1 2018-03-28 Martin Liska <mliska@suse.cz>
4 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5 chkp_type_bounds_count if MPX is not enabled.
7 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
14 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
15 function to create the function decl for complex long double
16 multiply and divide for -mabi=ieeelongdouble.
17 (init_float128_ieee): Call it.
19 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
22 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23 -fcf-protection=branch -mibt.
24 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
26 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
29 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
31 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
34 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
35 extern array declarations.
37 2018-03-27 Richard Biener <rguenther@suse.de>
40 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
41 explicit single_use checks.
43 2018-03-27 Richard Biener <rguenther@suse.de>
45 PR tree-optimization/85082
46 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
49 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
51 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
52 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
53 Turn on fasynchronous-unwind-tables and funwind-tables.
55 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
58 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
59 (*bmi_blsr_<mode>_ccz): Ditto.
61 2018-03-26 Tom de Vries <tom@codesourcery.com>
63 PR tree-optimization/85063
64 * omp-general.c (offloading_function_p): New function. Factor out
66 * omp-offload.c (pass_omp_target_link::gate): ... here.
67 * omp-general.h (offloading_function_p): Declare.
68 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
69 with attribute omp declare target for offloading functions.
71 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
73 PR tree-optimization/84005
74 * tree-data-ref.h (get_base_for_alignment): Declare.
75 * tree-data-ref.c (get_base_for_alignment_1): New function.
76 (get_base_for_alignment): Likewise.
77 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
78 get_base_for_alignment to find a suitable base object, instead
79 of always using drb->base_address.
81 2018-03-23 Jakub Jelinek <jakub@redhat.com>
84 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
85 known size by default.
87 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
90 * lra-constraints.c (process_alt_operands): Don't match BLKmode
91 and non BLKmode operands.
93 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
96 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
99 2018-03-23 Richard Biener <rguenther@suse.de>
102 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
103 we are going to emit early debug for LTO.
105 2018-03-23 Jakub Jelinek <jakub@redhat.com>
108 * function.c (match_asm_constraints_1): Don't optimize if input
109 doesn't satisfy general_operand predicate for output's mode.
112 * alias.c (write_dependence_p): Don't require for x_canonicalized
113 non-VOIDmode if x has VOIDmode.
116 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
117 just don't try to optimize it rather than assert it never happens.
119 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
121 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
122 macro expansions for definition of ST_INTERNAL_<mode> and
123 LD_INTERNAL_<mode> builtins.
124 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
126 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
128 (altivec_expand_st_builtin): Likewise.
129 (altivec_expand_builtin): Remove calls to deleted functions.
130 (rs6000_address_for_altivec): Delete this function.
131 * config/rs6000/vector.md: Remove expands for
132 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
134 2018-03-22 Sudakshina Das <sudi.das@arm.com>
137 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
138 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
139 re-computing once computed.
140 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
141 (arm_init_machine_status): Initialize
142 machine->static_chain_stack_bytes.
144 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
147 * doc/extend.texi: Add four new prototypes for vec_ld.
148 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
149 definitions for more logical presentation.
150 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
151 entries for V1TI variants of __builtin_altivec_ld builtin.
152 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
153 handling of V1TI variant of LVX icode pattern.
154 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
155 (rs6000_gimple_fold_builtin): Likewise.
156 (altivec_init_builtins): Add code to define
157 __builtin_altivec_lvx_v1ti function.
159 2018-03-22 Jakub Jelinek <jakub@redhat.com>
162 * function.c (match_asm_constraints_1): Don't do the optimization
163 if input isn't a REG, SUBREG, MEM or constant.
165 2018-03-22 Tom de Vries <tom@codesourcery.com>
167 PR tree-optimization/84956
168 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
169 bb_has_abnormal_pred.
171 2018-03-22 Jakub Jelinek <jakub@redhat.com>
174 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
175 DECL_INITIAL (decl) to decl at the end.
176 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
179 2018-03-21 Joseph Myers <joseph@codesourcery.com>
181 * doc/extend.texi (__builtin_tgmath): Document when complex
182 integer types are treated as _Complex _Float64.
184 2018-03-21 Tom de Vries <tom@codesourcery.com>
186 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
188 2018-03-21 Jakub Jelinek <jakub@redhat.com>
190 PR tree-optimization/84960
191 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
192 if it is ENTRY block, move them into single succ of ENTRY in that case.
194 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
196 PR tree-optimization/84811
197 * poly-int.h (poly_span_traits): Remove the T3 parameter and
198 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
199 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
200 (known_subrange_p): Update accordingly. Cast each value involved
201 in the size comparison, rather than casting the result of the
204 2018-03-21 Jakub Jelinek <jakub@redhat.com>
206 PR tree-optimization/84982
207 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
208 by flipping the least significant bit rather than all bits from
209 bitpos to bitpos + bitsize - 1.
211 2018-03-21 Nathan Sidwell <nathan@acm.org>
213 * doc/extend.texi (Deprecated Features): Remove mention of
214 long-deleted deprecations.
216 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
219 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
220 * configure: Regenerate.
222 2018-03-21 Tom de Vries <tom@codesourcery.com>
224 PR tree-optimization/83126
225 * tree-parloops.c (num_phis): New function.
226 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
228 2018-03-21 Nathan Sidwell <nathan@acm.org>
230 * doc/extend.texi (Deprecated Features): Update deprecated flags,
231 mention anon-struct/union members and trailing attributes.
233 2018-03-21 Bin Cheng <bin.cheng@arm.com>
235 PR tree-optimization/84969
236 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
237 builtin memset partitions if they set different rhs values.
239 2018-03-21 Jakub Jelinek <jakub@redhat.com>
241 PR rtl-optimization/84989
242 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
243 VEC_DUPLICATE with scalar result mode.
245 2018-03-21 Martin Liska <mliska@suse.cz>
248 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
249 not intended return statement.
251 2018-03-21 Martin Liska <mliska@suse.cz>
254 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
255 (chkp_find_bound_slots_1): Limit number of iterations.
257 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
260 * Minor grammar fixes for x86 options.
262 2018-03-20 Jakub Jelinek <jakub@redhat.com>
265 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
266 holding REG_CFA_RESTORE notes, instead turn them into a USE.
268 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
271 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
272 (altivec_lvx_<mode>_1op): Likewise.
273 (altivec_stvx_<mode>_2op): Likewise.
274 (altivec_stvx_<mode>_1op): Likewise.
275 (altivec_lvx_<VM2:mode>): New define_expand.
276 (altivec_stvx_<VM2:mode>): Likewise.
277 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
278 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
279 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
280 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
281 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
282 (rs6000_gen_lvx): Likewise.
283 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
284 (altivec_expand_stv_builtin): Likewise.
285 (altivec_expand_builtin): Likewise.
286 * config/rs6000/vector.md: Likewise.
288 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
291 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
294 2018-03-20 Richard Biener <rguenther@suse.de>
297 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
298 sign-conversions as zero, fall back to standard scalar_stmt
301 2018-03-20 Martin Liska <mliska@suse.cz>
304 * predict.c (rebuild_frequencies): Handle case when we have
305 PROFILE_ABSENT, but flag_guess_branch_prob is false.
307 2018-03-20 Jakub Jelinek <jakub@redhat.com>
310 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
311 flag_section_anchors.
312 * varasm.c (use_blocks_for_decl_p): Remove hack for
316 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
318 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
319 be created, use lowpart_subreg of operands[0] rather than operands[0]
321 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
322 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
323 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
324 and n constraint instead of aarch64_shift_imm_di and Usd.
325 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
326 (*aarch64_<optab>_reg_minus<mode>3): ... this.
328 2018-03-20 Sudakshina Das <sudi.das@arm.com>
331 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
332 to favor GPR over NEON registers.
333 (<shift>di3_neon): Likewise.
335 2018-03-20 Tom de Vries <tom@codesourcery.com>
338 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
339 (nvptx_process_pars): Emit bar.sync asap and alap.
341 2018-03-20 Tom de Vries <tom@codesourcery.com>
344 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
345 seen_label if seen_label is already set.
347 2018-03-20 Jakub Jelinek <jakub@redhat.com>
350 * config/i386/i386.c (fold_builtin_cpu): For features above 31
351 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
352 Use 1U instead of 1. Formatting fixes.
355 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
356 instead of TREE_TYPE (s1) for the return value.
358 2018-03-19 Jakub Jelinek <jakub@redhat.com>
360 PR tree-optimization/84946
361 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
362 bitsize + bitsize in poly_uint64 rather than poly_int64.
365 * dwarf2asm.c: Include fold-const.c.
366 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
367 of decl rather than decl itself.
369 PR rtl-optimization/84643
370 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
372 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
375 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
376 calling assemble_variable.
378 2018-03-19 Sudakshina Das <sudi.das@arm.com>
381 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
382 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
384 2018-03-19 Jim Wilson <jimw@sifive.com>
387 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
388 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
389 (riscv_first_stack_step): Likewise.
390 (riscv_option_override): Use STACK_BOUNDARY instead of
392 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
394 (BIGGEST_ALIGNMENT): Set to 128.
395 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
396 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
399 2018-03-19 Richard Biener <rguenther@suse.de>
401 PR tree-optimization/84933
402 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
403 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
405 2018-03-19 Richard Biener <rguenther@suse.de>
407 PR tree-optimization/84859
408 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
409 (cond_if_else_store_replacement): Perform sinking operation on
410 single-store BBs regardless of MAX_STORES_TO_SINK setting.
411 Generalize what a BB with a single eligible store is.
413 2018-03-19 Richard Biener <rguenther@suse.de>
415 PR tree-optimization/84929
416 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
417 chrec_is_positive against non-chrec arg.
419 2018-03-19 Tamar Christina <tamar.christina@arm.com>
422 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
424 2018-03-18 Martin Liska <mliska@suse.cz>
426 PR rtl-optimization/84635
427 * regrename.c (build_def_use): Use matches_mode only when
430 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
432 PR tree-optimization/84913
433 * tree-vect-loop.c (vectorizable_reduction): Don't try to
434 vectorize chains of COND_EXPRs.
436 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
438 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
440 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
442 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
444 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
446 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
448 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
449 Kito Cheng <kito.cheng@gmail.com>
451 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
452 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
453 (nds32_adjust_reg_alloc_order): New function.
454 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
456 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
458 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
459 nds32_print_operand, nds32_print_operand_address): Use
460 HOST_WIDE_INT_PRINT_DEC instead.
462 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
464 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
466 2018-03-17 Jakub Jelinek <jakub@redhat.com>
469 * config/i386/i386.c (initial_ix86_tune_features,
470 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
472 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
473 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
474 rather than 1u << ix86_tune. Formatting fix.
475 (ix86_option_override_internal): Change ix86_arch_mask from
476 unsigned int to unsigned HOST_WIDE_INT, initialize to
477 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
478 (ix86_function_specific_restore): Likewise.
480 2018-03-16 Jakub Jelinek <jakub@redhat.com>
483 * postreload.c (reload_combine_recognize_pattern): Perform
484 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
485 truncate_int_for_mode the result for the destination's mode.
488 * hsa-gen.c (mem_type_for_type): Fix comment typo.
489 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
491 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
494 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
497 * lra-assigns.c (lra_split_hard_reg_for): Don't use
498 regno_allocno_class_array and sorted_pseudos.
499 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
500 insns where regno is used.
502 2018-03-16 Martin Liska <mliska@suse.cz>
505 * multiple_target.c (create_dispatcher_calls): Redirect
506 reference in the symbol table.
508 2018-03-16 Martin Liska <mliska@suse.cz>
511 * multiple_target.c (create_dispatcher_calls): Redirect also
514 2018-03-16 Jakub Jelinek <jakub@redhat.com>
518 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
519 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
520 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
522 2018-03-16 Julia Koval <julia.koval@intel.com>
524 * doc/invoke.texi (Skylake Server): Add CLWB.
525 Cannonlake): Remove CLWB.
527 2018-03-16 Jakub Jelinek <jakub@redhat.com>
529 PR tree-optimization/84841
530 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
532 (FLOAT_ONE_CONST_TYPE): Define.
533 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
534 (sort_by_operand_rank): Put entries with higher constant_type last
535 rather than first to match comments.
537 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
539 * config/nios2/nios2.md (movsi_internal): Fix thinko in
542 2018-03-15 Jakub Jelinek <jakub@redhat.com>
545 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
546 check and use address of target always.
548 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
551 * config/i386/i386.c (indirect_thunk_needed): Update comments.
552 (indirect_thunk_bnd_needed): Likewise.
553 (indirect_thunks_used): Likewise.
554 (indirect_thunks_bnd_used): Likewise.
555 (indirect_return_needed): New.
556 (indirect_return_bnd_needed): Likewise.
557 (output_indirect_thunk_function): Add a bool argument for
559 (output_indirect_thunk_function): Don't generate alias for
560 function return thunk.
561 (ix86_code_end): Call output_indirect_thunk_function to generate
562 function return thunks.
563 (ix86_output_function_return): Set indirect_return_bnd_needed
564 and indirect_return_needed instead of indirect_thunk_bnd_needed
565 and indirect_thunk_needed.
567 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
569 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
570 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
571 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
573 2018-03-15 David Malcolm <dmalcolm@redhat.com>
574 Paul Hua <paul.hua.gm@gmail.com>
577 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
579 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
581 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
582 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
585 2018-03-15 Tamar Christina <tamar.christina@arm.com>
588 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
589 instead of GET_MODE_SIZE when comparing Units.
591 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
594 * varasm.c (hash_section): Return an unchangeble hash value
595 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
596 Return !aarch64_can_use_per_function_literal_pools_p ().
598 2018-03-15 Jakub Jelinek <jakub@redhat.com>
601 * optabs.c (emit_conditional_move): Pass address of cmode's copy
602 rather than address of cmode as last argument to prepare_cmp_insn.
604 2018-03-15 Julia Koval <julia.koval@intel.com>
606 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
607 F_AVX512VNNI, F_AVX512BITALG): New.
609 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
612 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
613 insn for floating-point loads and stores.
615 2018-03-14 Carl Love <cel@us.ibm.com>
617 * config/rs6000/rs6000-c.c: Add macro definitions for
618 ALTIVEC_BUILTIN_VEC_PERMXOR.
619 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
620 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
621 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
622 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
624 * config/doc/extend.texi: Add prototypes for vec_permxor.
626 2018-03-14 David Malcolm <dmalcolm@redhat.com>
629 * diagnostic-show-locus.c (class layout_point): Convert m_line
630 from int to linenum_type.
631 (line_span::comparator): Use linenum "compare" function when
632 comparing line numbers.
633 (test_line_span): New function.
634 (layout_range::contains_point): Convert param "row" from int to
636 (layout_range::intersects_line_p): Likewise.
637 (layout::will_show_line_p): Likewise.
638 (layout::print_source_line): Likewise.
639 (layout::should_print_annotation_line_p): Likewise.
640 (layout::print_annotation_line): Likewise.
641 (layout::print_leading_fixits): Likewise.
642 (layout::annotation_line_showed_range_p): Likewise.
643 (struct line_corrections): Likewise for field m_row.
644 (line_corrections::line_corrections): Likewise for param "row".
645 (layout::print_trailing_fixits): Likewise.
646 (layout::get_state_at_point): Likewise.
647 (layout::get_x_bound_for_row): Likewise.
648 (layout::print_line): Likewise.
649 (diagnostic_show_locus): Likewise for locals "last_line" and
651 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
652 * input.c (selftest::test_linenum_comparisons): New function.
653 (selftest::input_c_tests): Call it.
654 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
655 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
656 * selftest.h (ASSERT_GT): New macro.
657 (ASSERT_GT_AT): New macro.
658 (ASSERT_LT): New macro.
659 (ASSERT_LT_AT): New macro.
661 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
663 PR rtl-optimization/84780
664 * combine.c (distribute_links): Don't make a link based on pc_rtx.
666 2018-03-14 Martin Liska <mliska@suse.cz>
668 * tree.c (record_node_allocation_statistics): Use
670 (get_stats_node_kind): New function extracted from
671 record_node_allocation_statistics.
672 (free_node): Use get_stats_node_kind.
674 2018-03-14 Richard Biener <rguenther@suse.de>
676 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
677 that the value-set of ANTIC_IN doesn't grow.
680 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
682 (BB_VISITED_WITH_VISITED_SUCCS): New define.
683 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
685 2018-03-14 Julia Koval <julia.koval@intel.com>
687 * config.gcc (icelake-client, icelake-server): New.
689 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
690 (initial_ix86_arch_features): Ditto.
691 (PTA_SKYLAKE): Add SGX.
692 (PTA_ICELAKE): Remove.
693 (PTA_ICELAKE_CLIENT): New.
694 (PTA_ICELAKE_SERVER): New.
695 (ix86_option_override_internal): Split up icelake on icelake client and
697 (get_builtin_code_for_version): Ditto.
698 (fold_builtin_cpu): Ditto.
699 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
700 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
701 * config/i386/i386.h (processor_type): Ditto.
702 * doc/invoke.texi: Ditto.
704 2018-03-14 Jakub Jelinek <jakub@redhat.com>
707 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
708 INTEGER_CST offset, add it together with bitpos / 8 and
709 sign extend based on POINTER_SIZE.
713 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
716 * config/i386/constraints.md (Yc): New register constraint.
717 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
718 Use Yc constraint for alternative 2 of operand 0. Remove
719 preferred_for_speed attribute.
721 2018-03-14 Richard Biener <rguenther@suse.de>
723 PR tree-optimization/84830
724 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
725 with the old one to avoid oscillations.
727 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
730 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
732 (assign_by_spills): Return a flag of reload assignment failure.
733 Do not process the reload assignment failures. Do not spill other
734 reload pseudos if they has the same reg class. Update n if
736 (lra_assign): Add a return arg. Set up from the result of
737 assign_by_spills call.
738 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
739 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
740 usage_insns if it is not NULL.
741 (spill_hard_reg_in_range): New function.
742 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
743 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
745 (lra_assign): Change prototype.
746 * lra.c (lra): Add code to deal with fails by splitting hard reg
749 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
751 * config/riscv/riscv.opt (mrelax): New option.
752 * config/riscv/riscv.c (riscv_file_start): Emit ".option
753 "norelax" when riscv_mrelax is disabled.
754 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
756 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
759 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
760 reassociation for int modes.
762 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
764 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
765 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
767 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
768 * config/aarch64/aarch64-sve.md
769 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
770 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
771 (*extend<mode><Vwide>2): Rename to...
772 (aarch64_sve_extend<mode><Vwide>2): ...this.
773 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
774 renaming the old pattern to...
775 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
777 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
778 define_expand, renaming the old pattern to...
779 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
780 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
781 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
782 account when deciding which SVE instruction the optab should use.
783 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
785 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
787 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
788 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
789 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
790 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
791 (tlsdesc_small_<mode>): Turn a define_expand and use
792 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
793 (tlsdesc_small_advsimd_<mode>): ...this.
794 (tlsdesc_small_sve_<mode>): New pattern.
796 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
798 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
799 (UNSPEC_UMUL_HIGHPART): New constants.
800 (MUL_HIGHPART): New int iteraor.
801 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
802 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
804 (*<su>mul<mode>3_highpart): New define_insn.
806 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
809 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
812 2018-03-13 Martin Liska <mliska@suse.cz>
815 * (sem_item_optimizer::sem_item_optimizer): Initialize new
817 (sem_item_optimizer::~sem_item_optimizer): Release it.
818 (sem_item_optimizer::merge_classes): Register variable aliases.
819 (sem_item_optimizer::fixup_pt_set): New function.
820 (sem_item_optimizer::fixup_points_to_sets): Likewise.
821 * ipa-icf.h: Declare new variables and functions.
823 2018-03-13 Jakub Jelinek <jakub@redhat.com>
826 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
827 integer_pow2p@2 and test integer_pow2p in condition.
828 (A < 0 ? C : 0): Similarly for @1.
831 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
832 characters starting at p contain '\0' character, don't look beyond
836 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
837 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
840 * reg-stack.c (change_stack): Change update_end var from int to
841 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
842 also call set_block_for_insn on the newly added insns and rescan.
845 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
849 * tree.c (stabilize_reference_1): Return save_expr (e) for
850 STATEMENT_LIST even if it doesn't have side-effects.
852 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
854 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
856 2018-03-12 Renlin Li <renlin.li@arm.com>
858 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
859 aarch64_output_scalar_simd_mov_immediate.
861 2018-03-12 Martin Sebor <msebor@redhat.com>
863 PR tree-optimization/83456
864 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
865 for perfectly overlapping calls to memcpy.
866 (gimple_fold_builtin_memory_chk): Same.
867 (gimple_fold_builtin_strcpy): Handle no-warning.
868 (gimple_fold_builtin_stxcpy_chk): Same.
869 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
871 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
873 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
874 parameter. Use it for SFmode.
875 (rs6000_function_arg_advance_1): Adjust.
876 (rs6000_function_arg): Adjust.
877 (rs6000_gimplify_va_arg): Pass false for that new parameter.
879 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
881 PR rtl-optimization/84169
882 PR rtl-optimization/84780
883 * combine.c (can_combine_p): Check for a 2-insn combination whether
884 the destination register is used between the two insns, too.
886 2018-03-12 Richard Biener <rguenther@suse.de>
888 PR tree-optimization/84803
889 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
890 for refs DR analysis didn't process.
892 2018-03-12 Richard Biener <rguenther@suse.de>
894 PR tree-optimization/84777
895 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
896 force-vectorize loops ignore whether we are optimizing for size.
898 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
900 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
901 (TARGET_MD_ASM_ADJUST): Define.
903 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
904 Kito Cheng <kito.cheng@gmail.com>
905 Chung-Ju Wu <jasonwucj@gmail.com>
907 * config/nds32/nds32.c (nds32_compute_stack_frame,
908 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
909 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
910 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
911 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
912 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
913 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
914 * config/nds32/nds32.md (prologue, epilogue): Use macro
915 NDS32_V3PUSH_AVAILABLE_P to do checking.
917 2018-03-11 Jakub Jelinek <jakub@redhat.com>
920 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
921 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
922 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
923 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
924 addition of most attributes on !orig_type_die or the attribute not
925 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
927 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
928 Chung-Ju Wu <jasonwucj@gmail.com>
930 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
932 * config/nds32/nds32.opt (mvh): New option.
934 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
935 Chung-Ju Wu <jasonwucj@gmail.com>
937 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
939 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
940 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
943 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
944 Chung-Ju Wu <jasonwucj@gmail.com>
946 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
948 * config/nds32/nds32-multiple.md (strlensi): New pattern.
949 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
951 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
952 Kito Cheng <kito.cheng@gmail.com>
953 Chung-Ju Wu <jasonwucj@gmail.com>
955 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
956 UNSPEC_FFMISM and UNSPEC_FLMISM.
957 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
958 for ffb, ffmism and flmism.
959 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
960 (unspec_ffmism): Ditto.
961 (unspec_flmism): Ditto.
962 (nds32_expand_builtin_impl): Check if string extension is available.
