_simd_make_constant): Update call to
8038 aarch64_simd_valid_immediate.
8039 (aarch64_advsimd_valid_immediate_hs): New function.
8040 (aarch64_advsimd_valid_immediate): Likewise.
8041 (aarch64_simd_valid_immediate): Remove mode and inverse
8042 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8043 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8044 and aarch64_float_const_representable_p on the result.
8045 (aarch64_output_simd_mov_immediate): Remove mode argument.
8046 Update call to aarch64_simd_valid_immediate and use of
8047 simd_immediate_info.
8048 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8056 (mode_nunits): Likewise CONST_MODE_NUNITS.
8057 * machmode.def (ADJUST_NUNITS): Document.
8058 * genmodes.c (mode_data::need_nunits_adj): New field.
8059 (blank_mode): Update accordingly.
8060 (adj_nunits): New variable.
8061 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8063 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8064 listed in adj_nunits.
8065 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8066 listed in adj_nunits. Don't emit case statements for such modes.
8067 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8068 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8069 nothing if adj_nunits is nonnull.
8070 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8071 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8072 (emit_mode_fbit): Update use of print_maybe_const_decl.
8073 (emit_move_size): Likewise. Treat the array as non-const
8075 (emit_mode_adjustments): Handle adj_nunits.
8077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8079 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8080 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8081 (VECTOR_MODES): Use it.
8082 (make_vector_modes): Take the prefix as an argument.
8084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8085 Alan Hayward <alan.hayward@arm.com>
8086 David Sherwood <david.sherwood@arm.com>
8088 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8089 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8090 for MODE_VECTOR_BOOL.
8091 * machmode.def (VECTOR_BOOL_MODE): Document.
8092 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8093 (make_vector_bool_mode): New function.
8094 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8096 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8097 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8099 * stor-layout.c (int_mode_for_mode): Likewise.
8100 * tree.c (build_vector_type_for_mode): Likewise.
8101 * varasm.c (output_constant_pool_2): Likewise.
8102 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8103 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8104 for MODE_VECTOR_BOOL.
8105 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8106 of mode class checks.
8107 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8108 instead of a list of mode class checks.
8109 (expand_vector_scalar_condition): Likewise.
8110 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8113 Alan Hayward <alan.hayward@arm.com>
8114 David Sherwood <david.sherwood@arm.com>
8116 * machmode.h (mode_size): Change from unsigned short to
8118 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8119 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8120 or if measurement_type is not polynomial.
8121 (fixed_size_mode::includes_p): Check for constant-sized modes.
8122 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8123 return a poly_uint16 rather than an unsigned short.
8124 (emit_mode_size): Change the type of mode_size from unsigned short
8125 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8126 (emit_mode_adjustments): Cope with polynomial vector sizes.
8127 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8129 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8131 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8132 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8133 * caller-save.c (setup_save_areas): Likewise.
8134 (replace_reg_with_saved_mem): Likewise.
8135 * calls.c (emit_library_call_value_1): Likewise.
8136 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8137 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8138 (gen_lowpart_for_combine): Likewise.
8139 * convert.c (convert_to_integer_1): Likewise.
8140 * cse.c (equiv_constant, cse_insn): Likewise.
8141 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8142 (cselib_subst_to_values): Likewise.
8143 * dce.c (word_dce_process_block): Likewise.
8144 * df-problems.c (df_word_lr_mark_ref): Likewise.
8145 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8146 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8147 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8148 (rtl_for_decl_location): Likewise.
8149 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8150 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8151 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8152 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8153 (expand_expr_real_1): Likewise.
8154 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8155 (pad_below): Likewise.
8156 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8157 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8158 * ira.c (get_subreg_tracking_sizes): Likewise.
8159 * ira-build.c (ira_create_allocno_objects): Likewise.
8160 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8161 (ira_sort_regnos_for_alter_reg): Likewise.
8162 * ira-costs.c (record_operand_costs): Likewise.
8163 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8164 (resolve_simple_move): Likewise.
8165 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8166 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8167 (lra_constraints): Likewise.
8168 (CONST_POOL_OK_P): Reject variable-sized modes.
8169 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8170 (add_pseudo_to_slot, lra_spill): Likewise.
8171 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8172 * optabs-query.c (get_best_extraction_insn): Likewise.
8173 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8174 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8175 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8176 * recog.c (offsettable_address_addr_space_p): Likewise.
8177 * regcprop.c (maybe_mode_change): Likewise.
8178 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8179 * regrename.c (build_def_use): Likewise.
8180 * regstat.c (dump_reg_info): Likewise.
8181 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8182 (find_reloads, find_reloads_subreg_address): Likewise.
8183 * reload1.c (eliminate_regs_1): Likewise.
8184 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8185 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8186 (simplify_binary_operation_1, simplify_subreg): Likewise.
8187 * targhooks.c (default_function_arg_padding): Likewise.
8188 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8189 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8190 (verify_gimple_assign_ternary): Likewise.
8191 * tree-inline.c (estimate_move_cost): Likewise.
8192 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8193 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8194 (get_address_cost_ainc): Likewise.
8195 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8196 (vect_supportable_dr_alignment): Likewise.
8197 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8198 (vectorizable_reduction): Likewise.
8199 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8200 (vectorizable_operation, vectorizable_load): Likewise.
8201 * tree.c (build_same_sized_truth_vector_type): Likewise.
8202 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8203 * var-tracking.c (emit_note_insn_var_location): Likewise.
8204 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8205 (ADDR_VEC_ALIGN): Likewise.
8207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8208 Alan Hayward <alan.hayward@arm.com>
8209 David Sherwood <david.sherwood@arm.com>
8211 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8213 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8214 or if measurement_type is polynomial.
8215 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8216 * combine.c (make_extraction): Likewise.
8217 * dse.c (find_shift_sequence): Likewise.
8218 * dwarf2out.c (mem_loc_descriptor): Likewise.
8219 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8220 (extract_bit_field, extract_low_bits): Likewise.
