1 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9 function to create the function decl for complex long double
10 multiply and divide for -mabi=ieeelongdouble.
11 (init_float128_ieee): Call it.
13 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
17 -fcf-protection=branch -mibt.
18 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
23 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
25 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
28 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
29 extern array declarations.
31 2018-03-27 Richard Biener <rguenther@suse.de>
34 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
35 explicit single_use checks.
37 2018-03-27 Richard Biener <rguenther@suse.de>
39 PR tree-optimization/85082
40 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
43 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
45 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
46 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
47 Turn on fasynchronous-unwind-tables and funwind-tables.
49 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
52 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
53 (*bmi_blsr_<mode>_ccz): Ditto.
55 2018-03-26 Tom de Vries <tom@codesourcery.com>
57 PR tree-optimization/85063
58 * omp-general.c (offloading_function_p): New function. Factor out
60 * omp-offload.c (pass_omp_target_link::gate): ... here.
61 * omp-general.h (offloading_function_p): Declare.
62 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
63 with attribute omp declare target for offloading functions.
65 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
67 PR tree-optimization/84005
68 * tree-data-ref.h (get_base_for_alignment): Declare.
69 * tree-data-ref.c (get_base_for_alignment_1): New function.
70 (get_base_for_alignment): Likewise.
71 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
72 get_base_for_alignment to find a suitable base object, instead
73 of always using drb->base_address.
75 2018-03-23 Jakub Jelinek <jakub@redhat.com>
78 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
79 known size by default.
81 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
84 * lra-constraints.c (process_alt_operands): Don't match BLKmode
85 and non BLKmode operands.
87 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
90 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
93 2018-03-23 Richard Biener <rguenther@suse.de>
96 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
97 we are going to emit early debug for LTO.
99 2018-03-23 Jakub Jelinek <jakub@redhat.com>
102 * function.c (match_asm_constraints_1): Don't optimize if input
103 doesn't satisfy general_operand predicate for output's mode.
106 * alias.c (write_dependence_p): Don't require for x_canonicalized
107 non-VOIDmode if x has VOIDmode.
110 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
111 just don't try to optimize it rather than assert it never happens.
113 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
115 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
116 macro expansions for definition of ST_INTERNAL_<mode> and
117 LD_INTERNAL_<mode> builtins.
118 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
120 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
122 (altivec_expand_st_builtin): Likewise.
123 (altivec_expand_builtin): Remove calls to deleted functions.
124 (rs6000_address_for_altivec): Delete this function.
125 * config/rs6000/vector.md: Remove expands for
126 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
128 2018-03-22 Sudakshina Das <sudi.das@arm.com>
131 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
132 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
133 re-computing once computed.
134 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
135 (arm_init_machine_status): Initialize
136 machine->static_chain_stack_bytes.
138 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
141 * doc/extend.texi: Add four new prototypes for vec_ld.
142 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
143 definitions for more logical presentation.
144 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
145 entries for V1TI variants of __builtin_altivec_ld builtin.
146 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
147 handling of V1TI variant of LVX icode pattern.
148 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
149 (rs6000_gimple_fold_builtin): Likewise.
150 (altivec_init_builtins): Add code to define
151 __builtin_altivec_lvx_v1ti function.
153 2018-03-22 Jakub Jelinek <jakub@redhat.com>
156 * function.c (match_asm_constraints_1): Don't do the optimization
157 if input isn't a REG, SUBREG, MEM or constant.
159 2018-03-22 Tom de Vries <tom@codesourcery.com>
161 PR tree-optimization/84956
162 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
163 bb_has_abnormal_pred.
165 2018-03-22 Jakub Jelinek <jakub@redhat.com>
168 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
169 DECL_INITIAL (decl) to decl at the end.
170 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
173 2018-03-21 Joseph Myers <joseph@codesourcery.com>
175 * doc/extend.texi (__builtin_tgmath): Document when complex
176 integer types are treated as _Complex _Float64.
178 2018-03-21 Tom de Vries <tom@codesourcery.com>
180 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
182 2018-03-21 Jakub Jelinek <jakub@redhat.com>
184 PR tree-optimization/84960
185 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
186 if it is ENTRY block, move them into single succ of ENTRY in that case.
188 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
190 PR tree-optimization/84811
191 * poly-int.h (poly_span_traits): Remove the T3 parameter and
192 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
193 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
194 (known_subrange_p): Update accordingly. Cast each value involved
195 in the size comparison, rather than casting the result of the
198 2018-03-21 Jakub Jelinek <jakub@redhat.com>
200 PR tree-optimization/84982
201 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
202 by flipping the least significant bit rather than all bits from
203 bitpos to bitpos + bitsize - 1.
205 2018-03-21 Nathan Sidwell <nathan@acm.org>
207 * doc/extend.texi (Deprecated Features): Remove mention of
208 long-deleted deprecations.
210 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
213 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
214 * configure: Regenerate.
216 2018-03-21 Tom de Vries <tom@codesourcery.com>
218 PR tree-optimization/83126
219 * tree-parloops.c (num_phis): New function.
220 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
222 2018-03-21 Nathan Sidwell <nathan@acm.org>
224 * doc/extend.texi (Deprecated Features): Update deprecated flags,
225 mention anon-struct/union members and trailing attributes.
227 2018-03-21 Bin Cheng <bin.cheng@arm.com>
229 PR tree-optimization/84969
230 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
231 builtin memset partitions if they set different rhs values.
233 2018-03-21 Jakub Jelinek <jakub@redhat.com>
235 PR rtl-optimization/84989
236 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
237 VEC_DUPLICATE with scalar result mode.
239 2018-03-21 Martin Liska <mliska@suse.cz>
242 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
243 not intended return statement.
245 2018-03-21 Martin Liska <mliska@suse.cz>
248 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
249 (chkp_find_bound_slots_1): Limit number of iterations.
251 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
254 * Minor grammar fixes for x86 options.
256 2018-03-20 Jakub Jelinek <jakub@redhat.com>
259 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
260 holding REG_CFA_RESTORE notes, instead turn them into a USE.
262 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
265 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
266 (altivec_lvx_<mode>_1op): Likewise.
267 (altivec_stvx_<mode>_2op): Likewise.
268 (altivec_stvx_<mode>_1op): Likewise.
269 (altivec_lvx_<VM2:mode>): New define_expand.
270 (altivec_stvx_<VM2:mode>): Likewise.
271 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
272 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
273 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
274 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
275 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
276 (rs6000_gen_lvx): Likewise.
277 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
278 (altivec_expand_stv_builtin): Likewise.
279 (altivec_expand_builtin): Likewise.
280 * config/rs6000/vector.md: Likewise.
282 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
285 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
288 2018-03-20 Richard Biener <rguenther@suse.de>
291 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
292 sign-conversions as zero, fall back to standard scalar_stmt
295 2018-03-20 Martin Liska <mliska@suse.cz>
298 * predict.c (rebuild_frequencies): Handle case when we have
299 PROFILE_ABSENT, but flag_guess_branch_prob is false.
301 2018-03-20 Jakub Jelinek <jakub@redhat.com>
304 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
305 flag_section_anchors.
306 * varasm.c (use_blocks_for_decl_p): Remove hack for
310 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
312 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
313 be created, use lowpart_subreg of operands[0] rather than operands[0]
315 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
316 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
317 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
318 and n constraint instead of aarch64_shift_imm_di and Usd.
319 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
320 (*aarch64_<optab>_reg_minus<mode>3): ... this.
322 2018-03-20 Sudakshina Das <sudi.das@arm.com>
325 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
326 to favor GPR over NEON registers.
327 (<shift>di3_neon): Likewise.
329 2018-03-20 Tom de Vries <tom@codesourcery.com>
332 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
333 (nvptx_process_pars): Emit bar.sync asap and alap.
335 2018-03-20 Tom de Vries <tom@codesourcery.com>
338 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
339 seen_label if seen_label is already set.
341 2018-03-20 Jakub Jelinek <jakub@redhat.com>
344 * config/i386/i386.c (fold_builtin_cpu): For features above 31
345 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
346 Use 1U instead of 1. Formatting fixes.
349 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
350 instead of TREE_TYPE (s1) for the return value.
352 2018-03-19 Jakub Jelinek <jakub@redhat.com>
354 PR tree-optimization/84946
355 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
356 bitsize + bitsize in poly_uint64 rather than poly_int64.
359 * dwarf2asm.c: Include fold-const.c.
360 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
361 of decl rather than decl itself.
363 PR rtl-optimization/84643
364 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
366 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
369 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
370 calling assemble_variable.
372 2018-03-19 Sudakshina Das <sudi.das@arm.com>
375 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
376 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
378 2018-03-19 Jim Wilson <jimw@sifive.com>
381 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
382 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
383 (riscv_first_stack_step): Likewise.
384 (riscv_option_override): Use STACK_BOUNDARY instead of
386 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
388 (BIGGEST_ALIGNMENT): Set to 128.
389 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
390 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
393 2018-03-19 Richard Biener <rguenther@suse.de>
395 PR tree-optimization/84933
396 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
397 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
399 2018-03-19 Richard Biener <rguenther@suse.de>
401 PR tree-optimization/84859
402 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
403 (cond_if_else_store_replacement): Perform sinking operation on
404 single-store BBs regardless of MAX_STORES_TO_SINK setting.
405 Generalize what a BB with a single eligible store is.
407 2018-03-19 Richard Biener <rguenther@suse.de>
409 PR tree-optimization/84929
410 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
411 chrec_is_positive against non-chrec arg.
413 2018-03-19 Tamar Christina <tamar.christina@arm.com>
416 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
418 2018-03-18 Martin Liska <mliska@suse.cz>
420 PR rtl-optimization/84635
421 * regrename.c (build_def_use): Use matches_mode only when
424 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
426 PR tree-optimization/84913
427 * tree-vect-loop.c (vectorizable_reduction): Don't try to
428 vectorize chains of COND_EXPRs.
430 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
432 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
434 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
436 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
438 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
440 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
442 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
443 Kito Cheng <kito.cheng@gmail.com>
445 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
446 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
447 (nds32_adjust_reg_alloc_order): New function.
448 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
450 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
452 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
453 nds32_print_operand, nds32_print_operand_address): Use
454 HOST_WIDE_INT_PRINT_DEC instead.
456 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
458 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
460 2018-03-17 Jakub Jelinek <jakub@redhat.com>
463 * config/i386/i386.c (initial_ix86_tune_features,
464 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
466 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
467 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
468 rather than 1u << ix86_tune. Formatting fix.
469 (ix86_option_override_internal): Change ix86_arch_mask from
470 unsigned int to unsigned HOST_WIDE_INT, initialize to
471 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
472 (ix86_function_specific_restore): Likewise.
474 2018-03-16 Jakub Jelinek <jakub@redhat.com>
477 * postreload.c (reload_combine_recognize_pattern): Perform
478 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
479 truncate_int_for_mode the result for the destination's mode.
482 * hsa-gen.c (mem_type_for_type): Fix comment typo.
483 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
485 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
488 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
491 * lra-assigns.c (lra_split_hard_reg_for): Don't use
492 regno_allocno_class_array and sorted_pseudos.
493 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
494 insns where regno is used.
496 2018-03-16 Martin Liska <mliska@suse.cz>
499 * multiple_target.c (create_dispatcher_calls): Redirect
500 reference in the symbol table.
502 2018-03-16 Martin Liska <mliska@suse.cz>
505 * multiple_target.c (create_dispatcher_calls): Redirect also
508 2018-03-16 Jakub Jelinek <jakub@redhat.com>
512 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
513 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
514 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
516 2018-03-16 Julia Koval <julia.koval@intel.com>
518 * doc/invoke.texi (Skylake Server): Add CLWB.
519 Cannonlake): Remove CLWB.
521 2018-03-16 Jakub Jelinek <jakub@redhat.com>
523 PR tree-optimization/84841
524 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
526 (FLOAT_ONE_CONST_TYPE): Define.
527 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
528 (sort_by_operand_rank): Put entries with higher constant_type last
529 rather than first to match comments.
531 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
533 * config/nios2/nios2.md (movsi_internal): Fix thinko in
536 2018-03-15 Jakub Jelinek <jakub@redhat.com>
539 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
540 check and use address of target always.
542 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
545 * config/i386/i386.c (indirect_thunk_needed): Update comments.
546 (indirect_thunk_bnd_needed): Likewise.
547 (indirect_thunks_used): Likewise.
548 (indirect_thunks_bnd_used): Likewise.
549 (indirect_return_needed): New.
550 (indirect_return_bnd_needed): Likewise.
551 (output_indirect_thunk_function): Add a bool argument for
553 (output_indirect_thunk_function): Don't generate alias for
554 function return thunk.
555 (ix86_code_end): Call output_indirect_thunk_function to generate
556 function return thunks.
557 (ix86_output_function_return): Set indirect_return_bnd_needed
558 and indirect_return_needed instead of indirect_thunk_bnd_needed
559 and indirect_thunk_needed.
561 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
563 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
564 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
565 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
567 2018-03-15 David Malcolm <dmalcolm@redhat.com>
568 Paul Hua <paul.hua.gm@gmail.com>
571 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
573 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
575 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
576 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
579 2018-03-15 Tamar Christina <tamar.christina@arm.com>
582 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
583 instead of GET_MODE_SIZE when comparing Units.
585 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
588 * varasm.c (hash_section): Return an unchangeble hash value
589 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
590 Return !aarch64_can_use_per_function_literal_pools_p ().
592 2018-03-15 Jakub Jelinek <jakub@redhat.com>
595 * optabs.c (emit_conditional_move): Pass address of cmode's copy
596 rather than address of cmode as last argument to prepare_cmp_insn.
598 2018-03-15 Julia Koval <julia.koval@intel.com>
600 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
601 F_AVX512VNNI, F_AVX512BITALG): New.
603 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
606 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
607 insn for floating-point loads and stores.
609 2018-03-14 Carl Love <cel@us.ibm.com>
611 * config/rs6000/rs6000-c.c: Add macro definitions for
612 ALTIVEC_BUILTIN_VEC_PERMXOR.
613 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
614 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
615 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
616 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
618 * config/doc/extend.texi: Add prototypes for vec_permxor.
620 2018-03-14 David Malcolm <dmalcolm@redhat.com>
623 * diagnostic-show-locus.c (class layout_point): Convert m_line
624 from int to linenum_type.
625 (line_span::comparator): Use linenum "compare" function when
626 comparing line numbers.
627 (test_line_span): New function.
628 (layout_range::contains_point): Convert param "row" from int to
630 (layout_range::intersects_line_p): Likewise.
631 (layout::will_show_line_p): Likewise.
632 (layout::print_source_line): Likewise.
633 (layout::should_print_annotation_line_p): Likewise.
634 (layout::print_annotation_line): Likewise.
635 (layout::print_leading_fixits): Likewise.
636 (layout::annotation_line_showed_range_p): Likewise.
637 (struct line_corrections): Likewise for field m_row.
638 (line_corrections::line_corrections): Likewise for param "row".
639 (layout::print_trailing_fixits): Likewise.
640 (layout::get_state_at_point): Likewise.
641 (layout::get_x_bound_for_row): Likewise.
642 (layout::print_line): Likewise.
643 (diagnostic_show_locus): Likewise for locals "last_line" and
645 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
646 * input.c (selftest::test_linenum_comparisons): New function.
647 (selftest::input_c_tests): Call it.
648 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
649 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
650 * selftest.h (ASSERT_GT): New macro.
651 (ASSERT_GT_AT): New macro.
652 (ASSERT_LT): New macro.
653 (ASSERT_LT_AT): New macro.
655 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
657 PR rtl-optimization/84780
658 * combine.c (distribute_links): Don't make a link based on pc_rtx.
660 2018-03-14 Martin Liska <mliska@suse.cz>
662 * tree.c (record_node_allocation_statistics): Use
664 (get_stats_node_kind): New function extracted from
665 record_node_allocation_statistics.
666 (free_node): Use get_stats_node_kind.
668 2018-03-14 Richard Biener <rguenther@suse.de>
670 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
671 that the value-set of ANTIC_IN doesn't grow.
674 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
676 (BB_VISITED_WITH_VISITED_SUCCS): New define.
677 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
679 2018-03-14 Julia Koval <julia.koval@intel.com>
681 * config.gcc (icelake-client, icelake-server): New.
683 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
684 (initial_ix86_arch_features): Ditto.
685 (PTA_SKYLAKE): Add SGX.
686 (PTA_ICELAKE): Remove.
687 (PTA_ICELAKE_CLIENT): New.
688 (PTA_ICELAKE_SERVER): New.
689 (ix86_option_override_internal): Split up icelake on icelake client and
691 (get_builtin_code_for_version): Ditto.
692 (fold_builtin_cpu): Ditto.
693 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
694 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
695 * config/i386/i386.h (processor_type): Ditto.
696 * doc/invoke.texi: Ditto.
698 2018-03-14 Jakub Jelinek <jakub@redhat.com>
701 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
702 INTEGER_CST offset, add it together with bitpos / 8 and
703 sign extend based on POINTER_SIZE.
707 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
710 * config/i386/constraints.md (Yc): New register constraint.
711 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
712 Use Yc constraint for alternative 2 of operand 0. Remove
713 preferred_for_speed attribute.
715 2018-03-14 Richard Biener <rguenther@suse.de>
717 PR tree-optimization/84830
718 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
719 with the old one to avoid oscillations.
721 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
724 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
726 (assign_by_spills): Return a flag of reload assignment failure.
727 Do not process the reload assignment failures. Do not spill other
728 reload pseudos if they has the same reg class. Update n if
730 (lra_assign): Add a return arg. Set up from the result of
731 assign_by_spills call.
732 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
733 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
734 usage_insns if it is not NULL.
735 (spill_hard_reg_in_range): New function.
736 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
737 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
739 (lra_assign): Change prototype.
740 * lra.c (lra): Add code to deal with fails by splitting hard reg
743 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
745 * config/riscv/riscv.opt (mrelax): New option.
746 * config/riscv/riscv.c (riscv_file_start): Emit ".option
747 "norelax" when riscv_mrelax is disabled.
748 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
750 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
753 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
754 reassociation for int modes.
756 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
758 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
759 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
761 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
762 * config/aarch64/aarch64-sve.md
763 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
764 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
765 (*extend<mode><Vwide>2): Rename to...
766 (aarch64_sve_extend<mode><Vwide>2): ...this.
767 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
768 renaming the old pattern to...
769 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
771 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
772 define_expand, renaming the old pattern to...
773 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
774 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
775 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
776 account when deciding which SVE instruction the optab should use.
777 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
779 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
781 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
782 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
783 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
784 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
785 (tlsdesc_small_<mode>): Turn a define_expand and use
786 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
787 (tlsdesc_small_advsimd_<mode>): ...this.
788 (tlsdesc_small_sve_<mode>): New pattern.
790 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
792 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
793 (UNSPEC_UMUL_HIGHPART): New constants.
794 (MUL_HIGHPART): New int iteraor.
795 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
796 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
798 (*<su>mul<mode>3_highpart): New define_insn.
800 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
803 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
806 2018-03-13 Martin Liska <mliska@suse.cz>
809 * (sem_item_optimizer::sem_item_optimizer): Initialize new
811 (sem_item_optimizer::~sem_item_optimizer): Release it.
812 (sem_item_optimizer::merge_classes): Register variable aliases.
813 (sem_item_optimizer::fixup_pt_set): New function.
814 (sem_item_optimizer::fixup_points_to_sets): Likewise.
815 * ipa-icf.h: Declare new variables and functions.
817 2018-03-13 Jakub Jelinek <jakub@redhat.com>
820 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
821 integer_pow2p@2 and test integer_pow2p in condition.
822 (A < 0 ? C : 0): Similarly for @1.
825 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
826 characters starting at p contain '\0' character, don't look beyond
830 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
831 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
834 * reg-stack.c (change_stack): Change update_end var from int to
835 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
836 also call set_block_for_insn on the newly added insns and rescan.
839 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
843 * tree.c (stabilize_reference_1): Return save_expr (e) for
844 STATEMENT_LIST even if it doesn't have side-effects.
846 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
848 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
850 2018-03-12 Renlin Li <renlin.li@arm.com>
852 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
853 aarch64_output_scalar_simd_mov_immediate.
855 2018-03-12 Martin Sebor <msebor@redhat.com>
857 PR tree-optimization/83456
858 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
859 for perfectly overlapping calls to memcpy.
860 (gimple_fold_builtin_memory_chk): Same.
861 (gimple_fold_builtin_strcpy): Handle no-warning.
862 (gimple_fold_builtin_stxcpy_chk): Same.
863 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
865 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
867 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
868 parameter. Use it for SFmode.
869 (rs6000_function_arg_advance_1): Adjust.
870 (rs6000_function_arg): Adjust.
871 (rs6000_gimplify_va_arg): Pass false for that new parameter.
873 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
875 PR rtl-optimization/84169
876 PR rtl-optimization/84780
877 * combine.c (can_combine_p): Check for a 2-insn combination whether
878 the destination register is used between the two insns, too.
880 2018-03-12 Richard Biener <rguenther@suse.de>
882 PR tree-optimization/84803
883 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
884 for refs DR analysis didn't process.
886 2018-03-12 Richard Biener <rguenther@suse.de>
888 PR tree-optimization/84777
889 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
890 force-vectorize loops ignore whether we are optimizing for size.
892 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
894 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
895 (TARGET_MD_ASM_ADJUST): Define.
897 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
898 Kito Cheng <kito.cheng@gmail.com>
899 Chung-Ju Wu <jasonwucj@gmail.com>
901 * config/nds32/nds32.c (nds32_compute_stack_frame,
902 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
903 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
904 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
905 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
906 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
907 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
908 * config/nds32/nds32.md (prologue, epilogue): Use macro
909 NDS32_V3PUSH_AVAILABLE_P to do checking.
911 2018-03-11 Jakub Jelinek <jakub@redhat.com>
914 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
915 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
916 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
917 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
918 addition of most attributes on !orig_type_die or the attribute not
919 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
921 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
922 Chung-Ju Wu <jasonwucj@gmail.com>
924 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
926 * config/nds32/nds32.opt (mvh): New option.
928 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
929 Chung-Ju Wu <jasonwucj@gmail.com>
931 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
933 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
934 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
937 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
938 Chung-Ju Wu <jasonwucj@gmail.com>
940 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
942 * config/nds32/nds32-multiple.md (strlensi): New pattern.
943 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
945 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
946 Kito Cheng <kito.cheng@gmail.com>
947 Chung-Ju Wu <jasonwucj@gmail.com>
949 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
950 UNSPEC_FFMISM and UNSPEC_FLMISM.
951 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
952 for ffb, ffmism and flmism.
953 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
954 (unspec_ffmism): Ditto.
955 (unspec_flmism): Ditto.
956 (nds32_expand_builtin_impl): Check if string extension is available.
957 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
958 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
960 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
963 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
966 * lra-assigns.c (assign_by_spills): Return a flag of reload
967 assignment failure. Do not process the reload assignment
968 failures. Do not spill other reload pseudos if they has the same
970 (lra_assign): Add a return arg. Set up from the result of
971 assign_by_spills call.
972 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
973 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
974 usage_insns if it is not NULL.
975 (spill_hard_reg_in_range): New function.
976 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
977 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
979 (lra_assign): Change prototype.
980 * lra.c (lra): Add code to deal with fails by splitting hard reg
983 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
986 * config/i386/i386.opt: Replace Enforcment with Enforcement.
988 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
991 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
992 (dw_val_node): Add val_symbolic_view.
993 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
994 (symview_upper_bound): New.
995 (new_line_info_table): Initialize symviews_since_reset.
996 (dwarf2out_source_line): Count symviews_since_reset and set
998 (dw_val_equal_p): Handle symview.
999 (add_AT_symview): New.
1000 (print_dw_val): Handle symview.
1001 (attr_checksum, attr_checksum_ordered): Likewise.
1002 (same_dw_val_p, size_of_die): Likewise.
1003 (value_format, output_die): Likewise.
1004 (add_high_low_attributes): Use add_AT_symview for entry_view.
1005 (dwarf2out_finish): Reset symview_upper_bound, clear
1008 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1011 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1012 Add strict argument and use it.
1013 (rs6000_split_multireg_move): Update for new strict argument.
1014 (mem_operand_gpr): Disallow all non-offsettable addresses.
1015 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1017 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1020 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1021 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1022 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1025 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1026 decl, use remap_type if we want to use the type.
1028 2018-03-09 Martin Sebor <msebor@redhat.com>
1030 PR tree-optimization/84526
1031 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1033 (builtin_access::generic_overlap): Be prepared to handle non-array
1036 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1038 PR rtl-optimization/84682
1039 * lra-constraints.c (process_address_1): Check is_address flag
1040 for address constraints.
1041 (process_alt_operands): Likewise.
1042 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1043 preprocess_constraints.
1044 * recog.h (preprocess_constraints): Add oploc parameter.
1046 * recog.c (preprocess_constraints): Test address_operand for
1047 CT_ADDRESS constraints.
1049 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1052 * lra-assigns.c (assign_by_spills): Return a flag of reload
1053 assignment failure. Do not process the reload assignment
1054 failures. Do not spill other reload pseudos if they has the same
1056 (lra_assign): Add a return arg. Set up from the result of
1057 assign_by_spills call.
1058 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1059 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1060 usage_insns if it is not NULL.
1061 (spill_hard_reg_in_range): New function.
1062 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1063 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1064 function prototypes.
1065 (lra_assign): Change prototype.
1066 * lra.c (lra): Add code to deal with fails by splitting hard reg
1069 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1072 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1073 Accept complain bool parameter. Only emit errors if it is true.
1074 (arm_parse_cpu_option_name): Likewise.
1075 (arm_target_thumb_only): Adjust callers of the above.
1076 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1077 prototype to take a default true bool parameter.
1078 (arm_parse_arch_option_name): Likewise.
1080 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1081 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1085 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1086 * configure: Regenerate.
1087 * configure.ac ("linker --version-script option"): New.
1088 ("linker soname option"): New.
1090 2018-03-09 Richard Biener <rguenther@suse.de>
1092 PR tree-optimization/84775
1093 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1094 immediate uses of predicate stmts and mark them modified.
1097 PR tree-optimization/84178
1098 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1100 (version_loop_for_if_conversion): Delay update_ssa call.
1101 (tree_if_conversion): Delay update_ssa until after predicate
1104 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1107 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1108 when the function accesses prior frames.
1110 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1113 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1114 gen_llsym, otherwise call maybe_gen_llsym.
1117 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1118 has ',' character inside of it.
1120 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1123 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1124 as clobbering CC_REGNUM.
1126 2018-03-08 Richard Biener <rguenther@suse.de>
1129 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1130 (follow_copies_to_constant): Do not follow SSA names registered
1133 2018-03-08 Richard Biener <rguenther@suse.de>
1135 PR tree-optimization/84178
1136 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1138 (version_loop_for_if_conversion): Delay update_ssa call.
1139 (tree_if_conversion): Delay update_ssa until after predicate
1142 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1144 PR tree-optimization/84178
1145 * tree-if-conv.c (release_bb_predicate): Remove the
1146 the assertion that the stmts have NULL use_ops.
1147 Discard the statements, asserting that they haven't
1148 yet been added to a BB.
1150 2018-03-08 Richard Biener <rguenther@suse.de>
1152 PR tree-optimization/84746
1153 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1154 (phi_translate): Pass in destination ANTIC_OUT set.
1155 (phi_translate_1): Likewise. For a simplified result lookup
1156 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1157 (phi_translate_set): Adjust.
1158 (do_pre_regular_insertion): Likewise.
1159 (do_pre_partial_partial_insertion): Likewise.
1161 2018-03-08 Martin Liska <mliska@suse.cz>
1163 PR gcov-profile/84735
1164 * doc/gcov.texi: Document usage of profile files.
1165 * gcov-io.h: Document changes in the format.
1167 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1171 * dwarf2out.c (struct dw_line_info_table): Update comments for
1173 (FORCE_RESET_NEXT_VIEW): New.
1174 (FORCE_RESETTING_VIEW_P): New.
1175 (RESETTING_VIEW_P): Check for -1 too.
1176 (ZERO_VIEW_P): Likewise.
1177 (new_line_info_table): Force-reset next view.
1178 (dwarf2out_begin_function): Likewise.
1179 (dwarf2out_source_line): Simplify zero_view_p initialization.
1180 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1181 view directly. Omit view when omitting .loc at line 0.
1183 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1185 PR tree-optimization/84740
1186 * tree-switch-conversion.c (process_switch): Call build_constructors
1187 only if info.phi_count is non-zero.
1189 PR tree-optimization/84739
1190 * tree-tailcall.c (find_tail_calls): Check call arguments against
1191 DECL_ARGUMENTS (current_function_decl) rather than
1192 DECL_ARGUMENTS (func) when checking for tail recursion.
1194 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1196 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1197 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1198 Volker Reichelt's entry and add entries for people that perform
1199 GCC fuzzy testing and report numerous bugs.
1201 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1204 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1205 readonly data in sdata, if that is disabled.
1206 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1207 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1208 -mreadonly-in-sdata option.
1210 2018-03-07 Martin Sebor <msebor@redhat.com>
1212 PR tree-optimization/84468
1213 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1214 basic block when looking for nul assignment.
1216 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1219 * except.h (output_function_exception_table): Adjust prototype.
1220 * except.c (output_function_exception_table): Remove FNNAME parameter
1221 and add SECTION parameter. Ouput one part of the table at a time.
1222 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1223 the first part of the exception table and emit unwind directives.
1224 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1225 (i386_pe_seh_cold_init): Likewise.
1226 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1227 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1228 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1229 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1230 * config/i386/winnt.c: Include except.h.
1231 (struct seh_frame_state): Add reg_offset, after_prologue and
1232 in_cold_section fields.
1233 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1234 (i386_pe_seh_cold_init): New function.
1235 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1236 to seh->in_cold_section.
1237 (seh_emit_push): Record the offset of the push.
1238 (seh_emit_save): Record the offet of the save.
1239 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1240 Test seh->after_prologue to disregard the epilogue.
1241 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1242 (i386_pe_end_cold_function): New function.
1244 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1247 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1248 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1251 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1253 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1254 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1257 * multiple_target.c: Include tree-inline.h and intl.h.
1258 (expand_target_clones): Diagnose and fail if node->definition and
1259 !tree_versionable_function_p (node->decl).
1261 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1263 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1265 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1266 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1267 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1269 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1272 * combine.c (try_combine): Use reg_or_subregno instead of handling
1273 just paradoxical SUBREGs and REGs.
1275 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1277 * config/arc/arc.c (arc_finalize_pic): Remove function.
1278 (arc_must_save_register): We use single base PIC register, remove
1279 checks to save/restore the PIC register.
1280 (arc_expand_prologue): Likewise.
1281 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1283 (arc_verify_short): Likewise.
1284 (arc_attr_type): Likewise.
1285 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1286 (walk_stores): Likewise.
1287 (arc_address_cost): Make it static.
1288 (arc_verify_short): Likewise.
1289 (branch_dest): Likewise.
1290 (arc_attr_type): Likewise.
1291 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1292 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1293 (arc_final_prescan_insn): Remove inserting the nops due to
1294 hardware hazards. It is done in reorg step.
1295 (insn_length_variant_t): Remove.
1296 (insn_length_parameters_t): Likewise.
1297 (arc_insn_length_parameters): Likewise.
1298 (arc_get_insn_variants): Likewise.
1299 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1301 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1304 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1307 PR tree-optimization/84687
1308 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1310 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1312 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1314 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1315 Rename to ppc_speculation_barrier.
1316 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1317 __builtin_ppc_speculation_barrier.
1319 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1322 * combine.c (combine_simplify_rtx): Don't try to simplify if
1323 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1326 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1328 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1329 to 32 bytes when compiling for POWER9.
1331 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1334 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1335 regparm >= 3 with no arg reg available also for calls with
1336 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1339 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1341 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1343 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1346 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1348 2018-03-05 Richard Biener <rguenther@suse.de>
1350 PR tree-optimization/84486
1351 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1352 When inserting a __builtin_assume_aligned call set the LHS
1353 SSA name alignment info accordingly.
1355 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1357 PR tree-optimization/84114
1358 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1359 Avoid reassociation of FLOAT_MODE addition.