963 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
964 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
966 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
969 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
972 * lra-assigns.c (assign_by_spills): Return a flag of reload
973 assignment failure. Do not process the reload assignment
974 failures. Do not spill other reload pseudos if they has the same
976 (lra_assign): Add a return arg. Set up from the result of
977 assign_by_spills call.
978 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
979 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
980 usage_insns if it is not NULL.
981 (spill_hard_reg_in_range): New function.
982 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
983 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
985 (lra_assign): Change prototype.
986 * lra.c (lra): Add code to deal with fails by splitting hard reg
989 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
992 * config/i386/i386.opt: Replace Enforcment with Enforcement.
994 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
997 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
998 (dw_val_node): Add val_symbolic_view.
999 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1000 (symview_upper_bound): New.
1001 (new_line_info_table): Initialize symviews_since_reset.
1002 (dwarf2out_source_line): Count symviews_since_reset and set
1003 symview_upper_bound.
1004 (dw_val_equal_p): Handle symview.
1005 (add_AT_symview): New.
1006 (print_dw_val): Handle symview.
1007 (attr_checksum, attr_checksum_ordered): Likewise.
1008 (same_dw_val_p, size_of_die): Likewise.
1009 (value_format, output_die): Likewise.
1010 (add_high_low_attributes): Use add_AT_symview for entry_view.
1011 (dwarf2out_finish): Reset symview_upper_bound, clear
1014 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1017 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1018 Add strict argument and use it.
1019 (rs6000_split_multireg_move): Update for new strict argument.
1020 (mem_operand_gpr): Disallow all non-offsettable addresses.
1021 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1023 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1026 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1027 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1028 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1031 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1032 decl, use remap_type if we want to use the type.
1034 2018-03-09 Martin Sebor <msebor@redhat.com>
1036 PR tree-optimization/84526
1037 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1039 (builtin_access::generic_overlap): Be prepared to handle non-array
1042 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1044 PR rtl-optimization/84682
1045 * lra-constraints.c (process_address_1): Check is_address flag
1046 for address constraints.
1047 (process_alt_operands): Likewise.
1048 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1049 preprocess_constraints.
1050 * recog.h (preprocess_constraints): Add oploc parameter.
1052 * recog.c (preprocess_constraints): Test address_operand for
1053 CT_ADDRESS constraints.
1055 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1058 * lra-assigns.c (assign_by_spills): Return a flag of reload
1059 assignment failure. Do not process the reload assignment
1060 failures. Do not spill other reload pseudos if they has the same
1062 (lra_assign): Add a return arg. Set up from the result of
1063 assign_by_spills call.
1064 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1065 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1066 usage_insns if it is not NULL.
1067 (spill_hard_reg_in_range): New function.
1068 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1069 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1070 function prototypes.
1071 (lra_assign): Change prototype.
1072 * lra.c (lra): Add code to deal with fails by splitting hard reg
1075 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1078 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1079 Accept complain bool parameter. Only emit errors if it is true.
1080 (arm_parse_cpu_option_name): Likewise.
1081 (arm_target_thumb_only): Adjust callers of the above.
1082 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1083 prototype to take a default true bool parameter.
1084 (arm_parse_arch_option_name): Likewise.
1086 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1087 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1091 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1092 * configure: Regenerate.
1093 * configure.ac ("linker --version-script option"): New.
1094 ("linker soname option"): New.
1096 2018-03-09 Richard Biener <rguenther@suse.de>
1098 PR tree-optimization/84775
1099 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1100 immediate uses of predicate stmts and mark them modified.
1103 PR tree-optimization/84178
1104 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1106 (version_loop_for_if_conversion): Delay update_ssa call.
1107 (tree_if_conversion): Delay update_ssa until after predicate
1110 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1113 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1114 when the function accesses prior frames.
1116 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1119 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1120 gen_llsym, otherwise call maybe_gen_llsym.
1123 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1124 has ',' character inside of it.
1126 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1129 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1130 as clobbering CC_REGNUM.
1132 2018-03-08 Richard Biener <rguenther@suse.de>
1135 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1136 (follow_copies_to_constant): Do not follow SSA names registered
1139 2018-03-08 Richard Biener <rguenther@suse.de>
1141 PR tree-optimization/84178
1142 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1144 (version_loop_for_if_conversion): Delay update_ssa call.
1145 (tree_if_conversion): Delay update_ssa until after predicate
1148 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1150 PR tree-optimization/84178
1151 * tree-if-conv.c (release_bb_predicate): Remove the
1152 the assertion that the stmts have NULL use_ops.
1153 Discard the statements, asserting that they haven't
1154 yet been added to a BB.
1156 2018-03-08 Richard Biener <rguenther@suse.de>
1158 PR tree-optimization/84746
1159 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1160 (phi_translate): Pass in destination ANTIC_OUT set.
1161 (phi_translate_1): Likewise. For a simplified result lookup
1162 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1163 (phi_translate_set): Adjust.
1164 (do_pre_regular_insertion): Likewise.
1165 (do_pre_partial_partial_insertion): Likewise.
1167 2018-03-08 Martin Liska <mliska@suse.cz>
1169 PR gcov-profile/84735
1170 * doc/gcov.texi: Document usage of profile files.
1171 * gcov-io.h: Document changes in the format.
1173 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1177 * dwarf2out.c (struct dw_line_info_table): Update comments for
1179 (FORCE_RESET_NEXT_VIEW): New.
1180 (FORCE_RESETTING_VIEW_P): New.
1181 (RESETTING_VIEW_P): Check for -1 too.
1182 (ZERO_VIEW_P): Likewise.
1183 (new_line_info_table): Force-reset next view.
1184 (dwarf2out_begin_function): Likewise.
1185 (dwarf2out_source_line): Simplify zero_view_p initialization.
1186 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1187 view directly. Omit view when omitting .loc at line 0.
1189 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1191 PR tree-optimization/84740
1192 * tree-switch-conversion.c (process_switch): Call build_constructors
1193 only if info.phi_count is non-zero.
1195 PR tree-optimization/84739
1196 * tree-tailcall.c (find_tail_calls): Check call arguments against
1197 DECL_ARGUMENTS (current_function_decl) rather than
1198 DECL_ARGUMENTS (func) when checking for tail recursion.
1200 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1202 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1203 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1204 Volker Reichelt's entry and add entries for people that perform
1205 GCC fuzzy testing and report numerous bugs.
1207 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1210 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1211 readonly data in sdata, if that is disabled.
1212 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1213 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1214 -mreadonly-in-sdata option.
1216 2018-03-07 Martin Sebor <msebor@redhat.com>
1218 PR tree-optimization/84468
1219 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1220 basic block when looking for nul assignment.
1222 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1225 * except.h (output_function_exception_table): Adjust prototype.
1226 * except.c (output_function_exception_table): Remove FNNAME parameter
1227 and add SECTION parameter. Ouput one part of the table at a time.
1228 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1229 the first part of the exception table and emit unwind directives.
1230 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1231 (i386_pe_seh_cold_init): Likewise.
1232 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1233 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1234 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1235 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1236 * config/i386/winnt.c: Include except.h.
1237 (struct seh_frame_state): Add reg_offset, after_prologue and
1238 in_cold_section fields.
1239 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1240 (i386_pe_seh_cold_init): New function.
1241 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1242 to seh->in_cold_section.
1243 (seh_emit_push): Record the offset of the push.
1244 (seh_emit_save): Record the offet of the save.
1245 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1246 Test seh->after_prologue to disregard the epilogue.
1247 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1248 (i386_pe_end_cold_function): New function.
1250 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1253 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1254 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1257 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1259 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1260 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1263 * multiple_target.c: Include tree-inline.h and intl.h.
1264 (expand_target_clones): Diagnose and fail if node->definition and
1265 !tree_versionable_function_p (node->decl).
1267 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1269 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1271 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1272 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1273 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1275 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1278 * combine.c (try_combine): Use reg_or_subregno instead of handling
1279 just paradoxical SUBREGs and REGs.
1281 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1283 * config/arc/arc.c (arc_finalize_pic): Remove function.
1284 (arc_must_save_register): We use single base PIC register, remove
1285 checks to save/restore the PIC register.
1286 (arc_expand_prologue): Likewise.
1287 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1289 (arc_verify_short): Likewise.
1290 (arc_attr_type): Likewise.
1291 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1292 (walk_stores): Likewise.
1293 (arc_address_cost): Make it static.
1294 (arc_verify_short): Likewise.
1295 (branch_dest): Likewise.
1296 (arc_attr_type): Likewise.
1297 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1298 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1299 (arc_final_prescan_insn): Remove inserting the nops due to
1300 hardware hazards. It is done in reorg step.
1301 (insn_length_variant_t): Remove.
1302 (insn_length_parameters_t): Likewise.
1303 (arc_insn_length_parameters): Likewise.
1304 (arc_get_insn_variants): Likewise.
1305 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1307 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1310 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1313 PR tree-optimization/84687
1314 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1316 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1318 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1320 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1321 Rename to ppc_speculation_barrier.
1322 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1323 __builtin_ppc_speculation_barrier.
1325 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1328 * combine.c (combine_simplify_rtx): Don't try to simplify if
1329 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1332 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1334 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1335 to 32 bytes when compiling for POWER9.
1337 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1340 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1341 regparm >= 3 with no arg reg available also for calls with
1342 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1345 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1347 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1349 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1352 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1354 2018-03-05 Richard Biener <rguenther@suse.de>
1356 PR tree-optimization/84486
1357 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1358 When inserting a __builtin_assume_aligned call set the LHS
1359 SSA name alignment info accordingly.
1361 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1363 PR tree-optimization/84114
1364 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1365 Avoid reassociation of FLOAT_MODE addition.
1367 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1369 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1370 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1371 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1372 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1373 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1374 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1375 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1377 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1378 __builtin_ia32_wbinvd): New builtins.
1379 (SPECIAL_ARGS2): New.
1380 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1381 (SPECIAL_ARGS2): New.
1382 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1383 (ix86_valid_target_attribute_inner_p): Ditto.
1384 (ix86_init_mmx_sse_builtins): Add special_args2.
1385 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1386 TARGET_WBNOINVD_P): New.
1387 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1388 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1389 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1390 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1391 * config/i386/pconfigintrin.h: New file.
1392 * config/i386/wbnoinvdintrin.h: Ditto.
1393 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1394 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1396 2018-03-05 Richard Biener <rguenther@suse.de>
1398 PR tree-optimization/84670
1399 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1401 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1402 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1403 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1404 doesn't grow if all successors (recursively) were visited at least
1407 2018-03-05 Richard Biener <rguenther@suse.de>
1409 PR tree-optimization/84650
1410 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1411 if executed in the loop pipeline.
1413 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1415 * doc/configfiles.texi (Configuration Files): Move info about
1416 conditionalizing $target-protos.h to...
1417 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1418 differs from $target-protos.h.
1420 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1421 Chung-Ju Wu <jasonwucj@gmail.com>
1423 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1424 * config/nds32/nds32-multiple.md (setmemsi): Define.
1425 * config/nds32/nds32-memory-manipulation.c
1426 (nds32_gen_dup_4_byte_to_word_value): New.
1427 (emit_setmem_word_loop): New.
1428 (emit_setmem_byte_loop): New.
1429 (nds32_expand_setmem_loop): New.
1430 (nds32_expand_setmem_loop_v3m): New.
1431 (nds32_expand_setmem_unroll): New.
1432 (nds32_expand_setmem): New.
1434 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1435 Chung-Ju Wu <jasonwucj@gmail.com>
1437 * config/nds32/nds32-memory-manipulation.c
1438 (nds32_emit_load_store): New.
1439 (nds32_emit_post_inc_load_store): New.
1440 (nds32_emit_mem_move): New.
1441 (nds32_emit_mem_move_block): New.
1442 (nds32_expand_movmemsi_loop_unknown_size): New.
1443 (nds32_expand_movmemsi_loop_known_size): New.
1444 (nds32_expand_movmemsi_loop): New.
1445 (nds32_expand_movmemsi_unroll): New.
1446 (nds32_expand_movmemqi): Rename ...
1447 (nds32_expand_movmemsi): ... to this.
1448 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1449 (movmemsi): ... to this.
1450 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1451 (nds32_expand_movmemsi): ... to this.
1453 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1454 Monk Chiang <sh.chiang04@gmail.com>
1455 Chung-Ju Wu <jasonwucj@gmail.com>
1457 * config/nds32/nds32-protos.h
1458 (nds32_expand_load_multiple): New arguments.
1459 (nds32_expand_store_multiple): Ditto.
1460 (nds32_valid_multiple_load_store): Rename ...
1461 (nds32_valid_multiple_load_store_p): ... to this.
1462 * config/nds32/nds32-memory-manipulation.c
1463 (nds32_expand_load_multiple): Refine implementation.
1464 (nds32_expand_store_multiple): Ditto.
1465 * config/nds32/nds32-multiple.md
1466 (load_multiple): Update nds32_expand_load_multiple interface.
1467 (store_multiple): Update nds32_expand_store_multiple interface.
1468 * config/nds32/nds32-predicates.c
1469 (nds32_valid_multiple_load_store): Rename ...
1470 (nds32_valid_multiple_load_store_p): ... to this and refine
1472 * config/nds32/predicates.md
1473 (nds32_load_multiple_and_update_address_operation): New predicate.
1474 (nds32_store_multiple_and_update_address_operation): New predicate.
1476 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1477 Chung-Ju Wu <jasonwucj@gmail.com>
1479 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1480 (combo): New attribute.
1481 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1483 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1485 * config/nds32/nds32.opt: Change -mcmodel= default value.
1487 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1488 Monk Chiang <sh.chiang04@gmail.com>
1489 Chung-Ju Wu <jasonwucj@gmail.com>
1491 * config/nds32/constants.md (unspec_element): New enum.
1492 * config/nds32/constraints.md (Umw): New constraint.
1493 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1494 * config/nds32/nds32-intrinsic.md: Likewise.
1495 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1496 (nds32_valid_smw_lwm_base_p): New.
1497 (nds32_output_smw_single_word): New.
1498 (nds32_output_lmw_single_word): New.
1499 (nds32_expand_unaligned_load): New.
1500 (nds32_expand_unaligned_store): New.
1501 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1502 (nds32_output_smw_single_word): Declare.
1503 (nds32_output_lmw_single_word): Declare.
1504 (nds32_expand_unaligned_load): Declare.
1505 (nds32_expand_unaligned_store): Declare.
1506 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1507 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1508 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1509 NDS32_BUILTIN_UASTORE_DW.
1510 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1513 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1514 Kito Cheng <kito.cheng@gmail.com>
1515 Chung-Ju Wu <jasonwucj@gmail.com>
1517 * config/nds32/nds32-intrinsic.c
1518 (nds32_expand_builtin_null_ftype_reg): Delete.
1519 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1520 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1521 (nds32_read_argument): New.
1522 (nds32_legitimize_target): Ditto.
1523 (nds32_legitimize_argument): Ditto.
1524 (nds32_check_constant_argument): Ditto.
1525 (nds32_expand_unop_builtin): Ditto.
1526 (nds32_expand_unopimm_builtin): Ditto.
1527 (nds32_expand_binop_builtin): Ditto.
1528 (nds32_builtin_decl_impl): Ditto.
1529 (builtin_description): Ditto.
1530 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1531 (nds32_init_builtins_impl): Ditto.
1532 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1533 (nds32_builtin_decl): New.
1534 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1535 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1537 2018-03-02 Jeff Law <law@redhat.com>
1539 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1540 (redundant_insn, fill_simple_delay_slots): Likewise.
1541 (fill_slots_from_thread): Likewise.
1542 * resource.c (mark_referenced_resources): Likewise.
1543 (mark_set_resources, find_dead_or_set_registers): Likewise.
1545 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1547 * substring-locations.h (format_warning_va): Formatting fix for
1549 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1551 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1552 * substring-locations.c: Include intl.h.
1553 (format_warning_va): Turned into small wrapper around
1554 format_warning_n_va, renamed to ...
1555 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1556 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1558 (format_warning_at_substring_n): New function.
1559 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1560 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1561 format_warning_at_substring with just a shorter name instead of
1562 const function pointer.
1563 (fmtwarn_n): New function.
1564 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1565 appropriate, get rid of all the fmtstr temporaries, move conditionals
1566 with G_() wrapped string literals directly into fmtwarn arguments,
1567 cast dir.len to (int), formatting fixes.
1569 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1571 * doc/invoke.texi: Remove "Cilk Plus" references.
1573 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1574 Richard Biener <rguenther@suse.de>
1577 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1578 for error or warning attributes if CALL_FROM_THUNK_P is set.
1581 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1584 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1585 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1588 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1589 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1591 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1592 instead of -1U in last predictors element's probability member.
1594 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1597 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1598 arguments if they are comparable.
1600 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1602 PR tree-optimization/84634
1603 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1604 masks and masked_loop_p with a single loop_masks, making sure it's
1605 null for bb vectorization.
1607 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1609 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1610 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1611 loop->force_vectorize to check whether there is no alias.
1613 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1616 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1618 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1620 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1621 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1622 instead of a loop around prev_real_insn.
1623 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1627 * config/i386/i386.c (ix86_print_operand): Use conditional
1628 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1631 2018-03-02 Richard Biener <rguenther@suse.de>
1633 PR tree-optimization/84427
1634 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1635 (bitmap_set_subtract_values): Rewrite to handle multiple
1638 (prune_clobbered_mems): Likewise.
1639 (phi_translate): Take edge instead of pred/phiblock.
1640 (phi_translate_1): Likewise.
1641 (phi_translate_set): Likewise. Insert all translated
1642 exprs for a value into the set, keeping possibly multiple
1643 expressions per value.
1644 (compute_antic_aux): Adjust for phi_translate changes.
1645 When intersecting union the expressions and prune those
1646 not in the final value set, keeping possibly multiple
1647 expressions per value. Do not use value-insertion
1648 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1649 all expressions. Add verification that the value-sets
1650 only shrink during iteration.
1651 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1652 (do_pre_regular_insertion): Likewise.
1653 (do_pre_partial_partial_insertion): Likewise.
1655 2018-03-02 Richard Biener <rguenther@suse.de>
1658 * config/darwin.c (saved_debug_info_level): New static global.
1659 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1660 (darwin_asm_lto_end): Restore debug info generation settings.
1662 2018-03-01 Martin Liska <mliska@suse.cz>
1665 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1668 2018-03-01 Richard Biener <rguenther@suse.de>
1671 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1672 type annotation with LTO when debug was disabled at compile-time.
1674 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1676 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1678 (mips_final_postscan_insn): Likewise.
1680 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1682 PR rtl-optimization/84528
1683 * alias.c (init_alias_target): Add commentary.
1684 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1685 a unique base value if the frame pointer is not eliminated
1686 to the stack pointer.
1688 2018-03-01 Tom de Vries <tom@codesourcery.com>
1690 PR rtl-optimization/83327
1691 * lra-int.h (hard_regs_spilled_into): Declare.
1692 * lra.c (hard_regs_spilled_into): Define.
1693 (init_reg_info): Init hard_regs_spilled_into.
1694 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1695 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1696 (process_bb_lives): Handle hard_regs_spilled_into.
1697 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1698 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1700 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1702 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1703 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1704 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1705 * config/rs6000/aix72.h: New file.
1707 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1709 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1710 instead of warning_at with conditional singular and plural messages
1714 * stor-layout.c (update_alignment_for_field): For
1715 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1716 && !DECL_PACKED (field), do the alignment update, just use
1717 only desired_align instead of MAX (type_align, desired_align)
1719 (place_field): Don't do known_align < desired_align handling
1720 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1721 is non-NULL, instead do it after rli->prev_field handling and
1722 only if not within a bitfield word. For DECL_PACKED (field)
1723 use type_align of BITS_PER_UNIT.
1725 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1727 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1728 superfluous parentheses and trailing spaces.
1730 2018-02-28 Richard Biener <rguenther@suse.de>
1732 PR tree-optimization/84584
1733 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1734 SCoPs with fake exit edge.
1736 2018-02-28 Martin Liska <mliska@suse.cz>
1739 * timevar.c (timer::print): Fix format to properly print 100%
1742 2018-02-28 Richard Biener <rguenther@suse.de>
1745 * genmatch.c (capture_info::walk_match): Do not mark
1746 captured expressions without operands as expr_p given
1747 they act more like predicates and should be subject to
1748 "lost tail" side-effect preserving.
1750 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1752 PR rtl-optimization/81611
1753 * auto-inc-dec.c (attempt_change): Move dead note from
1754 mem_insn if it's the next use of regno
1755 (find_address): Take address use of reg holding
1756 non-incremented value. Add parm to limit search to the named
1758 (merge_in_block): Attempt to use a mem insn that is the next
1759 use of the original regno.
1761 2018-02-27 Martin Sebor <msebor@redhat.com>
1764 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1765 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1767 2018-02-27 Martin Sebor <msebor@redhat.com>
1769 PR translation/84207
1770 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1771 n argument to unsigned HOST_WIDE_INT.
1772 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1773 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1774 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1775 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1777 2018-02-27 Richard Biener <rguenther@suse.de>
1779 PR tree-optimization/84512
1780 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1781 Do not use the estimate returned from record_stmt_cost for
1782 the scalar iteration cost but sum properly using add_stmt_cost.
1784 2018-02-27 Richard Biener <rguenther@suse.de>
1786 PR tree-optimization/84466
1787 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1788 Adjust last change to less strictly validate use operands.
1790 2018-02-27 Martin Liska <mliska@suse.cz>
1792 PR gcov-profile/84548
1793 * gcov.c (process_file): Allow partial overlap and consider it
1794 also as group functions.
1795 (output_lines): Properly calculate range of lines for a group.
1797 2018-02-27 Martin Liska <mliska@suse.cz>
1799 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1800 'ggc' suffixes. Change first column width.
1801 (timer::print): Fix formatting of the column.
1803 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1805 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1806 preserve inline entry blocks for the sake of debug inline
1807 entry point markers alone.
1808 (remove_unused_locals): Suggest in comments a better place to
1809 force the preservation of inline entry blocks that are
1810 otherwise unused, but do not preserve them.
1812 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1814 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1816 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1819 * config/i386/constraints.md (Bs): Replace
1820 ix86_indirect_branch_register with
1821 TARGET_INDIRECT_BRANCH_REGISTER.
1823 * config/i386/i386.md (indirect_jump): Likewise.
1824 (tablejump): Likewise.
1825 (*sibcall_memory): Likewise.
1826 (*sibcall_value_memory): Likewise.
1827 Peepholes of indirect call and jump via memory: Likewise.
1828 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1829 (*sibcall_value_GOT_32): Likewise.
1830 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1831 (GOT_memory_operand): Likewise.
1832 (call_insn_operand): Likewise.
1833 (sibcall_insn_operand): Likewise.
1834 (GOT32_symbol_operand): Likewise.
1835 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1837 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1839 PR rtl-optimization/83496
1840 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1841 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1842 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1843 redundant insn, if any.
1844 (relax_delay_slots): Likewise.
1845 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1847 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1849 PR tree-optimization/83965
1850 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1851 that grouped statements are part of a reduction chain. Return
1852 true if the statement is not marked as a reduction itself but
1854 (vect_recog_dot_prod_pattern): Don't check whether the statement
1855 is part of a group here.