8221 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8222 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8223 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8224 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8225 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8226 * reload.c (find_reloads): Likewise.
8227 * reload1.c (alter_reg): Likewise.
8228 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8229 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8230 * tree-if-conv.c (predicate_mem_writes): Likewise.
8231 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8232 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8233 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8234 * valtrack.c (dead_debug_insert_temp): Likewise.
8235 * varasm.c (mergeable_constant_section): Likewise.
8236 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8239 Alan Hayward <alan.hayward@arm.com>
8240 David Sherwood <david.sherwood@arm.com>
8242 * expr.c (expand_assignment): Cope with polynomial mode sizes
8243 when assigning to a CONCAT.
8245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8246 Alan Hayward <alan.hayward@arm.com>
8247 David Sherwood <david.sherwood@arm.com>
8249 * machmode.h (mode_precision): Change from unsigned short to
8251 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8253 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8254 or if measurement_type is not polynomial.
8255 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8256 in which the mode is already known to be a scalar_int_mode.
8257 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8258 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8260 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8261 for GET_MODE_PRECISION.
8262 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8263 for GET_MODE_PRECISION.
8264 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8266 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8267 (expand_field_assignment, make_extraction): Likewise.
8268 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8269 (get_last_value): Likewise.
8270 * convert.c (convert_to_integer_1): Likewise.
8271 * cse.c (cse_insn): Likewise.
8272 * expr.c (expand_expr_real_1): Likewise.
8273 * lra-constraints.c (simplify_operand_subreg): Likewise.
8274 * optabs-query.c (can_atomic_load_p): Likewise.
8275 * optabs.c (expand_atomic_load): Likewise.
8276 (expand_atomic_store): Likewise.
8277 * ree.c (combine_reaching_defs): Likewise.
8278 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8279 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8280 * tree.h (type_has_mode_precision_p): Likewise.
8281 * ubsan.c (instrument_si_overflow): Likewise.
8283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8284 Alan Hayward <alan.hayward@arm.com>
8285 David Sherwood <david.sherwood@arm.com>
8287 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8288 polynomial numbers of units.
8289 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8290 (valid_vector_subparts_p): New function.
8291 (build_vector_type): Remove temporary shim and take the number
8292 of units as a poly_uint64 rather than an int.
8293 (build_opaque_vector_type): Take the number of units as a
8294 poly_uint64 rather than an int.
8295 * tree.c (build_vector_from_ctor): Handle polynomial
8296 TYPE_VECTOR_SUBPARTS.
8297 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8298 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8299 (build_vector_from_val): If the number of units is variable,
8300 use build_vec_duplicate_cst for constant operands and
8301 VEC_DUPLICATE_EXPR otherwise.
8302 (make_vector_type): Remove temporary is_constant ().
8303 (build_vector_type, build_opaque_vector_type): Take the number of
8304 units as a poly_uint64 rather than an int.
8305 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8307 * cfgexpand.c (expand_debug_expr): Likewise.
8308 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8309 (store_constructor, expand_expr_real_1): Likewise.
8310 (const_scalar_mask_from_tree): Likewise.
8311 * fold-const-call.c (fold_const_reduction): Likewise.
8312 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8313 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8314 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8315 (fold_relational_const): Likewise.
8316 (native_interpret_vector): Likewise. Change the size from an
8317 int to an unsigned int.
8318 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8319 TYPE_VECTOR_SUBPARTS.
8320 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8321 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8322 duplicating a non-constant operand into a variable-length vector.
8323 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8324 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8325 * ipa-icf.c (sem_variable::equals): Likewise.
8326 * match.pd: Likewise.
8327 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8328 * print-tree.c (print_node): Likewise.
8329 * stor-layout.c (layout_type): Likewise.
8330 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8331 * tree-cfg.c (verify_gimple_comparison): Likewise.
8332 (verify_gimple_assign_binary): Likewise.
8333 (verify_gimple_assign_ternary): Likewise.
8334 (verify_gimple_assign_single): Likewise.
8335 * tree-pretty-print.c (dump_generic_node): Likewise.
8336 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8337 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8338 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8339 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8340 (vect_shift_permute_load_chain): Likewise.
8341 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8342 (expand_vector_condition, optimize_vector_constructor): Likewise.
8343 (lower_vec_perm, get_compute_type): Likewise.
8344 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8345 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8346 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8347 (vect_recog_mask_conversion_pattern): Likewise.
8348 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8349 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8350 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8351 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8352 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8353 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8354 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8355 (supportable_widening_operation): Likewise.
8356 (supportable_narrowing_operation): Likewise.
8357 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8359 * varasm.c (output_constant): Likewise.
8361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8362 Alan Hayward <alan.hayward@arm.com>
8363 David Sherwood <david.sherwood@arm.com>
8365 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8366 so that both the length == 3 and length != 3 cases set up their
8367 own permute vectors. Add comments explaining why we know the
8368 number of elements is constant.
8369 (vect_permute_load_chain): Likewise.
8371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8372 Alan Hayward <alan.hayward@arm.com>
8373 David Sherwood <david.sherwood@arm.com>
8375 * machmode.h (mode_nunits): Change from unsigned char to
8377 (ONLY_FIXED_SIZE_MODES): New macro.
8378 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8379 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8380 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8382 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8383 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8384 or if measurement_type is not polynomial.
8385 * genmodes.c (ZERO_COEFFS): New macro.
8386 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8388 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8389 Use ZERO_COEFFS when emitting initializers.
8390 * data-streamer.h (bp_pack_poly_value): New function.
8391 (bp_unpack_poly_value): Likewise.
8392 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8393 for GET_MODE_NUNITS.
8394 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8395 for GET_MODE_NUNITS.
8396 * tree.c (make_vector_type): Remove temporary shim and make
8397 the real function take the number of units as a poly_uint64
8399 (build_vector_type_for_mode): Handle polynomial nunits.