1361 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1363 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1364 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1365 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1366 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1367 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1368 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1369 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1371 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1372 __builtin_ia32_wbinvd): New builtins.
1373 (SPECIAL_ARGS2): New.
1374 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1375 (SPECIAL_ARGS2): New.
1376 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1377 (ix86_valid_target_attribute_inner_p): Ditto.
1378 (ix86_init_mmx_sse_builtins): Add special_args2.
1379 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1380 TARGET_WBNOINVD_P): New.
1381 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1382 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1383 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1384 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1385 * config/i386/pconfigintrin.h: New file.
1386 * config/i386/wbnoinvdintrin.h: Ditto.
1387 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1388 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1390 2018-03-05 Richard Biener <rguenther@suse.de>
1392 PR tree-optimization/84670
1393 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1395 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1396 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1397 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1398 doesn't grow if all successors (recursively) were visited at least
1401 2018-03-05 Richard Biener <rguenther@suse.de>
1403 PR tree-optimization/84650
1404 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1405 if executed in the loop pipeline.
1407 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1409 * doc/configfiles.texi (Configuration Files): Move info about
1410 conditionalizing $target-protos.h to...
1411 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1412 differs from $target-protos.h.
1414 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1415 Chung-Ju Wu <jasonwucj@gmail.com>
1417 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1418 * config/nds32/nds32-multiple.md (setmemsi): Define.
1419 * config/nds32/nds32-memory-manipulation.c
1420 (nds32_gen_dup_4_byte_to_word_value): New.
1421 (emit_setmem_word_loop): New.
1422 (emit_setmem_byte_loop): New.
1423 (nds32_expand_setmem_loop): New.
1424 (nds32_expand_setmem_loop_v3m): New.
1425 (nds32_expand_setmem_unroll): New.
1426 (nds32_expand_setmem): New.
1428 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1429 Chung-Ju Wu <jasonwucj@gmail.com>
1431 * config/nds32/nds32-memory-manipulation.c
1432 (nds32_emit_load_store): New.
1433 (nds32_emit_post_inc_load_store): New.
1434 (nds32_emit_mem_move): New.
1435 (nds32_emit_mem_move_block): New.
1436 (nds32_expand_movmemsi_loop_unknown_size): New.
1437 (nds32_expand_movmemsi_loop_known_size): New.
1438 (nds32_expand_movmemsi_loop): New.
1439 (nds32_expand_movmemsi_unroll): New.
1440 (nds32_expand_movmemqi): Rename ...
1441 (nds32_expand_movmemsi): ... to this.
1442 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1443 (movmemsi): ... to this.
1444 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1445 (nds32_expand_movmemsi): ... to this.
1447 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1448 Monk Chiang <sh.chiang04@gmail.com>
1449 Chung-Ju Wu <jasonwucj@gmail.com>
1451 * config/nds32/nds32-protos.h
1452 (nds32_expand_load_multiple): New arguments.
1453 (nds32_expand_store_multiple): Ditto.
1454 (nds32_valid_multiple_load_store): Rename ...
1455 (nds32_valid_multiple_load_store_p): ... to this.
1456 * config/nds32/nds32-memory-manipulation.c
1457 (nds32_expand_load_multiple): Refine implementation.
1458 (nds32_expand_store_multiple): Ditto.
1459 * config/nds32/nds32-multiple.md
1460 (load_multiple): Update nds32_expand_load_multiple interface.
1461 (store_multiple): Update nds32_expand_store_multiple interface.
1462 * config/nds32/nds32-predicates.c
1463 (nds32_valid_multiple_load_store): Rename ...
1464 (nds32_valid_multiple_load_store_p): ... to this and refine
1466 * config/nds32/predicates.md
1467 (nds32_load_multiple_and_update_address_operation): New predicate.
1468 (nds32_store_multiple_and_update_address_operation): New predicate.
1470 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1471 Chung-Ju Wu <jasonwucj@gmail.com>
1473 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1474 (combo): New attribute.
1475 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1477 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1479 * config/nds32/nds32.opt: Change -mcmodel= default value.
1481 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1482 Monk Chiang <sh.chiang04@gmail.com>
1483 Chung-Ju Wu <jasonwucj@gmail.com>
1485 * config/nds32/constants.md (unspec_element): New enum.
1486 * config/nds32/constraints.md (Umw): New constraint.
1487 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1488 * config/nds32/nds32-intrinsic.md: Likewise.
1489 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1490 (nds32_valid_smw_lwm_base_p): New.
1491 (nds32_output_smw_single_word): New.
1492 (nds32_output_lmw_single_word): New.
1493 (nds32_expand_unaligned_load): New.
1494 (nds32_expand_unaligned_store): New.
1495 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1496 (nds32_output_smw_single_word): Declare.
1497 (nds32_output_lmw_single_word): Declare.
1498 (nds32_expand_unaligned_load): Declare.
1499 (nds32_expand_unaligned_store): Declare.
1500 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1501 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1502 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1503 NDS32_BUILTIN_UASTORE_DW.
1504 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1507 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1508 Kito Cheng <kito.cheng@gmail.com>
1509 Chung-Ju Wu <jasonwucj@gmail.com>
1511 * config/nds32/nds32-intrinsic.c
1512 (nds32_expand_builtin_null_ftype_reg): Delete.
1513 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1514 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1515 (nds32_read_argument): New.
1516 (nds32_legitimize_target): Ditto.
1517 (nds32_legitimize_argument): Ditto.
1518 (nds32_check_constant_argument): Ditto.
1519 (nds32_expand_unop_builtin): Ditto.
1520 (nds32_expand_unopimm_builtin): Ditto.
1521 (nds32_expand_binop_builtin): Ditto.
1522 (nds32_builtin_decl_impl): Ditto.
1523 (builtin_description): Ditto.
1524 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1525 (nds32_init_builtins_impl): Ditto.
1526 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1527 (nds32_builtin_decl): New.
1528 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1529 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1531 2018-03-02 Jeff Law <law@redhat.com>
1533 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1534 (redundant_insn, fill_simple_delay_slots): Likewise.
1535 (fill_slots_from_thread): Likewise.
1536 * resource.c (mark_referenced_resources): Likewise.
1537 (mark_set_resources, find_dead_or_set_registers): Likewise.
1539 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1541 * substring-locations.h (format_warning_va): Formatting fix for
1543 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1545 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1546 * substring-locations.c: Include intl.h.
1547 (format_warning_va): Turned into small wrapper around
1548 format_warning_n_va, renamed to ...
1549 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1550 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1552 (format_warning_at_substring_n): New function.
1553 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1554 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1555 format_warning_at_substring with just a shorter name instead of
1556 const function pointer.
1557 (fmtwarn_n): New function.
1558 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1559 appropriate, get rid of all the fmtstr temporaries, move conditionals
1560 with G_() wrapped string literals directly into fmtwarn arguments,
1561 cast dir.len to (int), formatting fixes.
1563 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1565 * doc/invoke.texi: Remove "Cilk Plus" references.
1567 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1568 Richard Biener <rguenther@suse.de>
1571 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1572 for error or warning attributes if CALL_FROM_THUNK_P is set.
1575 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1578 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1579 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1582 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1583 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1585 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1586 instead of -1U in last predictors element's probability member.
1588 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1591 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1592 arguments if they are comparable.
1594 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1596 PR tree-optimization/84634
1597 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1598 masks and masked_loop_p with a single loop_masks, making sure it's
1599 null for bb vectorization.
1601 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1603 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1604 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1605 loop->force_vectorize to check whether there is no alias.
1607 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1610 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1612 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1614 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1615 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1616 instead of a loop around prev_real_insn.
1617 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1621 * config/i386/i386.c (ix86_print_operand): Use conditional
1622 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1625 2018-03-02 Richard Biener <rguenther@suse.de>
1627 PR tree-optimization/84427
1628 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1629 (bitmap_set_subtract_values): Rewrite to handle multiple
1632 (prune_clobbered_mems): Likewise.
1633 (phi_translate): Take edge instead of pred/phiblock.
1634 (phi_translate_1): Likewise.
1635 (phi_translate_set): Likewise. Insert all translated
1636 exprs for a value into the set, keeping possibly multiple
1637 expressions per value.
1638 (compute_antic_aux): Adjust for phi_translate changes.
1639 When intersecting union the expressions and prune those
1640 not in the final value set, keeping possibly multiple
1641 expressions per value. Do not use value-insertion
1642 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1643 all expressions. Add verification that the value-sets
1644 only shrink during iteration.
1645 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1646 (do_pre_regular_insertion): Likewise.
1647 (do_pre_partial_partial_insertion): Likewise.
1649 2018-03-02 Richard Biener <rguenther@suse.de>
1652 * config/darwin.c (saved_debug_info_level): New static global.
1653 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1654 (darwin_asm_lto_end): Restore debug info generation settings.
1656 2018-03-01 Martin Liska <mliska@suse.cz>
1659 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1662 2018-03-01 Richard Biener <rguenther@suse.de>
1665 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1666 type annotation with LTO when debug was disabled at compile-time.
1668 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1670 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1672 (mips_final_postscan_insn): Likewise.
1674 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1676 PR rtl-optimization/84528
1677 * alias.c (init_alias_target): Add commentary.
1678 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1679 a unique base value if the frame pointer is not eliminated
1680 to the stack pointer.
1682 2018-03-01 Tom de Vries <tom@codesourcery.com>
1684 PR rtl-optimization/83327
1685 * lra-int.h (hard_regs_spilled_into): Declare.
1686 * lra.c (hard_regs_spilled_into): Define.
1687 (init_reg_info): Init hard_regs_spilled_into.
1688 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1689 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1690 (process_bb_lives): Handle hard_regs_spilled_into.
1691 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1692 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1694 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1696 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1697 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1698 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1699 * config/rs6000/aix72.h: New file.
1701 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1703 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1704 instead of warning_at with conditional singular and plural messages
1708 * stor-layout.c (update_alignment_for_field): For
1709 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1710 && !DECL_PACKED (field), do the alignment update, just use
1711 only desired_align instead of MAX (type_align, desired_align)
1713 (place_field): Don't do known_align < desired_align handling
1714 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1715 is non-NULL, instead do it after rli->prev_field handling and
1716 only if not within a bitfield word. For DECL_PACKED (field)
1717 use type_align of BITS_PER_UNIT.
1719 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1721 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1722 superfluous parentheses and trailing spaces.
1724 2018-02-28 Richard Biener <rguenther@suse.de>
1726 PR tree-optimization/84584
1727 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1728 SCoPs with fake exit edge.
1730 2018-02-28 Martin Liska <mliska@suse.cz>
1733 * timevar.c (timer::print): Fix format to properly print 100%
1736 2018-02-28 Richard Biener <rguenther@suse.de>
1739 * genmatch.c (capture_info::walk_match): Do not mark
1740 captured expressions without operands as expr_p given
1741 they act more like predicates and should be subject to
1742 "lost tail" side-effect preserving.
1744 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1746 PR rtl-optimization/81611
1747 * auto-inc-dec.c (attempt_change): Move dead note from
1748 mem_insn if it's the next use of regno
1749 (find_address): Take address use of reg holding
1750 non-incremented value. Add parm to limit search to the named
1752 (merge_in_block): Attempt to use a mem insn that is the next
1753 use of the original regno.
1755 2018-02-27 Martin Sebor <msebor@redhat.com>
1758 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1759 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1761 2018-02-27 Martin Sebor <msebor@redhat.com>
1763 PR translation/84207
1764 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1765 n argument to unsigned HOST_WIDE_INT.
1766 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1767 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1768 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1769 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1771 2018-02-27 Richard Biener <rguenther@suse.de>
1773 PR tree-optimization/84512
1774 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1775 Do not use the estimate returned from record_stmt_cost for
1776 the scalar iteration cost but sum properly using add_stmt_cost.
1778 2018-02-27 Richard Biener <rguenther@suse.de>
1780 PR tree-optimization/84466
1781 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1782 Adjust last change to less strictly validate use operands.
1784 2018-02-27 Martin Liska <mliska@suse.cz>
1786 PR gcov-profile/84548
1787 * gcov.c (process_file): Allow partial overlap and consider it
1788 also as group functions.
1789 (output_lines): Properly calculate range of lines for a group.
1791 2018-02-27 Martin Liska <mliska@suse.cz>
1793 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1794 'ggc' suffixes. Change first column width.
1795 (timer::print): Fix formatting of the column.
1797 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1799 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1800 preserve inline entry blocks for the sake of debug inline
1801 entry point markers alone.
1802 (remove_unused_locals): Suggest in comments a better place to
1803 force the preservation of inline entry blocks that are
1804 otherwise unused, but do not preserve them.
1806 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1808 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1810 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1813 * config/i386/constraints.md (Bs): Replace
1814 ix86_indirect_branch_register with
1815 TARGET_INDIRECT_BRANCH_REGISTER.
1817 * config/i386/i386.md (indirect_jump): Likewise.
1818 (tablejump): Likewise.
1819 (*sibcall_memory): Likewise.
1820 (*sibcall_value_memory): Likewise.
1821 Peepholes of indirect call and jump via memory: Likewise.
1822 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1823 (*sibcall_value_GOT_32): Likewise.
1824 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1825 (GOT_memory_operand): Likewise.
1826 (call_insn_operand): Likewise.
1827 (sibcall_insn_operand): Likewise.
1828 (GOT32_symbol_operand): Likewise.
1829 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1831 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1833 PR rtl-optimization/83496
1834 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1835 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1836 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1837 redundant insn, if any.
1838 (relax_delay_slots): Likewise.
1839 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1841 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1843 PR tree-optimization/83965
1844 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1845 that grouped statements are part of a reduction chain. Return
1846 true if the statement is not marked as a reduction itself but
1848 (vect_recog_dot_prod_pattern): Don't check whether the statement
1849 is part of a group here.
1850 (vect_recog_sad_pattern): Likewise.
1851 (vect_recog_widen_sum_pattern): Likewise.
1853 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1856 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1858 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1861 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1863 (ix86_output_indirect_function_return): New prototype.
1864 (ix86_split_simple_return_pop_internal): Likewise.
1865 * config/i386/i386.c (indirect_return_via_cx): New.
1866 (indirect_return_via_cx_bnd): Likewise.
1867 (indirect_thunk_name): Handle return va CX_REG.
1868 (output_indirect_thunk_function): Create alias for
1869 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1870 (ix86_output_indirect_jmp): Remove the bool argument.
1871 (ix86_output_indirect_function_return): New function.
1872 (ix86_split_simple_return_pop_internal): Likewise.
1873 * config/i386/i386.md (*indirect_jump): Don't pass false
1874 to ix86_output_indirect_jmp.
1875 (*tablejump_1): Likewise.
1876 (simple_return_pop_internal): Change it to define_insn_and_split.
1877 Call ix86_split_simple_return_pop_internal to split it for
1879 (simple_return_indirect_internal): Call
1880 ix86_output_indirect_function_return instead of
1881 ix86_output_indirect_jmp.
1883 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1886 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1887 memset and value initialization afterwards.
1889 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1891 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1893 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1896 * common/config/aarch64/aarch64-common.c
1897 (aarch_option_optimization_table[]): Switch
1898 off fomit-frame-pointer
1900 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1901 Chung-Ju Wu <jasonwucj@gmail.com>
1903 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1905 (store_multiple): Ditto.
1907 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1909 * config.gcc: Add --with-cpu support for nds32 target.
1910 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1911 * config/nds32/nds32.opt: Add -mcpu= option.
1913 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1915 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1916 isel=yes): Warn for these deprecated options.
1918 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1920 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1921 ISA_2_5_MASKS_EMBEDDED.
1923 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1925 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1926 p->max as pointers rather than using iterative_hash_expr.
1928 2018-02-23 Carl Love <cel@us.ibm.com>
1930 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1931 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1933 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1934 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1935 P8V_BUILTIN_VEC_VUNSIGNED2.
1937 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1940 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1941 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1943 * lra-constraints.c (curr_insn_transform): Set up
1944 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1946 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1947 * lra-eliminations.c (spill_pseudos): Ditto.
1948 (process_insn_for_elimination): Ditto.
1949 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1950 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1951 LRA_NON_CLOBBERED_ALT.
1953 2018-02-22 Martin Sebor <msebor@redhat.com>
1955 PR tree-optimization/84480
1956 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1957 to maybe_diag_stxncpy_trunc. Call it.
1958 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1959 from gimple_fold_builtin_strcpy. Print inlining stack.
1960 (handle_builtin_stxncpy): Print inlining stack.
1961 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1963 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1966 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1967 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1968 and -fcheck-pointer-bounds are used together.
1969 (indirect_thunk_prefix): New enum.
1970 (indirect_thunk_need_prefix): New function.
1971 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1972 "_nt" instead of "_bnd" for NOTRACK prefix.
1973 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1974 (output_indirect_thunk_function): Likewise.
1976 (ix86_code_end): Update output_indirect_thunk_function calls.
1977 (ix86_output_indirect_branch_via_reg): Replace
1978 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1979 (ix86_output_indirect_branch_via_push): Likewise.
1980 (ix86_output_function_return): Likewise.
1981 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1982 incompatible with -fcf-protection=branch and
1983 -fcheck-pointer-bounds.
1985 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1988 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1989 Change gcc_assert call to output_operand_lossage.
1991 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1993 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1995 2018-02-22 DJ Delorie <dj@redhat.com>
1996 Sebastian Perta <sebastian.perta@renesas.com>
1997 Oleg Endo <olegendo@gcc.gnu.org>
1999 * config/rx/rx.c (rx_rtx_costs): New function.
2000 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2002 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2004 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2006 2018-02-22 Martin Liska <mliska@suse.cz>
2009 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2010 Add "native" as a possible value.
2012 2018-02-22 Martin Liska <mliska@suse.cz>
2015 * config/i386/i386.c (ix86_option_override_internal):
2016 Add "native" as a possible value for -march and -mtune.
2018 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2021 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2022 to all type variants.
2024 PR tree-optimization/84503
2025 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2026 width as info->bitpos + info->bitsize - start.
2027 (merged_store_group::merge_overlapping): Simplify width computation.
2028 (check_no_overlap): New function.
2029 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2030 start + width and last_order of the group, fail if check_no_overlap
2032 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2033 to group if check_no_overlap fails.
2035 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2037 * config/rs6000/altivec.md: Delete contraint arguments to
2038 define_expand, define_split, and define_peephole2, and in
2039 define_insn_and_split if always unused.
2040 * config/rs6000/darwin.md: Ditto.
2041 * config/rs6000/dfp.md: Ditto.
2042 * config/rs6000/rs6000.md: Ditto.
2043 * config/rs6000/sync.md: Ditto.
2044 * config/rs6000/vector.md: Ditto.
2045 * config/rs6000/vsx.md: Ditto.
2047 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2049 * config/rs6000/altivec.md: Write output control strings as braced
2050 blocks instead of double-quoted strings.
2051 * config/rs6000/darwin.md: Ditto.
2052 * config/rs6000/rs6000.md: Ditto.
2053 * config/rs6000/vector.md: Ditto.
2054 * config/rs6000/vsx.md: Ditto.
2056 2018-02-21 Jason Merrill <jason@redhat.com>
2058 PR c++/84314 - ICE with templates and fastcall attribute.
2059 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2061 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2063 * ipa-cp.c (determine_versionability): Fix comment typos.
2065 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2068 * ipa-cp.c (determine_versionability): Do not version functions caling
2071 2018-02-21 Martin Liska <mliska@suse.cz>
2074 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2075 Add "native" as a possible value.
2076 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2077 the macro when native cpu detection is available.
2079 2018-02-21 Martin Liska <mliska@suse.cz>
2082 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2083 Add "native" as a possible value.
2084 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2085 when native cpu detection is available.
2087 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2088 Martin Sebor <msebor@redhat.com>
2090 PR tree-optimization/84478
2091 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2093 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2094 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2095 support which is conservatively correct, for 2 only stay conservative
2096 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2097 argument to the 2 argument get_range_strlen, adjust 6 arg
2098 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2100 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2101 (gimple_fold_builtin_strlen): Pass true as last argument to
2104 2018-02-20 Martin Sebor <msebor@redhat.com>
2107 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2108 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2109 (builtin_memref::builtin_memref): Factor out parts into
2110 set_base_and_offset and call it.
2112 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2115 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2116 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2117 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2118 search at the associated MODE_INT.
2120 2018-02-20 Jeff Law <law@redhat.com>
2123 PR tree-optimization/81592
2125 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2126 for range data rather than using global data.
2127 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2128 range data rather than using global data.
2129 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2130 pass it to children as needed.
2131 (struct directive::fmtresult): Similarly.
2132 (struct directive::set_width): Similarly.
2133 (struct directive::set_precision): Similarly.
2134 (format_integer, format_directive, parse_directive): Similarly.
2135 (format_none): Accept unnamed vr_values parameter.
2136 (format_percent, format_floating, format_character): Similarly.
2137 (format_string, format_plain): Similarly.
2138 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2139 the EVRP range analyzer for range data rather than using global data.
2140 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2141 gimple-ssa-evrp-analyze.h
2142 (class sprintf_dom_walker): Add after_dom_children member function.
2143 Add evrp_range_analyzer member.
2144 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2145 range analyzer as needed.
2146 (sprintf_dom_walker::after_dom_children): New member function.
2147 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2149 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2150 (evrp_range_analyzer::pop_to_marker): Likewise.
2152 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2154 PR tree-optimization/84419
2155 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2156 with the required type if its current type is compatible but
2159 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2162 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2163 after vectorization.
2165 2018-02-20 Martin Liska <mliska@suse.cz>
2168 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2169 possible values if we don't have a hint.
2171 2018-02-20 Martin Liska <mliska@suse.cz>
2175 * final.c (shorten_branches): Build align_tab array with one
2177 * opts.c (finish_options): Add alignment option limit check.
2178 (MAX_CODE_ALIGN): Likewise.
2179 (MAX_CODE_ALIGN_VALUE): Likewise.
2180 * doc/invoke.texi: Document maximum allowed option value for
2181 all -falign-* options.
2183 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2186 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2187 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2188 * var-tracking.c (emit_note_insn_var_location): Remove all references
2189 to NOTE_INSN_CALL_ARG_LOCATION.
2190 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2191 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2192 Use copy_rtx_if_shared.
2193 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2194 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2195 (dwarf2out_var_location): Remove handling of
2196 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2198 * final.c (final_scan_insn): Remove all references to
2199 NOTE_INSN_CALL_ARG_LOCATION.
2200 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2201 before dumping final insns.
2202 * except.c (emit_note_eh_region_end): Remove all references to
2203 NOTE_INSN_CALL_ARG_LOCATION.
2204 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2205 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2206 * config/arc/arc.c (hwloop_optimize): Likewise.
2207 * config/arm/arm.c (create_fix_barrier): Likewise.
2208 * config/s390/s390.c (s390_chunkify_start): Likewise.
2209 * config/sh/sh.c (find_barrier): Likewise.
2210 * config/i386/i386.c (rest_of_insert_endbranch,
2211 ix86_seh_fixup_eh_fallthru): Likewise.
2212 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2213 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2214 * config/frv/frv.c (frv_function_prologue): Likewise.
2215 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2217 (note_outside_basic_block_p): Remove all references to
2218 NOTE_INSN_CALL_ARG_LOCATION.
2219 * gengtype.c (adjust_field_rtx_def): Likewise.
2220 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2222 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2223 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2226 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2229 PR tree-optimization/84452
2230 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2231 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2234 2018-02-19 Martin Liska <mliska@suse.cz>
2237 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2239 2018-02-19 Martin Liska <mliska@suse.cz>
2240 Richard Sandiford <richard.sandiford@linaro.org>
2242 PR tree-optimization/82491
2243 * gimple-fold.c (get_base_constructor): Make earlier bail out
2246 2018-02-19 Carl Love <cel@us.ibm.com>
2248 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2249 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2251 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2252 P8V_BUILTIN_VEC_NEG.
2254 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2256 * config/rl78/rl78.md (movdf): New define expand.
2258 2018-02-19 Martin Liska <mliska@suse.cz>
2261 * doc/invoke.texi: Fix typo.
2262 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2264 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2266 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2267 handle rs6000_single_float and rs6000_double_float specially for
2270 2018-02-16 Jeff Law <law@redhat.com>
2272 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2275 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2278 * ipa-inline.c (inline_small_functions): Fix a typo.
2280 2018-02-16 Nathan Sidwell <nathan@acm.org>
2282 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2284 2018-02-16 Carl Love <cel@us.ibm.com>
2286 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2287 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2288 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2289 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2290 expansion to P8V_BUILTIN_VEC_FLOAT2.
2292 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2294 PR rtl-optimization/70023
2295 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2296 src_regno into account.
2298 2018-02-16 Carl Love <cel@us.ibm.com>
2300 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2301 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2302 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2303 * config/rs6000/rs6000.c: Remove case statements for
2304 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2305 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2306 and P9V_BUILTIN_VEC_VINSERT4B.
2307 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2308 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2309 * config/rs6000/vsx.md:
2310 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2313 2018-02-16 Carl Love <cel@us.ibm.com>
2315 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2317 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2319 * config/rs6000/rs6000-c.c: Add the definitions for
2320 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2321 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2322 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2323 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2324 definition for insert4b and define insn *insert3b_internal.
2325 * doc/extend.texi: Add documentation for vec_extract4b.
2327 2018-02-16 Nathan Sidwell <nathan@acm.org>
2329 * doc/extend.texi (Backwards Compatibility): Mention friend
2330 injection. Note for-scope is deprecated.
2331 * doc/invoke.texi (-ffriend-injection): Deprecate.
2333 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2335 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2336 that moved to I2, also allow destinations that are a paradoxical
2337 subreg (instead of a normal reg).
2339 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2342 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2345 2018-02-16 Richard Biener <rguenther@suse.de>
2347 PR tree-optimization/84037
2348 PR tree-optimization/84016
2350 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2351 Adjust vec_construct for the fact we need additional higher latency
2352 128bit inserts for AVX256 and AVX512 vector builds.
2353 (ix86_add_stmt_cost): Scale vector construction cost for
2356 2018-02-16 Richard Biener <rguenther@suse.de>
2358 PR tree-optimization/84417
2359 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2360 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2361 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2363 2018-02-16 Martin Liska <mliska@suse.cz>
2366 * internal-fn.def (ASAN_CHECK): Set proper flags.
2367 (ASAN_MARK): Likewise.
2369 2018-02-16 Julia Koval <julia.koval@intel.com>
2371 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2372 from PTA_CANNONLAKE.
2374 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2377 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2378 Use ++iter rather than iter++ for std::list iterators.
2379 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2380 defer deleting them until all nodes in the forest are processed. Do
2381 free even leaf nodes. Change to_process into auto_vec.
2384 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2385 * vec.h (vec_default_construct): Use memset instead of placement new
2386 if BROKEN_VALUE_INITIALIZATION is defined.
2387 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2388 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2391 PR rtl-optimization/83723
2392 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2393 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2394 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2396 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2398 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2400 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2402 PR rtl-optimization/81443
2403 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2404 from inner REGs to paradoxical SUBREGs.
2406 2018-02-16 Richard Biener <rguenther@suse.de>
2408 PR tree-optimization/84399
2409 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2410 For operands we can analyze at their definition make sure we can
2411 analyze them at each use as well.
2413 2018-02-16 Richard Biener <rguenther@suse.de>
2415 PR tree-optimization/84190
2416 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2417 volatile accesses if the decl isn't volatile.
2419 2018-02-15 Jason Merrill <jason@redhat.com>
2421 PR c++/84314 - ICE with templates and fastcall attribute.
2422 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2423 TYPE_CANONICAL on an existing type.
2425 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2427 PR tree-optimization/84383
2428 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2429 dstoff nor call operand_equal_p if dstbase is NULL.
2431 PR tree-optimization/84334
2432 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2433 also a CONSTANT_CLASS_P, punt.
2435 2018-02-14 Jim Wilson <jimw@sifive.com>
2437 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2438 first SMALL_OPERAND check. New local min_second_step. Move assert
2439 to where locals are set. Add TARGET_RVC support.
2440 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2442 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2444 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2446 2018-02-14 Martin Sebor <msebor@redhat.com>
2448 PR tree-optimization/83698
2449 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2450 arrays constrain the offset range to their bounds.
2451 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2452 (builtin_access::overlap): Avoid setting the size of overlap if it's
2454 (maybe_diag_overlap): Also consider arrays when deciding what values
2455 of offsets to include in diagnostics.
2457 2018-02-14 Martin Sebor <msebor@redhat.com>
2460 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2461 that correspond to the kind of a declaration.
2463 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2466 * config/pa/pa.md: Load address of PIC label using the linkage table
2467 if the label is nonlocal.
2469 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2471 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2472 warning message if user requests -maltivec=be.
2473 * doc/invoke.texi: Document deprecation of -maltivec=be.
2475 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2478 * config/rs6000/rs6000-c.c: Update definitions for
2479 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2480 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2482 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2485 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2486 add _get_ssp intrinsics. Remove argument from
2487 __builtin_ia32_rdssp[d|q].
2488 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2489 * config/i386/i386-builtin.def: Remove argument from
2490 __builtin_ia32_rdssp[d|q].
2491 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2492 ix86_expand_special_args_builtin for _rdssp[d|q].
2493 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2494 Clear register before usage.
2495 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2496 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2498 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2500 PR tree-optimization/84357
2501 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2502 operand 1 of an ARRAY_REF too.
2504 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2507 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2508 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2510 (set_of_reg): New struct.
2511 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2512 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2513 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2515 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2516 Split into bitclr, bitset, bitinvert patterns if appropriate.
2517 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2518 use rx_fuse_in_memory_bitop.
2519 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2520 to named insn, correct maximum insn length.
2522 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2525 * machmode.def: Define a complex mode for PARTIAL_INT.
2526 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2528 * doc/rtl.texi: Document CSPImode.
2529 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2531 (msp430_hard_regno_nregs_with_padding): Likewise.
2533 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2536 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2538 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2540 PR rtl-optimization/84169
2541 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2542 we generated a parallel as new i3 and we split that to new i2 and i3
2543 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2544 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2545 those to i2, not i1. Partially rewrite this scan code.
2547 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2550 * stor-layout.c (place_field): For variable length fields, adjust
2551 offset_align afterwards not just based on the field's alignment,
2552 but also on the size.
2555 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2556 of exps and logs in the use_exp2 case.
2558 2018-02-13 Jeff Law <law@redhat.com>
2560 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2563 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2566 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2570 * common.opt (gas-loc-support, gas-locview-support): New.
2571 (ginline-points, ginternal-reset-location-views): New.
2572 * doc/invoke.texi: Document them. Use @itemx where intended.
2573 (gvariable-location-views): Adjust.
2574 * target.def (reset_location_view): New.
2575 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2576 (TARGET_RESET_LOCATION_VIEW): New.
2577 * doc/tm.texi: Rebuilt.
2578 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2579 (dwarf2out_default_as_locview_support): New.
2580 (output_asm_line_debug_info): Use option variables.
2581 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2582 (output_loc_list): Likewise.
2583 (add_high_low_attributes): Check option variables.
2584 Don't output entry view attribute in strict mode.
2585 (gen_inlined_subroutine_die): Check option variables.
2586 (dwarf2out_inline_entry): Likewise.
2587 (init_sections_and_labels): Likewise.
2588 (dwarf2out_early_finish): Likewise.
2589 (maybe_reset_location_view): New, from...
2590 (dwarf2out_var_location): ... here. Call it.
2591 * debug.h (dwarf2out_default_as_loc_support): Declare.
2592 (dwarf2out_default_as_locview_support): Declare.
2593 * hooks.c (hook_int_rtx_insn_0): New.
2594 * hooks.h (hook_int_rtx_insn_0): Declare.
2595 * toplev.c (process_options): Take -gas-loc-support and
2596 -gas-locview-support from dwarf2out. Enable
2597 -gvariable-location-views by default only with locview
2598 assembler support. Enable -ginternal-reset-location-views by
2599 default only if the target defines the corresponding hook.
2600 Enable -ginline-points by default if location views are
2601 enabled; force it disabled if statement frontiers are
2603 * tree-inline.c (expand_call_inline): Check option variables.