1856 (vect_recog_sad_pattern): Likewise.
1857 (vect_recog_widen_sum_pattern): Likewise.
1859 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1862 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1864 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1867 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1869 (ix86_output_indirect_function_return): New prototype.
1870 (ix86_split_simple_return_pop_internal): Likewise.
1871 * config/i386/i386.c (indirect_return_via_cx): New.
1872 (indirect_return_via_cx_bnd): Likewise.
1873 (indirect_thunk_name): Handle return va CX_REG.
1874 (output_indirect_thunk_function): Create alias for
1875 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1876 (ix86_output_indirect_jmp): Remove the bool argument.
1877 (ix86_output_indirect_function_return): New function.
1878 (ix86_split_simple_return_pop_internal): Likewise.
1879 * config/i386/i386.md (*indirect_jump): Don't pass false
1880 to ix86_output_indirect_jmp.
1881 (*tablejump_1): Likewise.
1882 (simple_return_pop_internal): Change it to define_insn_and_split.
1883 Call ix86_split_simple_return_pop_internal to split it for
1885 (simple_return_indirect_internal): Call
1886 ix86_output_indirect_function_return instead of
1887 ix86_output_indirect_jmp.
1889 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1892 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1893 memset and value initialization afterwards.
1895 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1897 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1899 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1902 * common/config/aarch64/aarch64-common.c
1903 (aarch_option_optimization_table[]): Switch
1904 off fomit-frame-pointer
1906 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1907 Chung-Ju Wu <jasonwucj@gmail.com>
1909 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1911 (store_multiple): Ditto.
1913 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1915 * config.gcc: Add --with-cpu support for nds32 target.
1916 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1917 * config/nds32/nds32.opt: Add -mcpu= option.
1919 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1921 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1922 isel=yes): Warn for these deprecated options.
1924 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1926 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1927 ISA_2_5_MASKS_EMBEDDED.
1929 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1931 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1932 p->max as pointers rather than using iterative_hash_expr.
1934 2018-02-23 Carl Love <cel@us.ibm.com>
1936 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1937 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1939 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1940 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1941 P8V_BUILTIN_VEC_VUNSIGNED2.
1943 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1946 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1947 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1949 * lra-constraints.c (curr_insn_transform): Set up
1950 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1952 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1953 * lra-eliminations.c (spill_pseudos): Ditto.
1954 (process_insn_for_elimination): Ditto.
1955 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1956 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1957 LRA_NON_CLOBBERED_ALT.
1959 2018-02-22 Martin Sebor <msebor@redhat.com>
1961 PR tree-optimization/84480
1962 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1963 to maybe_diag_stxncpy_trunc. Call it.
1964 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1965 from gimple_fold_builtin_strcpy. Print inlining stack.
1966 (handle_builtin_stxncpy): Print inlining stack.
1967 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1969 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1972 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1973 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1974 and -fcheck-pointer-bounds are used together.
1975 (indirect_thunk_prefix): New enum.
1976 (indirect_thunk_need_prefix): New function.
1977 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1978 "_nt" instead of "_bnd" for NOTRACK prefix.
1979 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1980 (output_indirect_thunk_function): Likewise.
1982 (ix86_code_end): Update output_indirect_thunk_function calls.
1983 (ix86_output_indirect_branch_via_reg): Replace
1984 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1985 (ix86_output_indirect_branch_via_push): Likewise.
1986 (ix86_output_function_return): Likewise.
1987 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1988 incompatible with -fcf-protection=branch and
1989 -fcheck-pointer-bounds.
1991 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1994 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1995 Change gcc_assert call to output_operand_lossage.
1997 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1999 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2001 2018-02-22 DJ Delorie <dj@redhat.com>
2002 Sebastian Perta <sebastian.perta@renesas.com>
2003 Oleg Endo <olegendo@gcc.gnu.org>
2005 * config/rx/rx.c (rx_rtx_costs): New function.
2006 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2008 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2010 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2012 2018-02-22 Martin Liska <mliska@suse.cz>
2015 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2016 Add "native" as a possible value.
2018 2018-02-22 Martin Liska <mliska@suse.cz>
2021 * config/i386/i386.c (ix86_option_override_internal):
2022 Add "native" as a possible value for -march and -mtune.
2024 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2027 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2028 to all type variants.
2030 PR tree-optimization/84503
2031 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2032 width as info->bitpos + info->bitsize - start.
2033 (merged_store_group::merge_overlapping): Simplify width computation.
2034 (check_no_overlap): New function.
2035 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2036 start + width and last_order of the group, fail if check_no_overlap
2038 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2039 to group if check_no_overlap fails.
2041 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2043 * config/rs6000/altivec.md: Delete contraint arguments to
2044 define_expand, define_split, and define_peephole2, and in
2045 define_insn_and_split if always unused.
2046 * config/rs6000/darwin.md: Ditto.
2047 * config/rs6000/dfp.md: Ditto.
2048 * config/rs6000/rs6000.md: Ditto.
2049 * config/rs6000/sync.md: Ditto.
2050 * config/rs6000/vector.md: Ditto.
2051 * config/rs6000/vsx.md: Ditto.
2053 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2055 * config/rs6000/altivec.md: Write output control strings as braced
2056 blocks instead of double-quoted strings.
2057 * config/rs6000/darwin.md: Ditto.
2058 * config/rs6000/rs6000.md: Ditto.
2059 * config/rs6000/vector.md: Ditto.
2060 * config/rs6000/vsx.md: Ditto.
2062 2018-02-21 Jason Merrill <jason@redhat.com>
2064 PR c++/84314 - ICE with templates and fastcall attribute.
2065 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2067 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2069 * ipa-cp.c (determine_versionability): Fix comment typos.
2071 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2074 * ipa-cp.c (determine_versionability): Do not version functions caling
2077 2018-02-21 Martin Liska <mliska@suse.cz>
2080 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2081 Add "native" as a possible value.
2082 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2083 the macro when native cpu detection is available.
2085 2018-02-21 Martin Liska <mliska@suse.cz>
2088 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2089 Add "native" as a possible value.
2090 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2091 when native cpu detection is available.
2093 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2094 Martin Sebor <msebor@redhat.com>
2096 PR tree-optimization/84478
2097 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2099 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2100 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2101 support which is conservatively correct, for 2 only stay conservative
2102 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2103 argument to the 2 argument get_range_strlen, adjust 6 arg
2104 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2106 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2107 (gimple_fold_builtin_strlen): Pass true as last argument to
2110 2018-02-20 Martin Sebor <msebor@redhat.com>
2113 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2114 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2115 (builtin_memref::builtin_memref): Factor out parts into
2116 set_base_and_offset and call it.
2118 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2121 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2122 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2123 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2124 search at the associated MODE_INT.
2126 2018-02-20 Jeff Law <law@redhat.com>
2129 PR tree-optimization/81592
2131 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2132 for range data rather than using global data.
2133 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2134 range data rather than using global data.
2135 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2136 pass it to children as needed.
2137 (struct directive::fmtresult): Similarly.
2138 (struct directive::set_width): Similarly.
2139 (struct directive::set_precision): Similarly.
2140 (format_integer, format_directive, parse_directive): Similarly.
2141 (format_none): Accept unnamed vr_values parameter.
2142 (format_percent, format_floating, format_character): Similarly.
2143 (format_string, format_plain): Similarly.
2144 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2145 the EVRP range analyzer for range data rather than using global data.
2146 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2147 gimple-ssa-evrp-analyze.h
2148 (class sprintf_dom_walker): Add after_dom_children member function.
2149 Add evrp_range_analyzer member.
2150 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2151 range analyzer as needed.
2152 (sprintf_dom_walker::after_dom_children): New member function.
2153 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2155 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2156 (evrp_range_analyzer::pop_to_marker): Likewise.
2158 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2160 PR tree-optimization/84419
2161 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2162 with the required type if its current type is compatible but
2165 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2168 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2169 after vectorization.
2171 2018-02-20 Martin Liska <mliska@suse.cz>
2174 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2175 possible values if we don't have a hint.
2177 2018-02-20 Martin Liska <mliska@suse.cz>
2181 * final.c (shorten_branches): Build align_tab array with one
2183 * opts.c (finish_options): Add alignment option limit check.
2184 (MAX_CODE_ALIGN): Likewise.
2185 (MAX_CODE_ALIGN_VALUE): Likewise.
2186 * doc/invoke.texi: Document maximum allowed option value for
2187 all -falign-* options.
2189 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2192 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2193 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2194 * var-tracking.c (emit_note_insn_var_location): Remove all references
2195 to NOTE_INSN_CALL_ARG_LOCATION.
2196 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2197 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2198 Use copy_rtx_if_shared.
2199 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2200 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2201 (dwarf2out_var_location): Remove handling of
2202 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2204 * final.c (final_scan_insn): Remove all references to
2205 NOTE_INSN_CALL_ARG_LOCATION.
2206 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2207 before dumping final insns.
2208 * except.c (emit_note_eh_region_end): Remove all references to
2209 NOTE_INSN_CALL_ARG_LOCATION.
2210 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2211 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2212 * config/arc/arc.c (hwloop_optimize): Likewise.
2213 * config/arm/arm.c (create_fix_barrier): Likewise.
2214 * config/s390/s390.c (s390_chunkify_start): Likewise.
2215 * config/sh/sh.c (find_barrier): Likewise.
2216 * config/i386/i386.c (rest_of_insert_endbranch,
2217 ix86_seh_fixup_eh_fallthru): Likewise.
2218 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2219 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2220 * config/frv/frv.c (frv_function_prologue): Likewise.
2221 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2223 (note_outside_basic_block_p): Remove all references to
2224 NOTE_INSN_CALL_ARG_LOCATION.
2225 * gengtype.c (adjust_field_rtx_def): Likewise.
2226 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2228 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2229 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2232 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2235 PR tree-optimization/84452
2236 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2237 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2240 2018-02-19 Martin Liska <mliska@suse.cz>
2243 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2245 2018-02-19 Martin Liska <mliska@suse.cz>
2246 Richard Sandiford <richard.sandiford@linaro.org>
2248 PR tree-optimization/82491
2249 * gimple-fold.c (get_base_constructor): Make earlier bail out
2252 2018-02-19 Carl Love <cel@us.ibm.com>
2254 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2255 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2257 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2258 P8V_BUILTIN_VEC_NEG.
2260 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2262 * config/rl78/rl78.md (movdf): New define expand.
2264 2018-02-19 Martin Liska <mliska@suse.cz>
2267 * doc/invoke.texi: Fix typo.
2268 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2270 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2272 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2273 handle rs6000_single_float and rs6000_double_float specially for
2276 2018-02-16 Jeff Law <law@redhat.com>
2278 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2281 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2284 * ipa-inline.c (inline_small_functions): Fix a typo.
2286 2018-02-16 Nathan Sidwell <nathan@acm.org>
2288 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2290 2018-02-16 Carl Love <cel@us.ibm.com>
2292 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2293 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2294 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2295 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2296 expansion to P8V_BUILTIN_VEC_FLOAT2.
2298 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2300 PR rtl-optimization/70023
2301 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2302 src_regno into account.
2304 2018-02-16 Carl Love <cel@us.ibm.com>
2306 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2307 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2308 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2309 * config/rs6000/rs6000.c: Remove case statements for
2310 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2311 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2312 and P9V_BUILTIN_VEC_VINSERT4B.
2313 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2314 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2315 * config/rs6000/vsx.md:
2316 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2319 2018-02-16 Carl Love <cel@us.ibm.com>
2321 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2323 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2325 * config/rs6000/rs6000-c.c: Add the definitions for
2326 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2327 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2328 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2329 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2330 definition for insert4b and define insn *insert3b_internal.
2331 * doc/extend.texi: Add documentation for vec_extract4b.
2333 2018-02-16 Nathan Sidwell <nathan@acm.org>
2335 * doc/extend.texi (Backwards Compatibility): Mention friend
2336 injection. Note for-scope is deprecated.
2337 * doc/invoke.texi (-ffriend-injection): Deprecate.
2339 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2341 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2342 that moved to I2, also allow destinations that are a paradoxical
2343 subreg (instead of a normal reg).
2345 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2348 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2351 2018-02-16 Richard Biener <rguenther@suse.de>
2353 PR tree-optimization/84037
2354 PR tree-optimization/84016
2356 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2357 Adjust vec_construct for the fact we need additional higher latency
2358 128bit inserts for AVX256 and AVX512 vector builds.
2359 (ix86_add_stmt_cost): Scale vector construction cost for
2362 2018-02-16 Richard Biener <rguenther@suse.de>
2364 PR tree-optimization/84417
2365 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2366 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2367 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2369 2018-02-16 Martin Liska <mliska@suse.cz>
2372 * internal-fn.def (ASAN_CHECK): Set proper flags.
2373 (ASAN_MARK): Likewise.
2375 2018-02-16 Julia Koval <julia.koval@intel.com>
2377 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2378 from PTA_CANNONLAKE.
2380 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2383 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2384 Use ++iter rather than iter++ for std::list iterators.
2385 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2386 defer deleting them until all nodes in the forest are processed. Do
2387 free even leaf nodes. Change to_process into auto_vec.
2390 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2391 * vec.h (vec_default_construct): Use memset instead of placement new
2392 if BROKEN_VALUE_INITIALIZATION is defined.
2393 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2394 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2397 PR rtl-optimization/83723
2398 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2399 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2400 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2402 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2404 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2406 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2408 PR rtl-optimization/81443
2409 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2410 from inner REGs to paradoxical SUBREGs.
2412 2018-02-16 Richard Biener <rguenther@suse.de>
2414 PR tree-optimization/84399
2415 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2416 For operands we can analyze at their definition make sure we can
2417 analyze them at each use as well.
2419 2018-02-16 Richard Biener <rguenther@suse.de>
2421 PR tree-optimization/84190
2422 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2423 volatile accesses if the decl isn't volatile.
2425 2018-02-15 Jason Merrill <jason@redhat.com>
2427 PR c++/84314 - ICE with templates and fastcall attribute.
2428 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2429 TYPE_CANONICAL on an existing type.
2431 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2433 PR tree-optimization/84383
2434 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2435 dstoff nor call operand_equal_p if dstbase is NULL.
2437 PR tree-optimization/84334
2438 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2439 also a CONSTANT_CLASS_P, punt.
2441 2018-02-14 Jim Wilson <jimw@sifive.com>
2443 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2444 first SMALL_OPERAND check. New local min_second_step. Move assert
2445 to where locals are set. Add TARGET_RVC support.
2446 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2448 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2450 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2452 2018-02-14 Martin Sebor <msebor@redhat.com>
2454 PR tree-optimization/83698
2455 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2456 arrays constrain the offset range to their bounds.
2457 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2458 (builtin_access::overlap): Avoid setting the size of overlap if it's
2460 (maybe_diag_overlap): Also consider arrays when deciding what values
2461 of offsets to include in diagnostics.
2463 2018-02-14 Martin Sebor <msebor@redhat.com>
2466 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2467 that correspond to the kind of a declaration.
2469 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2472 * config/pa/pa.md: Load address of PIC label using the linkage table
2473 if the label is nonlocal.
2475 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2477 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2478 warning message if user requests -maltivec=be.
2479 * doc/invoke.texi: Document deprecation of -maltivec=be.
2481 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2484 * config/rs6000/rs6000-c.c: Update definitions for
2485 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2486 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2488 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2491 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2492 add _get_ssp intrinsics. Remove argument from
2493 __builtin_ia32_rdssp[d|q].
2494 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2495 * config/i386/i386-builtin.def: Remove argument from
2496 __builtin_ia32_rdssp[d|q].
2497 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2498 ix86_expand_special_args_builtin for _rdssp[d|q].
2499 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2500 Clear register before usage.
2501 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2502 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2504 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2506 PR tree-optimization/84357
2507 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2508 operand 1 of an ARRAY_REF too.
2510 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2513 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2514 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2516 (set_of_reg): New struct.
2517 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2518 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2519 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2521 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2522 Split into bitclr, bitset, bitinvert patterns if appropriate.
2523 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2524 use rx_fuse_in_memory_bitop.
2525 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2526 to named insn, correct maximum insn length.
2528 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2531 * machmode.def: Define a complex mode for PARTIAL_INT.
2532 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2534 * doc/rtl.texi: Document CSPImode.
2535 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2537 (msp430_hard_regno_nregs_with_padding): Likewise.
2539 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2542 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2544 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2546 PR rtl-optimization/84169
2547 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2548 we generated a parallel as new i3 and we split that to new i2 and i3
2549 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2550 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2551 those to i2, not i1. Partially rewrite this scan code.
2553 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2556 * stor-layout.c (place_field): For variable length fields, adjust
2557 offset_align afterwards not just based on the field's alignment,
2558 but also on the size.
2561 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2562 of exps and logs in the use_exp2 case.
2564 2018-02-13 Jeff Law <law@redhat.com>
2566 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2569 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2572 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2576 * common.opt (gas-loc-support, gas-locview-support): New.
2577 (ginline-points, ginternal-reset-location-views): New.
2578 * doc/invoke.texi: Document them. Use @itemx where intended.
2579 (gvariable-location-views): Adjust.
2580 * target.def (reset_location_view): New.
2581 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2582 (TARGET_RESET_LOCATION_VIEW): New.
2583 * doc/tm.texi: Rebuilt.
2584 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2585 (dwarf2out_default_as_locview_support): New.
2586 (output_asm_line_debug_info): Use option variables.
2587 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2588 (output_loc_list): Likewise.
2589 (add_high_low_attributes): Check option variables.
2590 Don't output entry view attribute in strict mode.
2591 (gen_inlined_subroutine_die): Check option variables.
2592 (dwarf2out_inline_entry): Likewise.
2593 (init_sections_and_labels): Likewise.
2594 (dwarf2out_early_finish): Likewise.
2595 (maybe_reset_location_view): New, from...
2596 (dwarf2out_var_location): ... here. Call it.
2597 * debug.h (dwarf2out_default_as_loc_support): Declare.
2598 (dwarf2out_default_as_locview_support): Declare.
2599 * hooks.c (hook_int_rtx_insn_0): New.
2600 * hooks.h (hook_int_rtx_insn_0): Declare.
2601 * toplev.c (process_options): Take -gas-loc-support and
2602 -gas-locview-support from dwarf2out. Enable
2603 -gvariable-location-views by default only with locview
2604 assembler support. Enable -ginternal-reset-location-views by
2605 default only if the target defines the corresponding hook.
2606 Enable -ginline-points by default if location views are
2607 enabled; force it disabled if statement frontiers are
2609 * tree-inline.c (expand_call_inline): Check option variables.
2610 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2612 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2614 PR tree-optimization/84321
2615 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2616 handling. Also check whether the anti-range contains any values
2617 that satisfy the mask; switch to a VR_RANGE if not.
2619 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2622 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2624 2018-02-13 Martin Jambor <mjambor@suse.cz>
2627 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2628 of call statements, also set location of a load to a temporary.
2630 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2632 * config/rl78/rl78.c (add_vector_labels): New function.
2633 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2634 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2635 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2636 which checks that no arguments are passed.
2637 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2638 * doc/extend.texi: Documentation for the new attribute.
2640 2018-02-13 Andreas Schwab <schwab@suse.de>
2642 * config/riscv/linux.h (CPP_SPEC): Define.
2644 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2647 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2648 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2649 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2650 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2651 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2652 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2653 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2654 temporarily for AES and PCLMUL builtins.
2656 PR tree-optimization/84339
2657 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2658 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2662 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2663 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2664 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2666 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2668 * omp-simd-clone.h: New file.
2669 * omp-simd-clone.c: Include omp-simd-clone.h.
2670 (expand_simd_clones): No longer static.
2671 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2672 cgraph.h and omp-simd-clone.h.
2673 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2674 (vect_recog_widen_shift_pattern): Formatting fix.
2675 (vect_pattern_recog_1): Don't check optab for calls.
2678 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2679 operands[2] into a REG before using gen_lowpart on it.
2681 2018-02-12 Jeff Law <law@redhat.com>
2684 * config/sh/sh.c (find_barrier): Consider a sibling call
2687 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2688 successfully back substituting a reg.
2690 2018-02-12 Richard Biener <rguenther@suse.de>
2692 PR tree-optimization/84037
2693 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2694 parameter, move visited init to caller.
2695 (vect_slp_analyze_operations): Separate cost from validity
2696 check, initialize visited once for all instances.
2697 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2699 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2701 (vect_model_promotion_demotion_cost): Likewise.
2702 (vectorizable_bswap): Guard cost modeling with !slp_node
2703 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2705 (vectorizable_call): Likewise.
2706 (vectorizable_conversion): Likewise.
2707 (vectorizable_assignment): Likewise.
2708 (vectorizable_shift): Likewise.
2709 (vectorizable_operation): Likewise.
2710 (vectorizable_store): Likewise.
2711 (vectorizable_load): Likewise.
2712 (vectorizable_condition): Likewise.
2713 (vectorizable_comparison): Likewise.
2715 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2718 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2719 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2720 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2722 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2725 * params.def (inline-min-speedup): Increase from 8 to 15.
2726 (max-inline-insns-auto): Decrease from 40 to 30.
2727 * ipa-split.c (consider_split): Add some buffer for function to
2728 be considered inlining candidate.
2729 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2732 2018-02-12 Richard Biener <rguenther@suse.de>
2734 PR tree-optimization/84037
2735 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2736 matched stmts if we cannot swap the non-matched ones.
2738 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2740 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2741 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2742 _mm_maskz_scalef_round_ss): New intrinsics.
2743 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2744 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2745 __builtin_ia32_scalefss_round): Remove.
2746 (__builtin_ia32_scalefsd_mask_round,
2747 __builtin_ia32_scalefss_mask_round): New intrinsics.
2748 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2749 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2750 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2751 "<round_constraint>")): Changed to ...
2752 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2753 "<round_scalar_constraint>")): ... this.
2754 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2755 %0, %1, %2<round_op3>}"): Changed to ...
2756 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2757 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2758 %2<round_scalar_mask_op3>}"): ... this.
2759 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2761 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2763 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2764 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2765 (_mm_maskz_sqrt_round_ss): New intrinsics.
2766 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2767 (__builtin_ia32_sqrtsd_mask_round)
2768 (__builtin_ia32_sqrtss_mask_round): New builtins.
2769 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2770 (__builtin_ia32_sqrtss_round): Remove.
2771 (__builtin_ia32_sqrtsd_mask_round)
2772 (__builtin_ia32_sqrtss_mask_round): New builtins.
2773 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2774 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2775 ((match_operand:VF_128 1 "vector_operand"
2776 "xBm,<round_constraint>")): Changed to ...
2777 ((match_operand:VF_128 1 "vector_operand"
2778 "xBm,<round_scalar_constraint>")): ... this.
2779 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2780 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2781 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2782 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2783 %<iptr>1<round_scalar_mask_op3>}): ... this.
2784 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2785 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2787 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2790 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2791 Cast vec_cmpeq result to correct type.
2792 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2793 Cast vec_cmpgt result to correct type.
2795 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2797 * final.c (final_scan_insn_1): Renamed from...