8400 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8401 * emit-rtl.c (const_vec_series_p_1): Likewise.
8402 (gen_rtx_CONST_VECTOR): Likewise.
8403 * fold-const.c (test_vec_duplicate_folding): Likewise.
8404 * genrecog.c (validate_pattern): Likewise.
8405 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8406 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8407 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8408 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8409 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8410 * rtlanal.c (subreg_get_info): Likewise.
8411 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8412 (vect_grouped_load_supported): Likewise.
8413 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8414 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8415 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8416 (simplify_const_unary_operation, simplify_binary_operation_1)
8417 (simplify_const_binary_operation, simplify_ternary_operation)
8418 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8419 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8420 instead of CONST_VECTOR_NUNITS.
8421 * varasm.c (output_constant_pool_2): Likewise.
8422 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8423 explicit-encoded elements in the XVEC for variable-length vectors.
8425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8427 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8430 Alan Hayward <alan.hayward@arm.com>
8431 David Sherwood <david.sherwood@arm.com>
8433 * coretypes.h (fixed_size_mode): Declare.
8434 (fixed_size_mode_pod): New typedef.
8435 * builtins.h (target_builtins::x_apply_args_mode)
8436 (target_builtins::x_apply_result_mode): Change type to
8437 fixed_size_mode_pod.
8438 * builtins.c (apply_args_size, apply_result_size, result_vector)
8439 (expand_builtin_apply_args_1, expand_builtin_apply)
8440 (expand_builtin_return): Update accordingly.
8442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8444 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8445 * cselib.c (cselib_hash_rtx): Likewise.
8446 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8447 CONST_VECTOR encoding.
8449 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8450 Jeff Law <law@redhat.com>
8453 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8454 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8455 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8456 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8459 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8460 explicitly probe *sp in a noreturn function if there were any callee
8461 register saves or frame pointer is needed.
8463 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8466 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8467 BLKmode for ternary, binary or unary expressions.
8470 * var-tracking.c (delete_vta_debug_insn): New inline function.
8471 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8472 insns from get_insns () to NULL instead of each bb separately.
8473 Use delete_vta_debug_insn. No longer static.
8474 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8475 delete_vta_debug_insns callers.
8476 * rtl.h (delete_vta_debug_insns): Declare.
8477 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8478 instead of variable_tracking_main.
8480 2018-01-03 Martin Sebor <msebor@redhat.com>
8482 PR tree-optimization/83603
8483 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8484 arguments past the endof the argument list in functions declared
8485 without a prototype.
8486 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8487 Avoid checking when arguments are null.
8489 2018-01-03 Martin Sebor <msebor@redhat.com>
8492 * doc/extend.texi (attribute const): Fix a typo.
8493 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8494 issuing -Wsuggest-attribute for void functions.
8496 2018-01-03 Martin Sebor <msebor@redhat.com>
8498 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8499 offset_int::from instead of wide_int::to_shwi.
8500 (maybe_diag_overlap): Remove assertion.
8501 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8502 * gimple-ssa-sprintf.c (format_directive): Same.
8503 (parse_directive): Same.
8504 (sprintf_dom_walker::compute_format_length): Same.
8505 (try_substitute_return_value): Same.
8507 2018-01-03 Jeff Law <law@redhat.com>
8510 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8511 non-constant residual for zero at runtime and avoid probing in
8512 that case. Reorganize code for trailing problem to mirror handling
8515 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8517 PR tree-optimization/83501
8518 * tree-ssa-strlen.c (get_string_cst): New.
8519 (handle_char_store): Call get_string_cst.
8521 2018-01-03 Martin Liska <mliska@suse.cz>
8523 PR tree-optimization/83593
8524 * tree-ssa-strlen.c: Include tree-cfg.h.
8525 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8526 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8527 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8529 (strlen_dom_walker::before_dom_children): Call
8530 gimple_purge_dead_eh_edges. Dump tranformation with details
8532 (strlen_dom_walker::before_dom_children): Update call by adding
8533 new argument cleanup_eh.
8534 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8536 2018-01-03 Martin Liska <mliska@suse.cz>
8539 * cif-code.def (VARIADIC_THUNK): New enum value.
8540 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8543 2018-01-03 Jan Beulich <jbeulich@suse.com>
8545 * sse.md (mov<mode>_internal): Tighten condition for when to use
8546 vmovdqu<ssescalarsize> for TI and OI modes.
8548 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8550 Update copyright years.
8552 2018-01-03 Martin Liska <mliska@suse.cz>
8555 * ipa-visibility.c (function_and_variable_visibility): Skip
8556 functions with noipa attribure.
8558 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8560 * gcc.c (process_command): Update copyright notice dates.
8561 * gcov-dump.c (print_version): Ditto.
8562 * gcov.c (print_version): Ditto.
8563 * gcov-tool.c (print_version): Ditto.
8564 * gengtype.c (create_file): Ditto.
8565 * doc/cpp.texi: Bump @copying's copyright year.
8566 * doc/cppinternals.texi: Ditto.
8567 * doc/gcc.texi: Ditto.
8568 * doc/gccint.texi: Ditto.
8569 * doc/gcov.texi: Ditto.
8570 * doc/install.texi: Ditto.
8571 * doc/invoke.texi: Ditto.
8573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8575 * vector-builder.h (vector_builder::m_full_nelts): Change from
8576 unsigned int to poly_uint64.
8577 (vector_builder::full_nelts): Update prototype accordingly.
8578 (vector_builder::new_vector): Likewise.
8579 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8580 (vector_builder::operator ==): Likewise.
8581 (vector_builder::finalize): Likewise.
8582 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8583 Take the number of elements as a poly_uint64 rather than an
8585 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8586 from unsigned int to poly_uint64.
8587 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8588 (vec_perm_indices::new_vector): Likewise.
8589 (vec_perm_indices::length): Likewise.
8590 (vec_perm_indices::nelts_per_input): Likewise.
8591 (vec_perm_indices::input_nelts): Likewise.