2604 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2606 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2608 PR tree-optimization/84321
2609 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2610 handling. Also check whether the anti-range contains any values
2611 that satisfy the mask; switch to a VR_RANGE if not.
2613 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2616 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2618 2018-02-13 Martin Jambor <mjambor@suse.cz>
2621 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2622 of call statements, also set location of a load to a temporary.
2624 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2626 * config/rl78/rl78.c (add_vector_labels): New function.
2627 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2628 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2629 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2630 which checks that no arguments are passed.
2631 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2632 * doc/extend.texi: Documentation for the new attribute.
2634 2018-02-13 Andreas Schwab <schwab@suse.de>
2636 * config/riscv/linux.h (CPP_SPEC): Define.
2638 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2641 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2642 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2643 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2644 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2645 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2646 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2647 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2648 temporarily for AES and PCLMUL builtins.
2650 PR tree-optimization/84339
2651 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2652 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2656 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2657 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2658 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2660 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2662 * omp-simd-clone.h: New file.
2663 * omp-simd-clone.c: Include omp-simd-clone.h.
2664 (expand_simd_clones): No longer static.
2665 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2666 cgraph.h and omp-simd-clone.h.
2667 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2668 (vect_recog_widen_shift_pattern): Formatting fix.
2669 (vect_pattern_recog_1): Don't check optab for calls.
2672 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2673 operands[2] into a REG before using gen_lowpart on it.
2675 2018-02-12 Jeff Law <law@redhat.com>
2678 * config/sh/sh.c (find_barrier): Consider a sibling call
2681 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2682 successfully back substituting a reg.
2684 2018-02-12 Richard Biener <rguenther@suse.de>
2686 PR tree-optimization/84037
2687 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2688 parameter, move visited init to caller.
2689 (vect_slp_analyze_operations): Separate cost from validity
2690 check, initialize visited once for all instances.
2691 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2693 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2695 (vect_model_promotion_demotion_cost): Likewise.
2696 (vectorizable_bswap): Guard cost modeling with !slp_node
2697 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2699 (vectorizable_call): Likewise.
2700 (vectorizable_conversion): Likewise.
2701 (vectorizable_assignment): Likewise.
2702 (vectorizable_shift): Likewise.
2703 (vectorizable_operation): Likewise.
2704 (vectorizable_store): Likewise.
2705 (vectorizable_load): Likewise.
2706 (vectorizable_condition): Likewise.
2707 (vectorizable_comparison): Likewise.
2709 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2712 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2713 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2714 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2716 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2719 * params.def (inline-min-speedup): Increase from 8 to 15.
2720 (max-inline-insns-auto): Decrease from 40 to 30.
2721 * ipa-split.c (consider_split): Add some buffer for function to
2722 be considered inlining candidate.
2723 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2726 2018-02-12 Richard Biener <rguenther@suse.de>
2728 PR tree-optimization/84037
2729 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2730 matched stmts if we cannot swap the non-matched ones.
2732 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2734 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2735 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2736 _mm_maskz_scalef_round_ss): New intrinsics.
2737 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2738 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2739 __builtin_ia32_scalefss_round): Remove.
2740 (__builtin_ia32_scalefsd_mask_round,
2741 __builtin_ia32_scalefss_mask_round): New intrinsics.
2742 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2743 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2744 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2745 "<round_constraint>")): Changed to ...
2746 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2747 "<round_scalar_constraint>")): ... this.
2748 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2749 %0, %1, %2<round_op3>}"): Changed to ...
2750 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2751 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2752 %2<round_scalar_mask_op3>}"): ... this.
2753 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2755 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2757 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2758 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2759 (_mm_maskz_sqrt_round_ss): New intrinsics.
2760 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2761 (__builtin_ia32_sqrtsd_mask_round)
2762 (__builtin_ia32_sqrtss_mask_round): New builtins.
2763 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2764 (__builtin_ia32_sqrtss_round): Remove.
2765 (__builtin_ia32_sqrtsd_mask_round)
2766 (__builtin_ia32_sqrtss_mask_round): New builtins.
2767 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2768 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2769 ((match_operand:VF_128 1 "vector_operand"
2770 "xBm,<round_constraint>")): Changed to ...
2771 ((match_operand:VF_128 1 "vector_operand"
2772 "xBm,<round_scalar_constraint>")): ... this.
2773 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2774 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2775 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2776 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2777 %<iptr>1<round_scalar_mask_op3>}): ... this.
2778 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2779 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2781 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2784 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2785 Cast vec_cmpeq result to correct type.
2786 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2787 Cast vec_cmpgt result to correct type.
2789 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2791 * final.c (final_scan_insn_1): Renamed from...
2792 (final_scan_insn): ... this. New wrapper, to recover
2793 seen from the outermost call in recursive ones.
2794 * config/sparc/sparc.c (output_return): Drop seen from call.
2795 (output_sibcall): Likewise.
2796 * config/visium/visium.c (output_branch): Likewise.
2798 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2800 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2803 2018-02-10 Alan Modra <amodra@gmail.com>
2806 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2807 Specify LR as an input.
2809 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2812 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2813 remove_member_access_dummy_vars): New functions.
2814 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2815 lower_omp_1, execute_lower_omp): Use them.
2817 PR rtl-optimization/84308
2818 * shrink-wrap.c (spread_components): Release todo vector.
2820 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2822 PR rtl-optimization/57193
2823 * ira-color.c (struct allocno_color_data): Add member
2824 conflict_allocno_hard_prefs.
2825 (update_conflict_allocno_hard_prefs): New.
2826 (bucket_allocno_compare_func): Add a preference based on
2827 conflict_allocno_hard_prefs.
2828 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2829 (color_allocnos): Remove a dead code. Initiate
2830 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2832 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2835 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2836 constraint from =wa to wa. Avoid a subreg on the output operand,
2837 instead use a pseudo and subreg it in a move.
2838 (p9_xxbrd_<mode>): Changed to ...
2839 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2840 (p9_xxbrd_v2df): New expander.
2841 (p9_xxbrw_<mode>): Changed to ...
2842 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2843 (p9_xxbrw_v4sf): New expander.
2845 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2847 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2849 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2852 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2853 multiply in 32-bit mode.
2854 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2855 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2858 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2860 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2861 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2862 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2863 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2865 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2868 * dwarf2out.c (is_trivial_indirect_ref): New function.
2869 (dwarf2out_late_global_decl): Do not generate a location
2870 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2871 and that are not defined in the current unit.
2873 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2875 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2876 instead of a libcall for UNORDERED.
2878 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2881 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2882 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2884 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2887 * config/s390/s390.c (s390_set_current_function): Invoke
2888 s390_indirect_branch_settings also if fndecl didn't change.
2890 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2892 * config/rs6000/rs6000.md (blockage): Set length to zero.
2894 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2896 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2898 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2901 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2902 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2906 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2907 PARALLEL incoming that failed vt_get_decl_and_offset check.
2910 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2911 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2913 (get_variable_section): For decls in named .bss* sections pass true as
2914 second argument to bss_initializer_p.
2916 2018-02-09 Marek Polacek <polacek@redhat.com>
2917 Jakub Jelinek <jakub@redhat.com>
2920 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2921 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2922 Sync some changes from cxx_fold_indirect_ref.
2924 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2926 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2928 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2929 (BLOCK_INLINE_ENTRY_LABEL): New.
2930 (dwarf2out_var_location): Disregard inline entry markers.
2931 (inline_entry_data): New struct.
2932 (inline_entry_data_hasher): New hashtable type.
2933 (inline_entry_data_hasher::hash): New.
2934 (inline_entry_data_hasher::equal): New.
2935 (inline_entry_data_table): New variable.
2936 (add_high_low_attributes): Add DW_AT_entry_pc and
2937 DW_AT_GNU_entry_view attributes if a pending entry is found
2938 in inline_entry_data_table. Add old entry_pc attribute only
2939 if debug nonbinding markers are disabled.
2940 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2941 markers are enabled.
2942 (block_within_block_p, dwarf2out_inline_entry): New.
2943 (dwarf2out_finish): Check that no entries remained in
2944 inline_entry_data_table.
2945 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2946 (final_scan_insn, notice_source_line): Likewise.
2947 (rest_of_clean_state): Skip inline entry markers.
2948 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2950 * gimple.c (gimple_build_debug_inline_entry): New.
2951 * gimple.h (enum gimple_debug_subcode): Add
2952 GIMPLE_DEBUG_INLINE_ENTRY.
2953 (gimple_build_debug_inline_entry): Declare.
2954 (gimple_debug_inline_entry_p): New.
2955 (gimple_debug_nonbind_marker_p): Adjust.
2956 * insn-notes.def (INLINE_ENTRY): New.
2957 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2958 inline entry marker notes.
2959 (print_insn): Likewise.
2960 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2961 (INSN_DEBUG_MARKER_KIND): Likewise.
2962 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2963 * tree-inline.c (expand_call_inline): Build and insert
2964 debug_inline_entry stmt.
2965 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2966 inline entry blocks early, if nonbind markers are enabled.
2967 (dump_scope_block): Dump fragment info.
2968 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2969 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2970 (gimple_build_debug_inline_entry): New.
2971 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2972 Enable/disable inline entry points too.
2973 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2974 (DEBUG_INSN): Describe inline entry markers.
2976 * common.opt (gvariable-location-views): New.
2977 (gvariable-location-views=incompat5): New.
2978 * config.in: Rebuilt.
2979 * configure: Rebuilt.
2980 * configure.ac: Test assembler for view support.
2981 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2982 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2983 * dwarf2out.c (var_loc_view): New typedef.
2984 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2985 (dwarf2out_locviews_in_attribute): New.
2986 (dwarf2out_locviews_in_loclist): New.
2987 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2988 (enum dw_line_info_opcode): Add LI_adv_address.
2989 (struct dw_line_info_table): Add view.
2990 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2991 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2992 (zero_view_p): New variable.
2993 (ZERO_VIEW_P): New macro.
2994 (output_asm_line_debug_info): New.
2995 (struct var_loc_node): Add view.
2996 (add_AT_view_list, AT_loc_list): New.
2997 (add_var_loc_to_decl): Add view param. Test it against last.
2998 (new_loc_list): Add view params. Record them.
2999 (AT_loc_list_ptr): Handle loc and view lists.
3000 (view_list_to_loc_list_val_node): New.
3001 (print_dw_val): Handle dw_val_class_view_list.
3002 (size_of_die): Likewise.
3003 (value_format): Likewise.
3004 (loc_list_has_views): New.
3005 (gen_llsym): Set vl_symbol too.
3006 (maybe_gen_llsym, skip_loc_list_entry): New.
3007 (dwarf2out_maybe_output_loclist_view_pair): New.
3008 (output_loc_list): Output view list or entries too.
3009 (output_view_list_offset): New.
3010 (output_die): Handle dw_val_class_view_list.
3011 (output_dwarf_version): New.
3012 (output_compilation_unit_header): Use it.
3013 (output_skeleton_debug_sections): Likewise.
3014 (output_rnglists, output_line_info): Likewise.
3015 (output_pubnames, output_aranges): Update version comments.
3016 (output_one_line_info_table): Output view numbers in asm comments.
3017 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3018 Call maybe_gen_llsym.
3019 (loc_list_from_tree_1): Adjust.
3020 (add_AT_location_description): Create view list attribute if
3021 needed, check it's absent otherwise.
3022 (convert_cfa_to_fb_loc_list): Adjust.
3023 (maybe_emit_file): Call output_asm_line_debug_info for test.
3024 (dwarf2out_var_location): Reset views as needed. Precompute
3025 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3026 attribute. Set view.
3027 (new_line_info_table): Reset next view.
3028 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3029 (dwarf2out_source_line): Likewise. Output view resets and labels to
3030 the assembler, or select appropriate line info opcodes.
3031 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3032 (optimize_string_length): Catch it. Adjust.
3033 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3034 dw_val_class_view_list, and remove it if no longer needed.
3035 (hash_loc_list): Hash view numbers.
3036 (loc_list_hasher::equal): Compare them.
3037 (optimize_location_lists): Check whether a view list symbol is
3038 needed, and whether the locview attribute is present, and
3039 whether they match. Remove the locview attribute if no longer
3041 (index_location_lists): Call skip_loc_list_entry for test.
3042 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3043 Use output_dwarf_version.
3044 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3045 (struct dw_val_node): Add val_view_list.
3046 * final.c (SEEN_NEXT_VIEW): New.
3047 (set_next_view_needed): New.
3048 (clear_next_view_needed): New.
3049 (maybe_output_next_view): New.
3050 (final_start_function): Rename to...
3051 (final_start_function_1): ... this. Take pointer to FIRST,
3052 add SEEN parameter. Emit param bindings in the initial view.
3053 (final_start_function): Reintroduce SEEN-less interface.
3054 (final): Rename to...
3055 (final_1): ... this. Take SEEN parameter. Output final pending
3056 next view at the end.
3057 (final): Reintroduce seen-less interface.
3058 (final_scan_insn): Output pending next view before switching
3059 sections or ending a block. Mark the next view as needed when
3060 outputting variable locations. Notify debug backend of section
3061 changes, and of location view changes.
3062 (rest_of_handle_final): Adjust.
3063 * toplev.c (process_options): Autodetect value for debug variable
3064 location views option. Warn on incompat5 without -gdwarf-5.
3065 * doc/invoke.texi (gvariable-location-views): New.
3066 (gvariable-location-views=incompat5): New.
3067 (gno-variable-location-views): New.
3069 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3071 PR tree-optimization/84136
3072 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3073 that the result of find_edge is non-NULL.
3075 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3078 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3079 storing integer register in SImode. Fix cost of 256 and 512
3080 byte aligned SSE register store.
3082 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3084 * config/i386/i386.c (ix86_multiplication_cost): Fix
3085 multiplication cost for TARGET_AVX512DQ.
3087 2018-02-08 Marek Polacek <polacek@redhat.com>
3089 PR tree-optimization/84238
3090 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3093 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3095 PR tree-optimization/84265
3096 * tree-vect-stmts.c (vectorizable_store): Don't treat
3097 VMAT_CONTIGUOUS accesses as grouped.
3098 (vectorizable_load): Likewise.
3100 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3102 PR tree-optimization/81635
3103 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3104 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3105 (test_round_for_mask): New functions.
3106 (wide_int_cc_tests): Call test_round_for_mask.
3107 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3108 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3109 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3110 range returned by get_range_info.
3112 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3115 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3116 * symtab.c: Include builtins.h
3117 (symtab_node::output_to_lto_symbol_table_p): Move here
3118 from lto-streamer-out.c:output_symbol_p.
3119 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3120 (output_symbol_p): Move all logic to symtab.c
3121 (produce_symtab): Update.
3123 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3125 * config/s390/s390-opts.h (enum indirect_branch): Define.
3126 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3127 (s390_indirect_branch_via_thunk)
3128 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3129 (enum s390_indirect_branch_type): Define.
3130 * config/s390/s390.c (struct s390_frame_layout, struct
3131 machine_function): Remove.
3132 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3133 (indirect_branch_table_label_no, indirect_branch_table_name):
3135 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3136 (enum s390_indirect_branch_option): Define.
3137 (s390_return_addr_from_memory): New function.
3138 (s390_handle_string_attribute): New function.
3139 (s390_attribute_table): Add new attribute handler.
3140 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3141 (s390_indirect_branch_via_thunk): New function.
3142 (s390_indirect_branch_via_inline_thunk): New function.
3143 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3144 sibling call optimization for non z10 compiles.
3145 (s390_emit_call): Force indirect branch target to be a single
3146 register. Add r1 clobber for non-z10 compiles.
3147 (s390_emit_epilogue): Emit return jump via return_use expander.
3148 (s390_reorg): Handle JUMP_INSNs as execute targets.
3149 (s390_option_override_internal): Perform validity checks for the
3150 new command line options.
3151 (s390_indirect_branch_attrvalue): New function.
3152 (s390_indirect_branch_settings): New function.
3153 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3154 (s390_output_indirect_thunk_function): New function.
3155 (s390_code_end): Implement target hook.
3156 (s390_case_values_threshold): Implement target hook.
3157 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3159 * config/s390/s390.h (struct s390_frame_layout)
3160 (struct machine_function): Move here from s390.c.
3161 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3162 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3163 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3164 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3165 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3166 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3167 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3168 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3169 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3170 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3171 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3172 (mnemonic attribute): Add values which aren't recognized
3174 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3175 pattern for branch conversion. Fix mnemonic attribute.
3176 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3177 indirect branch via thunk if requested.
3178 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3179 ("*indirect_jump"): Disable for branch conversion using out of
3181 ("indirect_jump_via_thunk<mode>_z10")
3182 ("indirect_jump_via_thunk<mode>")
3183 ("indirect_jump_via_inlinethunk<mode>_z10")
3184 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3185 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3186 ("casesi_jump_via_inlinethunk<mode>_z10")
3187 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3188 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3189 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3190 ("*indirect2_jump"): Disable for branch conversion.
3191 ("casesi_jump"): Turn into expander and expand patterns for branch
3193 ("return_use"): New expander.
3194 ("*return"): Emit return via thunk and rename it to ...
3195 ("*return<mode>"): ... this one.
3196 * config/s390/s390.opt: Add new options and and enum for the
3199 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3201 * lra-constraints.c (match_reload): Unconditionally use
3202 gen_lowpart_SUBREG, rather than selecting between that
3203 and equivalent gen_rtx_SUBREG code.
3205 2018-02-08 Richard Biener <rguenther@suse.de>
3207 PR tree-optimization/84233
3208 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3209 changed flag instead of boguously re-using phi_inserted.
3211 2018-02-08 Martin Jambor <mjambor@suse.cz>
3213 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3214 static local variables.
3216 2018-02-08 Richard Biener <rguenther@suse.de>
3218 PR tree-optimization/84278
3219 * tree-vect-stmts.c (vectorizable_store): When looking for
3220 smaller vector types to perform grouped strided loads/stores
3221 make sure the mode is supported by the target.
3222 (vectorizable_load): Likewise.
3224 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3226 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3227 Increase LDP/STP opportunities by adding adjacent callee-saves.
3229 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3231 PR rtl-optimization/84068
3232 PR rtl-optimization/83459
3233 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3235 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3237 PR tree-optimization/84224
3238 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3239 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3242 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3245 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3246 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3247 count, remove one USE.
3249 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3251 * doc/install.texi (Configuration): Document the
3252 --with-long-double-format={ibm,ieee} PowerPC configuration
3256 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3257 Convert from define_expand to be define_insn_and_split. Rework
3258 float/double/_Float128 conversions to QI/HI/SImode to work with
3259 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3260 conversions to QI/HImode types did a store and then a load to
3261 truncate the value. For conversions to VSX registers, don't split
3262 the insn, instead emit the code directly. Use the code iterator
3263 any_fix to combine signed and unsigned conversions.
3264 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3265 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3266 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3267 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3268 (fix_<mode>di2_hw): Likewise.
3269 (fixuns_<mode>di2_hw): Likewise.
3270 (fix_<mode>si2_hw): Likewise.
3271 (fixuns_<mode>si2_hw): Likewise.
3272 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3273 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3274 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3275 fix<uns>_trunc<SFDF:mode>si2_p8.
3276 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3278 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3279 (fix<uns>_<mode>_mem): Likewise.
3280 (fctiw<u>z_<mode>_mem): Likewise.
3281 (fix<uns>_<mode>_mem): Likewise.
3282 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3283 the register allocator from doing a direct move to the GPRs to do
3284 a store, and instead use the ISA 3.0 store byte/half-word from
3285 vector register instruction. For IEEE 128-bit floating point,
3286 also optimize stores of 32-bit ints.
3287 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3289 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3291 * genextract.c (push_pathstr_operand): New function to support
3293 (walk_rtx): Call push_pathstr_operand.
3294 (print_path): Support [a-zA-Z].
3296 2018-02-07 Richard Biener <rguenther@suse.de>
3298 PR tree-optimization/84037
3299 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3300 (cse_and_gimplify_to_preheader): Declare.
3301 (vect_get_place_in_interleaving_chain): Likewise.
3302 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3304 (_loop_vec_info::~_loop_vec_info): Delete it.
3305 (cse_and_gimplify_to_preheader): New function.
3306 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3307 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3308 (vectorizable_load): Likewise. For grouped stores always base
3309 the IV on the first element.
3310 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3311 condition before gimplifying.
3313 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3315 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3316 *DIV_EXPR and *MOD_EXPR.
3318 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3321 * config/i386/i386.c (ix86_option_override_internal): Mask out
3322 the CF_SET bit when checking -fcf-protection.
3324 2018-02-07 Tom de Vries <tom@codesourcery.com>
3327 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3330 2018-02-07 Richard Biener <rguenther@suse.de>
3332 PR tree-optimization/84204
3333 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3336 PR tree-optimization/84205
3337 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3338 special-case isl_ast_op_zdiv_r.
3340 PR tree-optimization/84223
3341 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3342 Only add conditions from within the region.
3343 (gather_bbs::after_dom_children): Adjust.
3345 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3348 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3349 * config/avr/avr.md: Only post-reload split REG-REG moves if
3350 either register is GENERAL_REG_P.
3352 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3354 PR tree-optimization/84235
3355 * tree-ssa-scopedtables.c
3356 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3357 if the subtraction is performed in floating point type where NaNs are
3358 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3359 build 1. Formatting fix.
3361 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3364 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3365 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3366 and skip it regardless of bb boundaries. Use CALL_P macro,
3367 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3368 unnecessarily, formatting fix.
3370 2018-02-06 Michael Collison <michael.collison@arm.com>
3372 * config/arm/thumb2.md:
3373 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3374 (*thumb_mov_notscc): Ditto.
3376 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3379 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3380 unsigned_fix, not "s".
3382 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3384 * configure.ac (gcc_fn_eh_frame_ro): New function.
3385 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3386 correct .eh_frame permissions.
3387 * configure: Regenerate.
3389 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3391 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3394 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3396 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3397 Display warning message for -mno-speculate-indirect-jumps.
3399 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3401 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3403 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3405 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3407 PR tree-optimization/84225
3408 * tree-eh.c (find_trapping_overflow): Only call
3409 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3411 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3414 * config/i386/i386.c: Reimplement the check of possible options
3415 -mibt/-mshstk conbination. Change error messages.
3416 * doc/invoke.texi: Fix a typo: remove extra '='.
3418 2018-02-06 Marek Polacek <polacek@redhat.com>
3420 PR tree-optimization/84228
3421 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3423 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3426 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3427 emitted arch directives.
3428 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3429 __ARM_FEATURE_COPROC before changing architectures.
3431 2018-02-06 Richard Biener <rguenther@suse.de>
3433 * config/i386/i386.c (print_reg): Fix typo.
3434 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3436 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3438 * configure: Regenerate.
3440 2018-02-05 Martin Sebor <msebor@redhat.com>
3442 PR tree-optimization/83369
3443 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3446 2018-02-05 Martin Liska <mliska@suse.cz>
3448 * doc/invoke.texi: Cherry-pick upstream r323995.
3450 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3452 * ira.c (ira_init_register_move_cost): Adjust comment.
3454 2018-02-05 Martin Liska <mliska@suse.cz>
3456 PR gcov-profile/84137
3457 * doc/gcov.texi: Fix typo in documentation.
3459 2018-02-05 Martin Liska <mliska@suse.cz>
3461 PR gcov-profile/83879
3462 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3463 using dlopen functionality.
3465 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3467 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3468 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3469 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3470 _mm_maskz_range_round_ss): New intrinsics.
3471 (__builtin_ia32_rangesd128_round)
3472 (__builtin_ia32_rangess128_round): Remove.
3473 (__builtin_ia32_rangesd128_mask_round,
3474 __builtin_ia32_rangess128_mask_round): New builtins.
3475 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3476 __builtin_ia32_rangess128_round): Remove.
3477 (__builtin_ia32_rangesd128_mask_round,
3478 __builtin_ia32_rangess128_mask_round): New builtins.
3479 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3480 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3481 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3482 "<round_saeonly_constraint>")): Changed to ...
3483 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3484 "<round_saeonly_scalar_constraint>")): ... this.
3485 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3486 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3487 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3488 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3489 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3491 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3493 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3495 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3496 Remove all values except native, 8540 and 8548.
3498 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3500 * config/i386/i386.c (ix86_output_function_return): Pass
3501 INVALID_REGNUM, instead of -1, as invalid register number to
3502 indirect_thunk_name and output_indirect_thunk.
3504 2018-02-02 Julia Koval <julia.koval@intel.com>
3506 * config.gcc: Add -march=icelake.
3507 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3508 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3509 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3510 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3511 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3512 (processor_target_table): Add icelake.
3513 (ix86_option_override_internal): Handle new PTAs.
3514 (get_builtin_code_for_version): Handle icelake.
3515 (M_INTEL_COREI7_ICELAKE): New.
3516 (fold_builtin_cpu): Handle icelake.
3517 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3518 * doc/invoke.texi: Add -march=icelake.
3520 2018-02-02 Julia Koval <julia.koval@intel.com>
3522 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3523 to wide_int_bitmask.
3524 * wide-int-bitmask.h: New.
3526 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3529 * config/i386/i386.md: Replace Pmode with word_mode in
3530 builtin_setjmp_setup and builtin_longjmp to support x32.
3532 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3536 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3538 (rs6000_supported_cpu_names): New static variable.
3539 (linux_cpu_translation_table): Likewise.
3540 (elf_platform) <cpu>: Define new static variable and use it.
3541 Translate kernel AT_PLATFORM name to canonical name if needed.
3542 Error if platform name is unknown.
3544 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3547 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3549 2018-02-01 Jeff Law <law@redhat.com>
3552 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3553 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3554 the scratch if RELEASE_VIA_POP is false.
3555 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3556 If we have to save a temporary register, decrement SIZE appropriately.
3557 Pass new arguments to release_scratch_register_on_entry.
3558 (ix86_adjust_stack_and_probe): Likewise.
3559 (ix86_emit_probe_stack_range): Pass new arguments to
3560 release_scratch_register_on_entry.
3562 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3564 PR rtl-optimization/84157
3565 * combine.c (change_zero_ext): Use REG_P predicate in
3566 front of HARD_REGISTER_P predicate.
3568 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3570 * config/avr/avr.c (avr_option_override): Move disabling of
3571 -fdelete-null-pointer-checks to...
3572 * common/config/avr/avr-common.c (avr_option_optimization_table):
3575 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3577 PR tree-optimization/81635
3578 * tree-data-ref.c (split_constant_offset_1): For types that
3579 wrap on overflow, try to use range info to prove that wrapping
3582 2018-02-01 Renlin Li <renlin.li@arm.com>
3585 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3587 (aarch64_register_move_cost): Likewise.
3588 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3590 (REG_CLASS_NAMES): Likewise.
3591 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3592 TAILCALL_ADDR_REGS. Remove IP registers.
3593 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3595 2018-02-01 Richard Biener <rguenther@suse.de>
3597 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3598 for specifying RPO order and allow NULL for that.
3599 * domwalk.c (dom_walker::dom_walker): Likewise.
3600 (dom_walker::walk): Handle NULL RPO order.
3601 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3603 (rewrite_update_dom_walker): Likewise.
3604 (mark_def_dom_walker): Likewise.
3606 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3608 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3609 (aarch64_maybe_expand_sve_subreg_move): Declare.
3610 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3611 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3613 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3614 that are semantically a reverse operation.
3615 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3616 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3617 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3619 (aarch64_can_change_mode_class): For big-endian, forbid changes
3620 between two SVE modes if they have different element sizes.
3622 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3624 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3625 the TImode handling for big-endian targets.
3627 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3629 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3630 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3632 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3633 Remove BSWAP handing for big-endian targets and use the form of
3634 LD1RQ appropariate for the mode.
3636 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3638 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3639 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3642 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3645 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3646 check for operands that need to go through aarch64_sve_reload_be.
3648 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3650 PR tree-optimization/81661
3651 PR tree-optimization/84117
3652 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3653 * tree-eh.c: Include gimplify.h.
3654 (find_trapping_overflow, replace_trapping_overflow,
3655 rewrite_to_non_trapping_overflow): New functions.
3656 * tree-vect-loop.c: Include tree-eh.h.
3657 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3658 * tree-data-ref.c: Include tree-eh.h.
3659 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3661 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3663 PR rtl-optimization/84123
3664 * combine.c (change_zero_ext): Check if hard register satisfies
3665 can_change_dest_mode before calling gen_lowpart_SUBREG.
3667 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3670 * ira.c (ira_init_register_move_cost): Remove assert.
3672 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3674 PR rtl-optimization/84071
3675 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3676 * doc/tm.texi: Regenerate.
3678 2018-01-31 Richard Biener <rguenther@suse.de>
3680 PR tree-optimization/84132
3681 * tree-data-ref.c (analyze_miv_subscript): Properly
3682 check whether evolution_function_is_affine_multivariate_p
3683 before calling gcd_of_steps_may_divide_p.
3685 2018-01-31 Julia Koval <julia.koval@intel.com>
3688 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3689 * config/i386/i386.md (rdpid_rex64) New.
3690 (rdpid): Make 32bit only.
3692 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3695 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3696 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3698 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3701 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3703 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3705 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3707 PR rtl-optimization/84071
3708 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3709 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3711 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3713 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3714 (arc_attribute_table): Add 'aux' attribute.
3715 (arc_in_small_data_p): Consider aux like variables.
3716 (arc_is_aux_reg_p): New function.
3717 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3718 (arc_get_aux_arg): New function.
3719 (prepare_move_operands): Handle aux-register access.
3720 (arc_handle_aux_attribute): New function.
3721 * doc/extend.texi (ARC Variable attributes): Add subsection.
3723 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3725 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3726 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3727 (arc_attribute_table): Add 'uncached' attribute.
3728 (arc_print_operand): Print '.di' flag for uncached memory
3730 (arc_in_small_data_p): Do not consider for small data the uncached
3732 (arc_is_uncached_mem_p): New function.
3733 * config/arc/predicates.md (compact_store_memory_operand): Check
3734 for uncached memory accesses.
3735 (nonvol_nonimm_operand): Likewise.
3736 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3738 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3741 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3742 falign-loops=): Add Optimization flag.
3744 2018-01-30 Jeff Law <law@redhat.com>
3747 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3748 INT_REGISTERS_SAVED. Check it prior to calling
3749 get_scratch_register_on_entry.
3750 (ix86_adjust_stack_and_probe): Similarly.
3751 (ix86_emit_probe_stack_range): Similarly.
3752 (ix86_expand_prologue): Corresponding changes.
3754 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3757 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3758 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3760 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3763 * lra-constraints.c (curr_insn_transform): Process AND in the
3766 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3768 PR rtl-optimization/83986
3769 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3770 dependence against last_pending_memory_flush in addition to
3773 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3775 PR tree-optimization/81611
3776 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3779 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3782 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3785 2018-01-30 Richard Biener <rguenther@suse.de>
3786 Jakub Jelinek <jakub@redhat.com>
3788 PR tree-optimization/84111
3789 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3790 inner loops added during recursion, as they don't have up-to-date
3793 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3796 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3797 (can_inline_edge_by_limits_p): ... here.
3798 (can_early_inline_edge_p, check_callers,
3799 update_caller_keys, update_callee_keys, recursive_inlining,
3800 add_new_edges_to_heap, speculation_useful_p,
3801 inline_small_functions,
3802 inline_small_functions, flatten_function,
3803 inline_to_all_callers_1): Update.
3805 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3807 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3810 2018-01-30 Richard Biener <rguenther@suse.de>
3812 PR tree-optimization/83008
3813 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3814 invariant and constant vector uses in stmts when they need
3817 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3820 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3821 * configure: Regenerate.
3823 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3825 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3827 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3828 Use gen_rtx_REG rather than gen_lowpart.
3830 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3832 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3833 rather than 0 when creating partial subregs.
3835 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3837 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3840 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3843 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3844 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3845 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3846 flags. This restores the settings used before the 2017-07-24.
3847 Turning off pre increment/decrement/modify allows IVOPTS to
3848 optimize DF/SF loops where the index is an int.