2798 (final_scan_insn): ... this. New wrapper, to recover
2799 seen from the outermost call in recursive ones.
2800 * config/sparc/sparc.c (output_return): Drop seen from call.
2801 (output_sibcall): Likewise.
2802 * config/visium/visium.c (output_branch): Likewise.
2804 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2806 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2809 2018-02-10 Alan Modra <amodra@gmail.com>
2812 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2813 Specify LR as an input.
2815 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2818 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2819 remove_member_access_dummy_vars): New functions.
2820 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2821 lower_omp_1, execute_lower_omp): Use them.
2823 PR rtl-optimization/84308
2824 * shrink-wrap.c (spread_components): Release todo vector.
2826 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2828 PR rtl-optimization/57193
2829 * ira-color.c (struct allocno_color_data): Add member
2830 conflict_allocno_hard_prefs.
2831 (update_conflict_allocno_hard_prefs): New.
2832 (bucket_allocno_compare_func): Add a preference based on
2833 conflict_allocno_hard_prefs.
2834 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2835 (color_allocnos): Remove a dead code. Initiate
2836 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2838 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2841 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2842 constraint from =wa to wa. Avoid a subreg on the output operand,
2843 instead use a pseudo and subreg it in a move.
2844 (p9_xxbrd_<mode>): Changed to ...
2845 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2846 (p9_xxbrd_v2df): New expander.
2847 (p9_xxbrw_<mode>): Changed to ...
2848 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2849 (p9_xxbrw_v4sf): New expander.
2851 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2853 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2855 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2858 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2859 multiply in 32-bit mode.
2860 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2861 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2864 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2866 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2867 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2868 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2869 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2871 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2874 * dwarf2out.c (is_trivial_indirect_ref): New function.
2875 (dwarf2out_late_global_decl): Do not generate a location
2876 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2877 and that are not defined in the current unit.
2879 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2881 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2882 instead of a libcall for UNORDERED.
2884 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2887 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2888 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2890 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2893 * config/s390/s390.c (s390_set_current_function): Invoke
2894 s390_indirect_branch_settings also if fndecl didn't change.
2896 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2898 * config/rs6000/rs6000.md (blockage): Set length to zero.
2900 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2902 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2904 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2907 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2908 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2912 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2913 PARALLEL incoming that failed vt_get_decl_and_offset check.
2916 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2917 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2919 (get_variable_section): For decls in named .bss* sections pass true as
2920 second argument to bss_initializer_p.
2922 2018-02-09 Marek Polacek <polacek@redhat.com>
2923 Jakub Jelinek <jakub@redhat.com>
2926 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2927 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2928 Sync some changes from cxx_fold_indirect_ref.
2930 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2932 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2934 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2935 (BLOCK_INLINE_ENTRY_LABEL): New.
2936 (dwarf2out_var_location): Disregard inline entry markers.
2937 (inline_entry_data): New struct.
2938 (inline_entry_data_hasher): New hashtable type.
2939 (inline_entry_data_hasher::hash): New.
2940 (inline_entry_data_hasher::equal): New.
2941 (inline_entry_data_table): New variable.
2942 (add_high_low_attributes): Add DW_AT_entry_pc and
2943 DW_AT_GNU_entry_view attributes if a pending entry is found
2944 in inline_entry_data_table. Add old entry_pc attribute only
2945 if debug nonbinding markers are disabled.
2946 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2947 markers are enabled.
2948 (block_within_block_p, dwarf2out_inline_entry): New.
2949 (dwarf2out_finish): Check that no entries remained in
2950 inline_entry_data_table.
2951 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2952 (final_scan_insn, notice_source_line): Likewise.
2953 (rest_of_clean_state): Skip inline entry markers.
2954 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2956 * gimple.c (gimple_build_debug_inline_entry): New.
2957 * gimple.h (enum gimple_debug_subcode): Add
2958 GIMPLE_DEBUG_INLINE_ENTRY.
2959 (gimple_build_debug_inline_entry): Declare.
2960 (gimple_debug_inline_entry_p): New.
2961 (gimple_debug_nonbind_marker_p): Adjust.
2962 * insn-notes.def (INLINE_ENTRY): New.
2963 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2964 inline entry marker notes.
2965 (print_insn): Likewise.
2966 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2967 (INSN_DEBUG_MARKER_KIND): Likewise.
2968 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2969 * tree-inline.c (expand_call_inline): Build and insert
2970 debug_inline_entry stmt.
2971 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2972 inline entry blocks early, if nonbind markers are enabled.
2973 (dump_scope_block): Dump fragment info.
2974 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2975 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2976 (gimple_build_debug_inline_entry): New.
2977 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2978 Enable/disable inline entry points too.
2979 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2980 (DEBUG_INSN): Describe inline entry markers.
2982 * common.opt (gvariable-location-views): New.
2983 (gvariable-location-views=incompat5): New.
2984 * config.in: Rebuilt.
2985 * configure: Rebuilt.
2986 * configure.ac: Test assembler for view support.
2987 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2988 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2989 * dwarf2out.c (var_loc_view): New typedef.
2990 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2991 (dwarf2out_locviews_in_attribute): New.
2992 (dwarf2out_locviews_in_loclist): New.
2993 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2994 (enum dw_line_info_opcode): Add LI_adv_address.
2995 (struct dw_line_info_table): Add view.
2996 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2997 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2998 (zero_view_p): New variable.
2999 (ZERO_VIEW_P): New macro.
3000 (output_asm_line_debug_info): New.
3001 (struct var_loc_node): Add view.
3002 (add_AT_view_list, AT_loc_list): New.
3003 (add_var_loc_to_decl): Add view param. Test it against last.
3004 (new_loc_list): Add view params. Record them.
3005 (AT_loc_list_ptr): Handle loc and view lists.
3006 (view_list_to_loc_list_val_node): New.
3007 (print_dw_val): Handle dw_val_class_view_list.
3008 (size_of_die): Likewise.
3009 (value_format): Likewise.
3010 (loc_list_has_views): New.
3011 (gen_llsym): Set vl_symbol too.
3012 (maybe_gen_llsym, skip_loc_list_entry): New.
3013 (dwarf2out_maybe_output_loclist_view_pair): New.
3014 (output_loc_list): Output view list or entries too.
3015 (output_view_list_offset): New.
3016 (output_die): Handle dw_val_class_view_list.
3017 (output_dwarf_version): New.
3018 (output_compilation_unit_header): Use it.
3019 (output_skeleton_debug_sections): Likewise.
3020 (output_rnglists, output_line_info): Likewise.
3021 (output_pubnames, output_aranges): Update version comments.
3022 (output_one_line_info_table): Output view numbers in asm comments.
3023 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3024 Call maybe_gen_llsym.
3025 (loc_list_from_tree_1): Adjust.
3026 (add_AT_location_description): Create view list attribute if
3027 needed, check it's absent otherwise.
3028 (convert_cfa_to_fb_loc_list): Adjust.
3029 (maybe_emit_file): Call output_asm_line_debug_info for test.
3030 (dwarf2out_var_location): Reset views as needed. Precompute
3031 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3032 attribute. Set view.
3033 (new_line_info_table): Reset next view.
3034 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3035 (dwarf2out_source_line): Likewise. Output view resets and labels to
3036 the assembler, or select appropriate line info opcodes.
3037 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3038 (optimize_string_length): Catch it. Adjust.
3039 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3040 dw_val_class_view_list, and remove it if no longer needed.
3041 (hash_loc_list): Hash view numbers.
3042 (loc_list_hasher::equal): Compare them.
3043 (optimize_location_lists): Check whether a view list symbol is
3044 needed, and whether the locview attribute is present, and
3045 whether they match. Remove the locview attribute if no longer
3047 (index_location_lists): Call skip_loc_list_entry for test.
3048 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3049 Use output_dwarf_version.
3050 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3051 (struct dw_val_node): Add val_view_list.
3052 * final.c (SEEN_NEXT_VIEW): New.
3053 (set_next_view_needed): New.
3054 (clear_next_view_needed): New.
3055 (maybe_output_next_view): New.
3056 (final_start_function): Rename to...
3057 (final_start_function_1): ... this. Take pointer to FIRST,
3058 add SEEN parameter. Emit param bindings in the initial view.
3059 (final_start_function): Reintroduce SEEN-less interface.
3060 (final): Rename to...
3061 (final_1): ... this. Take SEEN parameter. Output final pending
3062 next view at the end.
3063 (final): Reintroduce seen-less interface.
3064 (final_scan_insn): Output pending next view before switching
3065 sections or ending a block. Mark the next view as needed when
3066 outputting variable locations. Notify debug backend of section
3067 changes, and of location view changes.
3068 (rest_of_handle_final): Adjust.
3069 * toplev.c (process_options): Autodetect value for debug variable
3070 location views option. Warn on incompat5 without -gdwarf-5.
3071 * doc/invoke.texi (gvariable-location-views): New.
3072 (gvariable-location-views=incompat5): New.
3073 (gno-variable-location-views): New.
3075 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3077 PR tree-optimization/84136
3078 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3079 that the result of find_edge is non-NULL.
3081 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3084 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3085 storing integer register in SImode. Fix cost of 256 and 512
3086 byte aligned SSE register store.
3088 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3090 * config/i386/i386.c (ix86_multiplication_cost): Fix
3091 multiplication cost for TARGET_AVX512DQ.
3093 2018-02-08 Marek Polacek <polacek@redhat.com>
3095 PR tree-optimization/84238
3096 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3099 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3101 PR tree-optimization/84265
3102 * tree-vect-stmts.c (vectorizable_store): Don't treat
3103 VMAT_CONTIGUOUS accesses as grouped.
3104 (vectorizable_load): Likewise.
3106 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3108 PR tree-optimization/81635
3109 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3110 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3111 (test_round_for_mask): New functions.
3112 (wide_int_cc_tests): Call test_round_for_mask.
3113 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3114 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3115 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3116 range returned by get_range_info.
3118 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3121 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3122 * symtab.c: Include builtins.h
3123 (symtab_node::output_to_lto_symbol_table_p): Move here
3124 from lto-streamer-out.c:output_symbol_p.
3125 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3126 (output_symbol_p): Move all logic to symtab.c
3127 (produce_symtab): Update.
3129 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3131 * config/s390/s390-opts.h (enum indirect_branch): Define.
3132 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3133 (s390_indirect_branch_via_thunk)
3134 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3135 (enum s390_indirect_branch_type): Define.
3136 * config/s390/s390.c (struct s390_frame_layout, struct
3137 machine_function): Remove.
3138 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3139 (indirect_branch_table_label_no, indirect_branch_table_name):
3141 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3142 (enum s390_indirect_branch_option): Define.
3143 (s390_return_addr_from_memory): New function.
3144 (s390_handle_string_attribute): New function.
3145 (s390_attribute_table): Add new attribute handler.
3146 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3147 (s390_indirect_branch_via_thunk): New function.
3148 (s390_indirect_branch_via_inline_thunk): New function.
3149 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3150 sibling call optimization for non z10 compiles.
3151 (s390_emit_call): Force indirect branch target to be a single
3152 register. Add r1 clobber for non-z10 compiles.
3153 (s390_emit_epilogue): Emit return jump via return_use expander.
3154 (s390_reorg): Handle JUMP_INSNs as execute targets.
3155 (s390_option_override_internal): Perform validity checks for the
3156 new command line options.
3157 (s390_indirect_branch_attrvalue): New function.
3158 (s390_indirect_branch_settings): New function.
3159 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3160 (s390_output_indirect_thunk_function): New function.
3161 (s390_code_end): Implement target hook.
3162 (s390_case_values_threshold): Implement target hook.
3163 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3165 * config/s390/s390.h (struct s390_frame_layout)
3166 (struct machine_function): Move here from s390.c.
3167 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3168 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3169 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3170 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3171 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3172 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3173 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3174 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3175 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3176 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3177 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3178 (mnemonic attribute): Add values which aren't recognized
3180 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3181 pattern for branch conversion. Fix mnemonic attribute.
3182 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3183 indirect branch via thunk if requested.
3184 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3185 ("*indirect_jump"): Disable for branch conversion using out of
3187 ("indirect_jump_via_thunk<mode>_z10")
3188 ("indirect_jump_via_thunk<mode>")
3189 ("indirect_jump_via_inlinethunk<mode>_z10")
3190 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3191 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3192 ("casesi_jump_via_inlinethunk<mode>_z10")
3193 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3194 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3195 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3196 ("*indirect2_jump"): Disable for branch conversion.
3197 ("casesi_jump"): Turn into expander and expand patterns for branch
3199 ("return_use"): New expander.
3200 ("*return"): Emit return via thunk and rename it to ...
3201 ("*return<mode>"): ... this one.
3202 * config/s390/s390.opt: Add new options and and enum for the
3205 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3207 * lra-constraints.c (match_reload): Unconditionally use
3208 gen_lowpart_SUBREG, rather than selecting between that
3209 and equivalent gen_rtx_SUBREG code.
3211 2018-02-08 Richard Biener <rguenther@suse.de>
3213 PR tree-optimization/84233
3214 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3215 changed flag instead of boguously re-using phi_inserted.
3217 2018-02-08 Martin Jambor <mjambor@suse.cz>
3219 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3220 static local variables.
3222 2018-02-08 Richard Biener <rguenther@suse.de>
3224 PR tree-optimization/84278
3225 * tree-vect-stmts.c (vectorizable_store): When looking for
3226 smaller vector types to perform grouped strided loads/stores
3227 make sure the mode is supported by the target.
3228 (vectorizable_load): Likewise.
3230 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3232 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3233 Increase LDP/STP opportunities by adding adjacent callee-saves.
3235 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3237 PR rtl-optimization/84068
3238 PR rtl-optimization/83459
3239 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3241 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3243 PR tree-optimization/84224
3244 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3245 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3248 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3251 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3252 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3253 count, remove one USE.
3255 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3257 * doc/install.texi (Configuration): Document the
3258 --with-long-double-format={ibm,ieee} PowerPC configuration
3262 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3263 Convert from define_expand to be define_insn_and_split. Rework
3264 float/double/_Float128 conversions to QI/HI/SImode to work with
3265 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3266 conversions to QI/HImode types did a store and then a load to
3267 truncate the value. For conversions to VSX registers, don't split
3268 the insn, instead emit the code directly. Use the code iterator
3269 any_fix to combine signed and unsigned conversions.
3270 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3271 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3272 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3273 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3274 (fix_<mode>di2_hw): Likewise.
3275 (fixuns_<mode>di2_hw): Likewise.
3276 (fix_<mode>si2_hw): Likewise.
3277 (fixuns_<mode>si2_hw): Likewise.
3278 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3279 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3280 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3281 fix<uns>_trunc<SFDF:mode>si2_p8.
3282 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3284 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3285 (fix<uns>_<mode>_mem): Likewise.
3286 (fctiw<u>z_<mode>_mem): Likewise.
3287 (fix<uns>_<mode>_mem): Likewise.
3288 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3289 the register allocator from doing a direct move to the GPRs to do
3290 a store, and instead use the ISA 3.0 store byte/half-word from
3291 vector register instruction. For IEEE 128-bit floating point,
3292 also optimize stores of 32-bit ints.
3293 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3295 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3297 * genextract.c (push_pathstr_operand): New function to support
3299 (walk_rtx): Call push_pathstr_operand.
3300 (print_path): Support [a-zA-Z].
3302 2018-02-07 Richard Biener <rguenther@suse.de>
3304 PR tree-optimization/84037
3305 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3306 (cse_and_gimplify_to_preheader): Declare.
3307 (vect_get_place_in_interleaving_chain): Likewise.
3308 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3310 (_loop_vec_info::~_loop_vec_info): Delete it.
3311 (cse_and_gimplify_to_preheader): New function.
3312 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3313 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3314 (vectorizable_load): Likewise. For grouped stores always base
3315 the IV on the first element.
3316 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3317 condition before gimplifying.
3319 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3321 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3322 *DIV_EXPR and *MOD_EXPR.
3324 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3327 * config/i386/i386.c (ix86_option_override_internal): Mask out
3328 the CF_SET bit when checking -fcf-protection.
3330 2018-02-07 Tom de Vries <tom@codesourcery.com>
3333 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3336 2018-02-07 Richard Biener <rguenther@suse.de>
3338 PR tree-optimization/84204
3339 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3342 PR tree-optimization/84205
3343 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3344 special-case isl_ast_op_zdiv_r.
3346 PR tree-optimization/84223
3347 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3348 Only add conditions from within the region.
3349 (gather_bbs::after_dom_children): Adjust.
3351 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3354 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3355 * config/avr/avr.md: Only post-reload split REG-REG moves if
3356 either register is GENERAL_REG_P.
3358 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3360 PR tree-optimization/84235
3361 * tree-ssa-scopedtables.c
3362 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3363 if the subtraction is performed in floating point type where NaNs are
3364 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3365 build 1. Formatting fix.
3367 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3370 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3371 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3372 and skip it regardless of bb boundaries. Use CALL_P macro,
3373 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3374 unnecessarily, formatting fix.
3376 2018-02-06 Michael Collison <michael.collison@arm.com>
3378 * config/arm/thumb2.md:
3379 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3380 (*thumb_mov_notscc): Ditto.
3382 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3385 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3386 unsigned_fix, not "s".
3388 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3390 * configure.ac (gcc_fn_eh_frame_ro): New function.
3391 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3392 correct .eh_frame permissions.
3393 * configure: Regenerate.
3395 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3397 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3400 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3402 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3403 Display warning message for -mno-speculate-indirect-jumps.
3405 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3407 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3409 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3411 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3413 PR tree-optimization/84225
3414 * tree-eh.c (find_trapping_overflow): Only call
3415 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3417 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3420 * config/i386/i386.c: Reimplement the check of possible options
3421 -mibt/-mshstk conbination. Change error messages.
3422 * doc/invoke.texi: Fix a typo: remove extra '='.
3424 2018-02-06 Marek Polacek <polacek@redhat.com>
3426 PR tree-optimization/84228
3427 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3429 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3432 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3433 emitted arch directives.
3434 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3435 __ARM_FEATURE_COPROC before changing architectures.
3437 2018-02-06 Richard Biener <rguenther@suse.de>
3439 * config/i386/i386.c (print_reg): Fix typo.
3440 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3442 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3444 * configure: Regenerate.
3446 2018-02-05 Martin Sebor <msebor@redhat.com>
3448 PR tree-optimization/83369
3449 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3452 2018-02-05 Martin Liska <mliska@suse.cz>
3454 * doc/invoke.texi: Cherry-pick upstream r323995.
3456 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3458 * ira.c (ira_init_register_move_cost): Adjust comment.
3460 2018-02-05 Martin Liska <mliska@suse.cz>
3462 PR gcov-profile/84137
3463 * doc/gcov.texi: Fix typo in documentation.
3465 2018-02-05 Martin Liska <mliska@suse.cz>
3467 PR gcov-profile/83879
3468 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3469 using dlopen functionality.
3471 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3473 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3474 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3475 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3476 _mm_maskz_range_round_ss): New intrinsics.
3477 (__builtin_ia32_rangesd128_round)
3478 (__builtin_ia32_rangess128_round): Remove.
3479 (__builtin_ia32_rangesd128_mask_round,
3480 __builtin_ia32_rangess128_mask_round): New builtins.
3481 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3482 __builtin_ia32_rangess128_round): Remove.
3483 (__builtin_ia32_rangesd128_mask_round,
3484 __builtin_ia32_rangess128_mask_round): New builtins.
3485 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3486 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3487 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3488 "<round_saeonly_constraint>")): Changed to ...
3489 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3490 "<round_saeonly_scalar_constraint>")): ... this.
3491 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3492 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3493 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3494 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3495 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3497 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3499 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3501 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3502 Remove all values except native, 8540 and 8548.
3504 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3506 * config/i386/i386.c (ix86_output_function_return): Pass
3507 INVALID_REGNUM, instead of -1, as invalid register number to
3508 indirect_thunk_name and output_indirect_thunk.
3510 2018-02-02 Julia Koval <julia.koval@intel.com>
3512 * config.gcc: Add -march=icelake.
3513 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3514 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3515 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3516 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3517 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3518 (processor_target_table): Add icelake.
3519 (ix86_option_override_internal): Handle new PTAs.
3520 (get_builtin_code_for_version): Handle icelake.
3521 (M_INTEL_COREI7_ICELAKE): New.
3522 (fold_builtin_cpu): Handle icelake.
3523 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3524 * doc/invoke.texi: Add -march=icelake.
3526 2018-02-02 Julia Koval <julia.koval@intel.com>
3528 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3529 to wide_int_bitmask.
3530 * wide-int-bitmask.h: New.
3532 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3535 * config/i386/i386.md: Replace Pmode with word_mode in
3536 builtin_setjmp_setup and builtin_longjmp to support x32.
3538 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3542 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3544 (rs6000_supported_cpu_names): New static variable.
3545 (linux_cpu_translation_table): Likewise.
3546 (elf_platform) <cpu>: Define new static variable and use it.
3547 Translate kernel AT_PLATFORM name to canonical name if needed.
3548 Error if platform name is unknown.
3550 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3553 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3555 2018-02-01 Jeff Law <law@redhat.com>
3558 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3559 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3560 the scratch if RELEASE_VIA_POP is false.
3561 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3562 If we have to save a temporary register, decrement SIZE appropriately.
3563 Pass new arguments to release_scratch_register_on_entry.
3564 (ix86_adjust_stack_and_probe): Likewise.
3565 (ix86_emit_probe_stack_range): Pass new arguments to
3566 release_scratch_register_on_entry.
3568 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3570 PR rtl-optimization/84157
3571 * combine.c (change_zero_ext): Use REG_P predicate in
3572 front of HARD_REGISTER_P predicate.
3574 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3576 * config/avr/avr.c (avr_option_override): Move disabling of
3577 -fdelete-null-pointer-checks to...
3578 * common/config/avr/avr-common.c (avr_option_optimization_table):
3581 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3583 PR tree-optimization/81635
3584 * tree-data-ref.c (split_constant_offset_1): For types that
3585 wrap on overflow, try to use range info to prove that wrapping
3588 2018-02-01 Renlin Li <renlin.li@arm.com>
3591 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3593 (aarch64_register_move_cost): Likewise.
3594 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3596 (REG_CLASS_NAMES): Likewise.
3597 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3598 TAILCALL_ADDR_REGS. Remove IP registers.
3599 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3601 2018-02-01 Richard Biener <rguenther@suse.de>
3603 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3604 for specifying RPO order and allow NULL for that.
3605 * domwalk.c (dom_walker::dom_walker): Likewise.
3606 (dom_walker::walk): Handle NULL RPO order.
3607 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3609 (rewrite_update_dom_walker): Likewise.
3610 (mark_def_dom_walker): Likewise.
3612 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3614 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3615 (aarch64_maybe_expand_sve_subreg_move): Declare.
3616 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3617 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3619 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3620 that are semantically a reverse operation.
3621 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3622 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3623 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3625 (aarch64_can_change_mode_class): For big-endian, forbid changes
3626 between two SVE modes if they have different element sizes.
3628 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3630 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3631 the TImode handling for big-endian targets.