8592 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8593 number of elements per input as a poly_uint64 rather than an
8594 unsigned int. Use the original encoding for variable-length
8595 vectors, rather than clamping each individual element.
8596 For the second and subsequent elements in each pattern,
8597 clamp the step and base before clamping their sum.
8598 (vec_perm_indices::series_p): Handle polynomial element counts.
8599 (vec_perm_indices::all_in_range_p): Likewise.
8600 (vec_perm_indices_to_tree): Likewise.
8601 (vec_perm_indices_to_rtx): Likewise.
8602 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8603 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8604 (tree_vector_builder::new_binary_operation): Handle polynomial
8605 element counts. Return false if we need to know the number
8606 of elements at compile time.
8607 * fold-const.c (fold_vec_perm): Punt if the number of elements
8608 isn't known at compile time.
8610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8612 * vec-perm-indices.h (vec_perm_builder): Change element type
8613 from HOST_WIDE_INT to poly_int64.
8614 (vec_perm_indices::element_type): Update accordingly.
8615 (vec_perm_indices::clamp): Handle polynomial element_types.
8616 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8617 (vec_perm_indices::all_in_range_p): Likewise.
8618 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8620 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8621 polynomial vec_perm_indices element types.
8622 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8623 * fold-const.c (fold_vec_perm): Likewise.
8624 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8625 * tree-vect-generic.c (lower_vec_perm): Likewise.
8626 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8627 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8628 element type to HOST_WIDE_INT.
8630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8631 Alan Hayward <alan.hayward@arm.com>
8632 David Sherwood <david.sherwood@arm.com>
8634 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8635 rather than an int. Use plus_constant.
8636 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8637 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8640 Alan Hayward <alan.hayward@arm.com>
8641 David Sherwood <david.sherwood@arm.com>
8643 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8644 a HOST_WIDE_INT to a poly_int64.
8646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8647 Alan Hayward <alan.hayward@arm.com>
8648 David Sherwood <david.sherwood@arm.com>
8650 * calls.c (load_register_parameters): Cope with polynomial
8651 mode sizes. Require a constant size for BLKmode parameters
8652 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8653 forces a parameter to be padded at the lsb end in order to
8654 fill a complete number of words, require the parameter size
8655 to be ordered wrt UNITS_PER_WORD.
8657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8658 Alan Hayward <alan.hayward@arm.com>
8659 David Sherwood <david.sherwood@arm.com>
8661 * reload1.c (spill_stack_slot_width): Change element type
8662 from unsigned int to poly_uint64_pod.
8663 (alter_reg): Treat mode sizes as polynomial.
8665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8666 Alan Hayward <alan.hayward@arm.com>
8667 David Sherwood <david.sherwood@arm.com>
8669 * reload.c (complex_word_subreg_p): New function.
8670 (reload_inner_reg_of_subreg, push_reload): Use it.
8672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8673 Alan Hayward <alan.hayward@arm.com>
8674 David Sherwood <david.sherwood@arm.com>
8676 * lra-constraints.c (process_alt_operands): Reject matched
8677 operands whose sizes aren't ordered.
8678 (match_reload): Refer to this check here.
8680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8681 Alan Hayward <alan.hayward@arm.com>
8682 David Sherwood <david.sherwood@arm.com>
8684 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8685 that the mode size is in the set {1, 2, 4, 8, 16}.
8687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8688 Alan Hayward <alan.hayward@arm.com>
8689 David Sherwood <david.sherwood@arm.com>
8691 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8692 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8699 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8700 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8701 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8702 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8703 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8704 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8705 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8706 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8707 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8709 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8710 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8711 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8712 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8713 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8714 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8715 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8716 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8717 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8718 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8720 * expr.c (emit_move_resolve_push): Treat the input and result
8721 of PUSH_ROUNDING as a poly_int64.
8722 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8723 (emit_push_insn): Likewise.
8724 * lra-eliminations.c (mark_not_eliminable): Likewise.
8725 * recog.c (push_operand): Likewise.
8726 * reload1.c (elimination_effects): Likewise.
8727 * rtlanal.c (nonzero_bits1): Likewise.
8728 * calls.c (store_one_arg): Likewise. Require the padding to be
8729 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8736 Use plus_constant instead of gen_rtx_PLUS.
8738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8739 Alan Hayward <alan.hayward@arm.com>
8740 David Sherwood <david.sherwood@arm.com>
8742 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8746 Alan Hayward <alan.hayward@arm.com>
8747 David Sherwood <david.sherwood@arm.com>
8749 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8750 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8751 via stack temporaries. Treat the mode size as polynomial too.
8753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8754 Alan Hayward <alan.hayward@arm.com>
8755 David Sherwood <david.sherwood@arm.com>
8757 * expr.c (expand_expr_real_2): When handling conversions involving
8758 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8759 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8760 as a poly_uint64 too.
8762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8763 Alan Hayward <alan.hayward@arm.com>
8764 David Sherwood <david.sherwood@arm.com>
8766 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
8773 REGMODE_NATURAL_SIZE.
8774 * expmed.c (store_bit_field_1): Likewise.
8775 * expr.c (store_constructor): Likewise.
8776 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8777 and polynomial REGMODE_NATURAL_SIZE.
8778 (gen_lowpart_common): Likewise.
8779 * reginfo.c (record_subregs_of_mode): Likewise.
8780 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8787 numbers of elements.
8789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8790 Alan Hayward <alan.hayward@arm.com>
8791 David Sherwood <david.sherwood@arm.com>
8793 * match.pd: Cope with polynomial numbers of vector elements.
8795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8796 Alan Hayward <alan.hayward@arm.com>
8797 David Sherwood <david.sherwood@arm.com>
8799 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8800 in a POINTER_PLUS_EXPR.
8802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8803 Alan Hayward <alan.hayward@arm.com>
8804 David Sherwood <david.sherwood@arm.com>
8806 * omp-simd-clone.c (simd_clone_subparts): New function.