3850 2018-01-29 Richard Biener <rguenther@suse.de>
3851 Kelvin Nilsen <kelvin@gcc.gnu.org>
3854 * tree-vect-stmts.c (vectorizable_call): Don't call
3855 targetm.vectorize_builtin_md_vectorized_function if callee is
3858 2018-01-22 Carl Love <cel@us.ibm.com>
3860 * doc/extend.tex: Fix typo in second arg in
3861 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3863 2018-01-29 Richard Biener <rguenther@suse.de>
3865 PR tree-optimization/84086
3866 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3867 (flush_ssaname_freelist): When SSA names were released reset
3868 the SCEV hash table.
3870 2018-01-29 Richard Biener <rguenther@suse.de>
3872 PR tree-optimization/84057
3873 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3874 removed paths when removing edges.
3876 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3878 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3879 -mfunction-return=@var{choice}.
3881 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3884 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3886 (layout::print_source_line): Likewise.
3887 (test_get_line_width_without_trailing_whitespace): Add test cases.
3889 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3892 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3895 2018-01-26 Jim Wilson <jimw@sifive.com>
3897 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3899 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3902 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3904 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3905 and CMP + SUB-immediate -> SUBS.
3907 2018-01-26 Martin Sebor <msebor@redhat.com>
3909 PR tree-optimization/83896
3910 * tree-ssa-strlen.c (get_string_len): Rename...
3911 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3912 Avoid assuming length is constant.
3913 (handle_char_store): Use HOST_WIDE_INT for string length.
3915 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3918 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3919 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3921 2018-01-26 Richard Biener <rguenther@suse.de>
3923 PR rtl-optimization/84003
3924 * dse.c (record_store): Only record redundant stores when
3925 the earlier store aliases at least all accesses the later one does.
3927 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3929 PR rtl-optimization/83985
3930 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3931 REG_CFA_RESTORE insns.
3932 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3933 REG_CFA_RESTORE insns here.
3936 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3937 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3939 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3941 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3942 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3943 (arc_init): Likewise.
3944 (arc_override_options): Likewise.
3945 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3947 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3949 * config/arc/arc.h (TARGET_DBNZ): Define.
3950 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3951 properly set the tune attribute.
3952 (dbnz): Use TARGET_DBNZ guard.
3953 * config/arc/arc.opt (mtune): Add core3 option.
3955 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3957 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3958 recognize new pic like addresses.
3959 (arc_delegitimize_address): Clean up.
3961 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3963 * config/arc/arc-arches.def: Option mrf16 valid for all
3965 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3966 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3967 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3968 * config/arc/arc-tables.opt: Regenerate.
3969 * config/arc/arc.c (arc_conditional_register_usage): Handle
3970 reduced register file case.
3971 (arc_file_start): Set must have build attributes.
3972 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3974 * config/arc/arc.opt (mrf16): Add new option.
3975 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3976 * config/arc/genmultilib.awk: Handle new mrf16 option.
3977 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3978 * config/arc/t-multilib: Regenerate.
3979 * doc/invoke.texi (ARC Options): Document mrf16 option.
3981 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3983 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3984 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3985 (arc_attribute_table): Add 'secure_call' attribute.
3986 (arc_print_operand): Print secure call operand.
3987 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3989 (arc_is_secure_call_p): New function. * config/arc/arc.md
3990 (call_i): Add support for sjli instruction.
3991 (call_value_i): Likewise.
3992 * config/arc/constraints.md (Csc): New constraint.
3994 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3995 John Eric Martin <John.Martin@emmicro-us.com>
3997 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3998 * config/arc/arc.c (_arc_jli_section): New struct.
3999 (arc_jli_section): New type.
4000 (rc_jli_sections): New static variable.
4001 (arc_handle_jli_attribute): New function.
4002 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4003 (arc_file_end): New function.
4004 (TARGET_ASM_FILE_END): Define.
4005 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4006 (arc_add_jli_section): New function.
4007 (jli_call_scan): Likewise.
4008 (arc_reorg): Call jli_call_scan.
4009 (arc_output_addsi): Remove 'S' from printing asm operand.
4010 (arc_is_jli_call_p): New function.
4011 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4013 (movhi_insn): Likewise.
4014 (movsi_insn): Likewise.
4015 (movsi_set_cc_insn): Likewise.
4016 (loadqi_update): Likewise.
4017 (load_zeroextendqisi_update): Likewise.
4018 (load_signextendqisi_update): Likewise.
4019 (loadhi_update): Likewise.
4020 (load_zeroextendhisi_update): Likewise.
4021 (load_signextendhisi_update): Likewise.
4022 (loadsi_update): Likewise.
4023 (loadsf_update): Likewise.
4024 (movsicc_insn): Likewise.
4025 (bset_insn): Likewise.
4026 (bxor_insn): Likewise.
4027 (bclr_insn): Likewise.
4028 (bmsk_insn): Likewise.
4029 (bicsi3_insn): Likewise.
4030 (cmpsi_cc_c_insn): Likewise.
4031 (movsi_ne): Likewise.
4032 (movsi_cond_exec): Likewise.
4033 (clrsbsi2): Likewise.
4044 (call_i): Remove 'S' asm letter, add jli instruction.
4045 (call_value_i): Likewise.
4046 * config/arc/arc.op (mjli-always): New option.
4047 * config/arc/constraints.md (Cji): New constraint.
4048 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4050 (subsf3_fpx): Likewise.
4051 (mulsf3_fpx): Likewise.
4052 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4054 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4056 * doc/invoke.texi (ARC): Document mjli-always option.
4058 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4060 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4061 avoid addition with 0 and use incw and decw where possible.
4063 2018-01-26 Richard Biener <rguenther@suse.de>
4065 PR tree-optimization/81082
4066 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4067 association if it requires casting to unsigned.
4068 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4069 from fold_plusminus_mult_expr to catch important cases late when
4070 range info is available.
4072 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4074 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4075 * configure.ac (hidden_linkonce): New test.
4076 * configure: Regenerate.
4077 * config.in: Regenerate.
4079 2018-01-26 Julia Koval <julia.koval@intel.com>
4081 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4082 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4083 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4084 _mm_mask_bitshuffle_epi64_mask): Fix type.
4085 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4086 USI_FTYPE_V4DI_V4DI_USI): Remove.
4087 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4088 __builtin_ia32_vpshufbitqmb256_mask,
4089 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4090 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4091 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4093 2018-01-26 Alan Modra <amodra@gmail.com>
4096 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4097 UNSPEC_VBPERMQ. Sort other unspecs.
4099 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4101 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4103 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4106 * predict.c (drop_profile): Do not push/pop cfun; update also
4108 (handle_missing_profiles): Fix logic looking for zero profiles.
4110 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4113 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4114 on functions with #pragma omp declare simd or functions with simd
4116 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4117 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4118 Remove trailing \n from warning_at calls.
4120 2018-01-25 Tom de Vries <tom@codesourcery.com>
4123 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4124 for neutered workers.
4126 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4129 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4130 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4132 2018-01-24 Jeff Law <law@redhat.com>
4135 * i386.c (get_probe_interval): Move to earlier point.
4136 (ix86_compute_frame_layout): If -fstack-clash-protection and
4137 the frame is larger than the probe interval, then use pushes
4138 to save registers rather than reg->mem moves.
4139 (ix86_expand_prologue): Remove conditional for int_registers_saved
4142 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4145 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4146 min/max for never referenced object.
4148 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4151 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4153 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4154 attributes from DECL_ATTRIBUTES (decl) without affecting
4155 DECL_ATTRIBUTES (current_function_decl).
4156 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4157 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4159 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4161 PR tree-optimization/83979
4162 * fold-const.c (fold_comparison): Use constant_boolean_node
4163 instead of boolean_{true,false}_node.
4165 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4167 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4170 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4172 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4173 Simplify the clause that sets the length attribute.
4174 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4175 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4176 clause that sets the length attribute.
4177 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4179 2018-01-24 Tom de Vries <tom@codesourcery.com>
4182 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4183 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4184 Add strict parameter.
4185 (prevent_branch_around_nothing): Insert dummy insn between branch to
4186 label and label with no ptx insn inbetween.
4187 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4189 2018-01-24 Tom de Vries <tom@codesourcery.com>
4192 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4193 for neutered threads in warp.
4194 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4196 2018-01-24 Richard Biener <rguenther@suse.de>
4198 PR tree-optimization/83176
4199 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4202 2018-01-24 Richard Biener <rguenther@suse.de>
4204 PR tree-optimization/82819
4205 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4206 code generating pluses that are no-ops in the target precision.
4208 2018-01-24 Richard Biener <rguenther@suse.de>
4211 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4213 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4215 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4216 to merge probabilities.
4217 * predict.c (probably_never_executed): Also mark as cold functions
4218 with global 0 profile and guessed local profile.
4219 * profile-count.c (profile_probability::combine_with_count): New
4221 * profile-count.h (profile_probability::operator*,
4222 profile_probability::operator*=, profile_probability::operator/,
4223 profile_probability::operator/=): Reduce precision to adjusted
4224 and set value to guessed on contradictory divisions.
4225 (profile_probability::combine_with_freq): Remove.
4226 (profile_probability::combine_wiht_count): Declare.
4227 (profile_count::force_nonzero):: Set to adjusted.
4228 (profile_count::probability_in):: Set quality to adjusted.
4229 * tree-ssa-tail-merge.c (replace_block_by): Use
4232 2018-01-23 Andrew Waterman <andrew@sifive.com>
4233 Jim Wilson <jimw@sifive.com>
4235 * config/riscv/riscv.c (riscv_stack_boundary): New.
4236 (riscv_option_override): Set riscv_stack_boundary. Handle
4237 riscv_preferred_stack_boundary_arg.
4238 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4239 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4240 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4241 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4242 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4243 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4245 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4248 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4249 of struct ix86_frame.
4250 (ix86_expand_epilogue): Likewise. Add a local variable for
4251 the reg_save_offset field in struct ix86_frame.
4253 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4255 PR tree-optimization/82604
4256 * tree-loop-distribution.c (enum partition_kind): New enum item
4257 PKIND_PARTIAL_MEMSET.
4258 (partition_builtin_p): Support above new enum item.
4259 (generate_code_for_partition): Ditto.
4260 (compute_access_range): Differentiate cases that equality can be
4261 proven at all loops, the innermost loops or no loops.
4262 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4263 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4264 (finalize_partitions, distribute_loop): Don't fuse partition of
4265 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4266 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4269 2018-01-23 Martin Liska <mliska@suse.cz>
4271 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4272 order to ignore the predictor.
4273 (PRED_POLYMORPHIC_CALL): Likewise.
4274 (PRED_RECURSIVE_CALL): Likewise.
4276 2018-01-23 Martin Liska <mliska@suse.cz>
4278 * tree-profile.c (tree_profiling): Print function header to
4279 aware reader which function we are working on.
4280 * value-prof.c (gimple_find_values_to_profile): Do not print
4281 not interesting value histograms.
4283 2018-01-23 Martin Liska <mliska@suse.cz>
4285 * profile-count.h (enum profile_quality): Add
4286 profile_uninitialized as the first value. Do not number values
4287 as they are zero based.
4288 (profile_count::verify): Update sanity check.
4289 (profile_probability::verify): Likewise.
4291 2018-01-23 Nathan Sidwell <nathan@acm.org>
4293 * doc/invoke.texi (ffor-scope): Deprecate.
4295 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4297 PR tree-optimization/83510
4298 * domwalk.c (set_all_edges_as_executable): New function.
4299 (dom_walker::dom_walker): Convert bool param
4300 "skip_unreachable_blocks" to enum reachability. Move setup of
4301 edge flags to set_all_edges_as_executable and only do it when
4302 reachability is REACHABLE_BLOCKS.
4303 * domwalk.h (enum dom_walker::reachability): New enum.
4304 (dom_walker::dom_walker): Convert bool param
4305 "skip_unreachable_blocks" to enum reachability.
4306 (set_all_edges_as_executable): New decl.
4307 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4308 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4310 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4311 but converting true to REACHABLE_BLOCKS.
4312 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4314 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4315 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4316 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4318 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4319 if check_all_array_refs will be called.
4321 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4323 * tree.c (selftest::test_location_wrappers): Add more test
4326 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4328 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4329 (selftest::test_bit_in_range): Likewise.
4331 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4334 * doc/sourcebuild.texi (vect_float): Say that the selector
4335 only describes the situation when -funsafe-math-optimizations is on.
4336 (vect_float_strict): Document.
4338 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4340 PR tree-optimization/83965
4341 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4342 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4343 instead of checking only for a reduction.
4344 (vect_recog_widen_sum_pattern): Likewise.
4346 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4348 * predict.c (probably_never_executed): Only use precise profile info.
4349 (compute_function_frequency): Skip after inlining hack since we now
4350 have quality checking.
4352 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4354 * profile-count.h (profile_probability::very_unlikely,
4355 profile_probability::unlikely, profile_probability::even): Set
4356 precision to guessed.
4358 2018-01-23 Richard Biener <rguenther@suse.de>
4360 PR tree-optimization/83963
4361 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4362 Properly terminate dominator walk when crossing the exit edge not
4363 when visiting its source block.
4365 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4368 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4369 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4371 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4373 PR tree-optimization/83957
4374 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4375 semicolon after for body surrounded by braces.
4377 PR tree-optimization/83081
4378 * profile-count.h (profile_probability::split): New method.
4379 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4380 Use profile_probability::split.
4381 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4382 when splitting a single conditional jump into 2.
4384 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4386 PR tree-optimization/69452
4387 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4390 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4392 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4393 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4394 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4396 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4398 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4399 * config/rl78/rl78.md (movdi): New define_expand.
4400 * config/rl78/rl78.c (rl78_split_movdi): New function.
4402 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4405 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4407 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4408 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4409 128-bit to produce an UNSPEC move to get the double word with the
4410 signbit and then a shift directly to do signbit.
4411 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4412 implementation with a new version that just does either a direct
4413 move or a regular move. Move memory interface to separate insns.
4414 Move insns so they are next to the expander.
4415 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4416 with signbit move. Split big and little endian case.
4417 (signbit<mode>2_dm_mem_le): Likewise.
4418 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4419 (signbit<mode>2_dm2): Likewise.
4421 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4423 * config/rl78/rl78.md (anddi3): New define_expand.
4425 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4427 * config/rl78/rl78.md (umindi3): New define_expand.
4429 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4431 * config/rl78/rl78.md (smindi3): New define_expand.
4433 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4435 * config/rl78/rl78.md (smaxdi3): New define_expand.
4437 2018-01-22 Carl Love <cel@us.ibm.com>
4439 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4440 LVX_V1TI): Add macro expansion.
4441 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4442 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4443 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4444 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4445 Change check to determine if the instruction is a byte reversing
4446 entry. Fix typo in comment.
4447 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4448 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4449 Add def_builtin calls for new builtins.
4450 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4451 Add define_insn expansion.
4453 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4455 * config/rl78/rl78.md (umaxdi3): New define_expand.
4457 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4459 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4460 for non-QImode registers.
4462 2018-01-22 Richard Biener <rguenther@suse.de>
4464 PR tree-optimization/83963
4465 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4466 including the loop exit block.
4467 (scop_detection::merge_sese): Likewise.
4468 (scop_detection::add_scop): Do it here instead.
4470 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4472 * doc/sourcebuild.texi (arm_softfloat): Document.
4474 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4477 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4478 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4479 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4481 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4482 David Edelsohn <dje.gcc@gmail.com>
4485 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4486 Change "crset eq" to "crset 2".
4487 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4488 (*call_indirect_aix<mode>_nospec): Likewise.
4489 (*call_value_indirect_aix<mode>_nospec): Likewise.
4490 (*call_indirect_elfv2<mode>_nospec): Likewise.
4491 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4492 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4493 change assembly output from . to $.
4494 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4495 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4496 (*tablejump<mode>_internal1_nospec): Likewise.
4498 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4501 * config/sh/sh_optimize_sett_clrt.cc:
4502 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4504 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4506 PR tree-optimization/83940
4507 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4508 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4509 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4510 use it to pass back the definition type.
4511 (vect_check_store_rhs): Likewise rhs_dt_out.
4512 (vect_build_gather_load_calls): Add a mask_dt argument and use
4513 it instead of a call to vect_is_simple_use.
4514 (vectorizable_store): Update calls to vect_check_load_store_mask
4515 and vect_check_store_rhs. Use the dt returned by the latter instead
4516 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4517 instead of calls to vect_is_simple_use. Pass the scalar rather
4518 than the vector operand to vect_is_simple_use when handling
4519 second and subsequent copies of an rhs value.
4520 (vectorizable_load): Update calls to vect_check_load_store_mask
4521 and vect_build_gather_load_calls. Use the cached mask_dt and
4522 gs_info.offset_dt instead of calls to vect_is_simple_use.
4524 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4527 * tree-emutls.c: Include gimplify.h.
4528 (lower_emutls_2): New function.
4529 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4530 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4531 it before further processing.
4534 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4535 UINTVAL (trueop1) instead of INTVAL (op1).
4537 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4541 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4542 INCOMING_FRAME_SP_OFFSET if not defined.
4543 (scan_trace): Add ENTRY argument. If true and
4544 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4545 emit a note to adjust the CFA offset.
4546 (create_cfi_notes): Adjust scan_trace callers.
4547 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4548 INCOMING_FRAME_SP_OFFSET in the CIE.
4549 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4550 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4552 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4553 * doc/tm.texi: Regenerated.
4555 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4557 PR rtl-optimization/83147
4558 * lra-constraints.c (remove_inheritance_pseudos): Use
4559 lra_substitute_pseudo_within_insn.
4561 2018-01-19 Tom de Vries <tom@codesourcery.com>
4562 Cesar Philippidis <cesar@codesourcery.com>
4565 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4567 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4570 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4571 spaces for function labels.
4573 2018-01-19 Martin Liska <mliska@suse.cz>
4575 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4576 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4577 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4578 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4579 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4580 (PRED_CONST_RETURN): Change from 69 to 65.
4581 (PRED_NULL_RETURN): Change from 91 to 71.
4582 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4583 (PRED_LOOP_GUARD): Change from 66 to 73.
4585 2018-01-19 Martin Liska <mliska@suse.cz>
4587 * predict.c (predict_insn_def): Add new assert.
4588 (struct branch_predictor): Change type to signed integer.
4589 (test_prediction_value_range): Amend test to cover
4591 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4592 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4593 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4594 (PRED_LOOP_IV_COMPARE): Likewise.
4595 * predict.h (PROB_UNINITIALIZED): Define new constant.
4597 2018-01-19 Martin Liska <mliska@suse.cz>
4599 * predict.c (dump_prediction): Add new format for
4600 analyze_brprob.py script which is enabled with -details
4602 * profile-count.h (precise_p): New function.
4604 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4606 PR tree-optimization/83922
4607 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4608 there are no statements that need masking.
4609 (vect_active_double_reduction_p): New function.
4610 (vect_analyze_loop_operations): Use it when handling phis that
4611 are not in the loop header.
4613 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4615 PR tree-optimization/83914
4616 * tree-vect-loop.c (vectorizable_induction): Don't convert
4617 init_expr or apply the peeling adjustment for inductions
4618 that are nested within the vectorized loop.
4620 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4622 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4625 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4629 * function.h (gimplify_parameters): Add gimple_seq * argument.
4630 * function.c: Include gimple.h and options.h.
4631 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4632 for the added local temporaries if needed.
4633 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4634 if there are any parameter cleanups, wrap whole body into a
4635 try/finally with the cleanups.
4637 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4640 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4641 Use GET_MODE_CLASS for scalar floating point.
4643 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4647 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4648 Fix call of call_cgraph_insertion_hooks.
4650 2018-01-18 Martin Sebor <msebor@redhat.com>
4652 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4654 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4657 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4660 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4663 * common.opt: (-ffile-prefix-map): New option.
4664 * opts.c (common_handle_option): Defer it.
4665 * opts-global.c (handle_common_deferred_options): Handle it.
4666 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4667 * file-prefix-map.h: New file.
4668 (remap_debug_filename, add_debug_prefix_map): ...here.
4669 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4670 * final.c (debug_prefix_map, add_debug_prefix_map
4671 remap_debug_filename): Move to...
4672 * file-prefix-map.c: New file.
4673 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4674 generalize, get rid of alloca(), use strrchr() instead of strchr().
4675 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4676 Implement in terms of add_prefix_map().
4677 (remap_macro_filename, remap_debug_filename): Implement in term of
4679 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4680 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4681 * dbxout.c: Include file-prefix-map.h.
4682 * varasm.c: Likewise.
4683 * vmsdbgout.c: Likewise.
4684 * xcoffout.c: Likewise.
4685 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4686 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4687 * doc/invoke.texi (-ffile-prefix-map): Document.
4688 (-fdebug-prefix-map): Update description.
4690 2018-01-18 Martin Liska <mliska@suse.cz>
4692 * config/i386/i386.c (indirect_thunk_name): Document that also
4694 (output_indirect_thunk): Document why both instructions
4695 (pause and lfence) are generated.
4697 2018-01-18 Richard Biener <rguenther@suse.de>
4699 PR tree-optimization/83887
4700 * graphite-scop-detection.c
4701 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4702 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4703 (scop_detection::merge_sese): Re-implement with a flood-fill
4704 algorithm that properly finds a SESE region if it exists.
4706 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4709 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4710 pointer_diff optimizations use view_convert instead of convert.
4712 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4714 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4715 Generate different code for -mno-speculate-indirect-jumps.
4716 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4717 (*call_indirect_aix<mode>): Disable for
4718 -mno-speculate-indirect-jumps.
4719 (*call_indirect_aix<mode>_nospec): New define_insn.
4720 (*call_value_indirect_aix<mode>): Disable for
4721 -mno-speculate-indirect-jumps.
4722 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4723 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4724 -mno-speculate-indirect-jumps.
4725 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4727 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4729 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4730 long double type, set the flags for noting the default long double
4731 type, even if we don't pass or return a long double type.
4733 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4736 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4739 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4741 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4742 support for merge[hl].
4743 (fold_mergehl_helper): New helper function.
4744 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4745 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4746 (altivec_vmrglw_direct): Add xxmrglw insn.
4748 2018-01-17 Andrew Waterman <andrew@sifive.com>
4750 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4751 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4753 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4756 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4757 call the lto_location_cache before reading the
4758 DECL_SOURCE_LOCATION of the types.
4760 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4761 Richard Sandiford <richard.sandiford@linaro.org>
4763 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4764 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4765 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4766 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4767 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4769 * config/aarch64/constraints.md (aarch64_movti_operand):
4771 * config/aarch64/predicates.md (Uti): Add new constraint.
4773 2018-01-17 Carl Love <cel@us.ibm.com>
4775 * config/rs6000/vsx.md (define_expand xl_len_r,
4776 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4777 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4779 (define_expand, define_insn): Move the shift left from the
4780 define_insn to the define_expand for lxvl and stxvl instructions.
4781 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4782 and XL_LEN_R definitions to PURE.
4784 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4786 * config/i386/i386.c (indirect_thunk_name): Declare regno
4787 as unsigned int. Compare regno with INVALID_REGNUM.
4788 (output_indirect_thunk): Ditto.
4789 (output_indirect_thunk_function): Ditto.
4790 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4791 in the call to output_indirect_thunk_function.
4793 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4796 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4797 rather than the size of inner_type to determine the stack slot size
4798 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4800 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4803 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4806 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4808 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4809 endian Linux systems to optionally enable multilibs for selecting
4810 the long double type if the user configured an explicit type.
4811 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4812 have no long double multilibs if not defined.
4813 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4814 warn if the user used -mabi={ieee,ibm}longdouble and we built
4815 multilibs for long double.
4816 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4817 appropriate multilib option.
4818 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4820 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4821 for building long double multilibs.
4822 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4824 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4826 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4829 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4831 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4834 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4837 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4840 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4842 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4843 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4845 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4847 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4848 different rtl trees depending on TARGET_64BIT.
4849 (rs6000_gen_lvx): Likewise.
4851 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4853 * config/visium/visium.md (nop): Tweak comment.
4854 (hazard_nop): Likewise.
4856 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4858 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4859 -mspeculate-indirect-jumps.
4860 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4861 for -mno-speculate-indirect-jumps.
4862 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4863 (*call_value_indirect_elfv2<mode>): Disable for
4864 -mno-speculate-indirect-jumps.
4865 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4866 (indirect_jump): Emit different RTL for
4867 -mno-speculate-indirect-jumps.
4868 (*indirect_jump<mode>): Disable for
4869 -mno-speculate-indirect-jumps.
4870 (*indirect_jump<mode>_nospec): New define_insn.
4871 (tablejump): Emit different RTL for
4872 -mno-speculate-indirect-jumps.
4873 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4874 (tablejumpsi_nospec): New define_expand.
4875 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4876 (tablejumpdi_nospec): New define_expand.
4877 (*tablejump<mode>_internal1): Disable for
4878 -mno-speculate-indirect-jumps.
4879 (*tablejump<mode>_internal1_nospec): New define_insn.
4880 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4883 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4885 * caller-save.c (insert_save): Drop unnecessary parameter. All
4888 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4889 Richard Biener <rguenth@suse.de>
4892 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4893 return early, inline manually is_gimple_sizepos. Make sure if we
4894 call gimplify_expr we don't end up with a gimple constant.
4895 * tree.c (variably_modified_type_p): Don't return true for
4896 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4897 * gimplify.h (is_gimple_sizepos): Remove.
4899 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4901 PR tree-optimization/83857
4902 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4903 vectorizable_live_operation for pure SLP statements.
4904 (vectorizable_live_operation): Handle PHIs.
4906 2018-01-16 Richard Biener <rguenther@suse.de>
4908 PR tree-optimization/83867
4909 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4910 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4912 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4915 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4916 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4917 If off is not INTEGER_CST, issue a may not be aligned warning
4918 rather than isn't aligned. Use isn%'t rather than isn't.
4919 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4921 <case MULT_EXPR>: Improve the case when bottom and one of the
4922 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4923 operand, in that case check if the other operand is multiple of
4924 bottom divided by the INTEGER_CST operand.
4926 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4929 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4930 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4931 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4932 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4933 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4934 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4935 (pa_function_arg_size): New function.
4937 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4939 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4940 in a separate statement.
4942 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4944 PR tree-optimization/83847
4945 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4946 group gathers and scatters.
4948 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4950 PR rtl-optimization/86620
4951 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4953 PR rtl-optimization/83213
4954 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4955 to last if both are JUMP_INSNs.
4957 PR tree-optimization/83843
4958 * gimple-ssa-store-merging.c
4959 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4960 store_immediate_info for bswap/nop orig_stores.
4962 2018-01-15 Andrew Waterman <andrew@sifive.com>
4964 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4966 <UDIV>: Increase cost if !TARGET_DIV.
4968 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4970 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4971 (define_attr "cr_logical_3op"): New.
4972 (cceq_ior_compare): Adjust.
4973 (cceq_ior_compare_complement): Adjust.
4974 (*cceq_rev_compare): Adjust.
4975 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4976 (is_cracked_insn): Adjust.
4977 (insn_must_be_first_in_group): Adjust.
4978 * config/rs6000/40x.md: Adjust.
4979 * config/rs6000/440.md: Adjust.
4980 * config/rs6000/476.md: Adjust.
4981 * config/rs6000/601.md: Adjust.
4982 * config/rs6000/603.md: Adjust.
4983 * config/rs6000/6xx.md: Adjust.
4984 * config/rs6000/7450.md: Adjust.
4985 * config/rs6000/7xx.md: Adjust.
4986 * config/rs6000/8540.md: Adjust.
4987 * config/rs6000/cell.md: Adjust.
4988 * config/rs6000/e300c2c3.md: Adjust.
4989 * config/rs6000/e500mc.md: Adjust.
4990 * config/rs6000/e500mc64.md: Adjust.
4991 * config/rs6000/e5500.md: Adjust.
4992 * config/rs6000/e6500.md: Adjust.
4993 * config/rs6000/mpc.md: Adjust.
4994 * config/rs6000/power4.md: Adjust.
4995 * config/rs6000/power5.md: Adjust.
4996 * config/rs6000/power6.md: Adjust.
4997 * config/rs6000/power7.md: Adjust.
4998 * config/rs6000/power8.md: Adjust.
4999 * config/rs6000/power9.md: Adjust.
5000 * config/rs6000/rs64.md: Adjust.
5001 * config/rs6000/titan.md: Adjust.
5003 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5005 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5006 ix86_indirect_branch_register logic.
5008 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5010 * config/i386/constraints.md (Bs): Update
5011 ix86_indirect_branch_register check. Don't check
5012 ix86_indirect_branch_register with GOT_memory_operand.
5014 * config/i386/predicates.md (GOT_memory_operand): Don't check
5015 ix86_indirect_branch_register here.
5016 (GOT32_symbol_operand): Likewise.
5018 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5020 * config/i386/predicates.md (constant_call_address_operand):
5021 Rewrite ix86_indirect_branch_register logic.
5022 (sibcall_insn_operand): Likewise.
5024 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5026 * config/i386/constraints.md (Bs): Replace
5027 ix86_indirect_branch_thunk_register with
5028 ix86_indirect_branch_register.
5030 * config/i386/i386.md (indirect_jump): Likewise.
5031 (tablejump): Likewise.
5032 (*sibcall_memory): Likewise.
5033 (*sibcall_value_memory): Likewise.
5034 Peepholes of indirect call and jump via memory: Likewise.
5035 * config/i386/i386.opt: Likewise.
5036 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5037 (GOT_memory_operand): Likewise.
5038 (call_insn_operand): Likewise.
5039 (sibcall_insn_operand): Likewise.
5040 (GOT32_symbol_operand): Likewise.
5042 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5045 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5046 type rather than type addr's type points to.
5047 (expand_omp_atomic_mutex): Likewise.
5048 (expand_omp_atomic): Likewise.
5050 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5053 * config/i386/i386.c (output_indirect_thunk_function): Use
5054 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5055 for __x86_return_thunk.
5057 2018-01-15 Richard Biener <rguenther@suse.de>
5060 * expmed.c (extract_bit_field_1): Fix typo.
5062 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5065 * config/arm/iterators.md (VF): New mode iterator.
5066 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5067 Remove integer-related logic from pattern.
5068 (neon_vabd<mode>_3): Likewise.
5070 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5073 * common.opt (fstrict-overflow): No longer an alias.
5074 (fwrapv-pointer): New option.
5075 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5076 also for pointer types based on flag_wrapv_pointer.
5077 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5078 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5079 opts->x_flag_wrapv got set.
5080 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5081 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5082 POINTER_TYPE_OVERFLOW_UNDEFINED.
5083 * match.pd: Likewise in address comparison pattern.
5084 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5086 2018-01-15 Richard Biener <rguenther@suse.de>
5089 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5090 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5091 Reset type names to their identifier if their TYPE_DECL doesn't
5092 have linkage (and thus is used for ODR and devirt).
5093 (save_debug_info_for_decl): Remove.
5094 (save_debug_info_for_type): Likewise.
5095 (add_tree_to_fld_list): Adjust.
5096 * tree-pretty-print.c (dump_generic_node): Make dumping of
5097 type names more robust.
5099 2018-01-15 Richard Biener <rguenther@suse.de>
5101 * BASE-VER: Bump to 8.0.1.
5103 2018-01-14 Martin Sebor <msebor@redhat.com>
5106 * builtins.c (check_access): Avoid warning when the no-warning bit
5109 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5111 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5112 * ira-color (allocno_hard_regs_compare): Likewise.
5114 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5117 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5118 Use .pushsection/.popsection.
5120 2018-01-14 Martin Sebor <msebor@redhat.com>
5123 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5125 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5127 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5128 entry from extra_headers.
5129 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5130 extra_headers, make the list bitwise identical to the i?86-*-* one.
5132 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5134 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5135 -mcmodel=large with -mindirect-branch=thunk,
5136 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5137 -mfunction-return=thunk-extern.
5138 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5139 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5140 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5142 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5144 * config/i386/i386.c (print_reg): Print the name of the full
5145 integer register without '%'.
5146 (ix86_print_operand): Handle 'V'.
5147 * doc/extend.texi: Document 'V' modifier.
5149 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5151 * config/i386/constraints.md (Bs): Disallow memory operand for
5152 -mindirect-branch-register.