3633 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3635 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3636 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3638 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3639 Remove BSWAP handing for big-endian targets and use the form of
3640 LD1RQ appropariate for the mode.
3642 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3644 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3645 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3648 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3651 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3652 check for operands that need to go through aarch64_sve_reload_be.
3654 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3656 PR tree-optimization/81661
3657 PR tree-optimization/84117
3658 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3659 * tree-eh.c: Include gimplify.h.
3660 (find_trapping_overflow, replace_trapping_overflow,
3661 rewrite_to_non_trapping_overflow): New functions.
3662 * tree-vect-loop.c: Include tree-eh.h.
3663 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3664 * tree-data-ref.c: Include tree-eh.h.
3665 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3667 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3669 PR rtl-optimization/84123
3670 * combine.c (change_zero_ext): Check if hard register satisfies
3671 can_change_dest_mode before calling gen_lowpart_SUBREG.
3673 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3676 * ira.c (ira_init_register_move_cost): Remove assert.
3678 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3680 PR rtl-optimization/84071
3681 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3682 * doc/tm.texi: Regenerate.
3684 2018-01-31 Richard Biener <rguenther@suse.de>
3686 PR tree-optimization/84132
3687 * tree-data-ref.c (analyze_miv_subscript): Properly
3688 check whether evolution_function_is_affine_multivariate_p
3689 before calling gcd_of_steps_may_divide_p.
3691 2018-01-31 Julia Koval <julia.koval@intel.com>
3694 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3695 * config/i386/i386.md (rdpid_rex64) New.
3696 (rdpid): Make 32bit only.
3698 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3701 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3702 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3704 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3707 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3709 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3711 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3713 PR rtl-optimization/84071
3714 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3715 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3717 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3719 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3720 (arc_attribute_table): Add 'aux' attribute.
3721 (arc_in_small_data_p): Consider aux like variables.
3722 (arc_is_aux_reg_p): New function.
3723 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3724 (arc_get_aux_arg): New function.
3725 (prepare_move_operands): Handle aux-register access.
3726 (arc_handle_aux_attribute): New function.
3727 * doc/extend.texi (ARC Variable attributes): Add subsection.
3729 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3731 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3732 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3733 (arc_attribute_table): Add 'uncached' attribute.
3734 (arc_print_operand): Print '.di' flag for uncached memory
3736 (arc_in_small_data_p): Do not consider for small data the uncached
3738 (arc_is_uncached_mem_p): New function.
3739 * config/arc/predicates.md (compact_store_memory_operand): Check
3740 for uncached memory accesses.
3741 (nonvol_nonimm_operand): Likewise.
3742 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3744 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3747 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3748 falign-loops=): Add Optimization flag.
3750 2018-01-30 Jeff Law <law@redhat.com>
3753 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3754 INT_REGISTERS_SAVED. Check it prior to calling
3755 get_scratch_register_on_entry.
3756 (ix86_adjust_stack_and_probe): Similarly.
3757 (ix86_emit_probe_stack_range): Similarly.
3758 (ix86_expand_prologue): Corresponding changes.
3760 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3763 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3764 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3766 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3769 * lra-constraints.c (curr_insn_transform): Process AND in the
3772 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3774 PR rtl-optimization/83986
3775 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3776 dependence against last_pending_memory_flush in addition to
3779 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3781 PR tree-optimization/81611
3782 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3785 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3788 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3791 2018-01-30 Richard Biener <rguenther@suse.de>
3792 Jakub Jelinek <jakub@redhat.com>
3794 PR tree-optimization/84111
3795 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3796 inner loops added during recursion, as they don't have up-to-date
3799 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3802 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3803 (can_inline_edge_by_limits_p): ... here.
3804 (can_early_inline_edge_p, check_callers,
3805 update_caller_keys, update_callee_keys, recursive_inlining,
3806 add_new_edges_to_heap, speculation_useful_p,
3807 inline_small_functions,
3808 inline_small_functions, flatten_function,
3809 inline_to_all_callers_1): Update.
3811 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3813 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3816 2018-01-30 Richard Biener <rguenther@suse.de>
3818 PR tree-optimization/83008
3819 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3820 invariant and constant vector uses in stmts when they need
3823 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3826 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3827 * configure: Regenerate.
3829 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3831 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3833 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3834 Use gen_rtx_REG rather than gen_lowpart.
3836 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3838 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3839 rather than 0 when creating partial subregs.
3841 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3843 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3846 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3849 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3850 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3851 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3852 flags. This restores the settings used before the 2017-07-24.
3853 Turning off pre increment/decrement/modify allows IVOPTS to
3854 optimize DF/SF loops where the index is an int.
3856 2018-01-29 Richard Biener <rguenther@suse.de>
3857 Kelvin Nilsen <kelvin@gcc.gnu.org>
3860 * tree-vect-stmts.c (vectorizable_call): Don't call
3861 targetm.vectorize_builtin_md_vectorized_function if callee is
3864 2018-01-22 Carl Love <cel@us.ibm.com>
3866 * doc/extend.tex: Fix typo in second arg in
3867 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3869 2018-01-29 Richard Biener <rguenther@suse.de>
3871 PR tree-optimization/84086
3872 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3873 (flush_ssaname_freelist): When SSA names were released reset
3874 the SCEV hash table.
3876 2018-01-29 Richard Biener <rguenther@suse.de>
3878 PR tree-optimization/84057
3879 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3880 removed paths when removing edges.
3882 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3884 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3885 -mfunction-return=@var{choice}.
3887 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3890 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3892 (layout::print_source_line): Likewise.
3893 (test_get_line_width_without_trailing_whitespace): Add test cases.
3895 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3898 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3901 2018-01-26 Jim Wilson <jimw@sifive.com>
3903 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3905 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3908 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3910 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3911 and CMP + SUB-immediate -> SUBS.
3913 2018-01-26 Martin Sebor <msebor@redhat.com>
3915 PR tree-optimization/83896
3916 * tree-ssa-strlen.c (get_string_len): Rename...
3917 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3918 Avoid assuming length is constant.
3919 (handle_char_store): Use HOST_WIDE_INT for string length.
3921 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3924 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3925 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3927 2018-01-26 Richard Biener <rguenther@suse.de>
3929 PR rtl-optimization/84003
3930 * dse.c (record_store): Only record redundant stores when
3931 the earlier store aliases at least all accesses the later one does.
3933 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3935 PR rtl-optimization/83985
3936 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3937 REG_CFA_RESTORE insns.
3938 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3939 REG_CFA_RESTORE insns here.
3942 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3943 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3945 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3947 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3948 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3949 (arc_init): Likewise.
3950 (arc_override_options): Likewise.
3951 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3953 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3955 * config/arc/arc.h (TARGET_DBNZ): Define.
3956 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3957 properly set the tune attribute.
3958 (dbnz): Use TARGET_DBNZ guard.
3959 * config/arc/arc.opt (mtune): Add core3 option.
3961 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3963 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3964 recognize new pic like addresses.
3965 (arc_delegitimize_address): Clean up.
3967 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3969 * config/arc/arc-arches.def: Option mrf16 valid for all
3971 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3972 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3973 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3974 * config/arc/arc-tables.opt: Regenerate.
3975 * config/arc/arc.c (arc_conditional_register_usage): Handle
3976 reduced register file case.
3977 (arc_file_start): Set must have build attributes.
3978 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3980 * config/arc/arc.opt (mrf16): Add new option.
3981 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3982 * config/arc/genmultilib.awk: Handle new mrf16 option.
3983 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3984 * config/arc/t-multilib: Regenerate.
3985 * doc/invoke.texi (ARC Options): Document mrf16 option.
3987 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3989 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3990 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3991 (arc_attribute_table): Add 'secure_call' attribute.
3992 (arc_print_operand): Print secure call operand.
3993 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3995 (arc_is_secure_call_p): New function. * config/arc/arc.md
3996 (call_i): Add support for sjli instruction.
3997 (call_value_i): Likewise.
3998 * config/arc/constraints.md (Csc): New constraint.
4000 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4001 John Eric Martin <John.Martin@emmicro-us.com>
4003 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4004 * config/arc/arc.c (_arc_jli_section): New struct.
4005 (arc_jli_section): New type.
4006 (rc_jli_sections): New static variable.
4007 (arc_handle_jli_attribute): New function.
4008 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4009 (arc_file_end): New function.
4010 (TARGET_ASM_FILE_END): Define.
4011 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4012 (arc_add_jli_section): New function.
4013 (jli_call_scan): Likewise.
4014 (arc_reorg): Call jli_call_scan.
4015 (arc_output_addsi): Remove 'S' from printing asm operand.
4016 (arc_is_jli_call_p): New function.
4017 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4019 (movhi_insn): Likewise.
4020 (movsi_insn): Likewise.
4021 (movsi_set_cc_insn): Likewise.
4022 (loadqi_update): Likewise.
4023 (load_zeroextendqisi_update): Likewise.
4024 (load_signextendqisi_update): Likewise.
4025 (loadhi_update): Likewise.
4026 (load_zeroextendhisi_update): Likewise.
4027 (load_signextendhisi_update): Likewise.
4028 (loadsi_update): Likewise.
4029 (loadsf_update): Likewise.
4030 (movsicc_insn): Likewise.
4031 (bset_insn): Likewise.
4032 (bxor_insn): Likewise.
4033 (bclr_insn): Likewise.
4034 (bmsk_insn): Likewise.
4035 (bicsi3_insn): Likewise.
4036 (cmpsi_cc_c_insn): Likewise.
4037 (movsi_ne): Likewise.
4038 (movsi_cond_exec): Likewise.
4039 (clrsbsi2): Likewise.
4050 (call_i): Remove 'S' asm letter, add jli instruction.
4051 (call_value_i): Likewise.
4052 * config/arc/arc.op (mjli-always): New option.
4053 * config/arc/constraints.md (Cji): New constraint.
4054 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4056 (subsf3_fpx): Likewise.
4057 (mulsf3_fpx): Likewise.
4058 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4060 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4062 * doc/invoke.texi (ARC): Document mjli-always option.
4064 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4066 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4067 avoid addition with 0 and use incw and decw where possible.
4069 2018-01-26 Richard Biener <rguenther@suse.de>
4071 PR tree-optimization/81082
4072 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4073 association if it requires casting to unsigned.
4074 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4075 from fold_plusminus_mult_expr to catch important cases late when
4076 range info is available.
4078 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4080 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4081 * configure.ac (hidden_linkonce): New test.
4082 * configure: Regenerate.
4083 * config.in: Regenerate.
4085 2018-01-26 Julia Koval <julia.koval@intel.com>
4087 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4088 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4089 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4090 _mm_mask_bitshuffle_epi64_mask): Fix type.
4091 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4092 USI_FTYPE_V4DI_V4DI_USI): Remove.
4093 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4094 __builtin_ia32_vpshufbitqmb256_mask,
4095 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4096 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4097 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4099 2018-01-26 Alan Modra <amodra@gmail.com>
4102 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4103 UNSPEC_VBPERMQ. Sort other unspecs.
4105 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4107 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4109 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4112 * predict.c (drop_profile): Do not push/pop cfun; update also
4114 (handle_missing_profiles): Fix logic looking for zero profiles.
4116 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4119 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4120 on functions with #pragma omp declare simd or functions with simd
4122 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4123 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4124 Remove trailing \n from warning_at calls.
4126 2018-01-25 Tom de Vries <tom@codesourcery.com>
4129 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4130 for neutered workers.
4132 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4135 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4136 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4138 2018-01-24 Jeff Law <law@redhat.com>
4141 * i386.c (get_probe_interval): Move to earlier point.
4142 (ix86_compute_frame_layout): If -fstack-clash-protection and
4143 the frame is larger than the probe interval, then use pushes
4144 to save registers rather than reg->mem moves.
4145 (ix86_expand_prologue): Remove conditional for int_registers_saved
4148 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4151 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4152 min/max for never referenced object.
4154 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4157 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4159 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4160 attributes from DECL_ATTRIBUTES (decl) without affecting
4161 DECL_ATTRIBUTES (current_function_decl).
4162 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4163 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4165 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4167 PR tree-optimization/83979
4168 * fold-const.c (fold_comparison): Use constant_boolean_node
4169 instead of boolean_{true,false}_node.
4171 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4173 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4176 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4178 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4179 Simplify the clause that sets the length attribute.
4180 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4181 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4182 clause that sets the length attribute.
4183 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4185 2018-01-24 Tom de Vries <tom@codesourcery.com>
4188 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4189 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4190 Add strict parameter.
4191 (prevent_branch_around_nothing): Insert dummy insn between branch to
4192 label and label with no ptx insn inbetween.
4193 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4195 2018-01-24 Tom de Vries <tom@codesourcery.com>
4198 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4199 for neutered threads in warp.
4200 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4202 2018-01-24 Richard Biener <rguenther@suse.de>
4204 PR tree-optimization/83176
4205 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4208 2018-01-24 Richard Biener <rguenther@suse.de>
4210 PR tree-optimization/82819
4211 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4212 code generating pluses that are no-ops in the target precision.
4214 2018-01-24 Richard Biener <rguenther@suse.de>
4217 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4219 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4221 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4222 to merge probabilities.
4223 * predict.c (probably_never_executed): Also mark as cold functions
4224 with global 0 profile and guessed local profile.
4225 * profile-count.c (profile_probability::combine_with_count): New
4227 * profile-count.h (profile_probability::operator*,
4228 profile_probability::operator*=, profile_probability::operator/,
4229 profile_probability::operator/=): Reduce precision to adjusted
4230 and set value to guessed on contradictory divisions.
4231 (profile_probability::combine_with_freq): Remove.
4232 (profile_probability::combine_wiht_count): Declare.
4233 (profile_count::force_nonzero):: Set to adjusted.
4234 (profile_count::probability_in):: Set quality to adjusted.
4235 * tree-ssa-tail-merge.c (replace_block_by): Use
4238 2018-01-23 Andrew Waterman <andrew@sifive.com>
4239 Jim Wilson <jimw@sifive.com>
4241 * config/riscv/riscv.c (riscv_stack_boundary): New.
4242 (riscv_option_override): Set riscv_stack_boundary. Handle
4243 riscv_preferred_stack_boundary_arg.
4244 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4245 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4246 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4247 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4248 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4249 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4251 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4254 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4255 of struct ix86_frame.
4256 (ix86_expand_epilogue): Likewise. Add a local variable for
4257 the reg_save_offset field in struct ix86_frame.
4259 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4261 PR tree-optimization/82604
4262 * tree-loop-distribution.c (enum partition_kind): New enum item
4263 PKIND_PARTIAL_MEMSET.
4264 (partition_builtin_p): Support above new enum item.
4265 (generate_code_for_partition): Ditto.
4266 (compute_access_range): Differentiate cases that equality can be
4267 proven at all loops, the innermost loops or no loops.
4268 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4269 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4270 (finalize_partitions, distribute_loop): Don't fuse partition of
4271 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4272 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4275 2018-01-23 Martin Liska <mliska@suse.cz>
4277 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4278 order to ignore the predictor.
4279 (PRED_POLYMORPHIC_CALL): Likewise.
4280 (PRED_RECURSIVE_CALL): Likewise.
4282 2018-01-23 Martin Liska <mliska@suse.cz>
4284 * tree-profile.c (tree_profiling): Print function header to
4285 aware reader which function we are working on.
4286 * value-prof.c (gimple_find_values_to_profile): Do not print
4287 not interesting value histograms.
4289 2018-01-23 Martin Liska <mliska@suse.cz>
4291 * profile-count.h (enum profile_quality): Add
4292 profile_uninitialized as the first value. Do not number values
4293 as they are zero based.
4294 (profile_count::verify): Update sanity check.
4295 (profile_probability::verify): Likewise.
4297 2018-01-23 Nathan Sidwell <nathan@acm.org>
4299 * doc/invoke.texi (ffor-scope): Deprecate.
4301 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4303 PR tree-optimization/83510
4304 * domwalk.c (set_all_edges_as_executable): New function.
4305 (dom_walker::dom_walker): Convert bool param
4306 "skip_unreachable_blocks" to enum reachability. Move setup of
4307 edge flags to set_all_edges_as_executable and only do it when
4308 reachability is REACHABLE_BLOCKS.
4309 * domwalk.h (enum dom_walker::reachability): New enum.
4310 (dom_walker::dom_walker): Convert bool param
4311 "skip_unreachable_blocks" to enum reachability.
4312 (set_all_edges_as_executable): New decl.
4313 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4314 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4316 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4317 but converting true to REACHABLE_BLOCKS.
4318 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4320 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4321 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4322 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4324 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4325 if check_all_array_refs will be called.
4327 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4329 * tree.c (selftest::test_location_wrappers): Add more test
4332 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4334 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4335 (selftest::test_bit_in_range): Likewise.
4337 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4340 * doc/sourcebuild.texi (vect_float): Say that the selector
4341 only describes the situation when -funsafe-math-optimizations is on.
4342 (vect_float_strict): Document.
4344 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4346 PR tree-optimization/83965
4347 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4348 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4349 instead of checking only for a reduction.
4350 (vect_recog_widen_sum_pattern): Likewise.
4352 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4354 * predict.c (probably_never_executed): Only use precise profile info.
4355 (compute_function_frequency): Skip after inlining hack since we now
4356 have quality checking.
4358 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4360 * profile-count.h (profile_probability::very_unlikely,
4361 profile_probability::unlikely, profile_probability::even): Set
4362 precision to guessed.
4364 2018-01-23 Richard Biener <rguenther@suse.de>
4366 PR tree-optimization/83963
4367 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4368 Properly terminate dominator walk when crossing the exit edge not
4369 when visiting its source block.
4371 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4374 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4375 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4377 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4379 PR tree-optimization/83957
4380 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4381 semicolon after for body surrounded by braces.
4383 PR tree-optimization/83081
4384 * profile-count.h (profile_probability::split): New method.
4385 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4386 Use profile_probability::split.
4387 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4388 when splitting a single conditional jump into 2.
4390 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4392 PR tree-optimization/69452
4393 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4396 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4398 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4399 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4400 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4402 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4404 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4405 * config/rl78/rl78.md (movdi): New define_expand.
4406 * config/rl78/rl78.c (rl78_split_movdi): New function.
4408 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4411 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4413 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4414 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4415 128-bit to produce an UNSPEC move to get the double word with the
4416 signbit and then a shift directly to do signbit.
4417 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4418 implementation with a new version that just does either a direct
4419 move or a regular move. Move memory interface to separate insns.
4420 Move insns so they are next to the expander.
4421 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4422 with signbit move. Split big and little endian case.
4423 (signbit<mode>2_dm_mem_le): Likewise.
4424 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4425 (signbit<mode>2_dm2): Likewise.
4427 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4429 * config/rl78/rl78.md (anddi3): New define_expand.
4431 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4433 * config/rl78/rl78.md (umindi3): New define_expand.
4435 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4437 * config/rl78/rl78.md (smindi3): New define_expand.
4439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4441 * config/rl78/rl78.md (smaxdi3): New define_expand.
4443 2018-01-22 Carl Love <cel@us.ibm.com>
4445 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4446 LVX_V1TI): Add macro expansion.
4447 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4448 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4449 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4450 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4451 Change check to determine if the instruction is a byte reversing
4452 entry. Fix typo in comment.
4453 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4454 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4455 Add def_builtin calls for new builtins.
4456 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4457 Add define_insn expansion.
4459 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4461 * config/rl78/rl78.md (umaxdi3): New define_expand.
4463 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4465 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4466 for non-QImode registers.
4468 2018-01-22 Richard Biener <rguenther@suse.de>
4470 PR tree-optimization/83963
4471 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4472 including the loop exit block.
4473 (scop_detection::merge_sese): Likewise.
4474 (scop_detection::add_scop): Do it here instead.
4476 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4478 * doc/sourcebuild.texi (arm_softfloat): Document.
4480 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4483 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4484 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4485 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4487 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4488 David Edelsohn <dje.gcc@gmail.com>
4491 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4492 Change "crset eq" to "crset 2".
4493 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4494 (*call_indirect_aix<mode>_nospec): Likewise.
4495 (*call_value_indirect_aix<mode>_nospec): Likewise.
4496 (*call_indirect_elfv2<mode>_nospec): Likewise.
4497 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4498 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4499 change assembly output from . to $.
4500 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4501 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4502 (*tablejump<mode>_internal1_nospec): Likewise.
4504 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4507 * config/sh/sh_optimize_sett_clrt.cc:
4508 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4510 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4512 PR tree-optimization/83940
4513 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4514 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4515 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4516 use it to pass back the definition type.
4517 (vect_check_store_rhs): Likewise rhs_dt_out.
4518 (vect_build_gather_load_calls): Add a mask_dt argument and use
4519 it instead of a call to vect_is_simple_use.
4520 (vectorizable_store): Update calls to vect_check_load_store_mask
4521 and vect_check_store_rhs. Use the dt returned by the latter instead
4522 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4523 instead of calls to vect_is_simple_use. Pass the scalar rather
4524 than the vector operand to vect_is_simple_use when handling
4525 second and subsequent copies of an rhs value.
4526 (vectorizable_load): Update calls to vect_check_load_store_mask
4527 and vect_build_gather_load_calls. Use the cached mask_dt and
4528 gs_info.offset_dt instead of calls to vect_is_simple_use.
4530 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4533 * tree-emutls.c: Include gimplify.h.
4534 (lower_emutls_2): New function.
4535 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4536 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4537 it before further processing.
4540 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4541 UINTVAL (trueop1) instead of INTVAL (op1).
4543 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4547 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4548 INCOMING_FRAME_SP_OFFSET if not defined.
4549 (scan_trace): Add ENTRY argument. If true and
4550 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4551 emit a note to adjust the CFA offset.
4552 (create_cfi_notes): Adjust scan_trace callers.
4553 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4554 INCOMING_FRAME_SP_OFFSET in the CIE.
4555 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4556 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4558 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4559 * doc/tm.texi: Regenerated.
4561 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4563 PR rtl-optimization/83147
4564 * lra-constraints.c (remove_inheritance_pseudos): Use
4565 lra_substitute_pseudo_within_insn.
4567 2018-01-19 Tom de Vries <tom@codesourcery.com>
4568 Cesar Philippidis <cesar@codesourcery.com>
4571 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4573 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4576 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4577 spaces for function labels.
4579 2018-01-19 Martin Liska <mliska@suse.cz>
4581 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4582 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4583 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4584 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4585 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4586 (PRED_CONST_RETURN): Change from 69 to 65.
4587 (PRED_NULL_RETURN): Change from 91 to 71.
4588 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4589 (PRED_LOOP_GUARD): Change from 66 to 73.
4591 2018-01-19 Martin Liska <mliska@suse.cz>
4593 * predict.c (predict_insn_def): Add new assert.
4594 (struct branch_predictor): Change type to signed integer.
4595 (test_prediction_value_range): Amend test to cover
4597 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4598 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4599 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4600 (PRED_LOOP_IV_COMPARE): Likewise.
4601 * predict.h (PROB_UNINITIALIZED): Define new constant.
4603 2018-01-19 Martin Liska <mliska@suse.cz>
4605 * predict.c (dump_prediction): Add new format for
4606 analyze_brprob.py script which is enabled with -details
4608 * profile-count.h (precise_p): New function.