8807 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8808 (ipa_simd_modify_function_body): Likewise.
8810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8811 Alan Hayward <alan.hayward@arm.com>
8812 David Sherwood <david.sherwood@arm.com>
8814 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8815 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8816 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8817 (expand_vector_condition, vector_element): Likewise.
8818 (subparts_gt): New function.
8819 (get_compute_type): Use subparts_gt.
8820 (count_type_subparts): Delete.
8821 (expand_vector_operations_1): Use subparts_gt instead of
8822 count_type_subparts.
8824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8825 Alan Hayward <alan.hayward@arm.com>
8826 David Sherwood <david.sherwood@arm.com>
8828 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8829 (vect_compile_time_alias): ...this new function. Do the calculation
8830 on poly_ints rather than trees.
8831 (vect_prune_runtime_alias_test_list): Update call accordingly.
8833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8834 Alan Hayward <alan.hayward@arm.com>
8835 David Sherwood <david.sherwood@arm.com>
8837 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8839 (vect_schedule_slp_instance): Likewise.
8841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8842 Alan Hayward <alan.hayward@arm.com>
8843 David Sherwood <david.sherwood@arm.com>
8845 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8846 constant and extern definitions for variable-length vectors.
8847 (vect_get_constant_vectors): Note that the number of units
8848 is known to be constant.
8850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8851 Alan Hayward <alan.hayward@arm.com>
8852 David Sherwood <david.sherwood@arm.com>
8854 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8855 of units as polynomial. Choose between WIDE and NARROW based
8858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8859 Alan Hayward <alan.hayward@arm.com>
8860 David Sherwood <david.sherwood@arm.com>
8862 * tree-vect-stmts.c (simd_clone_subparts): New function.
8863 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8866 Alan Hayward <alan.hayward@arm.com>
8867 David Sherwood <david.sherwood@arm.com>
8869 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8870 vectors as polynomial. Use build_index_vector for
8873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8874 Alan Hayward <alan.hayward@arm.com>
8875 David Sherwood <david.sherwood@arm.com>
8877 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8878 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8879 for variable-length vectors.
8880 (vectorizable_mask_load_store): Treat the number of units as
8881 polynomial, asserting that it is constant if the condition has
8882 already been enforced.
8883 (vectorizable_store, vectorizable_load): Likewise.
8885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8886 Alan Hayward <alan.hayward@arm.com>
8887 David Sherwood <david.sherwood@arm.com>
8889 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8890 of units as polynomial. Punt if we can't tell at compile time
8891 which vector contains the final result.
8893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8894 Alan Hayward <alan.hayward@arm.com>
8895 David Sherwood <david.sherwood@arm.com>
8897 * tree-vect-loop.c (vectorizable_induction): Treat the number
8898 of units as polynomial. Punt on SLP inductions. Use an integer
8899 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8900 cast of such a series for variable-length floating-point
8903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8904 Alan Hayward <alan.hayward@arm.com>
8905 David Sherwood <david.sherwood@arm.com>
8907 * tree.h (build_index_vector): Declare.
8908 * tree.c (build_index_vector): New function.
8909 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8910 of units as polynomial, forcibly converting it to a constant if
8911 vectorizable_reduction has already enforced the condition.
8912 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8913 to create a {1,2,3,...} vector.
8914 (vectorizable_reduction): Treat the number of units as polynomial.
8915 Choose vectype_in based on the largest scalar element size rather
8916 than the smallest number of units. Enforce the restrictions
8919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8920 Alan Hayward <alan.hayward@arm.com>
8921 David Sherwood <david.sherwood@arm.com>
8923 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8924 number of units as polynomial.
8926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8927 Alan Hayward <alan.hayward@arm.com>
8928 David Sherwood <david.sherwood@arm.com>
8930 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8931 * target.def (autovectorize_vector_sizes): Return the vector sizes
8932 by pointer, using vector_sizes rather than a bitmask.
8933 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8934 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8935 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8937 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8938 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8939 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8940 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8941 * omp-general.c (omp_max_vf): Likewise.
8942 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8943 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8944 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8945 * tree-vect-slp.c (vect_slp_bb): Likewise.
8946 * doc/tm.texi: Regenerate.
8947 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8949 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8950 the vector size as a poly_uint64 rather than an unsigned int.
8951 (current_vector_size): Change from an unsigned int to a poly_uint64.
8952 (get_vectype_for_scalar_type): Update accordingly.
8953 * tree.h (build_truth_vector_type): Take the size and number of
8954 units as a poly_uint64 rather than an unsigned int.
8955 (build_vector_type): Add a temporary overload that takes
8956 the number of units as a poly_uint64 rather than an unsigned int.
8957 * tree.c (make_vector_type): Likewise.
8958 (build_truth_vector_type): Take the number of units as a poly_uint64
8959 rather than an unsigned int.
8961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8962 Alan Hayward <alan.hayward@arm.com>
8963 David Sherwood <david.sherwood@arm.com>
8965 * target.def (get_mask_mode): Take the number of units and length
8966 as poly_uint64s rather than unsigned ints.
8967 * targhooks.h (default_get_mask_mode): Update accordingly.
8968 * targhooks.c (default_get_mask_mode): Likewise.
8969 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8970 * doc/tm.texi: Regenerate.
8972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8973 Alan Hayward <alan.hayward@arm.com>
8974 David Sherwood <david.sherwood@arm.com>
8976 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8977 * omp-general.c (omp_max_vf): Likewise.
8978 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8979 (expand_omp_simd): Handle polynomial safelen.
8980 * omp-low.c (omplow_simd_context): Add a default constructor.
8981 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8982 (lower_rec_simd_input_clauses): Update accordingly.
8983 (lower_rec_input_clauses): Likewise.
8985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8986 Alan Hayward <alan.hayward@arm.com>
8987 David Sherwood <david.sherwood@arm.com>
8989 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8990 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8991 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8992 (vect_analyze_slp_cost): Likewise.