5154 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5155 (GOT_memory_operand): Likewise.
5156 (call_insn_operand): Likewise.
5157 (sibcall_insn_operand): Likewise.
5158 (GOT32_symbol_operand): Likewise.
5159 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5160 for -mindirect-branch-register.
5161 (tablejump): Likewise.
5162 (*sibcall_memory): Likewise.
5163 (*sibcall_value_memory): Likewise.
5164 Disallow peepholes of indirect call and jump via memory for
5165 -mindirect-branch-register.
5166 (*call_pop): Replace m with Bw.
5167 (*call_value_pop): Likewise.
5168 (*sibcall_pop_memory): Replace m with Bs.
5169 * config/i386/i386.opt (mindirect-branch-register): New option.
5170 * doc/invoke.texi: Document -mindirect-branch-register option.
5172 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5174 * config/i386/i386-protos.h (ix86_output_function_return): New.
5175 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5176 set function_return_type.
5177 (indirect_thunk_name): Add ret_p to indicate thunk for function
5179 (output_indirect_thunk_function): Pass false to
5180 indirect_thunk_name.
5181 (ix86_output_indirect_branch_via_reg): Likewise.
5182 (ix86_output_indirect_branch_via_push): Likewise.
5183 (output_indirect_thunk_function): Create alias for function
5184 return thunk if regno < 0.
5185 (ix86_output_function_return): New function.
5186 (ix86_handle_fndecl_attribute): Handle function_return.
5187 (ix86_attribute_table): Add function_return.
5188 * config/i386/i386.h (machine_function): Add
5189 function_return_type.
5190 * config/i386/i386.md (simple_return_internal): Use
5191 ix86_output_function_return.
5192 (simple_return_internal_long): Likewise.
5193 * config/i386/i386.opt (mfunction-return=): New option.
5194 (indirect_branch): Mention -mfunction-return=.
5195 * doc/extend.texi: Document function_return function attribute.
5196 * doc/invoke.texi: Document -mfunction-return= option.
5198 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5200 * config/i386/i386-opts.h (indirect_branch): New.
5201 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5202 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5203 with local indirect jump when converting indirect call and jump.
5204 (ix86_set_indirect_branch_type): New.
5205 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5206 (indirectlabelno): New.
5207 (indirect_thunk_needed): Likewise.
5208 (indirect_thunk_bnd_needed): Likewise.
5209 (indirect_thunks_used): Likewise.
5210 (indirect_thunks_bnd_used): Likewise.
5211 (INDIRECT_LABEL): Likewise.
5212 (indirect_thunk_name): Likewise.
5213 (output_indirect_thunk): Likewise.
5214 (output_indirect_thunk_function): Likewise.
5215 (ix86_output_indirect_branch_via_reg): Likewise.
5216 (ix86_output_indirect_branch_via_push): Likewise.
5217 (ix86_output_indirect_branch): Likewise.
5218 (ix86_output_indirect_jmp): Likewise.
5219 (ix86_code_end): Call output_indirect_thunk_function if needed.
5220 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5222 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5223 (ix86_attribute_table): Add indirect_branch.
5224 * config/i386/i386.h (machine_function): Add indirect_branch_type
5225 and has_local_indirect_jump.
5226 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5228 (tablejump): Likewise.
5229 (*indirect_jump): Use ix86_output_indirect_jmp.
5230 (*tablejump_1): Likewise.
5231 (simple_return_indirect_internal): Likewise.
5232 * config/i386/i386.opt (mindirect-branch=): New option.
5233 (indirect_branch): New.
5236 (thunk-inline): Likewise.
5237 (thunk-extern): Likewise.
5238 * doc/extend.texi: Document indirect_branch function attribute.
5239 * doc/invoke.texi: Document -mindirect-branch= option.
5241 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5244 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5246 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5248 * ipa-inline.c (want_inline_small_function_p): Return false if
5249 inlining has already failed with CIF_FINAL_ERROR.
5250 (update_caller_keys): Call want_inline_small_function_p before
5252 (update_callee_keys): Likewise.
5254 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5256 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5258 (rs6000_quadword_masked_address_p): Likewise.
5259 (quad_aligned_load_p): Likewise.
5260 (quad_aligned_store_p): Likewise.
5261 (const_load_sequence_p): Add comment to describe the outer-most loop.
5262 (mimic_memory_attributes_and_flags): New function.
5263 (rs6000_gen_stvx): Likewise.
5264 (replace_swapped_aligned_store): Likewise.
5265 (rs6000_gen_lvx): Likewise.
5266 (replace_swapped_aligned_load): Likewise.
5267 (replace_swapped_load_constant): Capitalize argument name in
5268 comment describing this function.
5269 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5270 and stores that access quad-word aligned addresses and replace
5271 with stvx or lvx instructions when appropriate.
5272 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5273 New function prototype.
5274 (rs6000_quadword_masked_address_p): Likewise.
5275 (rs6000_gen_lvx): Likewise.
5276 (rs6000_gen_stvx): Likewise.
5277 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5278 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5279 when memory address is aligned.
5280 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5281 this split to select lvx instruction when memory address is aligned.
5282 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5283 instruction when memory address is aligned.
5284 (*vsx_le_perm_load_v16qi): Likewise.
5285 (four unnamed splitters): Modify to select the stvx instruction
5286 when memory is aligned.
5288 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5290 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5291 which appears in the queue multiple times.
5293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5294 Alan Hayward <alan.hayward@arm.com>
5295 David Sherwood <david.sherwood@arm.com>
5297 * tree-vectorizer.h (vec_lower_bound): New structure.
5298 (_loop_vec_info): Add check_nonzero and lower_bounds.
5299 (LOOP_VINFO_CHECK_NONZERO): New macro.
5300 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5301 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5302 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5303 fields. Make seg_len the distance travelled, not including the
5305 (dr_direction_indicator): Declare.
5306 (dr_zero_step_indicator): Likewise.
5307 (dr_known_forward_stride_p): Likewise.
5308 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5310 (runtime_alias_check_p): Allow runtime alias checks with
5312 (operator ==): Compare access_size and align.
5313 (prune_runtime_alias_test_list): Rework for new distinction between
5314 the access_size and seg_len.
5315 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5317 (get_segment_min_max): New function.
5318 (create_intersect_range_checks): Use it.
5319 (dr_step_indicator): New function.
5320 (dr_direction_indicator): Likewise.
5321 (dr_zero_step_indicator): Likewise.
5322 (dr_known_forward_stride_p): Likewise.
5323 * tree-loop-distribution.c (data_ref_segment_size): Return
5324 DR_STEP * (niters - 1).
5325 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5327 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5328 (vect_preserves_scalar_order_p): New function, split out from...
5329 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5330 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5331 (vect_vfa_access_size): New function.
5332 (vect_vfa_align): Likewise.
5333 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5334 (dump_lower_bound): New function.
5335 (vect_check_lower_bound): Likewise.
5336 (vect_small_gap_p): Likewise.
5337 (vectorizable_with_step_bound_p): Likewise.
5338 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5339 depencies if the vectorization factor is 1. Convert the checks
5340 for nonzero steps into checks on the bounds of DR_STEP. Try using
5341 a bunds check for variable steps if the minimum required step is
5342 relatively small. Update calls to the dr_with_seg_len
5343 constructor and to vect_compile_time_alias.
5344 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5346 (vect_loop_versioning): Call it.
5347 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5349 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5352 Alan Hayward <alan.hayward@arm.com>
5353 David Sherwood <david.sherwood@arm.com>
5355 * doc/sourcebuild.texi (vect_scatter_store): Document.
5356 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5358 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5360 * genopinit.c (main): Add supports_vec_scatter_store and
5361 supports_vec_scatter_store_cached to target_optabs.
5362 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5363 IFN_MASK_SCATTER_STORE.
5364 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5366 * internal-fn.h (internal_store_fn_p): Declare.
5367 (internal_fn_stored_value_index): Likewise.
5368 * internal-fn.c (scatter_store_direct): New macro.
5369 (expand_scatter_store_optab_fn): New function.
5370 (direct_scatter_store_optab_supported_p): New macro.
5371 (internal_store_fn_p): New function.
5372 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5373 IFN_MASK_SCATTER_STORE.
5374 (internal_fn_mask_index): Likewise.
5375 (internal_fn_stored_value_index): New function.
5376 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5378 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5379 * optabs-query.c (supports_vec_scatter_store_p): New function.
5380 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5381 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5382 true for scatter stores.
5383 (vect_gather_scatter_fn_p): Handle scatter stores too.
5384 (vect_check_gather_scatter): Consider using scatter stores if
5385 supports_vec_scatter_store_p.
5386 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5388 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5389 internal_fn_stored_value_index.
5390 (check_load_store_masking): Handle scatter stores too.
5391 (vect_get_store_rhs): Make public.
5392 (vectorizable_call): Use internal_store_fn_p.
5393 (vectorizable_store): Handle scatter store internal functions.
5394 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5395 when deciding whether the end of the group has been reached.
5396 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5397 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5398 (mask_scatter_store<mode>): New insns.
5400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5401 Alan Hayward <alan.hayward@arm.com>
5402 David Sherwood <david.sherwood@arm.com>
5404 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5405 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5406 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5408 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5409 Use vect_truncate_gather_scatter_offset if we can't treat the
5410 operation as a normal gather load or scatter store.
5411 (get_group_load_store_type): Take the gather_scatter_info
5412 as argument. Try using a gather load or scatter store for
5413 single-element groups.
5414 (get_load_store_type): Update calls to get_group_load_store_type
5415 and vect_use_strided_gather_scatters_p.
5417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5418 Alan Hayward <alan.hayward@arm.com>
5419 David Sherwood <david.sherwood@arm.com>
5421 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5422 optional tree argument.
5423 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5425 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5426 but continue to use the current value as a fallback.
5427 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5428 to compare the updates.
5429 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5430 (get_load_store_type): Use it when handling a strided access.
5431 (vect_get_strided_load_store_ops): New function.
5432 (vect_get_data_ptr_increment): Likewise.
5433 (vectorizable_load): Handle strided gather loads. Always pass
5434 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5437 Alan Hayward <alan.hayward@arm.com>
5438 David Sherwood <david.sherwood@arm.com>
5440 * doc/md.texi (gather_load@var{m}): Document.
5441 (mask_gather_load@var{m}): Likewise.
5442 * genopinit.c (main): Add supports_vec_gather_load and
5443 supports_vec_gather_load_cached to target_optabs.
5444 * optabs-tree.c (init_tree_optimization_optabs): Use
5445 ggc_cleared_alloc to allocate target_optabs.
5446 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5447 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5449 * internal-fn.h (internal_load_fn_p): Declare.
5450 (internal_gather_scatter_fn_p): Likewise.
5451 (internal_fn_mask_index): Likewise.
5452 (internal_gather_scatter_fn_supported_p): Likewise.
5453 * internal-fn.c (gather_load_direct): New macro.
5454 (expand_gather_load_optab_fn): New function.
5455 (direct_gather_load_optab_supported_p): New macro.
5456 (direct_internal_fn_optab): New function.
5457 (internal_load_fn_p): Likewise.
5458 (internal_gather_scatter_fn_p): Likewise.
5459 (internal_fn_mask_index): Likewise.
5460 (internal_gather_scatter_fn_supported_p): Likewise.
5461 * optabs-query.c (supports_at_least_one_mode_p): New function.
5462 (supports_vec_gather_load_p): Likewise.
5463 * optabs-query.h (supports_vec_gather_load_p): Declare.
5464 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5465 and memory_type field.
5466 (NUM_PATTERNS): Bump to 15.
5467 * tree-vect-data-refs.c: Include internal-fn.h.
5468 (vect_gather_scatter_fn_p): New function.
5469 (vect_describe_gather_scatter_call): Likewise.
5470 (vect_check_gather_scatter): Try using internal functions for
5471 gather loads. Recognize existing calls to a gather load function.
5472 (vect_analyze_data_refs): Consider using gather loads if
5473 supports_vec_gather_load_p.
5474 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5475 (vect_get_gather_scatter_offset_type): Likewise.
5476 (vect_convert_mask_for_vectype): Likewise.
5477 (vect_add_conversion_to_patterm): Likewise.
5478 (vect_try_gather_scatter_pattern): Likewise.
5479 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5480 (vect_vect_recog_func_ptrs): Add it.
5481 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5482 internal_fn_mask_index and internal_gather_scatter_fn_p.
5483 (check_load_store_masking): Take the gather_scatter_info as an
5484 argument and handle gather loads.
5485 (vect_get_gather_scatter_ops): New function.
5486 (vectorizable_call): Check internal_load_fn_p.
5487 (vectorizable_load): Likewise. Handle gather load internal
5489 (vectorizable_store): Update call to check_load_store_masking.
5490 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5491 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5492 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5493 (aarch64_gather_scale_operand_d): New predicates.
5494 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5495 (mask_gather_load<mode>): New insns.
5497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5498 Alan Hayward <alan.hayward@arm.com>
5499 David Sherwood <david.sherwood@arm.com>
5501 * optabs.def (fold_left_plus_optab): New optab.
5502 * doc/md.texi (fold_left_plus_@var{m}): Document.
5503 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5504 * internal-fn.c (fold_left_direct): Define.
5505 (expand_fold_left_optab_fn): Likewise.
5506 (direct_fold_left_optab_supported_p): Likewise.
5507 * fold-const-call.c (fold_const_fold_left): New function.
5508 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5509 * tree-parloops.c (valid_reduction_p): New function.
5510 (gather_scalar_reductions): Use it.
5511 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5512 (vect_finish_replace_stmt): Declare.
5513 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5514 (needs_fold_left_reduction_p): New function, split out from...
5515 (vect_is_simple_reduction): ...here. Accept reductions that
5516 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5517 (vect_force_simple_reduction): Also store the reduction type in
5518 the assignment's STMT_VINFO_REDUC_TYPE.
5519 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5520 (merge_with_identity): New function.
5521 (vect_expand_fold_left): Likewise.
5522 (vectorize_fold_left_reduction): Likewise.
5523 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5524 scalar phi in place for it. Check for target support and reject
5525 cases that would reassociate the operation. Defer the transform
5526 phase to vectorize_fold_left_reduction.
5527 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5528 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5529 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5533 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5534 call to ifc_temp_var.
5536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5537 Alan Hayward <alan.hayward@arm.com>
5538 David Sherwood <david.sherwood@arm.com>
5540 * target.def (legitimize_address_displacement): Take the original
5541 offset as a poly_int.
5542 * targhooks.h (default_legitimize_address_displacement): Update
5544 * targhooks.c (default_legitimize_address_displacement): Likewise.
5545 * doc/tm.texi: Regenerate.
5546 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5547 as an argument, moving assert of ad->disp == ad->disp_term to...
5548 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5549 Try calling targetm.legitimize_address_displacement before expanding
5550 the address rather than afterwards, and adjust for the new interface.
5551 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5552 Match the new hook interface. Handle SVE addresses.
5553 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5558 * Makefile.in (OBJS): Add early-remat.o.
5559 * target.def (select_early_remat_modes): New hook.
5560 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5561 * doc/tm.texi: Regenerate.
5562 * targhooks.h (default_select_early_remat_modes): Declare.
5563 * targhooks.c (default_select_early_remat_modes): New function.
5564 * timevar.def (TV_EARLY_REMAT): New timevar.
5565 * passes.def (pass_early_remat): New pass.
5566 * tree-pass.h (make_pass_early_remat): Declare.
5567 * early-remat.c: New file.
5568 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5570 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5573 Alan Hayward <alan.hayward@arm.com>
5574 David Sherwood <david.sherwood@arm.com>
5576 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5577 vfm1 with a bound_epilog parameter.
5578 (vect_do_peeling): Update calls accordingly, and move the prologue
5579 call earlier in the function. Treat the base bound_epilog as 0 for
5580 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5581 this base when peeling for gaps.
5582 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5583 with fully-masked loops.
5584 (vect_estimate_min_profitable_iters): Handle the single peeled
5585 iteration in that case.
5587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5588 Alan Hayward <alan.hayward@arm.com>
5589 David Sherwood <david.sherwood@arm.com>
5591 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5592 single-element interleaving even if the size is not a power of 2.
5593 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5594 accesses for single-element interleaving if the group size is
5597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5598 Alan Hayward <alan.hayward@arm.com>
5599 David Sherwood <david.sherwood@arm.com>
5601 * doc/md.texi (fold_extract_last_@var{m}): Document.
5602 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5603 * optabs.def (fold_extract_last_optab): New optab.
5604 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5605 * internal-fn.c (fold_extract_direct): New macro.
5606 (expand_fold_extract_optab_fn): Likewise.
5607 (direct_fold_extract_optab_supported_p): Likewise.
5608 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5609 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5610 EXTRACT_LAST_REDUCTION.
5611 (get_initial_def_for_reduction): Do not create an initial vector
5612 for EXTRACT_LAST_REDUCTION reductions.
5613 (vectorizable_reduction): Leave the scalar phi in place for
5614 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5615 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5616 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5617 transform phase to vectorizable_condition.
5618 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5620 (vect_finish_stmt_generation): ...here.
5621 (vect_finish_replace_stmt): New function.
5622 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5623 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5625 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5627 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5628 Alan Hayward <alan.hayward@arm.com>
5629 David Sherwood <david.sherwood@arm.com>
5631 * doc/md.texi (extract_last_@var{m}): Document.
5632 * optabs.def (extract_last_optab): New optab.
5633 * internal-fn.def (EXTRACT_LAST): New internal function.
5634 * internal-fn.c (cond_unary_direct): New macro.
5635 (expand_cond_unary_optab_fn): Likewise.
5636 (direct_cond_unary_optab_supported_p): Likewise.
5637 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5638 loops using EXTRACT_LAST.
5639 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5640 (extract_last_<mode>): ...this optab.
5641 (vec_extract<mode><Vel>): Update accordingly.
5643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5644 Alan Hayward <alan.hayward@arm.com>
5645 David Sherwood <david.sherwood@arm.com>
5647 * target.def (empty_mask_is_expensive): New hook.
5648 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5649 * doc/tm.texi: Regenerate.
5650 * targhooks.h (default_empty_mask_is_expensive): Declare.
5651 * targhooks.c (default_empty_mask_is_expensive): New function.
5652 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5653 if the target says that empty masks are expensive.
5654 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5656 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5659 Alan Hayward <alan.hayward@arm.com>
5660 David Sherwood <david.sherwood@arm.com>
5662 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5663 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5664 (vect_use_loop_mask_for_alignment_p): New function.
5665 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5666 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5667 niters_skip argument. Make sure that the first niters_skip elements
5668 of the first iteration are inactive.
5669 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5670 Update call to vect_set_loop_masks_directly.
5671 (get_misalign_in_elems): New function, split out from...
5672 (vect_gen_prolog_loop_niters): ...here.
5673 (vect_update_init_of_dr): Take a code argument that specifies whether
5674 the adjustment should be added or subtracted.
5675 (vect_update_init_of_drs): Likewise.
5676 (vect_prepare_for_masked_peels): New function.
5677 (vect_do_peeling): Skip prologue peeling if we're using a mask
5678 instead. Update call to vect_update_inits_of_drs.
5679 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5681 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5682 alignment. Do not include the number of peeled iterations in
5683 the minimum threshold in that case.
5684 (vectorizable_induction): Adjust the start value down by
5685 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5686 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5687 Take the number of skipped iterations into account when calculating
5689 * tree-vect-stmts.c (vect_gen_while_not): New function.
5691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5692 Alan Hayward <alan.hayward@arm.com>
5693 David Sherwood <david.sherwood@arm.com>
5695 * doc/sourcebuild.texi (vect_fully_masked): Document.
5696 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5698 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5700 (vect_analyze_loop_2): ...here. Don't check the vectorization
5701 factor against the number of loop iterations if the loop is
5704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5705 Alan Hayward <alan.hayward@arm.com>
5706 David Sherwood <david.sherwood@arm.com>
5708 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5709 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5710 (dump_groups): Update accordingly.
5711 (iv_use::mem_type): New member variable.
5712 (address_p): New function.
5713 (record_use): Add a mem_type argument and initialize the new
5715 (record_group_use): Add a mem_type argument. Use address_p.
5716 Remove obsolete null checks of base_object. Update call to record_use.
5717 (find_interesting_uses_op): Update call to record_group_use.
5718 (find_interesting_uses_cond): Likewise.
5719 (find_interesting_uses_address): Likewise.
5720 (get_mem_type_for_internal_fn): New function.
5721 (find_address_like_use): Likewise.
5722 (find_interesting_uses_stmt): Try find_address_like_use before
5723 calling find_interesting_uses_op.
5724 (addr_offset_valid_p): Use the iv mem_type field as the type
5725 of the addressed memory.
5726 (add_autoinc_candidates): Likewise.
5727 (get_address_cost): Likewise.
5728 (split_small_address_groups_p): Use address_p.
5729 (split_address_groups): Likewise.
5730 (add_iv_candidate_for_use): Likewise.
5731 (autoinc_possible_for_pair): Likewise.
5732 (rewrite_groups): Likewise.
5733 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5734 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5735 (get_alias_ptr_type_for_ptr_address): New function.
5736 (rewrite_use_address): Rewrite address uses in calls that were
5737 identified by find_address_like_use.
5739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5740 Alan Hayward <alan.hayward@arm.com>
5741 David Sherwood <david.sherwood@arm.com>
5743 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5745 * gimple-expr.h (is_gimple_addressable: Likewise.
5746 * gimple-expr.c (is_gimple_address): Likewise.
5747 * internal-fn.c (expand_call_mem_ref): New function.
5748 (expand_mask_load_optab_fn): Use it.
5749 (expand_mask_store_optab_fn): Likewise.
5751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5752 Alan Hayward <alan.hayward@arm.com>
5753 David Sherwood <david.sherwood@arm.com>
5755 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5756 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5757 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5758 (cond_umax@var{mode}): Document.
5759 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5760 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5761 (cond_umin_optab, cond_umax_optab): New optabs.
5762 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5763 (COND_IOR, COND_XOR): New internal functions.
5764 * internal-fn.h (get_conditional_internal_fn): Declare.
5765 * internal-fn.c (cond_binary_direct): New macro.
5766 (expand_cond_binary_optab_fn): Likewise.
5767 (direct_cond_binary_optab_supported_p): Likewise.
5768 (get_conditional_internal_fn): New function.
5769 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5770 Cope with reduction statements that are vectorized as calls rather
5772 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5773 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5774 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5775 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5776 (UNSPEC_COND_EOR): New unspecs.
5777 (optab): Add mappings for them.
5778 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5779 (sve_int_op, sve_fp_op): New int attributes.
5781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5782 Alan Hayward <alan.hayward@arm.com>
5783 David Sherwood <david.sherwood@arm.com>
5785 * optabs.def (while_ult_optab): New optab.
5786 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5787 * internal-fn.def (WHILE_ULT): New internal function.
5788 * internal-fn.h (direct_internal_fn_supported_p): New override
5789 that takes two types as argument.
5790 * internal-fn.c (while_direct): New macro.
5791 (expand_while_optab_fn): New function.
5792 (convert_optab_supported_p): Likewise.
5793 (direct_while_optab_supported_p): New macro.
5794 * wide-int.h (wi::udiv_ceil): New function.
5795 * tree-vectorizer.h (rgroup_masks): New structure.
5796 (vec_loop_masks): New typedef.
5797 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5799 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5800 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5801 (vect_max_vf): New function.
5802 (slpeel_make_loop_iterate_ntimes): Delete.
5803 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5804 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5805 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5806 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5807 internal-fn.h, stor-layout.h and optabs-query.h.
5808 (vect_set_loop_mask): New function.
5809 (add_preheader_seq): Likewise.
5810 (add_header_seq): Likewise.
5811 (interleave_supported_p): Likewise.
5812 (vect_maybe_permute_loop_masks): Likewise.
5813 (vect_set_loop_masks_directly): Likewise.
5814 (vect_set_loop_condition_masked): Likewise.
5815 (vect_set_loop_condition_unmasked): New function, split out from
5816 slpeel_make_loop_iterate_ntimes.
5817 (slpeel_make_loop_iterate_ntimes): Rename to..
5818 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5819 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5820 (vect_do_peeling): Update call accordingly.
5821 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5823 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5824 mask_compare_type, can_fully_mask_p and fully_masked_p.
5825 (release_vec_loop_masks): New function.
5826 (_loop_vec_info): Use it to free the loop masks.
5827 (can_produce_all_loop_masks_p): New function.
5828 (vect_get_max_nscalars_per_iter): Likewise.
5829 (vect_verify_full_masking): Likewise.
5830 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5831 retries, and free the mask rgroups before retrying. Check loop-wide
5832 reasons for disallowing fully-masked loops. Make the final decision
5833 about whether use a fully-masked loop or not.
5834 (vect_estimate_min_profitable_iters): Do not assume that peeling
5835 for the number of iterations will be needed for fully-masked loops.
5836 (vectorizable_reduction): Disable fully-masked loops.
5837 (vectorizable_live_operation): Likewise.
5838 (vect_halve_mask_nunits): New function.
5839 (vect_double_mask_nunits): Likewise.
5840 (vect_record_loop_mask): Likewise.
5841 (vect_get_loop_mask): Likewise.
5842 (vect_transform_loop): Handle the case in which the final loop
5843 iteration might handle a partial vector. Call vect_set_loop_condition
5844 instead of slpeel_make_loop_iterate_ntimes.
5845 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5846 (check_load_store_masking): New function.
5847 (prepare_load_store_mask): Likewise.
5848 (vectorizable_store): Handle fully-masked loops.
5849 (vectorizable_load): Likewise.
5850 (supportable_widening_operation): Use vect_halve_mask_nunits for
5852 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5853 (vect_gen_while): New function.
5854 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5855 (aarch64_uqdec<mode>): New insn.
5857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5858 Alan Hayward <alan.hayward@arm.com>
5859 David Sherwood <david.sherwood@arm.com>
5861 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5862 (reduc_xor_scal_optab): New optabs.
5863 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5864 (reduc_xor_scal_@var{m}): Document.
5865 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5866 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5868 * fold-const-call.c (fold_const_call): Handle them.
5869 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5870 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5871 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5872 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5873 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5874 (UNSPEC_XORV): New unspecs.
5875 (optab): Add entries for them.
5876 (BITWISEV): New int iterator.
5877 (bit_reduc_op): New int attributes.
5879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5880 Alan Hayward <alan.hayward@arm.com>
5881 David Sherwood <david.sherwood@arm.com>
5883 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5884 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5885 * optabs.def (vec_shl_insert_optab): New optab.
5886 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5887 (duplicate_and_interleave): Likewise.
5888 * tree-vect-loop.c: Include internal-fn.h.
5889 (neutral_op_for_slp_reduction): New function, split out from
5890 get_initial_defs_for_reduction.
5891 (get_initial_def_for_reduction): Handle option 2 for variable-length
5892 vectors by loading the neutral value into a vector and then shifting
5893 the initial value into element 0.
5894 (get_initial_defs_for_reduction): Replace the code argument with
5895 the neutral value calculated by neutral_op_for_slp_reduction.
5896 Use gimple_build_vector for constant-length vectors.
5897 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5898 but the first group_size elements have a neutral value.
5899 Use duplicate_and_interleave otherwise.
5900 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5901 Update call to get_initial_defs_for_reduction. Handle SLP
5902 reductions for variable-length vectors by creating one vector
5903 result for each scalar result, with the elements associated
5904 with other scalar results stubbed out with the neutral value.
5905 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5906 Require IFN_VEC_SHL_INSERT for double reductions on
5907 variable-length vectors, or SLP reductions that have
5908 a neutral value. Require can_duplicate_and_interleave_p
5909 support for variable-length unchained SLP reductions if there
5910 is no neutral value, such as for MIN/MAX reductions. Also require
5911 the number of vector elements to be a multiple of the number of
5912 SLP statements when doing variable-length unchained SLP reductions.
5913 Update call to vect_create_epilog_for_reduction.
5914 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5915 and remove initial values.
5916 (duplicate_and_interleave): Make public.
5917 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5918 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5921 Alan Hayward <alan.hayward@arm.com>
5922 David Sherwood <david.sherwood@arm.com>
5924 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5925 (can_duplicate_and_interleave_p): New function.
5926 (vect_get_and_check_slp_defs): Take the vector of statements
5927 rather than just the current one. Remove excess parentheses.
5928 Restriction rejectinon of vect_constant_def and vect_external_def
5929 for variable-length vectors to boolean types, or types for which
5930 can_duplicate_and_interleave_p is false.
5931 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5932 (duplicate_and_interleave): New function.
5933 (vect_get_constant_vectors): Use gimple_build_vector for
5934 constant-length vectors and suitable variable-length constant
5935 vectors. Use duplicate_and_interleave for other variable-length
5936 vectors. Don't defer the update when inserting new statements.
5938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5939 Alan Hayward <alan.hayward@arm.com>
5940 David Sherwood <david.sherwood@arm.com>
5942 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5943 min_profitable_iters doesn't go negative.
5945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5946 Alan Hayward <alan.hayward@arm.com>
5947 David Sherwood <david.sherwood@arm.com>
5949 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5950 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5951 * optabs.def (vec_mask_load_lanes_optab): New optab.
5952 (vec_mask_store_lanes_optab): Likewise.
5953 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5954 (MASK_STORE_LANES): Likewise.
5955 * internal-fn.c (mask_load_lanes_direct): New macro.
5956 (mask_store_lanes_direct): Likewise.
5957 (expand_mask_load_optab_fn): Handle masked operations.
5958 (expand_mask_load_lanes_optab_fn): New macro.
5959 (expand_mask_store_optab_fn): Handle masked operations.
5960 (expand_mask_store_lanes_optab_fn): New macro.
5961 (direct_mask_load_lanes_optab_supported_p): Likewise.
5962 (direct_mask_store_lanes_optab_supported_p): Likewise.
5963 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5965 (vect_load_lanes_supported): Likewise.
5966 * tree-vect-data-refs.c (strip_conversion): New function.
5967 (can_group_stmts_p): Likewise.
5968 (vect_analyze_data_ref_accesses): Use it instead of checking
5969 for a pair of assignments.
5970 (vect_store_lanes_supported): Take a masked_p parameter.
5971 (vect_load_lanes_supported): Likewise.
5972 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5973 vect_store_lanes_supported and vect_load_lanes_supported.
5974 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5975 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5976 parameter. Don't allow gaps for masked accesses.
5977 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5978 and vect_load_lanes_supported.
5979 (get_load_store_type): Take a masked_p parameter and update
5980 call to get_group_load_store_type.
5981 (vectorizable_store): Update call to get_load_store_type.
5982 Handle IFN_MASK_STORE_LANES.
5983 (vectorizable_load): Update call to get_load_store_type.
5984 Handle IFN_MASK_LOAD_LANES.
5986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5987 Alan Hayward <alan.hayward@arm.com>
5988 David Sherwood <david.sherwood@arm.com>
5990 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5992 * config/aarch64/aarch64-protos.h
5993 (aarch64_sve_struct_memory_operand_p): Declare.
5994 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5995 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5996 (VPRED, vpred): Handle SVE structure modes.
5997 * config/aarch64/constraints.md (Utx): New constraint.
5998 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5999 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6000 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6001 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6002 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6003 structure modes. Split into pieces after RA.
6004 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6005 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6007 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6008 SVE structure modes.
6009 (aarch64_classify_address): Likewise.
6010 (sizetochar): Move earlier in file.
6011 (aarch64_print_operand): Handle SVE register lists.
6012 (aarch64_array_mode): New function.
6013 (aarch64_sve_struct_memory_operand_p): Likewise.
6014 (TARGET_ARRAY_MODE): Redefine.
6016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6017 Alan Hayward <alan.hayward@arm.com>
6018 David Sherwood <david.sherwood@arm.com>
6020 * target.def (array_mode): New target hook.
6021 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6022 * doc/tm.texi: Regenerate.
6023 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6024 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6025 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6027 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6031 Alan Hayward <alan.hayward@arm.com>
6032 David Sherwood <david.sherwood@arm.com>
6034 * fold-const.c (fold_binary_loc): Check the argument types
6035 rather than the result type when testing for a vector operation.