4610 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4612 PR tree-optimization/83922
4613 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4614 there are no statements that need masking.
4615 (vect_active_double_reduction_p): New function.
4616 (vect_analyze_loop_operations): Use it when handling phis that
4617 are not in the loop header.
4619 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4621 PR tree-optimization/83914
4622 * tree-vect-loop.c (vectorizable_induction): Don't convert
4623 init_expr or apply the peeling adjustment for inductions
4624 that are nested within the vectorized loop.
4626 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4628 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4631 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4635 * function.h (gimplify_parameters): Add gimple_seq * argument.
4636 * function.c: Include gimple.h and options.h.
4637 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4638 for the added local temporaries if needed.
4639 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4640 if there are any parameter cleanups, wrap whole body into a
4641 try/finally with the cleanups.
4643 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4646 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4647 Use GET_MODE_CLASS for scalar floating point.
4649 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4653 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4654 Fix call of call_cgraph_insertion_hooks.
4656 2018-01-18 Martin Sebor <msebor@redhat.com>
4658 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4660 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4663 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4666 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4669 * common.opt: (-ffile-prefix-map): New option.
4670 * opts.c (common_handle_option): Defer it.
4671 * opts-global.c (handle_common_deferred_options): Handle it.
4672 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4673 * file-prefix-map.h: New file.
4674 (remap_debug_filename, add_debug_prefix_map): ...here.
4675 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4676 * final.c (debug_prefix_map, add_debug_prefix_map
4677 remap_debug_filename): Move to...
4678 * file-prefix-map.c: New file.
4679 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4680 generalize, get rid of alloca(), use strrchr() instead of strchr().
4681 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4682 Implement in terms of add_prefix_map().
4683 (remap_macro_filename, remap_debug_filename): Implement in term of
4685 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4686 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4687 * dbxout.c: Include file-prefix-map.h.
4688 * varasm.c: Likewise.
4689 * vmsdbgout.c: Likewise.
4690 * xcoffout.c: Likewise.
4691 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4692 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4693 * doc/invoke.texi (-ffile-prefix-map): Document.
4694 (-fdebug-prefix-map): Update description.
4696 2018-01-18 Martin Liska <mliska@suse.cz>
4698 * config/i386/i386.c (indirect_thunk_name): Document that also
4700 (output_indirect_thunk): Document why both instructions
4701 (pause and lfence) are generated.
4703 2018-01-18 Richard Biener <rguenther@suse.de>
4705 PR tree-optimization/83887
4706 * graphite-scop-detection.c
4707 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4708 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4709 (scop_detection::merge_sese): Re-implement with a flood-fill
4710 algorithm that properly finds a SESE region if it exists.
4712 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4715 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4716 pointer_diff optimizations use view_convert instead of convert.
4718 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4720 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4721 Generate different code for -mno-speculate-indirect-jumps.
4722 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4723 (*call_indirect_aix<mode>): Disable for
4724 -mno-speculate-indirect-jumps.
4725 (*call_indirect_aix<mode>_nospec): New define_insn.
4726 (*call_value_indirect_aix<mode>): Disable for
4727 -mno-speculate-indirect-jumps.
4728 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4729 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4730 -mno-speculate-indirect-jumps.
4731 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4733 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4735 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4736 long double type, set the flags for noting the default long double
4737 type, even if we don't pass or return a long double type.
4739 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4742 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4745 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4747 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4748 support for merge[hl].
4749 (fold_mergehl_helper): New helper function.
4750 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4751 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4752 (altivec_vmrglw_direct): Add xxmrglw insn.
4754 2018-01-17 Andrew Waterman <andrew@sifive.com>
4756 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4757 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4759 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4762 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4763 call the lto_location_cache before reading the
4764 DECL_SOURCE_LOCATION of the types.
4766 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4767 Richard Sandiford <richard.sandiford@linaro.org>
4769 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4770 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4771 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4772 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4773 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4775 * config/aarch64/constraints.md (aarch64_movti_operand):
4777 * config/aarch64/predicates.md (Uti): Add new constraint.
4779 2018-01-17 Carl Love <cel@us.ibm.com>
4781 * config/rs6000/vsx.md (define_expand xl_len_r,
4782 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4783 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4785 (define_expand, define_insn): Move the shift left from the
4786 define_insn to the define_expand for lxvl and stxvl instructions.
4787 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4788 and XL_LEN_R definitions to PURE.
4790 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4792 * config/i386/i386.c (indirect_thunk_name): Declare regno
4793 as unsigned int. Compare regno with INVALID_REGNUM.
4794 (output_indirect_thunk): Ditto.
4795 (output_indirect_thunk_function): Ditto.
4796 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4797 in the call to output_indirect_thunk_function.
4799 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4802 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4803 rather than the size of inner_type to determine the stack slot size
4804 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4806 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4809 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4812 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4814 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4815 endian Linux systems to optionally enable multilibs for selecting
4816 the long double type if the user configured an explicit type.
4817 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4818 have no long double multilibs if not defined.
4819 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4820 warn if the user used -mabi={ieee,ibm}longdouble and we built
4821 multilibs for long double.
4822 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4823 appropriate multilib option.
4824 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4826 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4827 for building long double multilibs.
4828 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4830 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4832 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4835 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4837 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4840 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4843 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4846 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4848 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4849 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4851 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4853 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4854 different rtl trees depending on TARGET_64BIT.
4855 (rs6000_gen_lvx): Likewise.
4857 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4859 * config/visium/visium.md (nop): Tweak comment.
4860 (hazard_nop): Likewise.
4862 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4864 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4865 -mspeculate-indirect-jumps.
4866 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4867 for -mno-speculate-indirect-jumps.
4868 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4869 (*call_value_indirect_elfv2<mode>): Disable for
4870 -mno-speculate-indirect-jumps.
4871 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4872 (indirect_jump): Emit different RTL for
4873 -mno-speculate-indirect-jumps.
4874 (*indirect_jump<mode>): Disable for
4875 -mno-speculate-indirect-jumps.
4876 (*indirect_jump<mode>_nospec): New define_insn.
4877 (tablejump): Emit different RTL for
4878 -mno-speculate-indirect-jumps.
4879 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4880 (tablejumpsi_nospec): New define_expand.
4881 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4882 (tablejumpdi_nospec): New define_expand.
4883 (*tablejump<mode>_internal1): Disable for
4884 -mno-speculate-indirect-jumps.
4885 (*tablejump<mode>_internal1_nospec): New define_insn.
4886 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4889 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4891 * caller-save.c (insert_save): Drop unnecessary parameter. All
4894 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4895 Richard Biener <rguenth@suse.de>
4898 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4899 return early, inline manually is_gimple_sizepos. Make sure if we
4900 call gimplify_expr we don't end up with a gimple constant.
4901 * tree.c (variably_modified_type_p): Don't return true for
4902 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4903 * gimplify.h (is_gimple_sizepos): Remove.
4905 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4907 PR tree-optimization/83857
4908 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4909 vectorizable_live_operation for pure SLP statements.
4910 (vectorizable_live_operation): Handle PHIs.
4912 2018-01-16 Richard Biener <rguenther@suse.de>
4914 PR tree-optimization/83867
4915 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4916 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4918 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4921 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4922 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4923 If off is not INTEGER_CST, issue a may not be aligned warning
4924 rather than isn't aligned. Use isn%'t rather than isn't.
4925 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4927 <case MULT_EXPR>: Improve the case when bottom and one of the
4928 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4929 operand, in that case check if the other operand is multiple of
4930 bottom divided by the INTEGER_CST operand.
4932 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4935 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4936 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4937 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4938 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4939 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4940 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4941 (pa_function_arg_size): New function.
4943 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4945 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4946 in a separate statement.
4948 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4950 PR tree-optimization/83847
4951 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4952 group gathers and scatters.
4954 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4956 PR rtl-optimization/86620
4957 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4959 PR rtl-optimization/83213
4960 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4961 to last if both are JUMP_INSNs.
4963 PR tree-optimization/83843
4964 * gimple-ssa-store-merging.c
4965 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4966 store_immediate_info for bswap/nop orig_stores.
4968 2018-01-15 Andrew Waterman <andrew@sifive.com>
4970 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4972 <UDIV>: Increase cost if !TARGET_DIV.
4974 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4976 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4977 (define_attr "cr_logical_3op"): New.
4978 (cceq_ior_compare): Adjust.
4979 (cceq_ior_compare_complement): Adjust.
4980 (*cceq_rev_compare): Adjust.
4981 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4982 (is_cracked_insn): Adjust.
4983 (insn_must_be_first_in_group): Adjust.
4984 * config/rs6000/40x.md: Adjust.
4985 * config/rs6000/440.md: Adjust.
4986 * config/rs6000/476.md: Adjust.
4987 * config/rs6000/601.md: Adjust.
4988 * config/rs6000/603.md: Adjust.
4989 * config/rs6000/6xx.md: Adjust.
4990 * config/rs6000/7450.md: Adjust.
4991 * config/rs6000/7xx.md: Adjust.
4992 * config/rs6000/8540.md: Adjust.
4993 * config/rs6000/cell.md: Adjust.
4994 * config/rs6000/e300c2c3.md: Adjust.
4995 * config/rs6000/e500mc.md: Adjust.
4996 * config/rs6000/e500mc64.md: Adjust.
4997 * config/rs6000/e5500.md: Adjust.
4998 * config/rs6000/e6500.md: Adjust.
4999 * config/rs6000/mpc.md: Adjust.
5000 * config/rs6000/power4.md: Adjust.
5001 * config/rs6000/power5.md: Adjust.
5002 * config/rs6000/power6.md: Adjust.
5003 * config/rs6000/power7.md: Adjust.
5004 * config/rs6000/power8.md: Adjust.
5005 * config/rs6000/power9.md: Adjust.
5006 * config/rs6000/rs64.md: Adjust.
5007 * config/rs6000/titan.md: Adjust.
5009 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5011 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5012 ix86_indirect_branch_register logic.
5014 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5016 * config/i386/constraints.md (Bs): Update
5017 ix86_indirect_branch_register check. Don't check
5018 ix86_indirect_branch_register with GOT_memory_operand.
5020 * config/i386/predicates.md (GOT_memory_operand): Don't check
5021 ix86_indirect_branch_register here.
5022 (GOT32_symbol_operand): Likewise.
5024 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5026 * config/i386/predicates.md (constant_call_address_operand):
5027 Rewrite ix86_indirect_branch_register logic.
5028 (sibcall_insn_operand): Likewise.
5030 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5032 * config/i386/constraints.md (Bs): Replace
5033 ix86_indirect_branch_thunk_register with
5034 ix86_indirect_branch_register.
5036 * config/i386/i386.md (indirect_jump): Likewise.
5037 (tablejump): Likewise.
5038 (*sibcall_memory): Likewise.
5039 (*sibcall_value_memory): Likewise.
5040 Peepholes of indirect call and jump via memory: Likewise.
5041 * config/i386/i386.opt: Likewise.
5042 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5043 (GOT_memory_operand): Likewise.
5044 (call_insn_operand): Likewise.
5045 (sibcall_insn_operand): Likewise.
5046 (GOT32_symbol_operand): Likewise.
5048 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5051 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5052 type rather than type addr's type points to.
5053 (expand_omp_atomic_mutex): Likewise.
5054 (expand_omp_atomic): Likewise.
5056 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5059 * config/i386/i386.c (output_indirect_thunk_function): Use
5060 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5061 for __x86_return_thunk.
5063 2018-01-15 Richard Biener <rguenther@suse.de>
5066 * expmed.c (extract_bit_field_1): Fix typo.
5068 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5071 * config/arm/iterators.md (VF): New mode iterator.
5072 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5073 Remove integer-related logic from pattern.
5074 (neon_vabd<mode>_3): Likewise.
5076 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5079 * common.opt (fstrict-overflow): No longer an alias.
5080 (fwrapv-pointer): New option.
5081 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5082 also for pointer types based on flag_wrapv_pointer.
5083 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5084 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5085 opts->x_flag_wrapv got set.
5086 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5087 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5088 POINTER_TYPE_OVERFLOW_UNDEFINED.
5089 * match.pd: Likewise in address comparison pattern.
5090 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5092 2018-01-15 Richard Biener <rguenther@suse.de>
5095 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5096 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5097 Reset type names to their identifier if their TYPE_DECL doesn't
5098 have linkage (and thus is used for ODR and devirt).
5099 (save_debug_info_for_decl): Remove.
5100 (save_debug_info_for_type): Likewise.
5101 (add_tree_to_fld_list): Adjust.
5102 * tree-pretty-print.c (dump_generic_node): Make dumping of
5103 type names more robust.
5105 2018-01-15 Richard Biener <rguenther@suse.de>
5107 * BASE-VER: Bump to 8.0.1.
5109 2018-01-14 Martin Sebor <msebor@redhat.com>
5112 * builtins.c (check_access): Avoid warning when the no-warning bit
5115 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5117 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5118 * ira-color (allocno_hard_regs_compare): Likewise.
5120 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5123 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5124 Use .pushsection/.popsection.
5126 2018-01-14 Martin Sebor <msebor@redhat.com>
5129 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5131 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5133 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5134 entry from extra_headers.
5135 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5136 extra_headers, make the list bitwise identical to the i?86-*-* one.
5138 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5140 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5141 -mcmodel=large with -mindirect-branch=thunk,
5142 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5143 -mfunction-return=thunk-extern.
5144 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5145 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5146 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5148 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5150 * config/i386/i386.c (print_reg): Print the name of the full
5151 integer register without '%'.
5152 (ix86_print_operand): Handle 'V'.
5153 * doc/extend.texi: Document 'V' modifier.
5155 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5157 * config/i386/constraints.md (Bs): Disallow memory operand for
5158 -mindirect-branch-register.
5160 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5161 (GOT_memory_operand): Likewise.
5162 (call_insn_operand): Likewise.
5163 (sibcall_insn_operand): Likewise.
5164 (GOT32_symbol_operand): Likewise.
5165 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5166 for -mindirect-branch-register.
5167 (tablejump): Likewise.
5168 (*sibcall_memory): Likewise.
5169 (*sibcall_value_memory): Likewise.
5170 Disallow peepholes of indirect call and jump via memory for
5171 -mindirect-branch-register.
5172 (*call_pop): Replace m with Bw.
5173 (*call_value_pop): Likewise.
5174 (*sibcall_pop_memory): Replace m with Bs.
5175 * config/i386/i386.opt (mindirect-branch-register): New option.
5176 * doc/invoke.texi: Document -mindirect-branch-register option.
5178 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5180 * config/i386/i386-protos.h (ix86_output_function_return): New.
5181 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5182 set function_return_type.
5183 (indirect_thunk_name): Add ret_p to indicate thunk for function
5185 (output_indirect_thunk_function): Pass false to
5186 indirect_thunk_name.
5187 (ix86_output_indirect_branch_via_reg): Likewise.
5188 (ix86_output_indirect_branch_via_push): Likewise.
5189 (output_indirect_thunk_function): Create alias for function
5190 return thunk if regno < 0.
5191 (ix86_output_function_return): New function.
5192 (ix86_handle_fndecl_attribute): Handle function_return.
5193 (ix86_attribute_table): Add function_return.
5194 * config/i386/i386.h (machine_function): Add
5195 function_return_type.
5196 * config/i386/i386.md (simple_return_internal): Use
5197 ix86_output_function_return.
5198 (simple_return_internal_long): Likewise.
5199 * config/i386/i386.opt (mfunction-return=): New option.
5200 (indirect_branch): Mention -mfunction-return=.
5201 * doc/extend.texi: Document function_return function attribute.
5202 * doc/invoke.texi: Document -mfunction-return= option.
5204 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5206 * config/i386/i386-opts.h (indirect_branch): New.
5207 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5208 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5209 with local indirect jump when converting indirect call and jump.
5210 (ix86_set_indirect_branch_type): New.
5211 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5212 (indirectlabelno): New.
5213 (indirect_thunk_needed): Likewise.
5214 (indirect_thunk_bnd_needed): Likewise.
5215 (indirect_thunks_used): Likewise.
5216 (indirect_thunks_bnd_used): Likewise.
5217 (INDIRECT_LABEL): Likewise.
5218 (indirect_thunk_name): Likewise.
5219 (output_indirect_thunk): Likewise.
5220 (output_indirect_thunk_function): Likewise.
5221 (ix86_output_indirect_branch_via_reg): Likewise.
5222 (ix86_output_indirect_branch_via_push): Likewise.
5223 (ix86_output_indirect_branch): Likewise.
5224 (ix86_output_indirect_jmp): Likewise.
5225 (ix86_code_end): Call output_indirect_thunk_function if needed.
5226 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5228 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5229 (ix86_attribute_table): Add indirect_branch.
5230 * config/i386/i386.h (machine_function): Add indirect_branch_type
5231 and has_local_indirect_jump.
5232 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5234 (tablejump): Likewise.
5235 (*indirect_jump): Use ix86_output_indirect_jmp.
5236 (*tablejump_1): Likewise.
5237 (simple_return_indirect_internal): Likewise.
5238 * config/i386/i386.opt (mindirect-branch=): New option.
5239 (indirect_branch): New.
5242 (thunk-inline): Likewise.
5243 (thunk-extern): Likewise.
5244 * doc/extend.texi: Document indirect_branch function attribute.
5245 * doc/invoke.texi: Document -mindirect-branch= option.
5247 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5250 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5252 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5254 * ipa-inline.c (want_inline_small_function_p): Return false if
5255 inlining has already failed with CIF_FINAL_ERROR.
5256 (update_caller_keys): Call want_inline_small_function_p before
5258 (update_callee_keys): Likewise.
5260 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5262 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5264 (rs6000_quadword_masked_address_p): Likewise.
5265 (quad_aligned_load_p): Likewise.
5266 (quad_aligned_store_p): Likewise.
5267 (const_load_sequence_p): Add comment to describe the outer-most loop.
5268 (mimic_memory_attributes_and_flags): New function.
5269 (rs6000_gen_stvx): Likewise.
5270 (replace_swapped_aligned_store): Likewise.
5271 (rs6000_gen_lvx): Likewise.
5272 (replace_swapped_aligned_load): Likewise.
5273 (replace_swapped_load_constant): Capitalize argument name in
5274 comment describing this function.
5275 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5276 and stores that access quad-word aligned addresses and replace
5277 with stvx or lvx instructions when appropriate.
5278 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5279 New function prototype.
5280 (rs6000_quadword_masked_address_p): Likewise.
5281 (rs6000_gen_lvx): Likewise.
5282 (rs6000_gen_stvx): Likewise.
5283 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5284 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5285 when memory address is aligned.
5286 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5287 this split to select lvx instruction when memory address is aligned.
5288 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5289 instruction when memory address is aligned.
5290 (*vsx_le_perm_load_v16qi): Likewise.
5291 (four unnamed splitters): Modify to select the stvx instruction
5292 when memory is aligned.
5294 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5296 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5297 which appears in the queue multiple times.
5299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5300 Alan Hayward <alan.hayward@arm.com>
5301 David Sherwood <david.sherwood@arm.com>
5303 * tree-vectorizer.h (vec_lower_bound): New structure.
5304 (_loop_vec_info): Add check_nonzero and lower_bounds.
5305 (LOOP_VINFO_CHECK_NONZERO): New macro.
5306 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5307 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5308 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5309 fields. Make seg_len the distance travelled, not including the
5311 (dr_direction_indicator): Declare.
5312 (dr_zero_step_indicator): Likewise.
5313 (dr_known_forward_stride_p): Likewise.
5314 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5316 (runtime_alias_check_p): Allow runtime alias checks with
5318 (operator ==): Compare access_size and align.
5319 (prune_runtime_alias_test_list): Rework for new distinction between
5320 the access_size and seg_len.
5321 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5323 (get_segment_min_max): New function.
5324 (create_intersect_range_checks): Use it.
5325 (dr_step_indicator): New function.
5326 (dr_direction_indicator): Likewise.
5327 (dr_zero_step_indicator): Likewise.
5328 (dr_known_forward_stride_p): Likewise.
5329 * tree-loop-distribution.c (data_ref_segment_size): Return
5330 DR_STEP * (niters - 1).
5331 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5333 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5334 (vect_preserves_scalar_order_p): New function, split out from...
5335 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5336 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5337 (vect_vfa_access_size): New function.
5338 (vect_vfa_align): Likewise.
5339 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5340 (dump_lower_bound): New function.
5341 (vect_check_lower_bound): Likewise.
5342 (vect_small_gap_p): Likewise.
5343 (vectorizable_with_step_bound_p): Likewise.
5344 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5345 depencies if the vectorization factor is 1. Convert the checks
5346 for nonzero steps into checks on the bounds of DR_STEP. Try using
5347 a bunds check for variable steps if the minimum required step is
5348 relatively small. Update calls to the dr_with_seg_len
5349 constructor and to vect_compile_time_alias.
5350 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5352 (vect_loop_versioning): Call it.
5353 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5355 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5358 Alan Hayward <alan.hayward@arm.com>
5359 David Sherwood <david.sherwood@arm.com>
5361 * doc/sourcebuild.texi (vect_scatter_store): Document.
5362 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5364 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5366 * genopinit.c (main): Add supports_vec_scatter_store and
5367 supports_vec_scatter_store_cached to target_optabs.
5368 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5369 IFN_MASK_SCATTER_STORE.
5370 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5372 * internal-fn.h (internal_store_fn_p): Declare.
5373 (internal_fn_stored_value_index): Likewise.
5374 * internal-fn.c (scatter_store_direct): New macro.
5375 (expand_scatter_store_optab_fn): New function.
5376 (direct_scatter_store_optab_supported_p): New macro.
5377 (internal_store_fn_p): New function.
5378 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5379 IFN_MASK_SCATTER_STORE.
5380 (internal_fn_mask_index): Likewise.
5381 (internal_fn_stored_value_index): New function.
5382 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5384 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5385 * optabs-query.c (supports_vec_scatter_store_p): New function.
5386 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5387 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5388 true for scatter stores.
5389 (vect_gather_scatter_fn_p): Handle scatter stores too.
5390 (vect_check_gather_scatter): Consider using scatter stores if
5391 supports_vec_scatter_store_p.
5392 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5394 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5395 internal_fn_stored_value_index.
5396 (check_load_store_masking): Handle scatter stores too.
5397 (vect_get_store_rhs): Make public.
5398 (vectorizable_call): Use internal_store_fn_p.
5399 (vectorizable_store): Handle scatter store internal functions.
5400 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5401 when deciding whether the end of the group has been reached.
5402 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5403 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5404 (mask_scatter_store<mode>): New insns.
5406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5407 Alan Hayward <alan.hayward@arm.com>
5408 David Sherwood <david.sherwood@arm.com>
5410 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5411 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5412 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5414 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5415 Use vect_truncate_gather_scatter_offset if we can't treat the
5416 operation as a normal gather load or scatter store.
5417 (get_group_load_store_type): Take the gather_scatter_info
5418 as argument. Try using a gather load or scatter store for
5419 single-element groups.
5420 (get_load_store_type): Update calls to get_group_load_store_type
5421 and vect_use_strided_gather_scatters_p.