8993 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8994 (vect_model_load_cost): Likewise.
8996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8997 Alan Hayward <alan.hayward@arm.com>
8998 David Sherwood <david.sherwood@arm.com>
9000 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9001 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9002 from an unsigned int * to a poly_uint64_pod *.
9003 (calculate_unrolling_factor): New function.
9004 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9007 Alan Hayward <alan.hayward@arm.com>
9008 David Sherwood <david.sherwood@arm.com>
9010 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9011 from an unsigned int to a poly_uint64.
9012 (_loop_vec_info::slp_unrolling_factor): Likewise.
9013 (_loop_vec_info::vectorization_factor): Change from an int
9015 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9016 (vect_get_num_vectors): New function.
9017 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9018 (vect_get_num_copies): Use vect_get_num_vectors.
9019 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9020 to an unsigned int *.
9021 (vect_analyze_data_refs): Change min_vf from an int * to a
9023 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9024 than an unsigned HOST_WIDE_INT.
9025 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9026 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9027 to an unsigned int *.
9028 (vect_analyze_data_ref_dependences): Likewise.
9029 (vect_compute_data_ref_alignment): Handle polynomial vf.
9030 (vect_enhance_data_refs_alignment): Likewise.
9031 (vect_prune_runtime_alias_test_list): Likewise.
9032 (vect_shift_permute_load_chain): Likewise.
9033 (vect_supportable_dr_alignment): Likewise.
9034 (dependence_distance_ge_vf): Take the vectorization factor as a
9035 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9036 (vect_analyze_data_refs): Change min_vf from an int * to a
9038 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9039 vfm1 as a poly_uint64 rather than an int. Make the same change
9040 for the returned bound_scalar.
9041 (vect_gen_vector_loop_niters): Handle polynomial vf.
9042 (vect_do_peeling): Likewise. Update call to
9043 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9044 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9046 * tree-vect-loop.c (vect_determine_vectorization_factor)
9047 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9048 (vect_get_known_peeling_cost): Likewise.
9049 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9050 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9051 (vect_transform_loop): Likewise. Use the lowest possible VF when
9052 updating the upper bounds of the loop.
9053 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9055 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9056 polynomial unroll factors.
9057 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9058 (vect_make_slp_decision): Likewise.
9059 (vect_supported_load_permutation_p): Likewise, and polynomial
9061 (vect_analyze_slp_cost): Handle polynomial vf.
9062 (vect_slp_analyze_node_operations): Likewise.
9063 (vect_slp_analyze_bb_1): Likewise.
9064 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9065 than an unsigned HOST_WIDE_INT.
9066 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9067 (vectorizable_load): Handle polynomial vf.
9068 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9070 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9073 Alan Hayward <alan.hayward@arm.com>
9074 David Sherwood <david.sherwood@arm.com>
9076 * match.pd: Handle bit operations involving three constants
9077 and try to fold one pair.
9079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9081 * tree-vect-loop-manip.c: Include gimple-fold.h.
9082 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9083 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9084 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9085 Add a path that uses a step of VF instead of 1, but disable it
9087 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9088 and niters_no_overflow parameters. Update calls to
9089 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9090 Create a new SSA name if the latter choses to use a ste other
9091 than zero, and return it via niters_vector_mult_vf_var.
9092 * tree-vect-loop.c (vect_transform_loop): Update calls to
9093 vect_do_peeling, vect_gen_vector_loop_niters and
9094 slpeel_make_loop_iterate_ntimes.
9095 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9096 (vect_gen_vector_loop_niters): Update declarations after above changes.
9098 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9100 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9101 128-bit round to integer instructions.
9102 (ceil<mode>2): Likewise.
9103 (btrunc<mode>2): Likewise.
9104 (round<mode>2): Likewise.
9106 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9108 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9109 unaligned VSX load/store on P8/P9.
9110 (expand_block_clear): Allow the use of unaligned VSX
9111 load/store on P8/P9.
9113 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9115 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9117 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9118 swap associated with both a load and a store.
9120 2018-01-02 Andrew Waterman <andrew@sifive.com>
9122 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9123 * config/riscv/riscv.md (clear_cache): Use it.
9125 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9127 * web.c: Remove out-of-date comment.
9129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9131 * expr.c (fixup_args_size_notes): Check that any existing
9132 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9133 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9134 (emit_single_push_insn): ...here.
9136 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9138 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9139 (const_vector_encoded_nelts): New function.
9140 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9141 (const_vector_int_elt, const_vector_elt): Declare.
9142 * emit-rtl.c (const_vector_int_elt_1): New function.
9143 (const_vector_elt): Likewise.
9144 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9145 of CONST_VECTOR_ELT.
9147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9149 * expr.c: Include rtx-vector-builder.h.
9150 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9151 directly on the tree encoding.
9152 (const_vector_from_tree): Likewise.
9153 * optabs.c: Include rtx-vector-builder.h.
9154 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9155 sequence of "u" values.
9156 * vec-perm-indices.c: Include rtx-vector-builder.h.
9157 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9158 directly on the vec_perm_indices encoding.
9160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9162 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9163 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9164 * rtx-vector-builder.h: New file.
9165 * rtx-vector-builder.c: Likewise.
9166 * rtl.h (rtx_def::u2): Add a const_vector field.
9167 (CONST_VECTOR_NPATTERNS): New macro.
9168 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9169 (CONST_VECTOR_DUPLICATE_P): Likewise.
9170 (CONST_VECTOR_STEPPED_P): Likewise.
9171 (CONST_VECTOR_ENCODED_ELT): Likewise.
9172 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9173 (unwrap_const_vec_duplicate): Likewise.
9174 (const_vec_series_p): Check for a non-duplicated vector encoding.
9175 Say that the function only returns true for integer vectors.
9176 * emit-rtl.c: Include rtx-vector-builder.h.
9177 (gen_const_vec_duplicate_1): Delete.
9178 (gen_const_vector): Call gen_const_vec_duplicate instead of
9179 gen_const_vec_duplicate_1.