6037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6039 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6040 * doc/tm.texi: Regenerate.
6042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6043 Alan Hayward <alan.hayward@arm.com>
6044 David Sherwood <david.sherwood@arm.com>
6046 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6047 (sve): Document new AArch64 extension.
6048 * doc/md.texi (w): Extend the description of the AArch64
6049 constraint to include SVE vectors.
6050 (Upl, Upa): Document new AArch64 predicate constraints.
6051 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6053 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6054 (msve-vector-bits=): New option.
6055 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6056 SVE when these are disabled.
6057 (sve): New extension.
6058 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6059 modes. Adjust their number of units based on aarch64_sve_vg.
6060 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6061 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6062 aarch64_addr_query_type.
6063 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6064 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6065 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6066 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6067 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6068 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6069 (aarch64_simd_imm_zero_p): Delete.
6070 (aarch64_check_zero_based_sve_index_immediate): Declare.
6071 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6072 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6073 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6074 (aarch64_sve_float_mul_immediate_p): Likewise.
6075 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6077 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6078 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6079 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6080 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6081 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6082 (aarch64_regmode_natural_size): Likewise.
6083 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6084 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6086 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6087 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6088 for VG and the SVE predicate registers.
6089 (V_ALIASES): Add a "z"-prefixed alias.
6090 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6091 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6092 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6093 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6094 (REG_CLASS_NAMES): Add entries for them.
6095 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6096 and the predicate registers.
6097 (aarch64_sve_vg): Declare.
6098 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6099 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6100 (REGMODE_NATURAL_SIZE): Define.
6101 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6103 * config/aarch64/aarch64.c: Include cfgrtl.h.
6104 (simd_immediate_info): Add a constructor for series vectors,
6105 and an associated step field.
6106 (aarch64_sve_vg): New variable.
6107 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6108 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6109 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6110 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6111 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6112 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6113 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6114 (aarch64_get_mask_mode): New functions.
6115 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6116 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6117 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6118 predicate modes and predicate registers. Explicitly restrict
6119 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6120 to store a vector mode if it is recognized by
6121 aarch64_classify_vector_mode.
6122 (aarch64_regmode_natural_size): New function.
6123 (aarch64_hard_regno_caller_save_mode): Return the original mode
6125 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6126 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6127 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6128 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6130 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6131 does not overlap dest if the function is frame-related. Handle
6133 (aarch64_split_add_offset): New function.
6134 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6135 them aarch64_add_offset.
6136 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6137 and update call to aarch64_sub_sp.
6138 (aarch64_add_cfa_expression): New function.
6139 (aarch64_expand_prologue): Pass extra temporary registers to the
6140 functions above. Handle the case in which we need to emit new
6141 DW_CFA_expressions for registers that were originally saved
6142 relative to the stack pointer, but now have to be expressed
6143 relative to the frame pointer.
6144 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6146 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6147 IP0 and IP1 values for SVE frames.
6148 (aarch64_expand_vec_series): New function.
6149 (aarch64_expand_sve_widened_duplicate): Likewise.
6150 (aarch64_expand_sve_const_vector): Likewise.
6151 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6152 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6153 into the register, rather than emitting a SET directly.
6154 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6155 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6156 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6157 (offset_9bit_signed_scaled_p): New functions.
6158 (aarch64_replicate_bitmask_imm): New function.
6159 (aarch64_bitmask_imm): Use it.
6160 (aarch64_cannot_force_const_mem): Reject expressions involving
6161 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6162 (aarch64_classify_index): Handle SVE indices, by requiring
6163 a plain register index with a scale that matches the element size.
6164 (aarch64_classify_address): Handle SVE addresses. Assert that
6165 the mode of the address is VOIDmode or an integer mode.
6166 Update call to aarch64_classify_symbol.
6167 (aarch64_classify_symbolic_expression): Update call to
6168 aarch64_classify_symbol.
6169 (aarch64_const_vec_all_in_range_p): New function.
6170 (aarch64_print_vector_float_operand): Likewise.
6171 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6172 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6173 and the FP immediates 1.0 and 0.5.
6174 (aarch64_print_address_internal): Handle SVE addresses.
6175 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6176 (aarch64_regno_regclass): Handle predicate registers.
6177 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6179 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6180 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6181 (aarch64_convert_sve_vector_bits): New function.
6182 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6183 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6185 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6186 Handle SVE vector and predicate modes. Accept VL-based constants
6187 that need only one temporary register, and VL offsets that require
6188 no temporary registers.
6189 (aarch64_conditional_register_usage): Mark the predicate registers
6190 as fixed if SVE isn't available.
6191 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6192 Return true for SVE vector and predicate modes.
6193 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6194 rather than an unsigned int. Handle SVE modes.
6195 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6197 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6199 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6200 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6201 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6202 (aarch64_sve_float_mul_immediate_p): New functions.
6203 (aarch64_sve_valid_immediate): New function.
6204 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6205 Explicitly reject structure modes. Check for INDEX constants.
6206 Handle PTRUE and PFALSE constants.
6207 (aarch64_check_zero_based_sve_index_immediate): New function.
6208 (aarch64_simd_imm_zero_p): Delete.
6209 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6210 vector modes. Accept constants in the range of CNT[BHWD].
6211 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6212 ask for an Advanced SIMD mode.
6213 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6214 (aarch64_simd_vector_alignment): Handle SVE predicates.
6215 (aarch64_vectorize_preferred_vector_alignment): New function.
6216 (aarch64_simd_vector_alignment_reachable): Use it instead of
6218 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6219 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6221 (MAX_VECT_LEN): Delete.
6222 (expand_vec_perm_d): Add a vec_flags field.
6223 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6224 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6225 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6227 (aarch64_evpc_rev): Rename to...
6228 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6229 (aarch64_evpc_rev_global): New function.
6230 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6231 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6233 (aarch64_evpc_sve_tbl): New function.
6234 (aarch64_expand_vec_perm_const_1): Update after rename of
6235 aarch64_evpc_rev. Handle SVE permutes too, trying
6236 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6237 than aarch64_evpc_tbl.
6238 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6239 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6240 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6241 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6242 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6243 (aarch64_expand_sve_vcond): New functions.
6244 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6245 of aarch64_vector_mode_p.
6246 (aarch64_dwarf_poly_indeterminate_value): New function.
6247 (aarch64_compute_pressure_classes): Likewise.
6248 (aarch64_can_change_mode_class): Likewise.
6249 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6250 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6251 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6252 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6253 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6254 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6255 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6256 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6258 (Dn, Dl, Dr): Accept const as well as const_vector.
6259 (Dz): Likewise. Compare against CONST0_RTX.
6260 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6261 of "vector" where appropriate.
6262 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6263 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6264 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6265 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6266 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6267 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6268 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6269 (v_int_equiv): Extend to SVE modes.
6270 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6272 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6273 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6274 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6275 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6276 (SVE_COND_FP_CMP): New int iterators.
6277 (perm_hilo): Handle the new unpack unspecs.
6278 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6280 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6281 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6282 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6283 (aarch64_equality_operator, aarch64_constant_vector_operand)
6284 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6285 (aarch64_sve_nonimmediate_operand): Likewise.
6286 (aarch64_sve_general_operand): Likewise.
6287 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6288 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6289 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6290 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6291 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6292 (aarch64_sve_float_arith_immediate): Likewise.
6293 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6294 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6295 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6296 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6297 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6298 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6299 (aarch64_sve_float_arith_operand): Likewise.
6300 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6301 (aarch64_sve_float_mul_operand): Likewise.
6302 (aarch64_sve_vec_perm_operand): Likewise.
6303 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6304 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6305 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6306 as well as const_vector.
6307 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6308 in file. Use CONST0_RTX and CONSTM1_RTX.
6309 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6310 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6311 Use aarch64_simd_imm_zero.
6312 * config/aarch64/aarch64-sve.md: New file.
6313 * config/aarch64/aarch64.md: Include it.
6314 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6315 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6316 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6317 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6318 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6319 (sve): New attribute.
6320 (enabled): Disable instructions with the sve attribute unless
6322 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6323 aarch64_expand_mov_immediate.
6324 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6325 CNT[BHSD] immediates.
6326 (movti): Split CONST_POLY_INT moves into two halves.
6327 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6328 Split additions that need a temporary here if the destination
6329 is the stack pointer.
6330 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6331 (*add<mode>3_poly_1): New instruction.
6332 (set_clobber_cc): New expander.
6334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6336 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6337 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6338 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6339 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6340 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6341 Change innermode from fixed_mode_size to machine_mode.
6342 (simplify_subreg): Update call accordingly. Handle a constant-sized
6343 subreg of a variable-length CONST_VECTOR.
6345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6346 Alan Hayward <alan.hayward@arm.com>
6347 David Sherwood <david.sherwood@arm.com>
6349 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6350 (add_offset_to_base): New function, split out from...
6351 (create_mem_ref): ...here. When handling a scale other than 1,
6352 check first whether the address is valid without the offset.
6353 Add it into the base if so, leaving the index and scale as-is.
6355 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6358 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6359 fold_for_warn before checking if arg2 is INTEGER_CST.
6361 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6363 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6364 (store_multiple_operation): Delete.
6365 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6366 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6367 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6368 guarded by TARGET_STRING.
6369 (rs6000_output_load_multiple): Delete.
6370 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6371 OPTION_MASK_STRING / TARGET_STRING handling.
6372 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6373 (const rs6000_opt_masks) <"string">: Change mask to 0.
6374 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6375 (MASK_STRING): Delete.
6376 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6378 (load_multiple): Delete.
6385 (store_multiple): Delete.
6392 (movmemsi_8reg): Delete.
6393 (corresponding unnamed define_insn): Delete.
6394 (movmemsi_6reg): Delete.
6395 (corresponding unnamed define_insn): Delete.
6396 (movmemsi_4reg): Delete.
6397 (corresponding unnamed define_insn): Delete.
6398 (movmemsi_2reg): Delete.
6399 (corresponding unnamed define_insn): Delete.
6400 (movmemsi_1reg): Delete.
6401 (corresponding unnamed define_insn): Delete.
6402 * config/rs6000/rs6000.opt (mno-string): New.
6403 (mstring): Replace by deprecation warning stub.
6404 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6406 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6408 * regrename.c (regrename_do_replace): If replacing the same
6409 reg multiple times, try to reuse last created gen_raw_REG.
6412 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6413 main to workaround a bug in GDB.
6415 2018-01-12 Tom de Vries <tom@codesourcery.com>
6418 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6420 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6422 PR rtl-optimization/80481
6423 * ira-color.c (get_cap_member): New function.
6424 (allocnos_conflict_by_live_ranges_p): Use it.
6425 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6426 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6428 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6431 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6432 (*saddl_se_1): Ditto.
6434 (*ssubl_se_1): Ditto.
6436 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6438 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6439 rather than wi::to_widest for DR_INITs.
6440 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6441 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6442 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6444 (vect_analyze_group_access_1): Note that here.
6446 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6448 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6449 polynomial type sizes.
6451 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6453 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6454 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6455 (gimple_add_tmp_var): Likewise.
6457 2018-01-12 Martin Liska <mliska@suse.cz>
6459 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6460 (gimple_alloc_sizes): Likewise.
6461 (dump_gimple_statistics): Use PRIu64 in printf format.
6462 * gimple.h: Change uint64_t to int.
6464 2018-01-12 Martin Liska <mliska@suse.cz>
6466 * tree-core.h: Use uint64_t instead of int.
6467 * tree.c (tree_node_counts): Likewise.
6468 (tree_node_sizes): Likewise.
6469 (dump_tree_statistics): Use PRIu64 in printf format.
6471 2018-01-12 Martin Liska <mliska@suse.cz>
6473 * Makefile.in: As qsort_chk is implemented in vec.c, add
6474 vec.o to linkage of gencfn-macros.
6475 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6476 passing the info to record_node_allocation_statistics.
6477 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6479 * ggc-common.c (struct ggc_usage): Add operator== and use
6480 it in operator< and compare function.
6481 * mem-stats.h (struct mem_usage): Likewise.
6482 * vec.c (struct vec_usage): Remove operator< and compare
6483 function. Can be simply inherited.
6485 2018-01-12 Martin Jambor <mjambor@suse.cz>
6488 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6489 * tree-ssa-math-opts.c: Include domwalk.h.
6490 (convert_mult_to_fma_1): New function.
6491 (fma_transformation_info): New type.
6492 (fma_deferring_state): Likewise.
6493 (cancel_fma_deferring): New function.
6494 (result_of_phi): Likewise.
6495 (last_fma_candidate_feeds_initial_phi): Likewise.
6496 (convert_mult_to_fma): Added deferring logic, split actual
6497 transformation to convert_mult_to_fma_1.
6498 (math_opts_dom_walker): New type.
6499 (math_opts_dom_walker::after_dom_children): New method, body moved
6500 here from pass_optimize_widening_mul::execute, added deferring logic
6502 (pass_optimize_widening_mul::execute): Moved most of code to
6503 math_opts_dom_walker::after_dom_children.
6504 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6505 * config/i386/i386.c (ix86_option_override_internal): Added
6506 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6508 2018-01-12 Richard Biener <rguenther@suse.de>
6511 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6512 inline instance vars.
6514 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6517 * config/rx/rx.c (rx_is_restricted_memory_address):
6520 2018-01-12 Richard Biener <rguenther@suse.de>
6522 PR tree-optimization/80846
6523 * target.def (split_reduction): New target hook.
6524 * targhooks.c (default_split_reduction): New function.
6525 * targhooks.h (default_split_reduction): Declare.
6526 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6527 target requests first reduce vectors by combining low and high
6529 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6530 (get_vectype_for_scalar_type_and_size): Export.
6531 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6532 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6533 * doc/tm.texi: Regenerate.
6534 * config/i386/i386.c (ix86_split_reduction): Implement
6535 TARGET_VECTORIZE_SPLIT_REDUCTION.
6537 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6540 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6541 in PIC mode except for TARGET_VXWORKS_RTP.
6542 * config/sparc/sparc.c: Include cfgrtl.h.
6543 (TARGET_INIT_PIC_REG): Define.
6544 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6545 (sparc_pic_register_p): New predicate.
6546 (sparc_legitimate_address_p): Use it.
6547 (sparc_legitimize_pic_address): Likewise.
6548 (sparc_delegitimize_address): Likewise.
6549 (sparc_mode_dependent_address_p): Likewise.
6550 (gen_load_pcrel_sym): Remove 4th parameter.
6551 (load_got_register): Adjust call to above. Remove obsolete stuff.
6552 (sparc_expand_prologue): Do not call load_got_register here.
6553 (sparc_flat_expand_prologue): Likewise.
6554 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6555 (sparc_use_pseudo_pic_reg): New function.
6556 (sparc_init_pic_reg): Likewise.
6557 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6558 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6560 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6562 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6563 Add item for branch_cost.
6565 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6567 PR rtl-optimization/83565
6568 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6569 not extend the result to a larger mode for rotate operations.
6570 (num_sign_bit_copies1): Likewise.
6572 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6575 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6577 Use values-Xc.o for -pedantic.
6578 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6580 2018-01-12 Martin Liska <mliska@suse.cz>
6583 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6585 (possible_polymorphic_call_targets): Use it.
6586 (ipa_devirt): Likewise.
6588 2018-01-12 Martin Liska <mliska@suse.cz>
6590 * profile-count.h (enum profile_quality): Use 0 as invalid
6591 enum value of profile_quality.
6593 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6595 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6596 -mext-string options.
6598 2018-01-12 Richard Biener <rguenther@suse.de>
6600 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6601 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6602 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6604 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6606 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6608 * configure.ac (--with-long-double-format): Add support for the
6609 configuration option to change the default long double format on
6611 * config.gcc (powerpc*-linux*-*): Likewise.
6612 * configure: Regenerate.
6613 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6614 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6615 used without modification.
6617 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6619 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6620 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6621 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6622 MISC_BUILTIN_SPEC_BARRIER.
6623 (rs6000_init_builtins): Likewise.
6624 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6626 (speculation_barrier): New define_insn.
6627 * doc/extend.texi: Document __builtin_speculation_barrier.
6629 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6632 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6633 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6634 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6636 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6637 integral modes instead of "ss" and "sd".
6638 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6639 vectors with 32-bit and 64-bit elements.
6640 (vecdupssescalarmodesuffix): New mode attribute.
6641 (vec_dup<mode>): Use it.
6643 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6646 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6647 frame if argument is passed on stack.
6649 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6652 * ree.c (combine_reaching_defs): Optimize also
6653 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6654 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6656 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6659 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6661 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6664 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6665 after they are computed.
6667 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6669 PR tree-optimization/83695
6670 * gimple-loop-linterchange.cc
6671 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6672 reset cached scev information after interchange.
6673 (pass_linterchange::execute): Remove call to scev_reset_htab.
6675 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6677 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6678 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6679 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6680 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6681 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6682 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6683 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6684 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6685 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6686 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6687 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6688 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6689 (V_lane_reg): Likewise.
6690 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6692 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6693 (vfmal_lane_low<mode>_intrinsic,
6694 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6695 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6696 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6697 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6698 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6699 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6701 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6703 * config/arm/arm-cpus.in (fp16fml): New feature.
6704 (ALL_SIMD): Add fp16fml.
6705 (armv8.2-a): Add fp16fml as an option.
6706 (armv8.3-a): Likewise.
6707 (armv8.4-a): Add fp16fml as part of fp16.
6708 * config/arm/arm.h (TARGET_FP16FML): Define.
6709 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6711 * config/arm/arm-modes.def (V2HF): Define.
6712 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6713 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6714 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6715 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6716 vfmsl_low, vfmsl_high): New set of builtins.
6717 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6718 (vfml_op): New code attribute.
6719 (VFMLHALVES): New int iterator.
6720 (VFML, VFMLSEL): New mode attributes.
6721 (V_reg): Define mapping for V2HF.
6722 (V_hi, V_lo): New mode attributes.
6723 (VF_constraint): Likewise.
6724 (vfml_half, vfml_half_selector): New int attributes.
6725 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6727 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6728 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6730 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6731 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6732 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6733 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6735 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6736 Document new effective target and option set.
6738 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6740 * config/arm/arm-cpus.in (armv8_4): New feature.
6741 (ARMv8_4a): New fgroup.
6742 (armv8.4-a): New arch.
6743 * config/arm/arm-tables.opt: Regenerate.
6744 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6745 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6746 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6747 Add matching rules for -march=armv8.4-a and extensions.
6748 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6750 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6753 * config/rx/rx.md (BW): New mode attribute.
6754 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6756 2018-01-11 Richard Biener <rguenther@suse.de>
6758 PR tree-optimization/83435
6759 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6760 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6761 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6763 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6764 Alan Hayward <alan.hayward@arm.com>
6765 David Sherwood <david.sherwood@arm.com>
6767 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6769 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6770 (aarch64_print_address_internal): Use it to check for a zero offset.
6772 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6773 Alan Hayward <alan.hayward@arm.com>
6774 David Sherwood <david.sherwood@arm.com>
6776 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6777 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6778 Return a poly_int64 rather than a HOST_WIDE_INT.
6779 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6780 rather than a HOST_WIDE_INT.
6781 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6782 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6783 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6784 final_offset from HOST_WIDE_INT to poly_int64.
6785 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6786 to_constant when getting the number of units in an Advanced SIMD
6788 (aarch64_builtin_vectorized_function): Check for a constant number
6790 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6792 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6793 attribute instead of GET_MODE_NUNITS.
6794 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6795 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6796 GET_MODE_SIZE for fixed-size registers.
6797 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6798 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6799 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6800 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6801 (aarch64_print_operand, aarch64_print_address_internal)
6802 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6803 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6804 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6805 Handle polynomial GET_MODE_SIZE.
6806 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6807 wider than SImode without modification.
6808 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6809 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6810 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6811 passing and returning SVE modes.
6812 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6813 rather than GEN_INT.
6814 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6815 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6816 (aarch64_allocate_and_probe_stack_space): Likewise.
6817 (aarch64_layout_frame): Cope with polynomial offsets.
6818 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6819 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6821 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6822 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6823 poly_int64 rather than a HOST_WIDE_INT.
6824 (aarch64_get_separate_components, aarch64_process_components)
6825 (aarch64_expand_prologue, aarch64_expand_epilogue)
6826 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6827 (aarch64_anchor_offset): New function, split out from...
6828 (aarch64_legitimize_address): ...here.
6829 (aarch64_builtin_vectorization_cost): Handle polynomial
6830 TYPE_VECTOR_SUBPARTS.
6831 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6833 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6834 number of elements from the PARALLEL rather than the mode.
6835 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6836 rather than GET_MODE_BITSIZE.
6837 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6838 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6839 (aarch64_expand_vec_perm_const_1): Handle polynomial
6840 d->perm.length () and d->perm elements.
6841 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6842 Apply to_constant to d->perm elements.
6843 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6844 polynomial CONST_VECTOR_NUNITS.
6845 (aarch64_move_pointer): Take amount as a poly_int64 rather
6847 (aarch64_progress_pointer): Avoid temporary variable.
6848 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6849 the mode attribute instead of GET_MODE.
6851 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6852 Alan Hayward <alan.hayward@arm.com>
6853 David Sherwood <david.sherwood@arm.com>
6855 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6856 x exists before using it.
6857 (aarch64_add_constant_internal): Rename to...
6858 (aarch64_add_offset_1): ...this. Replace regnum with separate
6859 src and dest rtxes. Handle the case in which they're different,
6860 including when the offset is zero. Replace scratchreg with an rtx.
6861 Use 2 additions if there is no spare register into which we can
6862 move a 16-bit constant.
6863 (aarch64_add_constant): Delete.
6864 (aarch64_add_offset): Replace reg with separate src and dest
6865 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6866 Use aarch64_add_offset_1.
6867 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6868 an rtx rather than an int. Take the delta as a poly_int64
6869 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6870 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6871 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6872 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6873 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6875 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6876 aarch64_add_constant.
6878 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6880 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6881 Use scalar_float_mode.
6883 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6885 * config/aarch64/aarch64-simd.md
6886 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6887 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6888 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6889 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6890 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6891 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6892 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6893 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6894 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6895 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6897 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6900 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6901 targ_options->x_arm_arch_string is non NULL.
6903 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6905 * config/aarch64/aarch64.h
6906 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6908 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6911 * expmed.c (emit_store_flag_force): Swap if const op0
6912 and change VOIDmode to mode of op0.
6914 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6916 PR rtl-optimization/83761
6917 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6918 than bytes to mode_for_size.
6920 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6923 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6924 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6927 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6930 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6931 when in layout mode.
6932 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6933 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6936 2018-01-10 Michael Collison <michael.collison@arm.com>
6938 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6939 * config/aarch64/aarch64-option-extension.def: Add
6940 AARCH64_OPT_EXTENSION of 'fp16fml'.
6941 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6942 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6943 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6944 * config/aarch64/constraints.md (Ui7): New constraint.
6945 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6946 (VFMLA_SEL_W): Ditto.
6949 (VFMLA16_LOW): New int iterator.
6950 (VFMLA16_HIGH): Ditto.
6951 (UNSPEC_FMLAL): New unspec.
6952 (UNSPEC_FMLSL): Ditto.
6953 (UNSPEC_FMLAL2): Ditto.
6954 (UNSPEC_FMLSL2): Ditto.
6955 (f16mac): New code attribute.
6956 * config/aarch64/aarch64-simd-builtins.def
6957 (aarch64_fmlal_lowv2sf): Ditto.
6958 (aarch64_fmlsl_lowv2sf): Ditto.
6959 (aarch64_fmlalq_lowv4sf): Ditto.
6960 (aarch64_fmlslq_lowv4sf): Ditto.
6961 (aarch64_fmlal_highv2sf): Ditto.
6962 (aarch64_fmlsl_highv2sf): Ditto.
6963 (aarch64_fmlalq_highv4sf): Ditto.
6964 (aarch64_fmlslq_highv4sf): Ditto.
6965 (aarch64_fmlal_lane_lowv2sf): Ditto.
6966 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6967 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6968 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6969 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6970 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6971 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6972 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6973 (aarch64_fmlal_lane_highv2sf): Ditto.
6974 (aarch64_fmlsl_lane_highv2sf): Ditto.
6975 (aarch64_fmlal_laneq_highv2sf): Ditto.
6976 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6977 (aarch64_fmlalq_lane_highv4sf): Ditto.
6978 (aarch64_fmlsl_lane_highv4sf): Ditto.
6979 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6980 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6981 * config/aarch64/aarch64-simd.md:
6982 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6983 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6984 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6985 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6986 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6987 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6988 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6989 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6990 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6991 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6992 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6993 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6994 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6995 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6996 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6997 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6998 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6999 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7000 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7001 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7002 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7003 (vfmlsl_low_u32): Ditto.
7004 (vfmlalq_low_u32): Ditto.
7005 (vfmlslq_low_u32): Ditto.
7006 (vfmlal_high_u32): Ditto.
7007 (vfmlsl_high_u32): Ditto.
7008 (vfmlalq_high_u32): Ditto.
7009 (vfmlslq_high_u32): Ditto.
7010 (vfmlal_lane_low_u32): Ditto.
7011 (vfmlsl_lane_low_u32): Ditto.
7012 (vfmlal_laneq_low_u32): Ditto.
7013 (vfmlsl_laneq_low_u32): Ditto.
7014 (vfmlalq_lane_low_u32): Ditto.
7015 (vfmlslq_lane_low_u32): Ditto.
7016 (vfmlalq_laneq_low_u32): Ditto.
7017 (vfmlslq_laneq_low_u32): Ditto.
7018 (vfmlal_lane_high_u32): Ditto.
7019 (vfmlsl_lane_high_u32): Ditto.
7020 (vfmlal_laneq_high_u32): Ditto.
7021 (vfmlsl_laneq_high_u32): Ditto.
7022 (vfmlalq_lane_high_u32): Ditto.
7023 (vfmlslq_lane_high_u32): Ditto.
7024 (vfmlalq_laneq_high_u32): Ditto.
7025 (vfmlslq_laneq_high_u32): Ditto.
7026 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7027 (AARCH64_FL_FOR_ARCH8_4): New.
7028 (AARCH64_ISA_F16FML): New ISA flag.
7029 (TARGET_F16FML): New feature flag for fp16fml.
7030 (doc/invoke.texi): Document new fp16fml option.
7032 2018-01-10 Michael Collison <michael.collison@arm.com>
7034 * config/aarch64/aarch64-builtins.c:
7035 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7036 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7037 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7038 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7039 (AARCH64_ISA_SHA3): New ISA flag.
7040 (TARGET_SHA3): New feature flag for sha3.
7041 * config/aarch64/iterators.md (sha512_op): New int attribute.
7042 (CRYPTO_SHA512): New int iterator.
7043 (UNSPEC_SHA512H): New unspec.
7044 (UNSPEC_SHA512H2): Ditto.
7045 (UNSPEC_SHA512SU0): Ditto.
7046 (UNSPEC_SHA512SU1): Ditto.
7047 * config/aarch64/aarch64-simd-builtins.def
7048 (aarch64_crypto_sha512hqv2di): New builtin.
7049 (aarch64_crypto_sha512h2qv2di): Ditto.
7050 (aarch64_crypto_sha512su0qv2di): Ditto.
7051 (aarch64_crypto_sha512su1qv2di): Ditto.
7052 (aarch64_eor3qv8hi): Ditto.
7053 (aarch64_rax1qv2di): Ditto.
7054 (aarch64_xarqv2di): Ditto.
7055 (aarch64_bcaxqv8hi): Ditto.
7056 * config/aarch64/aarch64-simd.md:
7057 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7058 (aarch64_crypto_sha512su0qv2di): Ditto.
7059 (aarch64_crypto_sha512su1qv2di): Ditto.
7060 (aarch64_eor3qv8hi): Ditto.
7061 (aarch64_rax1qv2di): Ditto.
7062 (aarch64_xarqv2di): Ditto.
7063 (aarch64_bcaxqv8hi): Ditto.
7064 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7065 (vsha512h2q_u64): Ditto.
7066 (vsha512su0q_u64): Ditto.
7067 (vsha512su1q_u64): Ditto.
7068 (veor3q_u16): Ditto.
7069 (vrax1q_u64): Ditto.
7071 (vbcaxq_u16): Ditto.
7072 * config/arm/types.md (crypto_sha512): New type attribute.
7073 (crypto_sha3): Ditto.
7074 (doc/invoke.texi): Document new sha3 option.
7076 2018-01-10 Michael Collison <michael.collison@arm.com>
7078 * config/aarch64/aarch64-builtins.c:
7079 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7080 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7081 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7082 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7083 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7084 (AARCH64_ISA_SM4): New ISA flag.
7085 (TARGET_SM4): New feature flag for sm4.
7086 * config/aarch64/aarch64-simd-builtins.def
7087 (aarch64_sm3ss1qv4si): Ditto.
7088 (aarch64_sm3tt1aq4si): Ditto.
7089 (aarch64_sm3tt1bq4si): Ditto.
7090 (aarch64_sm3tt2aq4si): Ditto.
7091 (aarch64_sm3tt2bq4si): Ditto.
7092 (aarch64_sm3partw1qv4si): Ditto.
7093 (aarch64_sm3partw2qv4si): Ditto.
7094 (aarch64_sm4eqv4si): Ditto.
7095 (aarch64_sm4ekeyqv4si): Ditto.
7096 * config/aarch64/aarch64-simd.md:
7097 (aarch64_sm3ss1qv4si): Ditto.
7098 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7099 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7100 (aarch64_sm4eqv4si): Ditto.
7101 (aarch64_sm4ekeyqv4si): Ditto.
7102 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7103 (sm3part_op): Ditto.
7104 (CRYPTO_SM3TT): Ditto.
7105 (CRYPTO_SM3PART): Ditto.
7106 (UNSPEC_SM3SS1): New unspec.
7107 (UNSPEC_SM3TT1A): Ditto.
7108 (UNSPEC_SM3TT1B): Ditto.
7109 (UNSPEC_SM3TT2A): Ditto.
7110 (UNSPEC_SM3TT2B): Ditto.
7111 (UNSPEC_SM3PARTW1): Ditto.
7112 (UNSPEC_SM3PARTW2): Ditto.
7113 (UNSPEC_SM4E): Ditto.
7114 (UNSPEC_SM4EKEY): Ditto.
7115 * config/aarch64/constraints.md (Ui2): New constraint.
7116 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7117 * config/arm/types.md (crypto_sm3): New type attribute.
7118 (crypto_sm4): Ditto.
7119 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7120 (vsm3tt1aq_u32): Ditto.
7121 (vsm3tt1bq_u32): Ditto.
7122 (vsm3tt2aq_u32): Ditto.
7123 (vsm3tt2bq_u32): Ditto.
7124 (vsm3partw1q_u32): Ditto.
7125 (vsm3partw2q_u32): Ditto.
7126 (vsm4eq_u32): Ditto.
7127 (vsm4ekeyq_u32): Ditto.
7128 (doc/invoke.texi): Document new sm4 option.
7130 2018-01-10 Michael Collison <michael.collison@arm.com>
7132 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7133 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7134 (AARCH64_FL_FOR_ARCH8_4): New.
7135 (AARCH64_FL_V8_4): New flag.
7136 (doc/invoke.texi): Document new armv8.4-a option.
7138 2018-01-10 Michael Collison <michael.collison@arm.com>
7140 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7141 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7142 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7143 * config/aarch64/aarch64-option-extension.def: Add
7144 AARCH64_OPT_EXTENSION of 'sha2'.
7145 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7146 (crypto): Disable sha2 and aes if crypto disabled.
7147 (crypto): Enable aes and sha2 if enabled.
7148 (simd): Disable sha2 and aes if simd disabled.
7149 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7151 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7152 (TARGET_SHA2): New feature flag for sha2.
7153 (TARGET_AES): New feature flag for aes.
7154 * config/aarch64/aarch64-simd.md:
7155 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7156 conditional on TARGET_AES.
7157 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7158 (aarch64_crypto_sha1hsi): Make pattern conditional
7160 (aarch64_crypto_sha1hv4si): Ditto.
7161 (aarch64_be_crypto_sha1hv4si): Ditto.