5423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5424 Alan Hayward <alan.hayward@arm.com>
5425 David Sherwood <david.sherwood@arm.com>
5427 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5428 optional tree argument.
5429 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5431 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5432 but continue to use the current value as a fallback.
5433 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5434 to compare the updates.
5435 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5436 (get_load_store_type): Use it when handling a strided access.
5437 (vect_get_strided_load_store_ops): New function.
5438 (vect_get_data_ptr_increment): Likewise.
5439 (vectorizable_load): Handle strided gather loads. Always pass
5440 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5443 Alan Hayward <alan.hayward@arm.com>
5444 David Sherwood <david.sherwood@arm.com>
5446 * doc/md.texi (gather_load@var{m}): Document.
5447 (mask_gather_load@var{m}): Likewise.
5448 * genopinit.c (main): Add supports_vec_gather_load and
5449 supports_vec_gather_load_cached to target_optabs.
5450 * optabs-tree.c (init_tree_optimization_optabs): Use
5451 ggc_cleared_alloc to allocate target_optabs.
5452 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5453 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5455 * internal-fn.h (internal_load_fn_p): Declare.
5456 (internal_gather_scatter_fn_p): Likewise.
5457 (internal_fn_mask_index): Likewise.
5458 (internal_gather_scatter_fn_supported_p): Likewise.
5459 * internal-fn.c (gather_load_direct): New macro.
5460 (expand_gather_load_optab_fn): New function.
5461 (direct_gather_load_optab_supported_p): New macro.
5462 (direct_internal_fn_optab): New function.
5463 (internal_load_fn_p): Likewise.
5464 (internal_gather_scatter_fn_p): Likewise.
5465 (internal_fn_mask_index): Likewise.
5466 (internal_gather_scatter_fn_supported_p): Likewise.
5467 * optabs-query.c (supports_at_least_one_mode_p): New function.
5468 (supports_vec_gather_load_p): Likewise.
5469 * optabs-query.h (supports_vec_gather_load_p): Declare.
5470 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5471 and memory_type field.
5472 (NUM_PATTERNS): Bump to 15.
5473 * tree-vect-data-refs.c: Include internal-fn.h.
5474 (vect_gather_scatter_fn_p): New function.
5475 (vect_describe_gather_scatter_call): Likewise.
5476 (vect_check_gather_scatter): Try using internal functions for
5477 gather loads. Recognize existing calls to a gather load function.
5478 (vect_analyze_data_refs): Consider using gather loads if
5479 supports_vec_gather_load_p.
5480 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5481 (vect_get_gather_scatter_offset_type): Likewise.
5482 (vect_convert_mask_for_vectype): Likewise.
5483 (vect_add_conversion_to_patterm): Likewise.
5484 (vect_try_gather_scatter_pattern): Likewise.
5485 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5486 (vect_vect_recog_func_ptrs): Add it.
5487 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5488 internal_fn_mask_index and internal_gather_scatter_fn_p.
5489 (check_load_store_masking): Take the gather_scatter_info as an
5490 argument and handle gather loads.
5491 (vect_get_gather_scatter_ops): New function.
5492 (vectorizable_call): Check internal_load_fn_p.
5493 (vectorizable_load): Likewise. Handle gather load internal
5495 (vectorizable_store): Update call to check_load_store_masking.
5496 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5497 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5498 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5499 (aarch64_gather_scale_operand_d): New predicates.
5500 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5501 (mask_gather_load<mode>): New insns.
5503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5504 Alan Hayward <alan.hayward@arm.com>
5505 David Sherwood <david.sherwood@arm.com>
5507 * optabs.def (fold_left_plus_optab): New optab.
5508 * doc/md.texi (fold_left_plus_@var{m}): Document.
5509 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5510 * internal-fn.c (fold_left_direct): Define.
5511 (expand_fold_left_optab_fn): Likewise.
5512 (direct_fold_left_optab_supported_p): Likewise.
5513 * fold-const-call.c (fold_const_fold_left): New function.
5514 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5515 * tree-parloops.c (valid_reduction_p): New function.
5516 (gather_scalar_reductions): Use it.
5517 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5518 (vect_finish_replace_stmt): Declare.
5519 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5520 (needs_fold_left_reduction_p): New function, split out from...
5521 (vect_is_simple_reduction): ...here. Accept reductions that
5522 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5523 (vect_force_simple_reduction): Also store the reduction type in
5524 the assignment's STMT_VINFO_REDUC_TYPE.
5525 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5526 (merge_with_identity): New function.
5527 (vect_expand_fold_left): Likewise.
5528 (vectorize_fold_left_reduction): Likewise.
5529 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5530 scalar phi in place for it. Check for target support and reject
5531 cases that would reassociate the operation. Defer the transform
5532 phase to vectorize_fold_left_reduction.
5533 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5534 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5535 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5539 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5540 call to ifc_temp_var.
5542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5543 Alan Hayward <alan.hayward@arm.com>
5544 David Sherwood <david.sherwood@arm.com>
5546 * target.def (legitimize_address_displacement): Take the original
5547 offset as a poly_int.
5548 * targhooks.h (default_legitimize_address_displacement): Update
5550 * targhooks.c (default_legitimize_address_displacement): Likewise.
5551 * doc/tm.texi: Regenerate.
5552 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5553 as an argument, moving assert of ad->disp == ad->disp_term to...
5554 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5555 Try calling targetm.legitimize_address_displacement before expanding
5556 the address rather than afterwards, and adjust for the new interface.
5557 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5558 Match the new hook interface. Handle SVE addresses.
5559 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5564 * Makefile.in (OBJS): Add early-remat.o.
5565 * target.def (select_early_remat_modes): New hook.
5566 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5567 * doc/tm.texi: Regenerate.
5568 * targhooks.h (default_select_early_remat_modes): Declare.
5569 * targhooks.c (default_select_early_remat_modes): New function.
5570 * timevar.def (TV_EARLY_REMAT): New timevar.
5571 * passes.def (pass_early_remat): New pass.
5572 * tree-pass.h (make_pass_early_remat): Declare.
5573 * early-remat.c: New file.
5574 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5576 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5579 Alan Hayward <alan.hayward@arm.com>
5580 David Sherwood <david.sherwood@arm.com>
5582 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5583 vfm1 with a bound_epilog parameter.
5584 (vect_do_peeling): Update calls accordingly, and move the prologue
5585 call earlier in the function. Treat the base bound_epilog as 0 for
5586 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5587 this base when peeling for gaps.
5588 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5589 with fully-masked loops.
5590 (vect_estimate_min_profitable_iters): Handle the single peeled
5591 iteration in that case.
5593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5594 Alan Hayward <alan.hayward@arm.com>
5595 David Sherwood <david.sherwood@arm.com>
5597 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5598 single-element interleaving even if the size is not a power of 2.
5599 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5600 accesses for single-element interleaving if the group size is
5603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5604 Alan Hayward <alan.hayward@arm.com>
5605 David Sherwood <david.sherwood@arm.com>
5607 * doc/md.texi (fold_extract_last_@var{m}): Document.
5608 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5609 * optabs.def (fold_extract_last_optab): New optab.
5610 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5611 * internal-fn.c (fold_extract_direct): New macro.
5612 (expand_fold_extract_optab_fn): Likewise.
5613 (direct_fold_extract_optab_supported_p): Likewise.
5614 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5615 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5616 EXTRACT_LAST_REDUCTION.
5617 (get_initial_def_for_reduction): Do not create an initial vector
5618 for EXTRACT_LAST_REDUCTION reductions.
5619 (vectorizable_reduction): Leave the scalar phi in place for
5620 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5621 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5622 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5623 transform phase to vectorizable_condition.
5624 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5626 (vect_finish_stmt_generation): ...here.
5627 (vect_finish_replace_stmt): New function.
5628 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5629 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5631 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5634 Alan Hayward <alan.hayward@arm.com>
5635 David Sherwood <david.sherwood@arm.com>
5637 * doc/md.texi (extract_last_@var{m}): Document.
5638 * optabs.def (extract_last_optab): New optab.
5639 * internal-fn.def (EXTRACT_LAST): New internal function.
5640 * internal-fn.c (cond_unary_direct): New macro.
5641 (expand_cond_unary_optab_fn): Likewise.
5642 (direct_cond_unary_optab_supported_p): Likewise.
5643 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5644 loops using EXTRACT_LAST.
5645 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5646 (extract_last_<mode>): ...this optab.
5647 (vec_extract<mode><Vel>): Update accordingly.
5649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5650 Alan Hayward <alan.hayward@arm.com>
5651 David Sherwood <david.sherwood@arm.com>
5653 * target.def (empty_mask_is_expensive): New hook.
5654 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5655 * doc/tm.texi: Regenerate.
5656 * targhooks.h (default_empty_mask_is_expensive): Declare.
5657 * targhooks.c (default_empty_mask_is_expensive): New function.
5658 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5659 if the target says that empty masks are expensive.
5660 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5662 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5665 Alan Hayward <alan.hayward@arm.com>
5666 David Sherwood <david.sherwood@arm.com>
5668 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5669 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5670 (vect_use_loop_mask_for_alignment_p): New function.
5671 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5672 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5673 niters_skip argument. Make sure that the first niters_skip elements
5674 of the first iteration are inactive.
5675 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5676 Update call to vect_set_loop_masks_directly.
5677 (get_misalign_in_elems): New function, split out from...
5678 (vect_gen_prolog_loop_niters): ...here.
5679 (vect_update_init_of_dr): Take a code argument that specifies whether
5680 the adjustment should be added or subtracted.
5681 (vect_update_init_of_drs): Likewise.
5682 (vect_prepare_for_masked_peels): New function.
5683 (vect_do_peeling): Skip prologue peeling if we're using a mask
5684 instead. Update call to vect_update_inits_of_drs.
5685 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5687 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5688 alignment. Do not include the number of peeled iterations in
5689 the minimum threshold in that case.
5690 (vectorizable_induction): Adjust the start value down by
5691 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5692 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5693 Take the number of skipped iterations into account when calculating
5695 * tree-vect-stmts.c (vect_gen_while_not): New function.
5697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5698 Alan Hayward <alan.hayward@arm.com>
5699 David Sherwood <david.sherwood@arm.com>
5701 * doc/sourcebuild.texi (vect_fully_masked): Document.
5702 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5704 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5706 (vect_analyze_loop_2): ...here. Don't check the vectorization
5707 factor against the number of loop iterations if the loop is
5710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5711 Alan Hayward <alan.hayward@arm.com>
5712 David Sherwood <david.sherwood@arm.com>
5714 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5715 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5716 (dump_groups): Update accordingly.
5717 (iv_use::mem_type): New member variable.
5718 (address_p): New function.
5719 (record_use): Add a mem_type argument and initialize the new
5721 (record_group_use): Add a mem_type argument. Use address_p.
5722 Remove obsolete null checks of base_object. Update call to record_use.
5723 (find_interesting_uses_op): Update call to record_group_use.
5724 (find_interesting_uses_cond): Likewise.
5725 (find_interesting_uses_address): Likewise.
5726 (get_mem_type_for_internal_fn): New function.
5727 (find_address_like_use): Likewise.
5728 (find_interesting_uses_stmt): Try find_address_like_use before
5729 calling find_interesting_uses_op.
5730 (addr_offset_valid_p): Use the iv mem_type field as the type
5731 of the addressed memory.
5732 (add_autoinc_candidates): Likewise.
5733 (get_address_cost): Likewise.
5734 (split_small_address_groups_p): Use address_p.
5735 (split_address_groups): Likewise.
5736 (add_iv_candidate_for_use): Likewise.
5737 (autoinc_possible_for_pair): Likewise.
5738 (rewrite_groups): Likewise.
5739 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5740 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5741 (get_alias_ptr_type_for_ptr_address): New function.
5742 (rewrite_use_address): Rewrite address uses in calls that were
5743 identified by find_address_like_use.
5745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5746 Alan Hayward <alan.hayward@arm.com>
5747 David Sherwood <david.sherwood@arm.com>
5749 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5751 * gimple-expr.h (is_gimple_addressable: Likewise.
5752 * gimple-expr.c (is_gimple_address): Likewise.
5753 * internal-fn.c (expand_call_mem_ref): New function.
5754 (expand_mask_load_optab_fn): Use it.
5755 (expand_mask_store_optab_fn): Likewise.
5757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5758 Alan Hayward <alan.hayward@arm.com>
5759 David Sherwood <david.sherwood@arm.com>
5761 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5762 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5763 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5764 (cond_umax@var{mode}): Document.
5765 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5766 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5767 (cond_umin_optab, cond_umax_optab): New optabs.
5768 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5769 (COND_IOR, COND_XOR): New internal functions.
5770 * internal-fn.h (get_conditional_internal_fn): Declare.
5771 * internal-fn.c (cond_binary_direct): New macro.
5772 (expand_cond_binary_optab_fn): Likewise.
5773 (direct_cond_binary_optab_supported_p): Likewise.
5774 (get_conditional_internal_fn): New function.
5775 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5776 Cope with reduction statements that are vectorized as calls rather
5778 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5779 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5780 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5781 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5782 (UNSPEC_COND_EOR): New unspecs.
5783 (optab): Add mappings for them.
5784 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5785 (sve_int_op, sve_fp_op): New int attributes.
5787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5788 Alan Hayward <alan.hayward@arm.com>
5789 David Sherwood <david.sherwood@arm.com>
5791 * optabs.def (while_ult_optab): New optab.
5792 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5793 * internal-fn.def (WHILE_ULT): New internal function.
5794 * internal-fn.h (direct_internal_fn_supported_p): New override
5795 that takes two types as argument.
5796 * internal-fn.c (while_direct): New macro.
5797 (expand_while_optab_fn): New function.
5798 (convert_optab_supported_p): Likewise.
5799 (direct_while_optab_supported_p): New macro.
5800 * wide-int.h (wi::udiv_ceil): New function.
5801 * tree-vectorizer.h (rgroup_masks): New structure.
5802 (vec_loop_masks): New typedef.
5803 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5805 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5806 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5807 (vect_max_vf): New function.
5808 (slpeel_make_loop_iterate_ntimes): Delete.
5809 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5810 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5811 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5812 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5813 internal-fn.h, stor-layout.h and optabs-query.h.
5814 (vect_set_loop_mask): New function.
5815 (add_preheader_seq): Likewise.
5816 (add_header_seq): Likewise.
5817 (interleave_supported_p): Likewise.
5818 (vect_maybe_permute_loop_masks): Likewise.
5819 (vect_set_loop_masks_directly): Likewise.
5820 (vect_set_loop_condition_masked): Likewise.
5821 (vect_set_loop_condition_unmasked): New function, split out from
5822 slpeel_make_loop_iterate_ntimes.
5823 (slpeel_make_loop_iterate_ntimes): Rename to..
5824 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5825 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5826 (vect_do_peeling): Update call accordingly.
5827 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5829 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5830 mask_compare_type, can_fully_mask_p and fully_masked_p.
5831 (release_vec_loop_masks): New function.
5832 (_loop_vec_info): Use it to free the loop masks.
5833 (can_produce_all_loop_masks_p): New function.
5834 (vect_get_max_nscalars_per_iter): Likewise.
5835 (vect_verify_full_masking): Likewise.
5836 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5837 retries, and free the mask rgroups before retrying. Check loop-wide
5838 reasons for disallowing fully-masked loops. Make the final decision
5839 about whether use a fully-masked loop or not.
5840 (vect_estimate_min_profitable_iters): Do not assume that peeling
5841 for the number of iterations will be needed for fully-masked loops.
5842 (vectorizable_reduction): Disable fully-masked loops.
5843 (vectorizable_live_operation): Likewise.
5844 (vect_halve_mask_nunits): New function.
5845 (vect_double_mask_nunits): Likewise.
5846 (vect_record_loop_mask): Likewise.
5847 (vect_get_loop_mask): Likewise.
5848 (vect_transform_loop): Handle the case in which the final loop
5849 iteration might handle a partial vector. Call vect_set_loop_condition
5850 instead of slpeel_make_loop_iterate_ntimes.
5851 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5852 (check_load_store_masking): New function.
5853 (prepare_load_store_mask): Likewise.
5854 (vectorizable_store): Handle fully-masked loops.
5855 (vectorizable_load): Likewise.
5856 (supportable_widening_operation): Use vect_halve_mask_nunits for
5858 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5859 (vect_gen_while): New function.
5860 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5861 (aarch64_uqdec<mode>): New insn.
5863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5864 Alan Hayward <alan.hayward@arm.com>
5865 David Sherwood <david.sherwood@arm.com>
5867 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5868 (reduc_xor_scal_optab): New optabs.
5869 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5870 (reduc_xor_scal_@var{m}): Document.
5871 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5872 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5874 * fold-const-call.c (fold_const_call): Handle them.
5875 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5876 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5877 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5878 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5879 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5880 (UNSPEC_XORV): New unspecs.
5881 (optab): Add entries for them.
5882 (BITWISEV): New int iterator.
5883 (bit_reduc_op): New int attributes.
5885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5886 Alan Hayward <alan.hayward@arm.com>
5887 David Sherwood <david.sherwood@arm.com>
5889 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5890 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5891 * optabs.def (vec_shl_insert_optab): New optab.
5892 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5893 (duplicate_and_interleave): Likewise.
5894 * tree-vect-loop.c: Include internal-fn.h.
5895 (neutral_op_for_slp_reduction): New function, split out from
5896 get_initial_defs_for_reduction.
5897 (get_initial_def_for_reduction): Handle option 2 for variable-length
5898 vectors by loading the neutral value into a vector and then shifting
5899 the initial value into element 0.
5900 (get_initial_defs_for_reduction): Replace the code argument with
5901 the neutral value calculated by neutral_op_for_slp_reduction.
5902 Use gimple_build_vector for constant-length vectors.
5903 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5904 but the first group_size elements have a neutral value.
5905 Use duplicate_and_interleave otherwise.
5906 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5907 Update call to get_initial_defs_for_reduction. Handle SLP
5908 reductions for variable-length vectors by creating one vector
5909 result for each scalar result, with the elements associated
5910 with other scalar results stubbed out with the neutral value.
5911 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5912 Require IFN_VEC_SHL_INSERT for double reductions on
5913 variable-length vectors, or SLP reductions that have
5914 a neutral value. Require can_duplicate_and_interleave_p
5915 support for variable-length unchained SLP reductions if there
5916 is no neutral value, such as for MIN/MAX reductions. Also require
5917 the number of vector elements to be a multiple of the number of
5918 SLP statements when doing variable-length unchained SLP reductions.
5919 Update call to vect_create_epilog_for_reduction.
5920 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5921 and remove initial values.
5922 (duplicate_and_interleave): Make public.
5923 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5924 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5927 Alan Hayward <alan.hayward@arm.com>
5928 David Sherwood <david.sherwood@arm.com>
5930 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5931 (can_duplicate_and_interleave_p): New function.
5932 (vect_get_and_check_slp_defs): Take the vector of statements
5933 rather than just the current one. Remove excess parentheses.
5934 Restriction rejectinon of vect_constant_def and vect_external_def
5935 for variable-length vectors to boolean types, or types for which
5936 can_duplicate_and_interleave_p is false.
5937 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5938 (duplicate_and_interleave): New function.
5939 (vect_get_constant_vectors): Use gimple_build_vector for
5940 constant-length vectors and suitable variable-length constant
5941 vectors. Use duplicate_and_interleave for other variable-length
5942 vectors. Don't defer the update when inserting new statements.
5944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5945 Alan Hayward <alan.hayward@arm.com>
5946 David Sherwood <david.sherwood@arm.com>
5948 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5949 min_profitable_iters doesn't go negative.
5951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5952 Alan Hayward <alan.hayward@arm.com>
5953 David Sherwood <david.sherwood@arm.com>
5955 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5956 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5957 * optabs.def (vec_mask_load_lanes_optab): New optab.
5958 (vec_mask_store_lanes_optab): Likewise.
5959 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5960 (MASK_STORE_LANES): Likewise.
5961 * internal-fn.c (mask_load_lanes_direct): New macro.
5962 (mask_store_lanes_direct): Likewise.
5963 (expand_mask_load_optab_fn): Handle masked operations.
5964 (expand_mask_load_lanes_optab_fn): New macro.
5965 (expand_mask_store_optab_fn): Handle masked operations.
5966 (expand_mask_store_lanes_optab_fn): New macro.
5967 (direct_mask_load_lanes_optab_supported_p): Likewise.
5968 (direct_mask_store_lanes_optab_supported_p): Likewise.
5969 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5971 (vect_load_lanes_supported): Likewise.
5972 * tree-vect-data-refs.c (strip_conversion): New function.
5973 (can_group_stmts_p): Likewise.
5974 (vect_analyze_data_ref_accesses): Use it instead of checking
5975 for a pair of assignments.
5976 (vect_store_lanes_supported): Take a masked_p parameter.
5977 (vect_load_lanes_supported): Likewise.
5978 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5979 vect_store_lanes_supported and vect_load_lanes_supported.
5980 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5981 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5982 parameter. Don't allow gaps for masked accesses.
5983 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5984 and vect_load_lanes_supported.
5985 (get_load_store_type): Take a masked_p parameter and update
5986 call to get_group_load_store_type.
5987 (vectorizable_store): Update call to get_load_store_type.
5988 Handle IFN_MASK_STORE_LANES.
5989 (vectorizable_load): Update call to get_load_store_type.
5990 Handle IFN_MASK_LOAD_LANES.
5992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5993 Alan Hayward <alan.hayward@arm.com>
5994 David Sherwood <david.sherwood@arm.com>
5996 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5998 * config/aarch64/aarch64-protos.h
5999 (aarch64_sve_struct_memory_operand_p): Declare.
6000 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6001 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6002 (VPRED, vpred): Handle SVE structure modes.
6003 * config/aarch64/constraints.md (Utx): New constraint.
6004 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6005 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6006 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6007 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6008 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6009 structure modes. Split into pieces after RA.
6010 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6011 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6013 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6014 SVE structure modes.
6015 (aarch64_classify_address): Likewise.
6016 (sizetochar): Move earlier in file.
6017 (aarch64_print_operand): Handle SVE register lists.
6018 (aarch64_array_mode): New function.
6019 (aarch64_sve_struct_memory_operand_p): Likewise.
6020 (TARGET_ARRAY_MODE): Redefine.
6022 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6023 Alan Hayward <alan.hayward@arm.com>
6024 David Sherwood <david.sherwood@arm.com>
6026 * target.def (array_mode): New target hook.
6027 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6028 * doc/tm.texi: Regenerate.
6029 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6030 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6031 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6033 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6037 Alan Hayward <alan.hayward@arm.com>
6038 David Sherwood <david.sherwood@arm.com>
6040 * fold-const.c (fold_binary_loc): Check the argument types
6041 rather than the result type when testing for a vector operation.
6043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6045 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6046 * doc/tm.texi: Regenerate.
6048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6049 Alan Hayward <alan.hayward@arm.com>
6050 David Sherwood <david.sherwood@arm.com>
6052 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6053 (sve): Document new AArch64 extension.
6054 * doc/md.texi (w): Extend the description of the AArch64
6055 constraint to include SVE vectors.