9180 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9181 (gen_const_vec_duplicate): Use rtx_vector_builder.
9182 (gen_const_vec_series): Likewise.
9183 (gen_rtx_CONST_VECTOR): Likewise.
9184 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9185 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9186 Build a new vector rather than modifying a CONST_VECTOR in-place.
9187 (handle_special_swappables): Update call accordingly.
9188 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9189 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9190 Build a new vector rather than modifying a CONST_VECTOR in-place.
9191 (handle_special_swappables): Update call accordingly.
9193 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9195 * simplify-rtx.c (simplify_const_binary_operation): Use
9196 CONST_VECTOR_ELT instead of XVECEXP.
9198 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9200 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9201 the selector elements to be different from the data elements
9202 if the selector is a VECTOR_CST.
9203 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9204 ssizetype for the selector.
9206 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9208 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9209 before testing each element individually.
9210 * tree-vect-generic.c (lower_vec_perm): Likewise.
9212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9214 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9215 * selftest-run-tests.c (selftest::run_tests): Call it.
9216 * vector-builder.h (vector_builder::operator ==): New function.
9217 (vector_builder::operator !=): Likewise.
9218 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9219 (vec_perm_indices::all_from_input_p): New function.
9220 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9221 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9222 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9223 instead of reading the VECTOR_CST directly. Detect whether both
9224 vector inputs are the same before constructing the vec_perm_indices,
9225 and update the number of inputs argument accordingly. Use the
9226 utility functions added above. Only construct sel2 if we need to.
9228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9230 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9231 the broadcast of the low byte.
9232 (expand_mult_highpart): Use an explicit encoding for the permutes.
9233 * optabs-query.c (can_mult_highpart_p): Likewise.
9234 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9235 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9236 (vectorizable_bswap): Likewise.
9237 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9238 explicit encoding for the power-of-2 permutes.
9239 (vect_permute_store_chain): Likewise.
9240 (vect_grouped_load_supported): Likewise.
9241 (vect_permute_load_chain): Likewise.
9243 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9245 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9246 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9247 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9248 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9249 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9250 (vect_gen_perm_mask_any): Likewise.
9252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9254 * int-vector-builder.h: New file.
9255 * vec-perm-indices.h: Include int-vector-builder.h.
9256 (vec_perm_indices): Redefine as an int_vector_builder.
9257 (auto_vec_perm_indices): Delete.
9258 (vec_perm_builder): Redefine as a stand-alone class.
9259 (vec_perm_indices::vec_perm_indices): New function.
9260 (vec_perm_indices::clamp): Likewise.
9261 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9262 (vec_perm_indices::new_vector): New function.
9263 (vec_perm_indices::new_expanded_vector): Update for new
9264 vec_perm_indices class.
9265 (vec_perm_indices::rotate_inputs): New function.
9266 (vec_perm_indices::all_in_range_p): Operate directly on the
9267 encoded form, without computing elided elements.
9268 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9269 encoding. Update for new vec_perm_indices class.
9270 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9271 the given vec_perm_builder.
9272 (expand_vec_perm_var): Update vec_perm_builder constructor.
9273 (expand_mult_highpart): Use vec_perm_builder instead of
9274 auto_vec_perm_indices.
9275 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9276 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9277 or double series encoding as appropriate.
9278 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9279 vec_perm_indices instead of auto_vec_perm_indices.
9280 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9281 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9282 (vect_permute_store_chain): Likewise.
9283 (vect_grouped_load_supported): Likewise.
9284 (vect_permute_load_chain): Likewise.
9285 (vect_shift_permute_load_chain): Likewise.
9286 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9287 (vect_transform_slp_perm_load): Likewise.
9288 (vect_schedule_slp_instance): Likewise.
9289 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9290 (vectorizable_mask_load_store): Likewise.
9291 (vectorizable_bswap): Likewise.
9292 (vectorizable_store): Likewise.
9293 (vectorizable_load): Likewise.
9294 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9295 vec_perm_indices instead of auto_vec_perm_indices. Use
9296 tree_to_vec_perm_builder to read the vector from a tree.
9297 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9298 vec_perm_builder instead of a vec_perm_indices.
9299 (have_whole_vector_shift): Use vec_perm_builder and
9300 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9301 truncation to calc_vec_perm_mask_for_shift.
9302 (vect_create_epilog_for_reduction): Likewise.
9303 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9304 from auto_vec_perm_indices to vec_perm_indices.
9305 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9306 instead of changing individual elements.
9307 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9308 the vector in d.perm.
9309 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9310 from auto_vec_perm_indices to vec_perm_indices.
9311 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9312 instead of changing individual elements.
9313 (arm_vectorize_vec_perm_const): Use new_vector to install
9314 the vector in d.perm.
9315 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9316 Update vec_perm_builder constructor.
9317 (rs6000_expand_interleave): Likewise.
9318 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9319 (rs6000_expand_interleave): Likewise.
9321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9323 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9324 to qimode could truncate the indices.
9325 * optabs.c (expand_vec_perm_var): Likewise.
9327 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9329 * Makefile.in (OBJS): Add vec-perm-indices.o.
9330 * vec-perm-indices.h: New file.
9331 * vec-perm-indices.c: Likewise.
9332 * target.h (vec_perm_indices): Replace with a forward class
9334 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9335 * optabs.h: Include vec-perm-indices.h.
9336 (expand_vec_perm): Delete.
9337 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9338 (expand_vec_perm_const): Declare.
9339 * target.def (vec_perm_const_ok): Replace with...
9340 (vec_perm_const): ...this new hook.
9341 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9342 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9343 * doc/tm.texi: Regenerate.
9344 * optabs.def (vec_perm_const): Delete.
9345 * doc/md.texi (vec_perm_const): Likewise.
9346 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9347 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9348 expand_vec_perm for constant permutation vectors. Assert that
9349 the mode of variable permutation vectors is the integer equivalent
9350 of the mode that is being permuted.