7162 (aarch64_crypto_sha1su1v4si): Ditto.
7163 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7164 (aarch64_crypto_sha1su0v4si): Ditto.
7165 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7166 (aarch64_crypto_sha256su0v4si): Ditto.
7167 (aarch64_crypto_sha256su1v4si): Ditto.
7168 (doc/invoke.texi): Document new aes and sha2 options.
7170 2018-01-10 Martin Sebor <msebor@redhat.com>
7172 PR tree-optimization/83781
7173 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7176 2018-01-11 Martin Sebor <msebor@gmail.com>
7177 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7179 PR tree-optimization/83501
7180 PR tree-optimization/81703
7182 * tree-ssa-strlen.c (get_string_cst): Rename...
7183 (get_string_len): ...to this. Handle global constants.
7184 (handle_char_store): Adjust.
7186 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7187 Jim Wilson <jimw@sifive.com>
7189 * config/riscv/riscv-protos.h (riscv_output_return): New.
7190 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7191 (riscv_attribute_table, riscv_output_return),
7192 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7193 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7194 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7196 (riscv_expand_prologue): Add early return for naked function.
7197 (riscv_expand_epilogue): Likewise.
7198 (riscv_function_ok_for_sibcall): Return false for naked function.
7199 (riscv_set_current_function): New.
7200 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7201 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7202 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7203 * doc/extend.texi (RISC-V Function Attributes): New.
7205 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7207 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7208 check for 128-bit long double before checking TCmode.
7209 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7210 128-bit long doubles before checking TFmode or TCmode.
7211 (FLOAT128_IBM_P): Likewise.
7213 2018-01-10 Martin Sebor <msebor@redhat.com>
7215 PR tree-optimization/83671
7216 * builtins.c (c_strlen): Unconditionally return zero for the empty
7218 Use -Warray-bounds for warnings.
7219 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7220 for non-constant array indices with COMPONENT_REF, arrays of
7221 arrays, and pointers to arrays.
7222 (gimple_fold_builtin_strlen): Determine and set length range for
7223 non-constant character arrays.
7225 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7228 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7231 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7233 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7235 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7238 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7239 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7240 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7241 indexed_or_indirect_operand predicate.
7242 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7243 (*vsx_le_perm_load_v8hi): Likewise.
7244 (*vsx_le_perm_load_v16qi): Likewise.
7245 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7246 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7247 (*vsx_le_perm_store_v8hi): Likewise.
7248 (*vsx_le_perm_store_v16qi): Likewise.
7249 (eight unnamed splitters): Likewise.
7251 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7253 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7254 * config/rs6000/emmintrin.h: Likewise.
7255 * config/rs6000/mmintrin.h: Likewise.
7256 * config/rs6000/xmmintrin.h: Likewise.
7258 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7261 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7263 * tree.c (tree_nop_conversion): Return true for location wrapper
7265 (maybe_wrap_with_location): New function.
7266 (selftest::check_strip_nops): New function.
7267 (selftest::test_location_wrappers): New function.
7268 (selftest::tree_c_tests): Call it.
7269 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7270 (maybe_wrap_with_location): New decl.
7271 (EXPR_LOCATION_WRAPPER_P): New macro.
7272 (location_wrapper_p): New inline function.
7273 (tree_strip_any_location_wrapper): New inline function.
7275 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7278 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7279 stack_realign_offset for the largest alignment of stack slot
7281 (ix86_find_max_used_stack_alignment): New function.
7282 (ix86_finalize_stack_frame_flags): Use it. Set
7283 max_used_stack_alignment if we don't realign stack.
7284 * config/i386/i386.h (machine_function): Add
7285 max_used_stack_alignment.
7287 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7289 * config/arm/arm.opt (-mbranch-cost): New option.
7290 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7293 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7296 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7297 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7299 2018-01-10 Richard Biener <rguenther@suse.de>
7302 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7303 early out so it also covers the case where we have a non-NULL
7306 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7308 PR tree-optimization/83753
7309 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7310 for non-strided grouped accesses if the number of elements is 1.
7312 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7315 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7316 * i386.h (TARGET_USE_GATHER): Define.
7317 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7319 2018-01-10 Martin Liska <mliska@suse.cz>
7322 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7323 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7325 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7326 CLEANUP_NO_PARTITIONING is not set.
7328 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7330 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7331 for vectors, as a partial revert of r254296.
7332 * rtl.h (const_vec_p): Delete.
7333 (const_vec_duplicate_p): Don't test for vector CONSTs.
7334 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7335 * expmed.c (make_tree): Likewise.
7338 * common.md (E, F): Use CONSTANT_P instead of checking for
7340 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7341 checking for CONST_VECTOR.
7343 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7346 * predict.c (force_edge_cold): Handle in more sane way edges
7349 2018-01-09 Carl Love <cel@us.ibm.com>
7351 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7353 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7354 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7355 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7356 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7357 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7358 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7359 * config/rs6000/rs6000-protos.h: Add extern defition for
7360 rs6000_generate_float2_double_code.
7361 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7363 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7364 (float2_v2df): Add define_expand.
7366 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7369 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7370 op_mode in the force_to_mode call.
7372 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7374 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7375 instead of checking each element individually.
7376 (aarch64_evpc_uzp): Likewise.
7377 (aarch64_evpc_zip): Likewise.
7378 (aarch64_evpc_ext): Likewise.
7379 (aarch64_evpc_rev): Likewise.
7380 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7381 instead of checking each element individually. Return true without
7383 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7384 whether all selected elements come from the same input, instead of
7385 checking each element individually. Remove calls to gen_rtx_REG,
7386 start_sequence and end_sequence and instead assert that no rtl is
7389 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7391 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7392 order of HIGH and CONST checks.
7394 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7396 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7397 if the destination isn't an SSA_NAME.
7399 2018-01-09 Richard Biener <rguenther@suse.de>
7401 PR tree-optimization/83668
7402 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7404 (canonicalize_loop_form): ... here, renamed from ...
7405 (canonicalize_loop_closed_ssa_form): ... this and amended to
7406 swap successor edges for loop exit blocks to make us use
7407 the RPO order we need for initial schedule generation.
7409 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7411 PR tree-optimization/64811
7412 * match.pd: When optimizing comparisons with Inf, avoid
7413 introducing or losing exceptions from comparisons with NaN.
7415 2018-01-09 Martin Liska <mliska@suse.cz>
7418 * asan.c (shadow_mem_size): Add gcc_assert.
7420 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7422 Don't save registers in main().
7425 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7426 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7427 * config/avr/avr.c (avr_set_current_function): Don't error if
7428 naked, OS_task or OS_main are specified at the same time.
7429 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7431 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7433 * common/config/avr/avr-common.c (avr_option_optimization_table):
7434 Switch on -mmain-is-OS_task for optimizing compilations.
7436 2018-01-09 Richard Biener <rguenther@suse.de>
7438 PR tree-optimization/83572
7439 * graphite.c: Include cfganal.h.
7440 (graphite_transform_loops): Connect infinite loops to exit
7441 and remove fake edges at the end.
7443 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7445 * ipa-inline.c (edge_badness): Revert accidental checkin.
7447 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7450 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7451 symbols; not inline clones.
7453 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7456 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7457 hard registers. Formatting fixes.
7459 PR preprocessor/83722
7460 * gcc.c (try_generate_repro): Pass
7461 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7462 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7465 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7466 Kito Cheng <kito.cheng@gmail.com>
7468 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7469 (riscv_leaf_function_p): Delete.
7470 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7472 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7474 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7476 (do_ifelse): New function.
7477 (do_isel): New function.
7478 (do_sub3): New function.
7479 (do_add3): New function.
7480 (do_load_mask_compare): New function.
7481 (do_overlap_load_compare): New function.
7482 (expand_compare_loop): New function.
7483 (expand_block_compare): Call expand_compare_loop() when appropriate.
7484 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7486 (-mblock-compare-inline-loop-limit): New option.
7488 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7491 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7492 Reverse order of second and third operands in first alternative.
7493 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7494 of first and second elements in UNSPEC_VPERMR vector.
7495 (altivec_expand_vec_perm_le): Likewise.
7497 2018-01-08 Jeff Law <law@redhat.com>
7499 PR rtl-optimizatin/81308
7500 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7501 (process_switch): If group_case_labels makes a change, then set
7503 (pass_convert_switch::execute): If a switch is converted, then
7504 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7506 PR rtl-optimization/81308
7507 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7510 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7512 PR target/83663 - Revert r255946
7513 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7514 generation for cases where splatting a value is not useful.
7515 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7516 across a vec_duplicate and a paradoxical subreg forming a vector
7517 mode to a vec_concat.
7519 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7521 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7522 -march=armv8.3-a variants.
7523 * config/arm/t-multilib: Likewise.
7524 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7526 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7528 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7530 (cceq_ior_compare_complement): Give it a name so I can use it, and
7531 change boolean_or_operator predicate to boolean_operator so it can
7532 be used to generate a crand.
7533 (eqne): New code iterator.
7534 (bd/bd_neg): New code_attrs.
7535 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7536 a single define_insn.
7537 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7538 decrement (bdnzt/bdnzf/bdzt/bdzf).
7539 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7540 with the new names of the branch decrement patterns, and added the
7541 names of the branch decrement conditional patterns.
7543 2018-01-08 Richard Biener <rguenther@suse.de>
7545 PR tree-optimization/83563
7546 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7549 2018-01-08 Richard Biener <rguenther@suse.de>
7552 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7554 2018-01-08 Richard Biener <rguenther@suse.de>
7556 PR tree-optimization/83685
7557 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7558 references to abnormals.
7560 2018-01-08 Richard Biener <rguenther@suse.de>
7563 * dwarf2out.c (output_indirect_strings): Handle empty
7564 skeleton_debug_str_hash.
7565 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7567 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7569 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7570 (emit_store_direct): Likewise.
7571 (arc_trampoline_adjust_address): Likewise.
7572 (arc_asm_trampoline_template): New function.
7573 (arc_initialize_trampoline): Use asm_trampoline_template.
7574 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7575 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7576 * config/arc/arc.md (flush_icache): Delete pattern.
7578 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7580 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7581 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7584 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7587 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7588 by not USED_FOR_TARGET.
7589 (make_pass_resolve_sw_modes): Likewise.
7591 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7593 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7596 2018-01-08 Richard Biener <rguenther@suse.de>
7599 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7601 2018-01-08 Richard Biener <rguenther@suse.de>
7604 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7606 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7609 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7610 basic blocks with a small number of successors.
7611 (convert_control_dep_chain_into_preds): Improve handling of
7613 (dump_predicates): Split apart into...
7614 (dump_pred_chain): ...here...
7615 (dump_pred_info): ...and here.
7616 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7617 (can_chain_union_be_invalidated_p): Improve check for invalidation
7619 (uninit_uses_cannot_happen): Avoid unnecessary if
7620 convert_control_dep_chain_into_preds yielded nothing.
7622 2018-01-06 Martin Sebor <msebor@redhat.com>
7624 PR tree-optimization/83640
7625 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7626 subtracting negative offset from size.
7627 (builtin_access::overlap): Adjust offset bounds of the access to fall
7628 within the size of the object if possible.
7630 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7632 PR rtl-optimization/83699
7633 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7634 extract_bit_field_as_subreg to cases in which the extracted
7635 value is also a vector.
7637 * lra-constraints.c (process_alt_operands): Test for the equivalence
7638 substitutions when detecting a possible reload cycle.
7640 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7643 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7644 by default if flag_selective_schedling{,2}. Formatting fixes.
7646 PR rtl-optimization/83682
7647 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7648 if it has non-VECTOR_MODE element mode.
7649 (vec_duplicate_p): Likewise.
7652 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7653 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7655 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7658 * config/i386/i386-builtin.def
7659 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7660 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7661 Require also OPTION_MASK_ISA_AVX512F in addition to
7662 OPTION_MASK_ISA_GFNI.
7663 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7664 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7665 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7666 to OPTION_MASK_ISA_GFNI.
7667 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7668 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7669 OPTION_MASK_ISA_AVX512BW.
7670 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7671 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7672 addition to OPTION_MASK_ISA_GFNI.
7673 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7674 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7675 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7676 to OPTION_MASK_ISA_GFNI.
7677 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7678 a requirement for all ISAs rather than any of them with a few
7680 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7682 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7683 bitmasks to be enabled with 3 exceptions, instead of requiring any
7684 enabled ISA with lots of exceptions.
7685 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7686 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7687 Change avx512bw in isa attribute to avx512f.
7688 * config/i386/sgxintrin.h: Add license boilerplate.
7689 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7690 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7691 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7692 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7694 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7695 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7696 temporarily sse2 rather than sse if not enabled already.
7699 * config/i386/sse.md (VI248_VLBW): Rename to ...
7700 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7701 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7702 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7703 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7704 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7705 mode iterator instead of VI248_VLBW.
7707 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7709 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7710 (record_modified): Skip clobbers; add debug output.
7711 (param_change_prob): Use sreal frequencies.
7713 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7715 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7716 punt for user-aligned variables.
7718 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7720 * tree-chrec.c (chrec_contains_symbols): Return true for
7723 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7726 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7727 of (x|y) == x for BICS pattern.
7729 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7731 PR tree-optimization/83605
7732 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7733 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7736 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7738 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7739 * config/epiphany/rtems.h: New file.
7741 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7742 Uros Bizjak <ubizjak@gmail.com>
7745 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7746 QIreg_operand instead of register_operand predicate.
7747 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7748 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7749 comments instead of -fmitigate[-_]rop.
7751 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7754 * cgraphunit.c (symbol_table::compile): Switch to text_section
7755 before calling assembly_start debug hook.
7756 * run-rtl-passes.c (run_rtl_passes): Likewise.
7759 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7761 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7762 range_int_cst_p rather than !symbolic_range_p before calling
7763 extract_range_from_multiplicative_op_1.
7765 2018-01-04 Jeff Law <law@redhat.com>
7767 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7768 redundant test in assertion.
7770 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7772 * doc/rtl.texi: Document machine_mode wrapper classes.
7774 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7776 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7779 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7781 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7782 the VEC_PERM_EXPR fold to fail.
7784 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7787 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7788 to switched_sections.
7790 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7793 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7796 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7799 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7800 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7802 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7805 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7806 is BLKmode and bitpos not zero or mode change is needed.
7808 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7811 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7814 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7817 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7818 instead of MULT rtx. Update all corresponding splitters.
7820 (*ssub<modesuffix>): Ditto.
7822 (*cmp_sadd_di): Update split patterns.
7823 (*cmp_sadd_si): Ditto.
7824 (*cmp_sadd_sidi): Ditto.
7825 (*cmp_ssub_di): Ditto.
7826 (*cmp_ssub_si): Ditto.
7827 (*cmp_ssub_sidi): Ditto.
7828 * config/alpha/predicates.md (const23_operand): New predicate.
7829 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7830 Look for ASHIFT, not MULT inner operand.
7831 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7833 2018-01-04 Martin Liska <mliska@suse.cz>
7835 PR gcov-profile/83669
7836 * gcov.c (output_intermediate_file): Add version to intermediate
7838 * doc/gcov.texi: Document new field 'version' in intermediate
7839 file format. Fix location of '-k' option of gcov command.
7841 2018-01-04 Martin Liska <mliska@suse.cz>
7844 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7846 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7848 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7850 2018-01-03 Martin Sebor <msebor@redhat.com>
7852 PR tree-optimization/83655
7853 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7854 checking calls with invalid arguments.
7856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7858 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7859 (vectorizable_mask_load_store): Delete.
7860 (vectorizable_call): Return false for masked loads and stores.
7861 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7862 instead of gimple_assign_rhs1.
7863 (vectorizable_load): Handle IFN_MASK_LOAD.
7864 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7868 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7870 (vectorizable_mask_load_store): ...here.
7871 (vectorizable_load): ...and here.
7873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7875 * tree-vect-stmts.c (vect_build_all_ones_mask)
7876 (vect_build_zero_merge_argument): New functions, split out from...
7877 (vectorizable_load): ...here.
7879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7881 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7883 (vectorizable_mask_load_store): ...here.
7884 (vectorizable_store): ...and here.
7886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7888 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7890 (vectorizable_mask_load_store): ...here.
7892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7894 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7895 (vect_model_store_cost): Take a vec_load_store_type instead of a
7897 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7898 (vect_model_store_cost): Take a vec_load_store_type instead of a
7900 (vectorizable_mask_load_store): Update accordingly.
7901 (vectorizable_store): Likewise.
7902 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7906 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7907 IFN_MASK_LOAD calls here rather than...
7908 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7911 Alan Hayward <alan.hayward@arm.com>
7912 David Sherwood <david.sherwood@arm.com>
7914 * expmed.c (extract_bit_field_1): For vector extracts,
7915 fall back to extract_bit_field_as_subreg if vec_extract
7918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7919 Alan Hayward <alan.hayward@arm.com>
7920 David Sherwood <david.sherwood@arm.com>
7922 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7923 they are variable or constant sized.
7924 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7925 slots for constant-sized data.
7927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7928 Alan Hayward <alan.hayward@arm.com>
7929 David Sherwood <david.sherwood@arm.com>
7931 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7932 handling COND_EXPRs with boolean comparisons, try to find a better
7933 basis for the mask type than the boolean itself.
7935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7937 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7938 is calculated and how it can be overridden.
7939 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7940 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7942 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7946 Alan Hayward <alan.hayward@arm.com>
7947 David Sherwood <david.sherwood@arm.com>
7949 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7950 Remove the mode argument.
7951 (aarch64_simd_valid_immediate): Remove the mode and inverse
7953 * config/aarch64/iterators.md (bitsize): New iterator.
7954 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7955 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7956 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7957 aarch64_simd_valid_immediate.
7958 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7959 (aarch64_reg_or_bic_imm): Likewise.
7960 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7961 with an insn_type enum and msl with a modifier_type enum.
7962 Replace element_width with a scalar_mode. Change the shift
7963 to unsigned int. Add constructors for scalar_float_mode and
7964 scalar_int_mode elements.
7965 (aarch64_vect_float_const_representable_p): Delete.
7966 (aarch64_can_const_movi_rtx_p)
7967 (aarch64_simd_scalar_immediate_valid_for_move)
7968 (aarch64_simd_make_constant): Update call to
7969 aarch64_simd_valid_immediate.
7970 (aarch64_advsimd_valid_immediate_hs): New function.
7971 (aarch64_advsimd_valid_immediate): Likewise.
7972 (aarch64_simd_valid_immediate): Remove mode and inverse
7973 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7974 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7975 and aarch64_float_const_representable_p on the result.
7976 (aarch64_output_simd_mov_immediate): Remove mode argument.
7977 Update call to aarch64_simd_valid_immediate and use of
7978 simd_immediate_info.
7979 (aarch64_output_scalar_simd_mov_immediate): Update call
7982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7983 Alan Hayward <alan.hayward@arm.com>
7984 David Sherwood <david.sherwood@arm.com>
7986 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7987 (mode_nunits): Likewise CONST_MODE_NUNITS.
7988 * machmode.def (ADJUST_NUNITS): Document.
7989 * genmodes.c (mode_data::need_nunits_adj): New field.
7990 (blank_mode): Update accordingly.
7991 (adj_nunits): New variable.
7992 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7994 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7995 listed in adj_nunits.
7996 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7997 listed in adj_nunits. Don't emit case statements for such modes.
7998 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7999 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8000 nothing if adj_nunits is nonnull.
8001 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8002 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8003 (emit_mode_fbit): Update use of print_maybe_const_decl.
8004 (emit_move_size): Likewise. Treat the array as non-const
8006 (emit_mode_adjustments): Handle adj_nunits.
8008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8010 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8011 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8012 (VECTOR_MODES): Use it.
8013 (make_vector_modes): Take the prefix as an argument.
8015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8016 Alan Hayward <alan.hayward@arm.com>
8017 David Sherwood <david.sherwood@arm.com>
8019 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8020 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8021 for MODE_VECTOR_BOOL.
8022 * machmode.def (VECTOR_BOOL_MODE): Document.
8023 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8024 (make_vector_bool_mode): New function.
8025 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8027 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8028 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8030 * stor-layout.c (int_mode_for_mode): Likewise.
8031 * tree.c (build_vector_type_for_mode): Likewise.
8032 * varasm.c (output_constant_pool_2): Likewise.
8033 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8034 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8035 for MODE_VECTOR_BOOL.
8036 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8037 of mode class checks.
8038 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8039 instead of a list of mode class checks.
8040 (expand_vector_scalar_condition): Likewise.
8041 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8044 Alan Hayward <alan.hayward@arm.com>
8045 David Sherwood <david.sherwood@arm.com>
8047 * machmode.h (mode_size): Change from unsigned short to
8049 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8050 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8051 or if measurement_type is not polynomial.
8052 (fixed_size_mode::includes_p): Check for constant-sized modes.
8053 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8054 return a poly_uint16 rather than an unsigned short.
8055 (emit_mode_size): Change the type of mode_size from unsigned short
8056 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8057 (emit_mode_adjustments): Cope with polynomial vector sizes.
8058 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8060 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8062 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8063 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8064 * caller-save.c (setup_save_areas): Likewise.
8065 (replace_reg_with_saved_mem): Likewise.
8066 * calls.c (emit_library_call_value_1): Likewise.
8067 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8068 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8069 (gen_lowpart_for_combine): Likewise.
8070 * convert.c (convert_to_integer_1): Likewise.
8071 * cse.c (equiv_constant, cse_insn): Likewise.
8072 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8073 (cselib_subst_to_values): Likewise.
8074 * dce.c (word_dce_process_block): Likewise.
8075 * df-problems.c (df_word_lr_mark_ref): Likewise.
8076 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8077 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8078 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8079 (rtl_for_decl_location): Likewise.
8080 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8081 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8082 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8083 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8084 (expand_expr_real_1): Likewise.
8085 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8086 (pad_below): Likewise.
8087 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8088 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8089 * ira.c (get_subreg_tracking_sizes): Likewise.
8090 * ira-build.c (ira_create_allocno_objects): Likewise.
8091 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8092 (ira_sort_regnos_for_alter_reg): Likewise.
8093 * ira-costs.c (record_operand_costs): Likewise.
8094 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8095 (resolve_simple_move): Likewise.
8096 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8097 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8098 (lra_constraints): Likewise.
8099 (CONST_POOL_OK_P): Reject variable-sized modes.
8100 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8101 (add_pseudo_to_slot, lra_spill): Likewise.
8102 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8103 * optabs-query.c (get_best_extraction_insn): Likewise.
8104 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8105 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8106 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8107 * recog.c (offsettable_address_addr_space_p): Likewise.
8108 * regcprop.c (maybe_mode_change): Likewise.
8109 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8110 * regrename.c (build_def_use): Likewise.
8111 * regstat.c (dump_reg_info): Likewise.
8112 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8113 (find_reloads, find_reloads_subreg_address): Likewise.
8114 * reload1.c (eliminate_regs_1): Likewise.
8115 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8116 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8117 (simplify_binary_operation_1, simplify_subreg): Likewise.
8118 * targhooks.c (default_function_arg_padding): Likewise.
8119 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8120 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8121 (verify_gimple_assign_ternary): Likewise.
8122 * tree-inline.c (estimate_move_cost): Likewise.
8123 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8124 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8125 (get_address_cost_ainc): Likewise.
8126 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8127 (vect_supportable_dr_alignment): Likewise.
8128 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8129 (vectorizable_reduction): Likewise.
8130 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8131 (vectorizable_operation, vectorizable_load): Likewise.
8132 * tree.c (build_same_sized_truth_vector_type): Likewise.
8133 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8134 * var-tracking.c (emit_note_insn_var_location): Likewise.
8135 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8136 (ADDR_VEC_ALIGN): Likewise.
8138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8139 Alan Hayward <alan.hayward@arm.com>
8140 David Sherwood <david.sherwood@arm.com>
8142 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8144 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8145 or if measurement_type is polynomial.
8146 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8147 * combine.c (make_extraction): Likewise.
8148 * dse.c (find_shift_sequence): Likewise.
8149 * dwarf2out.c (mem_loc_descriptor): Likewise.
8150 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8151 (extract_bit_field, extract_low_bits): Likewise.
8152 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8153 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8154 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8155 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8156 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8157 * reload.c (find_reloads): Likewise.
8158 * reload1.c (alter_reg): Likewise.
8159 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8160 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8161 * tree-if-conv.c (predicate_mem_writes): Likewise.
8162 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8163 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8164 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8165 * valtrack.c (dead_debug_insert_temp): Likewise.
8166 * varasm.c (mergeable_constant_section): Likewise.
8167 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8170 Alan Hayward <alan.hayward@arm.com>
8171 David Sherwood <david.sherwood@arm.com>
8173 * expr.c (expand_assignment): Cope with polynomial mode sizes
8174 when assigning to a CONCAT.
8176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8177 Alan Hayward <alan.hayward@arm.com>
8178 David Sherwood <david.sherwood@arm.com>
8180 * machmode.h (mode_precision): Change from unsigned short to
8182 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8184 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8185 or if measurement_type is not polynomial.
8186 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8187 in which the mode is already known to be a scalar_int_mode.
8188 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8189 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8191 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8192 for GET_MODE_PRECISION.
8193 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8194 for GET_MODE_PRECISION.
8195 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8197 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8198 (expand_field_assignment, make_extraction): Likewise.
8199 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8200 (get_last_value): Likewise.
8201 * convert.c (convert_to_integer_1): Likewise.
8202 * cse.c (cse_insn): Likewise.
8203 * expr.c (expand_expr_real_1): Likewise.
8204 * lra-constraints.c (simplify_operand_subreg): Likewise.
8205 * optabs-query.c (can_atomic_load_p): Likewise.
8206 * optabs.c (expand_atomic_load): Likewise.
8207 (expand_atomic_store): Likewise.
8208 * ree.c (combine_reaching_defs): Likewise.
8209 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8210 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8211 * tree.h (type_has_mode_precision_p): Likewise.
8212 * ubsan.c (instrument_si_overflow): Likewise.
8214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8215 Alan Hayward <alan.hayward@arm.com>
8216 David Sherwood <david.sherwood@arm.com>
8218 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8219 polynomial numbers of units.
8220 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8221 (valid_vector_subparts_p): New function.
8222 (build_vector_type): Remove temporary shim and take the number
8223 of units as a poly_uint64 rather than an int.
8224 (build_opaque_vector_type): Take the number of units as a
8225 poly_uint64 rather than an int.
8226 * tree.c (build_vector_from_ctor): Handle polynomial
8227 TYPE_VECTOR_SUBPARTS.
8228 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8229 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8230 (build_vector_from_val): If the number of units is variable,
8231 use build_vec_duplicate_cst for constant operands and
8232 VEC_DUPLICATE_EXPR otherwise.
8233 (make_vector_type): Remove temporary is_constant ().
8234 (build_vector_type, build_opaque_vector_type): Take the number of
8235 units as a poly_uint64 rather than an int.
8236 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8238 * cfgexpand.c (expand_debug_expr): Likewise.
8239 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8240 (store_constructor, expand_expr_real_1): Likewise.
8241 (const_scalar_mask_from_tree): Likewise.
8242 * fold-const-call.c (fold_const_reduction): Likewise.
8243 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8244 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8245 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8246 (fold_relational_const): Likewise.
8247 (native_interpret_vector): Likewise. Change the size from an
8248 int to an unsigned int.
8249 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8250 TYPE_VECTOR_SUBPARTS.
8251 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8252 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8253 duplicating a non-constant operand into a variable-length vector.
8254 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8255 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8256 * ipa-icf.c (sem_variable::equals): Likewise.
8257 * match.pd: Likewise.
8258 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8259 * print-tree.c (print_node): Likewise.
8260 * stor-layout.c (layout_type): Likewise.
8261 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8262 * tree-cfg.c (verify_gimple_comparison): Likewise.
8263 (verify_gimple_assign_binary): Likewise.
8264 (verify_gimple_assign_ternary): Likewise.
8265 (verify_gimple_assign_single): Likewise.
8266 * tree-pretty-print.c (dump_generic_node): Likewise.
8267 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8268 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8269 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8270 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8271 (vect_shift_permute_load_chain): Likewise.
8272 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8273 (expand_vector_condition, optimize_vector_constructor): Likewise.
8274 (lower_vec_perm, get_compute_type): Likewise.
8275 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8276 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8277 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8278 (vect_recog_mask_conversion_pattern): Likewise.
8279 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8280 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8281 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8282 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8283 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8284 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8285 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8286 (supportable_widening_operation): Likewise.
8287 (supportable_narrowing_operation): Likewise.
8288 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8290 * varasm.c (output_constant): Likewise.
8292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8293 Alan Hayward <alan.hayward@arm.com>
8294 David Sherwood <david.sherwood@arm.com>
8296 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8297 so that both the length == 3 and length != 3 cases set up their
8298 own permute vectors. Add comments explaining why we know the
8299 number of elements is constant.
8300 (vect_permute_load_chain): Likewise.
8302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8303 Alan Hayward <alan.hayward@arm.com>
8304 David Sherwood <david.sherwood@arm.com>
8306 * machmode.h (mode_nunits): Change from unsigned char to
8308 (ONLY_FIXED_SIZE_MODES): New macro.
8309 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8310 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8311 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8313 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8314 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8315 or if measurement_type is not polynomial.
8316 * genmodes.c (ZERO_COEFFS): New macro.
8317 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8319 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8320 Use ZERO_COEFFS when emitting initializers.
8321 * data-streamer.h (bp_pack_poly_value): New function.
8322 (bp_unpack_poly_value): Likewise.
8323 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8324 for GET_MODE_NUNITS.
8325 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8326 for GET_MODE_NUNITS.
8327 * tree.c (make_vector_type): Remove temporary shim and make
8328 the real function take the number of units as a poly_uint64
8330 (build_vector_type_for_mode): Handle polynomial nunits.
8331 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8332 * emit-rtl.c (const_vec_series_p_1): Likewise.
8333 (gen_rtx_CONST_VECTOR): Likewise.
8334 * fold-const.c (test_vec_duplicate_folding): Likewise.
8335 * genrecog.c (validate_pattern): Likewise.
8336 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8337 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8338 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8339 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8340 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8341 * rtlanal.c (subreg_get_info): Likewise.
8342 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8343 (vect_grouped_load_supported): Likewise.
8344 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8345 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8346 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8347 (simplify_const_unary_operation, simplify_binary_operation_1)
8348 (simplify_const_binary_operation, simplify_ternary_operation)
8349 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8350 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8351 instead of CONST_VECTOR_NUNITS.
8352 * varasm.c (output_constant_pool_2): Likewise.
8353 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8354 explicit-encoded elements in the XVEC for variable-length vectors.
8356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8358 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8361 Alan Hayward <alan.hayward@arm.com>
8362 David Sherwood <david.sherwood@arm.com>
8364 * coretypes.h (fixed_size_mode): Declare.
8365 (fixed_size_mode_pod): New typedef.
8366 * builtins.h (target_builtins::x_apply_args_mode)
8367 (target_builtins::x_apply_result_mode): Change type to
8368 fixed_size_mode_pod.
8369 * builtins.c (apply_args_size, apply_result_size, result_vector)
8370 (expand_builtin_apply_args_1, expand_builtin_apply)
8371 (expand_builtin_return): Update accordingly.
8373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8375 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8376 * cselib.c (cselib_hash_rtx): Likewise.
8377 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8378 CONST_VECTOR encoding.
8380 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8381 Jeff Law <law@redhat.com>
8384 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8385 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8386 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8387 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8390 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8391 explicitly probe *sp in a noreturn function if there were any callee
8392 register saves or frame pointer is needed.
8394 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8397 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8398 BLKmode for ternary, binary or unary expressions.
8401 * var-tracking.c (delete_vta_debug_insn): New inline function.
8402 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8403 insns from get_insns () to NULL instead of each bb separately.
8404 Use delete_vta_debug_insn. No longer static.
8405 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8406 delete_vta_debug_insns callers.
8407 * rtl.h (delete_vta_debug_insns): Declare.
8408 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8409 instead of variable_tracking_main.
8411 2018-01-03 Martin Sebor <msebor@redhat.com>
8413 PR tree-optimization/83603
8414 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8415 arguments past the endof the argument list in functions declared
8416 without a prototype.