9351 * optabs-query.h (selector_fits_mode_p): Declare.
9352 * optabs-query.c: Include vec-perm-indices.h.
9353 (selector_fits_mode_p): New function.
9354 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9355 is defined, instead of checking whether the vec_perm_const_optab
9356 exists. Use targetm.vectorize.vec_perm_const instead of
9357 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9358 fit in the vector mode before using a variable permute.
9359 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9360 vec_perm_indices instead of an rtx.
9361 (expand_vec_perm): Replace with...
9362 (expand_vec_perm_const): ...this new function. Take the selector
9363 as a vec_perm_indices rather than an rtx. Also take the mode of
9364 the selector. Update call to shift_amt_for_vec_perm_mask.
9365 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9366 Use vec_perm_indices::new_expanded_vector to expand the original
9367 selector into bytes. Check whether the indices fit in the vector
9368 mode before using a variable permute.
9369 (expand_vec_perm_var): Make global.
9370 (expand_mult_highpart): Use expand_vec_perm_const.
9371 * fold-const.c: Includes vec-perm-indices.h.
9372 * tree-ssa-forwprop.c: Likewise.
9373 * tree-vect-data-refs.c: Likewise.
9374 * tree-vect-generic.c: Likewise.
9375 * tree-vect-loop.c: Likewise.
9376 * tree-vect-slp.c: Likewise.
9377 * tree-vect-stmts.c: Likewise.
9378 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9380 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9381 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9382 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9383 (aarch64_vectorize_vec_perm_const): ...this new function.
9384 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9385 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9386 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9387 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9388 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9389 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9390 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9392 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9393 check for NEON modes.
9394 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9395 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9396 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9397 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9399 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9400 the old VEC_PERM_CONST conditions.
9401 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9402 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9403 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9404 (ia64_vectorize_vec_perm_const_ok): Merge into...
9405 (ia64_vectorize_vec_perm_const): ...this new function.
9406 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9407 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9408 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9409 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9410 * config/mips/mips.c (mips_expand_vec_perm_const)
9411 (mips_vectorize_vec_perm_const_ok): Merge into...
9412 (mips_vectorize_vec_perm_const): ...this new function.
9413 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9414 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9415 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9416 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9417 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9418 (rs6000_expand_vec_perm_const): Delete.
9419 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9421 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9422 (altivec_expand_vec_perm_const_le): Take each operand individually.
9423 Operate on constant selectors rather than rtxes.
9424 (altivec_expand_vec_perm_const): Likewise. Update call to
9425 altivec_expand_vec_perm_const_le.
9426 (rs6000_expand_vec_perm_const): Delete.
9427 (rs6000_vectorize_vec_perm_const_ok): Delete.
9428 (rs6000_vectorize_vec_perm_const): New function.
9429 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9430 an element count and rtx array.
9431 (rs6000_expand_extract_even): Update call accordingly.
9432 (rs6000_expand_interleave): Likewise.
9433 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9434 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9435 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9436 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9437 (rs6000_expand_vec_perm_const): Delete.
9438 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9439 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9440 (altivec_expand_vec_perm_const_le): Take each operand individually.
9441 Operate on constant selectors rather than rtxes.
9442 (altivec_expand_vec_perm_const): Likewise. Update call to
9443 altivec_expand_vec_perm_const_le.
9444 (rs6000_expand_vec_perm_const): Delete.
9445 (rs6000_vectorize_vec_perm_const_ok): Delete.
9446 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9447 reference to the SPE evmerge intructions.
9448 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9449 an element count and rtx array.
9450 (rs6000_expand_extract_even): Update call accordingly.
9451 (rs6000_expand_interleave): Likewise.
9452 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9453 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9455 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9459 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9460 vector mode and that that mode matches the mode of the data
9462 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9463 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9464 directly using expand_vec_perm_1 when forcing selectors into
9466 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9468 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9470 * optabs-query.h (can_vec_perm_p): Delete.
9471 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9472 * optabs-query.c (can_vec_perm_p): Split into...
9473 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9474 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9475 particular selector is valid.
9476 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9477 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9478 (vect_grouped_load_supported): Likewise.
9479 (vect_shift_permute_load_chain): Likewise.
9480 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9481 (vect_transform_slp_perm_load): Likewise.
9482 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9483 (vectorizable_bswap): Likewise.
9484 (vect_gen_perm_mask_checked): Likewise.
9485 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9486 implementations of variable permutation vectors into account
9487 when deciding which selector to use.
9488 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9489 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9490 with a false third argument.
9491 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9492 to test whether the constant selector is valid and can_vec_perm_var_p
9493 to test whether a variable selector is valid.
9495 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9497 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9498 * optabs-query.c (can_vec_perm_p): Likewise.
9499 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9500 instead of vec_perm_indices.
9501 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9502 (vect_gen_perm_mask_checked): Likewise,
9503 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9504 (vect_gen_perm_mask_checked): Likewise,
9506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9508 * optabs-query.h (qimode_for_vec_perm): Declare.
9509 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9510 (qimode_for_vec_perm): ...this new function.
9511 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9513 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9515 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9516 does not have a conditional at the top.
9518 2018-01-02 Richard Biener <rguenther@suse.de>
9520 * ipa-inline.c (big_speedup_p): Fix expression.
9522 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9525 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9528 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9532 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9533 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9534 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9535 cond_taken_branch_cost 3->4.
9537 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9539 PR tree-optimization/83581
9540 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9541 TODO_cleanup_cfg if any changes have been made.
9544 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9545 convert_modes if target mode has the right side, but different mode
9549 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9550 last argument when extracting from CONCAT. If either from_real or
9551 from_imag is NULL, use expansion through memory. If result is not
9552 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9553 the parts directly to inner mode, if even that fails, use expansion
9557 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9558 check for bswap in mode rather than HImode and use that in expand_unop
9561 Copyright (C) 2018 Free Software Foundation, Inc.
9563 Copying and distribu