8417 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8418 Avoid checking when arguments are null.
8420 2018-01-03 Martin Sebor <msebor@redhat.com>
8423 * doc/extend.texi (attribute const): Fix a typo.
8424 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8425 issuing -Wsuggest-attribute for void functions.
8427 2018-01-03 Martin Sebor <msebor@redhat.com>
8429 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8430 offset_int::from instead of wide_int::to_shwi.
8431 (maybe_diag_overlap): Remove assertion.
8432 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8433 * gimple-ssa-sprintf.c (format_directive): Same.
8434 (parse_directive): Same.
8435 (sprintf_dom_walker::compute_format_length): Same.
8436 (try_substitute_return_value): Same.
8438 2018-01-03 Jeff Law <law@redhat.com>
8441 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8442 non-constant residual for zero at runtime and avoid probing in
8443 that case. Reorganize code for trailing problem to mirror handling
8446 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8448 PR tree-optimization/83501
8449 * tree-ssa-strlen.c (get_string_cst): New.
8450 (handle_char_store): Call get_string_cst.
8452 2018-01-03 Martin Liska <mliska@suse.cz>
8454 PR tree-optimization/83593
8455 * tree-ssa-strlen.c: Include tree-cfg.h.
8456 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8457 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8458 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8460 (strlen_dom_walker::before_dom_children): Call
8461 gimple_purge_dead_eh_edges. Dump tranformation with details
8463 (strlen_dom_walker::before_dom_children): Update call by adding
8464 new argument cleanup_eh.
8465 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8467 2018-01-03 Martin Liska <mliska@suse.cz>
8470 * cif-code.def (VARIADIC_THUNK): New enum value.
8471 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8474 2018-01-03 Jan Beulich <jbeulich@suse.com>
8476 * sse.md (mov<mode>_internal): Tighten condition for when to use
8477 vmovdqu<ssescalarsize> for TI and OI modes.
8479 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8481 Update copyright years.
8483 2018-01-03 Martin Liska <mliska@suse.cz>
8486 * ipa-visibility.c (function_and_variable_visibility): Skip
8487 functions with noipa attribure.
8489 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8491 * gcc.c (process_command): Update copyright notice dates.
8492 * gcov-dump.c (print_version): Ditto.
8493 * gcov.c (print_version): Ditto.
8494 * gcov-tool.c (print_version): Ditto.
8495 * gengtype.c (create_file): Ditto.
8496 * doc/cpp.texi: Bump @copying's copyright year.
8497 * doc/cppinternals.texi: Ditto.
8498 * doc/gcc.texi: Ditto.
8499 * doc/gccint.texi: Ditto.
8500 * doc/gcov.texi: Ditto.
8501 * doc/install.texi: Ditto.
8502 * doc/invoke.texi: Ditto.
8504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8506 * vector-builder.h (vector_builder::m_full_nelts): Change from
8507 unsigned int to poly_uint64.
8508 (vector_builder::full_nelts): Update prototype accordingly.
8509 (vector_builder::new_vector): Likewise.
8510 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8511 (vector_builder::operator ==): Likewise.
8512 (vector_builder::finalize): Likewise.
8513 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8514 Take the number of elements as a poly_uint64 rather than an
8516 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8517 from unsigned int to poly_uint64.
8518 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8519 (vec_perm_indices::new_vector): Likewise.
8520 (vec_perm_indices::length): Likewise.
8521 (vec_perm_indices::nelts_per_input): Likewise.
8522 (vec_perm_indices::input_nelts): Likewise.
8523 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8524 number of elements per input as a poly_uint64 rather than an
8525 unsigned int. Use the original encoding for variable-length
8526 vectors, rather than clamping each individual element.
8527 For the second and subsequent elements in each pattern,
8528 clamp the step and base before clamping their sum.
8529 (vec_perm_indices::series_p): Handle polynomial element counts.
8530 (vec_perm_indices::all_in_range_p): Likewise.
8531 (vec_perm_indices_to_tree): Likewise.
8532 (vec_perm_indices_to_rtx): Likewise.
8533 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8534 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8535 (tree_vector_builder::new_binary_operation): Handle polynomial
8536 element counts. Return false if we need to know the number
8537 of elements at compile time.
8538 * fold-const.c (fold_vec_perm): Punt if the number of elements
8539 isn't known at compile time.
8541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8543 * vec-perm-indices.h (vec_perm_builder): Change element type
8544 from HOST_WIDE_INT to poly_int64.
8545 (vec_perm_indices::element_type): Update accordingly.
8546 (vec_perm_indices::clamp): Handle polynomial element_types.
8547 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8548 (vec_perm_indices::all_in_range_p): Likewise.
8549 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8551 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8552 polynomial vec_perm_indices element types.
8553 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8554 * fold-const.c (fold_vec_perm): Likewise.
8555 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8556 * tree-vect-generic.c (lower_vec_perm): Likewise.
8557 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8558 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8559 element type to HOST_WIDE_INT.
8561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8562 Alan Hayward <alan.hayward@arm.com>
8563 David Sherwood <david.sherwood@arm.com>
8565 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8566 rather than an int. Use plus_constant.
8567 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8568 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8571 Alan Hayward <alan.hayward@arm.com>
8572 David Sherwood <david.sherwood@arm.com>
8574 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8575 a HOST_WIDE_INT to a poly_int64.
8577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8578 Alan Hayward <alan.hayward@arm.com>
8579 David Sherwood <david.sherwood@arm.com>
8581 * calls.c (load_register_parameters): Cope with polynomial
8582 mode sizes. Require a constant size for BLKmode parameters
8583 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8584 forces a parameter to be padded at the lsb end in order to
8585 fill a complete number of words, require the parameter size
8586 to be ordered wrt UNITS_PER_WORD.
8588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8589 Alan Hayward <alan.hayward@arm.com>
8590 David Sherwood <david.sherwood@arm.com>
8592 * reload1.c (spill_stack_slot_width): Change element type
8593 from unsigned int to poly_uint64_pod.
8594 (alter_reg): Treat mode sizes as polynomial.
8596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8597 Alan Hayward <alan.hayward@arm.com>
8598 David Sherwood <david.sherwood@arm.com>
8600 * reload.c (complex_word_subreg_p): New function.
8601 (reload_inner_reg_of_subreg, push_reload): Use it.
8603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8604 Alan Hayward <alan.hayward@arm.com>
8605 David Sherwood <david.sherwood@arm.com>
8607 * lra-constraints.c (process_alt_operands): Reject matched
8608 operands whose sizes aren't ordered.
8609 (match_reload): Refer to this check here.
8611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8612 Alan Hayward <alan.hayward@arm.com>
8613 David Sherwood <david.sherwood@arm.com>
8615 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8616 that the mode size is in the set {1, 2, 4, 8, 16}.
8618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8619 Alan Hayward <alan.hayward@arm.com>
8620 David Sherwood <david.sherwood@arm.com>
8622 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8623 Use plus_constant instead of gen_rtx_PLUS.
8625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8626 Alan Hayward <alan.hayward@arm.com>
8627 David Sherwood <david.sherwood@arm.com>
8629 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8630 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8631 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8632 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8633 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8634 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8635 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8636 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8637 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8638 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8640 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8641 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8642 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8643 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8644 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8645 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8646 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8647 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8648 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8649 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8651 * expr.c (emit_move_resolve_push): Treat the input and result
8652 of PUSH_ROUNDING as a poly_int64.
8653 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8654 (emit_push_insn): Likewise.
8655 * lra-eliminations.c (mark_not_eliminable): Likewise.
8656 * recog.c (push_operand): Likewise.
8657 * reload1.c (elimination_effects): Likewise.
8658 * rtlanal.c (nonzero_bits1): Likewise.
8659 * calls.c (store_one_arg): Likewise. Require the padding to be
8660 known at compile time.
8662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8663 Alan Hayward <alan.hayward@arm.com>
8664 David Sherwood <david.sherwood@arm.com>
8666 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8667 Use plus_constant instead of gen_rtx_PLUS.
8669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8670 Alan Hayward <alan.hayward@arm.com>
8671 David Sherwood <david.sherwood@arm.com>
8673 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8677 Alan Hayward <alan.hayward@arm.com>
8678 David Sherwood <david.sherwood@arm.com>
8680 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8681 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8682 via stack temporaries. Treat the mode size as polynomial too.
8684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8685 Alan Hayward <alan.hayward@arm.com>
8686 David Sherwood <david.sherwood@arm.com>
8688 * expr.c (expand_expr_real_2): When handling conversions involving
8689 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8690 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8691 as a poly_uint64 too.
8693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8694 Alan Hayward <alan.hayward@arm.com>
8695 David Sherwood <david.sherwood@arm.com>
8697 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8700 Alan Hayward <alan.hayward@arm.com>
8701 David Sherwood <david.sherwood@arm.com>
8703 * combine.c (can_change_dest_mode): Handle polynomial
8704 REGMODE_NATURAL_SIZE.
8705 * expmed.c (store_bit_field_1): Likewise.
8706 * expr.c (store_constructor): Likewise.
8707 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8708 and polynomial REGMODE_NATURAL_SIZE.
8709 (gen_lowpart_common): Likewise.
8710 * reginfo.c (record_subregs_of_mode): Likewise.
8711 * rtlanal.c (read_modify_subreg_p): Likewise.
8713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8714 Alan Hayward <alan.hayward@arm.com>
8715 David Sherwood <david.sherwood@arm.com>
8717 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8718 numbers of elements.
8720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8721 Alan Hayward <alan.hayward@arm.com>
8722 David Sherwood <david.sherwood@arm.com>
8724 * match.pd: Cope with polynomial numbers of vector elements.
8726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8727 Alan Hayward <alan.hayward@arm.com>
8728 David Sherwood <david.sherwood@arm.com>
8730 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8731 in a POINTER_PLUS_EXPR.
8733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8734 Alan Hayward <alan.hayward@arm.com>
8735 David Sherwood <david.sherwood@arm.com>
8737 * omp-simd-clone.c (simd_clone_subparts): New function.
8738 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8739 (ipa_simd_modify_function_body): Likewise.
8741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8742 Alan Hayward <alan.hayward@arm.com>
8743 David Sherwood <david.sherwood@arm.com>
8745 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8746 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8747 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8748 (expand_vector_condition, vector_element): Likewise.
8749 (subparts_gt): New function.
8750 (get_compute_type): Use subparts_gt.
8751 (count_type_subparts): Delete.
8752 (expand_vector_operations_1): Use subparts_gt instead of
8753 count_type_subparts.
8755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8756 Alan Hayward <alan.hayward@arm.com>
8757 David Sherwood <david.sherwood@arm.com>
8759 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8760 (vect_compile_time_alias): ...this new function. Do the calculation
8761 on poly_ints rather than trees.
8762 (vect_prune_runtime_alias_test_list): Update call accordingly.
8764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8765 Alan Hayward <alan.hayward@arm.com>
8766 David Sherwood <david.sherwood@arm.com>
8768 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8770 (vect_schedule_slp_instance): Likewise.
8772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8773 Alan Hayward <alan.hayward@arm.com>
8774 David Sherwood <david.sherwood@arm.com>
8776 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8777 constant and extern definitions for variable-length vectors.
8778 (vect_get_constant_vectors): Note that the number of units
8779 is known to be constant.
8781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8782 Alan Hayward <alan.hayward@arm.com>
8783 David Sherwood <david.sherwood@arm.com>
8785 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8786 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
8794 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8797 Alan Hayward <alan.hayward@arm.com>
8798 David Sherwood <david.sherwood@arm.com>
8800 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8801 vectors as polynomial. Use build_index_vector for
8804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8805 Alan Hayward <alan.hayward@arm.com>
8806 David Sherwood <david.sherwood@arm.com>
8808 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8809 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8810 for variable-length vectors.
8811 (vectorizable_mask_load_store): Treat the number of units as
8812 polynomial, asserting that it is constant if the condition has
8813 already been enforced.
8814 (vectorizable_store, vectorizable_load): Likewise.
8816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8817 Alan Hayward <alan.hayward@arm.com>
8818 David Sherwood <david.sherwood@arm.com>
8820 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8821 of units as polynomial. Punt if we can't tell at compile time
8822 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
8829 of units as polynomial. Punt on SLP inductions. Use an integer
8830 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8831 cast of such a series for variable-length floating-point
8834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8835 Alan Hayward <alan.hayward@arm.com>
8836 David Sherwood <david.sherwood@arm.com>
8838 * tree.h (build_index_vector): Declare.
8839 * tree.c (build_index_vector): New function.
8840 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8841 of units as polynomial, forcibly converting it to a constant if
8842 vectorizable_reduction has already enforced the condition.
8843 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8844 to create a {1,2,3,...} vector.
8845 (vectorizable_reduction): Treat the number of units as polynomial.
8846 Choose vectype_in based on the largest scalar element size rather
8847 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
8855 number of units as polynomial.
8857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8858 Alan Hayward <alan.hayward@arm.com>
8859 David Sherwood <david.sherwood@arm.com>
8861 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8862 * target.def (autovectorize_vector_sizes): Return the vector sizes
8863 by pointer, using vector_sizes rather than a bitmask.
8864 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8865 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8866 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8868 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8869 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8870 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8871 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8872 * omp-general.c (omp_max_vf): Likewise.
8873 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8874 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8875 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8876 * tree-vect-slp.c (vect_slp_bb): Likewise.
8877 * doc/tm.texi: Regenerate.
8878 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8880 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8881 the vector size as a poly_uint64 rather than an unsigned int.
8882 (current_vector_size): Change from an unsigned int to a poly_uint64.
8883 (get_vectype_for_scalar_type): Update accordingly.
8884 * tree.h (build_truth_vector_type): Take the size and number of
8885 units as a poly_uint64 rather than an unsigned int.
8886 (build_vector_type): Add a temporary overload that takes
8887 the number of units as a poly_uint64 rather than an unsigned int.
8888 * tree.c (make_vector_type): Likewise.
8889 (build_truth_vector_type): Take the number of units as a poly_uint64
8890 rather than an unsigned int.
8892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8893 Alan Hayward <alan.hayward@arm.com>
8894 David Sherwood <david.sherwood@arm.com>
8896 * target.def (get_mask_mode): Take the number of units and length
8897 as poly_uint64s rather than unsigned ints.
8898 * targhooks.h (default_get_mask_mode): Update accordingly.
8899 * targhooks.c (default_get_mask_mode): Likewise.
8900 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8901 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8908 * omp-general.c (omp_max_vf): Likewise.
8909 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8910 (expand_omp_simd): Handle polynomial safelen.
8911 * omp-low.c (omplow_simd_context): Add a default constructor.
8912 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8913 (lower_rec_simd_input_clauses): Update accordingly.
8914 (lower_rec_input_clauses): Likewise.
8916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8917 Alan Hayward <alan.hayward@arm.com>
8918 David Sherwood <david.sherwood@arm.com>
8920 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8921 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8922 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8923 (vect_analyze_slp_cost): Likewise.
8924 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8925 (vect_model_load_cost): Likewise.
8927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8928 Alan Hayward <alan.hayward@arm.com>
8929 David Sherwood <david.sherwood@arm.com>
8931 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8932 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8933 from an unsigned int * to a poly_uint64_pod *.
8934 (calculate_unrolling_factor): New function.
8935 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8938 Alan Hayward <alan.hayward@arm.com>
8939 David Sherwood <david.sherwood@arm.com>
8941 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8942 from an unsigned int to a poly_uint64.
8943 (_loop_vec_info::slp_unrolling_factor): Likewise.
8944 (_loop_vec_info::vectorization_factor): Change from an int
8946 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8947 (vect_get_num_vectors): New function.
8948 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8949 (vect_get_num_copies): Use vect_get_num_vectors.
8950 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8951 to an unsigned int *.
8952 (vect_analyze_data_refs): Change min_vf from an int * to a
8954 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8955 than an unsigned HOST_WIDE_INT.
8956 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8957 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8958 to an unsigned int *.
8959 (vect_analyze_data_ref_dependences): Likewise.
8960 (vect_compute_data_ref_alignment): Handle polynomial vf.
8961 (vect_enhance_data_refs_alignment): Likewise.
8962 (vect_prune_runtime_alias_test_list): Likewise.
8963 (vect_shift_permute_load_chain): Likewise.
8964 (vect_supportable_dr_alignment): Likewise.
8965 (dependence_distance_ge_vf): Take the vectorization factor as a
8966 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8967 (vect_analyze_data_refs): Change min_vf from an int * to a
8969 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8970 vfm1 as a poly_uint64 rather than an int. Make the same change
8971 for the returned bound_scalar.
8972 (vect_gen_vector_loop_niters): Handle polynomial vf.
8973 (vect_do_peeling): Likewise. Update call to
8974 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8975 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8977 * tree-vect-loop.c (vect_determine_vectorization_factor)
8978 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8979 (vect_get_known_peeling_cost): Likewise.
8980 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8981 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8982 (vect_transform_loop): Likewise. Use the lowest possible VF when
8983 updating the upper bounds of the loop.
8984 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8986 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8987 polynomial unroll factors.
8988 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8989 (vect_make_slp_decision): Likewise.
8990 (vect_supported_load_permutation_p): Likewise, and polynomial
8992 (vect_analyze_slp_cost): Handle polynomial vf.
8993 (vect_slp_analyze_node_operations): Likewise.
8994 (vect_slp_analyze_bb_1): Likewise.
8995 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8996 than an unsigned HOST_WIDE_INT.
8997 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8998 (vectorizable_load): Handle polynomial vf.
8999 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9001 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9004 Alan Hayward <alan.hayward@arm.com>
9005 David Sherwood <david.sherwood@arm.com>
9007 * match.pd: Handle bit operations involving three constants
9008 and try to fold one pair.
9010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9012 * tree-vect-loop-manip.c: Include gimple-fold.h.
9013 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9014 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9015 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9016 Add a path that uses a step of VF instead of 1, but disable it
9018 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9019 and niters_no_overflow parameters. Update calls to
9020 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9021 Create a new SSA name if the latter choses to use a ste other
9022 than zero, and return it via niters_vector_mult_vf_var.
9023 * tree-vect-loop.c (vect_transform_loop): Update calls to
9024 vect_do_peeling, vect_gen_vector_loop_niters and
9025 slpeel_make_loop_iterate_ntimes.
9026 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9027 (vect_gen_vector_loop_niters): Update declarations after above changes.
9029 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9031 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9032 128-bit round to integer instructions.
9033 (ceil<mode>2): Likewise.
9034 (btrunc<mode>2): Likewise.
9035 (round<mode>2): Likewise.
9037 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9039 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9040 unaligned VSX load/store on P8/P9.
9041 (expand_block_clear): Allow the use of unaligned VSX
9042 load/store on P8/P9.
9044 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9046 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9048 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9049 swap associated with both a load and a store.
9051 2018-01-02 Andrew Waterman <andrew@sifive.com>
9053 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9054 * config/riscv/riscv.md (clear_cache): Use it.
9056 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9058 * web.c: Remove out-of-date comment.
9060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9062 * expr.c (fixup_args_size_notes): Check that any existing
9063 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9064 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9065 (emit_single_push_insn): ...here.
9067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9069 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9070 (const_vector_encoded_nelts): New function.
9071 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9072 (const_vector_int_elt, const_vector_elt): Declare.
9073 * emit-rtl.c (const_vector_int_elt_1): New function.
9074 (const_vector_elt): Likewise.
9075 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9076 of CONST_VECTOR_ELT.
9078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9080 * expr.c: Include rtx-vector-builder.h.
9081 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9082 directly on the tree encoding.
9083 (const_vector_from_tree): Likewise.
9084 * optabs.c: Include rtx-vector-builder.h.
9085 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9086 sequence of "u" values.
9087 * vec-perm-indices.c: Include rtx-vector-builder.h.
9088 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9089 directly on the vec_perm_indices encoding.
9091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9093 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9094 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9095 * rtx-vector-builder.h: New file.
9096 * rtx-vector-builder.c: Likewise.
9097 * rtl.h (rtx_def::u2): Add a const_vector field.
9098 (CONST_VECTOR_NPATTERNS): New macro.
9099 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9100 (CONST_VECTOR_DUPLICATE_P): Likewise.
9101 (CONST_VECTOR_STEPPED_P): Likewise.
9102 (CONST_VECTOR_ENCODED_ELT): Likewise.
9103 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9104 (unwrap_const_vec_duplicate): Likewise.
9105 (const_vec_series_p): Check for a non-duplicated vector encoding.
9106 Say that the function only returns true for integer vectors.
9107 * emit-rtl.c: Include rtx-vector-builder.h.
9108 (gen_const_vec_duplicate_1): Delete.
9109 (gen_const_vector): Call gen_const_vec_duplicate instead of
9110 gen_const_vec_duplicate_1.
9111 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9112 (gen_const_vec_duplicate): Use rtx_vector_builder.
9113 (gen_const_vec_series): Likewise.
9114 (gen_rtx_CONST_VECTOR): Likewise.
9115 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9116 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9117 Build a new vector rather than modifying a CONST_VECTOR in-place.
9118 (handle_special_swappables): Update call accordingly.
9119 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9120 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9121 Build a new vector rather than modifying a CONST_VECTOR in-place.
9122 (handle_special_swappables): Update call accordingly.
9124 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9126 * simplify-rtx.c (simplify_const_binary_operation): Use
9127 CONST_VECTOR_ELT instead of XVECEXP.
9129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9131 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9132 the selector elements to be different from the data elements
9133 if the selector is a VECTOR_CST.
9134 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9135 ssizetype for the selector.
9137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9139 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9140 before testing each element individually.
9141 * tree-vect-generic.c (lower_vec_perm): Likewise.
9143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9145 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9146 * selftest-run-tests.c (selftest::run_tests): Call it.
9147 * vector-builder.h (vector_builder::operator ==): New function.
9148 (vector_builder::operator !=): Likewise.
9149 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9150 (vec_perm_indices::all_from_input_p): New function.
9151 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9152 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9153 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9154 instead of reading the VECTOR_CST directly. Detect whether both
9155 vector inputs are the same before constructing the vec_perm_indices,
9156 and update the number of inputs argument accordingly. Use the
9157 utility functions added above. Only construct sel2 if we need to.
9159 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9161 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9162 the broadcast of the low byte.
9163 (expand_mult_highpart): Use an explicit encoding for the permutes.
9164 * optabs-query.c (can_mult_highpart_p): Likewise.
9165 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9166 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9167 (vectorizable_bswap): Likewise.
9168 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9169 explicit encoding for the power-of-2 permutes.
9170 (vect_permute_store_chain): Likewise.
9171 (vect_grouped_load_supported): Likewise.
9172 (vect_permute_load_chain): Likewise.
9174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9176 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9177 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9178 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9179 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9180 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9181 (vect_gen_perm_mask_any): Likewise.
9183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9185 * int-vector-builder.h: New file.
9186 * vec-perm-indices.h: Include int-vector-builder.h.
9187 (vec_perm_indices): Redefine as an int_vector_builder.
9188 (auto_vec_perm_indices): Delete.
9189 (vec_perm_builder): Redefine as a stand-alone class.
9190 (vec_perm_indices::vec_perm_indices): New function.
9191 (vec_perm_indices::clamp): Likewise.
9192 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9193 (vec_perm_indices::new_vector): New function.
9194 (vec_perm_indices::new_expanded_vector): Update for new
9195 vec_perm_indices class.
9196 (vec_perm_indices::rotate_inputs): New function.
9197 (vec_perm_indices::all_in_range_p): Operate directly on the
9198 encoded form, without computing elided elements.
9199 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9200 encoding. Update for new vec_perm_indices class.
9201 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9202 the given vec_perm_builder.
9203 (expand_vec_perm_var): Update vec_perm_builder constructor.
9204 (expand_mult_highpart): Use vec_perm_builder instead of
9205 auto_vec_perm_indices.
9206 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9207 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9208 or double series encoding as appropriate.
9209 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9210 vec_perm_indices instead of auto_vec_perm_indices.
9211 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9212 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9213 (vect_permute_store_chain): Likewise.
9214 (vect_grouped_load_supported): Likewise.
9215 (vect_permute_load_chain): Likewise.
9216 (vect_shift_permute_load_chain): Likewise.
9217 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9218 (vect_transform_slp_perm_load): Likewise.
9219 (vect_schedule_slp_instance): Likewise.
9220 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9221 (vectorizable_mask_load_store): Likewise.
9222 (vectorizable_bswap): Likewise.
9223 (vectorizable_store): Likewise.
9224 (vectorizable_load): Likewise.
9225 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9226 vec_perm_indices instead of auto_vec_perm_indices. Use
9227 tree_to_vec_perm_builder to read the vector from a tree.
9228 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9229 vec_perm_builder instead of a vec_perm_indices.
9230 (have_whole_vector_shift): Use vec_perm_builder and
9231 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9232 truncation to calc_vec_perm_mask_for_shift.
9233 (vect_create_epilog_for_reduction): Likewise.
9234 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9235 from auto_vec_perm_indices to vec_perm_indices.
9236 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9237 instead of changing individual elements.
9238 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9239 the vector in d.perm.
9240 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9241 from auto_vec_perm_indices to vec_perm_indices.
9242 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9243 instead of changing individual elements.
9244 (arm_vectorize_vec_perm_const): Use new_vector to install
9245 the vector in d.perm.
9246 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9247 Update vec_perm_builder constructor.
9248 (rs6000_expand_interleave): Likewise.
9249 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9250 (rs6000_expand_interleave): Likewise.
9252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9254 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9255 to qimode could truncate the indices.
9256 * optabs.c (expand_vec_perm_var): Likewise.
9258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9260 * Makefile.in (OBJS): Add vec-perm-indices.o.
9261 * vec-perm-indices.h: New file.
9262 * vec-perm-indices.c: Likewise.
9263 * target.h (vec_perm_indices): Replace with a forward class
9265 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9266 * optabs.h: Include vec-perm-indices.h.
9267 (expand_vec_perm): Delete.
9268 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9269 (expand_vec_perm_const): Declare.
9270 * target.def (vec_perm_const_ok): Replace with...
9271 (vec_perm_const): ...this new hook.
9272 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9273 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9274 * doc/tm.texi: Regenerate.
9275 * optabs.def (vec_perm_const): Delete.
9276 * doc/md.texi (vec_perm_const): Likewise.
9277 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9278 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9279 expand_vec_perm for constant permutation vectors. Assert that
9280 the mode of variable permutation vectors is the integer equivalent
9281 of the mode that is being permuted.
9282 * optabs-query.h (selector_fits_mode_p): Declare.
9283 * optabs-query.c: Include vec-perm-indices.h.
9284 (selector_fits_mode_p): New function.
9285 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9286 is defined, instead of checking whether the vec_perm_const_optab
9287 exists. Use targetm.vectorize.vec_perm_const instead of
9288 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9289 fit in the vector mode before using a variable permute.
9290 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9291 vec_perm_indices instead of an rtx.
9292 (expand_vec_perm): Replace with...
9293 (expand_vec_perm_const): ...this new function. Take the selector
9294 as a vec_perm_indices rather than an rtx. Also take the mode of
9295 the selector. Update call to shift_amt_for_vec_perm_mask.
9296 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9297 Use vec_perm_indices::new_expanded_vector to expand the original
9298 selector into bytes. Check whether the indices fit in the vector
9299 mode before using a variable permute.
9300 (expand_vec_perm_var): Make global.
9301 (expand_mult_highpart): Use expand_vec_perm_const.
9302 * fold-const.c: Includes vec-perm-indices.h.
9303 * tree-ssa-forwprop.c: Likewise.
9304 * tree-vect-data-refs.c: Likewise.
9305 * tree-vect-generic.c: Likewise.
9306 * tree-vect-loop.c: Likewise.
9307 * tree-vect-slp.c: Likewise.
9308 * tree-vect-stmts.c: Likewise.
9309 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9311 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9312 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9313 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9314 (aarch64_vectorize_vec_perm_const): ...this new function.
9315 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9316 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9317 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9318 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9319 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9320 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9321 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9323 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9324 check for NEON modes.
9325 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9326 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9327 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9328 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9330 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9331 the old VEC_PERM_CONST conditions.
9332 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9333 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9334 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9335 (ia64_vectorize_vec_perm_const_ok): Merge into...
9336 (ia64_vectorize_vec_perm_const): ...this new function.
9337 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9338 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9339 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9340 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9341 * config/mips/mips.c (mips_expand_vec_perm_const)
9342 (mips_vectorize_vec_perm_const_ok): Merge into...
9343 (mips_vectorize_vec_perm_const): ...this new function.
9344 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9345 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9346 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9347 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9348 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9349 (rs6000_expand_vec_perm_const): Delete.
9350 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9352 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9353 (altivec_expand_vec_perm_const_le): Take each operand individually.
9354 Operate on constant selectors rather than rtxes.
9355 (altivec_expand_vec_perm_const): Likewise. Update call to
9356 altivec_expand_vec_perm_const_le.
9357 (rs6000_expand_vec_perm_const): Delete.
9358 (rs6000_vectorize_vec_perm_const_ok): Delete.
9359 (rs6000_vectorize_vec_perm_const): New function.
9360 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9361 an element count and rtx array.
9362 (rs6000_expand_extract_even): Update call accordingly.
9363 (rs6000_expand_interleave): Likewise.
9364 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9365 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9366 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9367 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9368 (rs6000_expand_vec_perm_const): Delete.
9369 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9370 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9371 (altivec_expand_vec_perm_const_le): Take each operand individually.
9372 Operate on constant selectors rather than rtxes.
9373 (altivec_expand_vec_perm_const): Likewise. Update call to
9374 altivec_expand_vec_perm_const_le.
9375 (rs6000_expand_vec_perm_const): Delete.
9376 (rs6000_vectorize_vec_perm_const_ok): Delete.
9377 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9378 reference to the SPE evmerge intructions.
9379 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9380 an element count and rtx array.
9381 (rs6000_expand_extract_even): Update call accordingly.
9382 (rs6000_expand_interleave): Likewise.
9383 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9384 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9386 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9390 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9391 vector mode and that that mode matches the mode of the data
9393 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9394 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9395 directly using expand_vec_perm_1 when forcing selectors into
9397 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9399 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9401 * optabs-query.h (can_vec_perm_p): Delete.
9402 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9403 * optabs-query.c (can_vec_perm_p): Split into...
9404 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9405 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9406 particular selector is valid.
9407 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9408 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9409 (vect_grouped_load_supported): Likewise.
9410 (vect_shift_permute_load_chain): Likewise.
9411 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9412 (vect_transform_slp_perm_load): Likewise.
9413 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9414 (vectorizable_bswap): Likewise.
9415 (vect_gen_perm_mask_checked): Likewise.
9416 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9417 implementations of variable permutation vectors into account
9418 when deciding which selector to use.
9419 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9420 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9421 with a false third argument.
9422 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9423 to test whether the constant selector is valid and can_vec_perm_var_p
9424 to test whether a variable selector is valid.
9426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9428 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9429 * optabs-query.c (can_vec_perm_p): Likewise.
9430 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9431 instead of vec_perm_indices.
9432 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9433 (vect_gen_perm_mask_checked): Likewise,
9434 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9435 (vect_gen_perm_mask_checked): Likewise,
9437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9439 * optabs-query.h (qimode_for_vec_perm): Declare.
9440 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9441 (qimode_for_vec_perm): ...this new function.
9442 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9444 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9446 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9447 does not have a conditional at the top.
9449 2018-01-02 Richard Biener <rguenther@suse.de>
9451 * ipa-inline.c (big_speedup_p): Fix expression.
9453 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9456 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9459 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9463 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9464 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9465 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9466 cond_taken_branch_cost 3->4.
9468 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9470 PR tree-optimization/83581
9471 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9472 TODO_cleanup_cfg if any changes have been made.
9475 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9476 convert_modes if target mode has the right side, but different mode
9480 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9481 last argument when extracting from CONCAT. If either from_real or
9482 from_imag is NULL, use expansion through memory. If result is not
9483 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9484 the parts directly to inner mode, if even that fails, use expansion
9488 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9489 check for bswap in mode rather than HImode and use that in expand_unop
9492 Copyright (C) 2018 Free Software Foundation, Inc.
9494 Copying and distribu