6056 (Upl, Upa): Document new AArch64 predicate constraints.
6057 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6059 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6060 (msve-vector-bits=): New option.
6061 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6062 SVE when these are disabled.
6063 (sve): New extension.
6064 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6065 modes. Adjust their number of units based on aarch64_sve_vg.
6066 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6067 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6068 aarch64_addr_query_type.
6069 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6070 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6071 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6072 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6073 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6074 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6075 (aarch64_simd_imm_zero_p): Delete.
6076 (aarch64_check_zero_based_sve_index_immediate): Declare.
6077 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6078 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6079 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6080 (aarch64_sve_float_mul_immediate_p): Likewise.
6081 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6083 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6084 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6085 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6086 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6087 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6088 (aarch64_regmode_natural_size): Likewise.
6089 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6090 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6092 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6093 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6094 for VG and the SVE predicate registers.
6095 (V_ALIASES): Add a "z"-prefixed alias.
6096 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6097 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6098 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6099 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6100 (REG_CLASS_NAMES): Add entries for them.
6101 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6102 and the predicate registers.
6103 (aarch64_sve_vg): Declare.
6104 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6105 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6106 (REGMODE_NATURAL_SIZE): Define.
6107 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6109 * config/aarch64/aarch64.c: Include cfgrtl.h.
6110 (simd_immediate_info): Add a constructor for series vectors,
6111 and an associated step field.
6112 (aarch64_sve_vg): New variable.
6113 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6114 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6115 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6116 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6117 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6118 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6119 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6120 (aarch64_get_mask_mode): New functions.
6121 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6122 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6123 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6124 predicate modes and predicate registers. Explicitly restrict
6125 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6126 to store a vector mode if it is recognized by
6127 aarch64_classify_vector_mode.
6128 (aarch64_regmode_natural_size): New function.
6129 (aarch64_hard_regno_caller_save_mode): Return the original mode
6131 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6132 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6133 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6134 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6136 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6137 does not overlap dest if the function is frame-related. Handle
6139 (aarch64_split_add_offset): New function.
6140 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6141 them aarch64_add_offset.
6142 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6143 and update call to aarch64_sub_sp.
6144 (aarch64_add_cfa_expression): New function.
6145 (aarch64_expand_prologue): Pass extra temporary registers to the
6146 functions above. Handle the case in which we need to emit new
6147 DW_CFA_expressions for registers that were originally saved
6148 relative to the stack pointer, but now have to be expressed
6149 relative to the frame pointer.
6150 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6152 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6153 IP0 and IP1 values for SVE frames.
6154 (aarch64_expand_vec_series): New function.
6155 (aarch64_expand_sve_widened_duplicate): Likewise.
6156 (aarch64_expand_sve_const_vector): Likewise.
6157 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6158 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6159 into the register, rather than emitting a SET directly.
6160 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6161 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6162 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6163 (offset_9bit_signed_scaled_p): New functions.
6164 (aarch64_replicate_bitmask_imm): New function.
6165 (aarch64_bitmask_imm): Use it.
6166 (aarch64_cannot_force_const_mem): Reject expressions involving
6167 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6168 (aarch64_classify_index): Handle SVE indices, by requiring
6169 a plain register index with a scale that matches the element size.
6170 (aarch64_classify_address): Handle SVE addresses. Assert that
6171 the mode of the address is VOIDmode or an integer mode.
6172 Update call to aarch64_classify_symbol.
6173 (aarch64_classify_symbolic_expression): Update call to
6174 aarch64_classify_symbol.
6175 (aarch64_const_vec_all_in_range_p): New function.
6176 (aarch64_print_vector_float_operand): Likewise.
6177 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6178 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6179 and the FP immediates 1.0 and 0.5.
6180 (aarch64_print_address_internal): Handle SVE addresses.
6181 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6182 (aarch64_regno_regclass): Handle predicate registers.
6183 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6185 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6186 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6187 (aarch64_convert_sve_vector_bits): New function.
6188 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6189 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6191 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6192 Handle SVE vector and predicate modes. Accept VL-based constants
6193 that need only one temporary register, and VL offsets that require
6194 no temporary registers.
6195 (aarch64_conditional_register_usage): Mark the predicate registers
6196 as fixed if SVE isn't available.
6197 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6198 Return true for SVE vector and predicate modes.
6199 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6200 rather than an unsigned int. Handle SVE modes.
6201 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6203 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6205 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6206 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6207 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6208 (aarch64_sve_float_mul_immediate_p): New functions.
6209 (aarch64_sve_valid_immediate): New function.
6210 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6211 Explicitly reject structure modes. Check for INDEX constants.
6212 Handle PTRUE and PFALSE constants.
6213 (aarch64_check_zero_based_sve_index_immediate): New function.
6214 (aarch64_simd_imm_zero_p): Delete.
6215 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6216 vector modes. Accept constants in the range of CNT[BHWD].
6217 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6218 ask for an Advanced SIMD mode.
6219 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6220 (aarch64_simd_vector_alignment): Handle SVE predicates.
6221 (aarch64_vectorize_preferred_vector_alignment): New function.
6222 (aarch64_simd_vector_alignment_reachable): Use it instead of
6224 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6225 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6227 (MAX_VECT_LEN): Delete.
6228 (expand_vec_perm_d): Add a vec_flags field.
6229 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6230 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6231 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6233 (aarch64_evpc_rev): Rename to...
6234 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6235 (aarch64_evpc_rev_global): New function.
6236 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6237 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6239 (aarch64_evpc_sve_tbl): New function.
6240 (aarch64_expand_vec_perm_const_1): Update after rename of
6241 aarch64_evpc_rev. Handle SVE permutes too, trying
6242 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6243 than aarch64_evpc_tbl.
6244 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6245 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6246 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6247 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6248 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6249 (aarch64_expand_sve_vcond): New functions.
6250 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6251 of aarch64_vector_mode_p.
6252 (aarch64_dwarf_poly_indeterminate_value): New function.
6253 (aarch64_compute_pressure_classes): Likewise.
6254 (aarch64_can_change_mode_class): Likewise.
6255 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6256 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6257 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6258 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6259 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6260 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6261 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6262 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6264 (Dn, Dl, Dr): Accept const as well as const_vector.
6265 (Dz): Likewise. Compare against CONST0_RTX.
6266 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6267 of "vector" where appropriate.
6268 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6269 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6270 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6271 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6272 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6273 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6274 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6275 (v_int_equiv): Extend to SVE modes.
6276 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6278 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6279 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6280 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6281 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6282 (SVE_COND_FP_CMP): New int iterators.
6283 (perm_hilo): Handle the new unpack unspecs.
6284 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6286 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6287 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6288 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6289 (aarch64_equality_operator, aarch64_constant_vector_operand)
6290 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6291 (aarch64_sve_nonimmediate_operand): Likewise.
6292 (aarch64_sve_general_operand): Likewise.
6293 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6294 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6295 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6296 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6297 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6298 (aarch64_sve_float_arith_immediate): Likewise.
6299 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6300 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6301 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6302 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6303 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6304 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6305 (aarch64_sve_float_arith_operand): Likewise.
6306 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6307 (aarch64_sve_float_mul_operand): Likewise.
6308 (aarch64_sve_vec_perm_operand): Likewise.
6309 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6310 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6311 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6312 as well as const_vector.
6313 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6314 in file. Use CONST0_RTX and CONSTM1_RTX.
6315 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6316 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6317 Use aarch64_simd_imm_zero.
6318 * config/aarch64/aarch64-sve.md: New file.
6319 * config/aarch64/aarch64.md: Include it.
6320 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6321 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6322 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6323 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6324 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6325 (sve): New attribute.
6326 (enabled): Disable instructions with the sve attribute unless
6328 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6329 aarch64_expand_mov_immediate.
6330 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6331 CNT[BHSD] immediates.
6332 (movti): Split CONST_POLY_INT moves into two halves.
6333 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6334 Split additions that need a temporary here if the destination
6335 is the stack pointer.
6336 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6337 (*add<mode>3_poly_1): New instruction.
6338 (set_clobber_cc): New expander.
6340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6342 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6343 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6344 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6345 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6346 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6347 Change innermode from fixed_mode_size to machine_mode.
6348 (simplify_subreg): Update call accordingly. Handle a constant-sized
6349 subreg of a variable-length CONST_VECTOR.
6351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6352 Alan Hayward <alan.hayward@arm.com>
6353 David Sherwood <david.sherwood@arm.com>
6355 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6356 (add_offset_to_base): New function, split out from...
6357 (create_mem_ref): ...here. When handling a scale other than 1,
6358 check first whether the address is valid without the offset.
6359 Add it into the base if so, leaving the index and scale as-is.
6361 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6364 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6365 fold_for_warn before checking if arg2 is INTEGER_CST.
6367 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6369 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6370 (store_multiple_operation): Delete.
6371 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6372 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6373 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6374 guarded by TARGET_STRING.
6375 (rs6000_output_load_multiple): Delete.
6376 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6377 OPTION_MASK_STRING / TARGET_STRING handling.
6378 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6379 (const rs6000_opt_masks) <"string">: Change mask to 0.
6380 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6381 (MASK_STRING): Delete.
6382 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6384 (load_multiple): Delete.
6391 (store_multiple): Delete.
6398 (movmemsi_8reg): Delete.
6399 (corresponding unnamed define_insn): Delete.
6400 (movmemsi_6reg): Delete.
6401 (corresponding unnamed define_insn): Delete.
6402 (movmemsi_4reg): Delete.
6403 (corresponding unnamed define_insn): Delete.
6404 (movmemsi_2reg): Delete.
6405 (corresponding unnamed define_insn): Delete.
6406 (movmemsi_1reg): Delete.
6407 (corresponding unnamed define_insn): Delete.
6408 * config/rs6000/rs6000.opt (mno-string): New.
6409 (mstring): Replace by deprecation warning stub.
6410 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6412 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6414 * regrename.c (regrename_do_replace): If replacing the same
6415 reg multiple times, try to reuse last created gen_raw_REG.
6418 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6419 main to workaround a bug in GDB.
6421 2018-01-12 Tom de Vries <tom@codesourcery.com>
6424 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6426 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6428 PR rtl-optimization/80481
6429 * ira-color.c (get_cap_member): New function.
6430 (allocnos_conflict_by_live_ranges_p): Use it.
6431 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6432 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6434 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6437 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6438 (*saddl_se_1): Ditto.
6440 (*ssubl_se_1): Ditto.
6442 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6444 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6445 rather than wi::to_widest for DR_INITs.
6446 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6447 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6448 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6450 (vect_analyze_group_access_1): Note that here.
6452 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6454 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6455 polynomial type sizes.
6457 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6459 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6460 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6461 (gimple_add_tmp_var): Likewise.
6463 2018-01-12 Martin Liska <mliska@suse.cz>
6465 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6466 (gimple_alloc_sizes): Likewise.
6467 (dump_gimple_statistics): Use PRIu64 in printf format.
6468 * gimple.h: Change uint64_t to int.
6470 2018-01-12 Martin Liska <mliska@suse.cz>
6472 * tree-core.h: Use uint64_t instead of int.
6473 * tree.c (tree_node_counts): Likewise.
6474 (tree_node_sizes): Likewise.
6475 (dump_tree_statistics): Use PRIu64 in printf format.
6477 2018-01-12 Martin Liska <mliska@suse.cz>
6479 * Makefile.in: As qsort_chk is implemented in vec.c, add
6480 vec.o to linkage of gencfn-macros.
6481 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6482 passing the info to record_node_allocation_statistics.
6483 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6485 * ggc-common.c (struct ggc_usage): Add operator== and use
6486 it in operator< and compare function.
6487 * mem-stats.h (struct mem_usage): Likewise.
6488 * vec.c (struct vec_usage): Remove operator< and compare
6489 function. Can be simply inherited.
6491 2018-01-12 Martin Jambor <mjambor@suse.cz>
6494 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6495 * tree-ssa-math-opts.c: Include domwalk.h.
6496 (convert_mult_to_fma_1): New function.
6497 (fma_transformation_info): New type.
6498 (fma_deferring_state): Likewise.
6499 (cancel_fma_deferring): New function.
6500 (result_of_phi): Likewise.
6501 (last_fma_candidate_feeds_initial_phi): Likewise.
6502 (convert_mult_to_fma): Added deferring logic, split actual
6503 transformation to convert_mult_to_fma_1.
6504 (math_opts_dom_walker): New type.
6505 (math_opts_dom_walker::after_dom_children): New method, body moved
6506 here from pass_optimize_widening_mul::execute, added deferring logic
6508 (pass_optimize_widening_mul::execute): Moved most of code to
6509 math_opts_dom_walker::after_dom_children.
6510 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6511 * config/i386/i386.c (ix86_option_override_internal): Added
6512 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6514 2018-01-12 Richard Biener <rguenther@suse.de>
6517 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6518 inline instance vars.
6520 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6523 * config/rx/rx.c (rx_is_restricted_memory_address):
6526 2018-01-12 Richard Biener <rguenther@suse.de>
6528 PR tree-optimization/80846
6529 * target.def (split_reduction): New target hook.
6530 * targhooks.c (default_split_reduction): New function.
6531 * targhooks.h (default_split_reduction): Declare.
6532 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6533 target requests first reduce vectors by combining low and high
6535 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6536 (get_vectype_for_scalar_type_and_size): Export.
6537 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6538 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6539 * doc/tm.texi: Regenerate.
6540 * config/i386/i386.c (ix86_split_reduction): Implement
6541 TARGET_VECTORIZE_SPLIT_REDUCTION.
6543 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6546 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6547 in PIC mode except for TARGET_VXWORKS_RTP.
6548 * config/sparc/sparc.c: Include cfgrtl.h.
6549 (TARGET_INIT_PIC_REG): Define.
6550 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6551 (sparc_pic_register_p): New predicate.
6552 (sparc_legitimate_address_p): Use it.
6553 (sparc_legitimize_pic_address): Likewise.
6554 (sparc_delegitimize_address): Likewise.
6555 (sparc_mode_dependent_address_p): Likewise.
6556 (gen_load_pcrel_sym): Remove 4th parameter.
6557 (load_got_register): Adjust call to above. Remove obsolete stuff.
6558 (sparc_expand_prologue): Do not call load_got_register here.
6559 (sparc_flat_expand_prologue): Likewise.
6560 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6561 (sparc_use_pseudo_pic_reg): New function.
6562 (sparc_init_pic_reg): Likewise.
6563 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6564 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6566 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6568 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6569 Add item for branch_cost.
6571 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6573 PR rtl-optimization/83565
6574 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6575 not extend the result to a larger mode for rotate operations.
6576 (num_sign_bit_copies1): Likewise.
6578 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6581 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6583 Use values-Xc.o for -pedantic.
6584 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6586 2018-01-12 Martin Liska <mliska@suse.cz>
6589 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6591 (possible_polymorphic_call_targets): Use it.
6592 (ipa_devirt): Likewise.
6594 2018-01-12 Martin Liska <mliska@suse.cz>
6596 * profile-count.h (enum profile_quality): Use 0 as invalid
6597 enum value of profile_quality.
6599 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6601 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6602 -mext-string options.
6604 2018-01-12 Richard Biener <rguenther@suse.de>
6606 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6607 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6608 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6610 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6612 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6614 * configure.ac (--with-long-double-format): Add support for the
6615 configuration option to change the default long double format on
6617 * config.gcc (powerpc*-linux*-*): Likewise.
6618 * configure: Regenerate.
6619 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6620 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6621 used without modification.
6623 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6625 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6626 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6627 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6628 MISC_BUILTIN_SPEC_BARRIER.
6629 (rs6000_init_builtins): Likewise.
6630 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6632 (speculation_barrier): New define_insn.
6633 * doc/extend.texi: Document __builtin_speculation_barrier.
6635 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6638 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6639 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6640 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6642 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6643 integral modes instead of "ss" and "sd".
6644 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6645 vectors with 32-bit and 64-bit elements.
6646 (vecdupssescalarmodesuffix): New mode attribute.
6647 (vec_dup<mode>): Use it.
6649 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6652 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6653 frame if argument is passed on stack.
6655 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6658 * ree.c (combine_reaching_defs): Optimize also
6659 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6660 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6662 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6665 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6667 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6670 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6671 after they are computed.
6673 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6675 PR tree-optimization/83695
6676 * gimple-loop-linterchange.cc
6677 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6678 reset cached scev information after interchange.
6679 (pass_linterchange::execute): Remove call to scev_reset_htab.
6681 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6683 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6684 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6685 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6686 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6687 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6688 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6689 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6690 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6691 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6692 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6693 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6694 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6695 (V_lane_reg): Likewise.
6696 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6698 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6699 (vfmal_lane_low<mode>_intrinsic,
6700 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6701 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6702 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6703 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6704 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6705 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6707 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6709 * config/arm/arm-cpus.in (fp16fml): New feature.
6710 (ALL_SIMD): Add fp16fml.
6711 (armv8.2-a): Add fp16fml as an option.
6712 (armv8.3-a): Likewise.
6713 (armv8.4-a): Add fp16fml as part of fp16.
6714 * config/arm/arm.h (TARGET_FP16FML): Define.
6715 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6717 * config/arm/arm-modes.def (V2HF): Define.
6718 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6719 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6720 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6721 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6722 vfmsl_low, vfmsl_high): New set of builtins.
6723 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6724 (vfml_op): New code attribute.
6725 (VFMLHALVES): New int iterator.
6726 (VFML, VFMLSEL): New mode attributes.
6727 (V_reg): Define mapping for V2HF.
6728 (V_hi, V_lo): New mode attributes.
6729 (VF_constraint): Likewise.
6730 (vfml_half, vfml_half_selector): New int attributes.
6731 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6733 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6734 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6736 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6737 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6738 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6739 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6741 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6742 Document new effective target and option set.
6744 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6746 * config/arm/arm-cpus.in (armv8_4): New feature.
6747 (ARMv8_4a): New fgroup.
6748 (armv8.4-a): New arch.
6749 * config/arm/arm-tables.opt: Regenerate.
6750 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6751 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6752 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6753 Add matching rules for -march=armv8.4-a and extensions.
6754 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6756 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6759 * config/rx/rx.md (BW): New mode attribute.
6760 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6762 2018-01-11 Richard Biener <rguenther@suse.de>
6764 PR tree-optimization/83435
6765 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6766 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6767 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6769 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6770 Alan Hayward <alan.hayward@arm.com>
6771 David Sherwood <david.sherwood@arm.com>
6773 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6775 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6776 (aarch64_print_address_internal): Use it to check for a zero offset.
6778 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6779 Alan Hayward <alan.hayward@arm.com>
6780 David Sherwood <david.sherwood@arm.com>
6782 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6783 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6784 Return a poly_int64 rather than a HOST_WIDE_INT.
6785 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6786 rather than a HOST_WIDE_INT.
6787 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6788 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6789 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6790 final_offset from HOST_WIDE_INT to poly_int64.
6791 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6792 to_constant when getting the number of units in an Advanced SIMD
6794 (aarch64_builtin_vectorized_function): Check for a constant number
6796 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6798 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6799 attribute instead of GET_MODE_NUNITS.
6800 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6801 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6802 GET_MODE_SIZE for fixed-size registers.
6803 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6804 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6805 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6806 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6807 (aarch64_print_operand, aarch64_print_address_internal)
6808 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6809 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6810 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6811 Handle polynomial GET_MODE_SIZE.
6812 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6813 wider than SImode without modification.
6814 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6815 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6816 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6817 passing and returning SVE modes.
6818 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6819 rather than GEN_INT.
6820 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6821 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6822 (aarch64_allocate_and_probe_stack_space): Likewise.
6823 (aarch64_layout_frame): Cope with polynomial offsets.
6824 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6825 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6827 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6828 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6829 poly_int64 rather than a HOST_WIDE_INT.
6830 (aarch64_get_separate_components, aarch64_process_components)
6831 (aarch64_expand_prologue, aarch64_expand_epilogue)
6832 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6833 (aarch64_anchor_offset): New function, split out from...
6834 (aarch64_legitimize_address): ...here.
6835 (aarch64_builtin_vectorization_cost): Handle polynomial
6836 TYPE_VECTOR_SUBPARTS.
6837 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6839 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6840 number of elements from the PARALLEL rather than the mode.
6841 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6842 rather than GET_MODE_BITSIZE.
6843 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6844 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6845 (aarch64_expand_vec_perm_const_1): Handle polynomial
6846 d->perm.length () and d->perm elements.
6847 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6848 Apply to_constant to d->perm elements.
6849 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6850 polynomial CONST_VECTOR_NUNITS.
6851 (aarch64_move_pointer): Take amount as a poly_int64 rather
6853 (aarch64_progress_pointer): Avoid temporary variable.
6854 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6855 the mode attribute instead of GET_MODE.
6857 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6858 Alan Hayward <alan.hayward@arm.com>
6859 David Sherwood <david.sherwood@arm.com>
6861 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6862 x exists before using it.
6863 (aarch64_add_constant_internal): Rename to...
6864 (aarch64_add_offset_1): ...this. Replace regnum with separate
6865 src and dest rtxes. Handle the case in which they're different,
6866 including when the offset is zero. Replace scratchreg with an rtx.
6867 Use 2 additions if there is no spare register into which we can
6868 move a 16-bit constant.
6869 (aarch64_add_constant): Delete.
6870 (aarch64_add_offset): Replace reg with separate src and dest
6871 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6872 Use aarch64_add_offset_1.
6873 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6874 an rtx rather than an int. Take the delta as a poly_int64
6875 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6876 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6877 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6878 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6879 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6881 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6882 aarch64_add_constant.
6884 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6886 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6887 Use scalar_float_mode.
6889 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6891 * config/aarch64/aarch64-simd.md
6892 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6893 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6894 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6895 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6896 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6897 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6898 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6899 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6900 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6901 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6903 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6906 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6907 targ_options->x_arm_arch_string is non NULL.
6909 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6911 * config/aarch64/aarch64.h
6912 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6914 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6917 * expmed.c (emit_store_flag_force): Swap if const op0
6918 and change VOIDmode to mode of op0.
6920 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6922 PR rtl-optimization/83761
6923 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6924 than bytes to mode_for_size.
6926 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6929 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6930 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6933 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6936 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6937 when in layout mode.
6938 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6939 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6942 2018-01-10 Michael Collison <michael.collison@arm.com>
6944 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6945 * config/aarch64/aarch64-option-extension.def: Add
6946 AARCH64_OPT_EXTENSION of 'fp16fml'.
6947 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6948 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6949 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6950 * config/aarch64/constraints.md (Ui7): New constraint.
6951 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6952 (VFMLA_SEL_W): Ditto.
6955 (VFMLA16_LOW): New int iterator.
6956 (VFMLA16_HIGH): Ditto.
6957 (UNSPEC_FMLAL): New unspec.
6958 (UNSPEC_FMLSL): Ditto.
6959 (UNSPEC_FMLAL2): Ditto.
6960 (UNSPEC_FMLSL2): Ditto.
6961 (f16mac): New code attribute.
6962 * config/aarch64/aarch64-simd-builtins.def
6963 (aarch64_fmlal_lowv2sf): Ditto.
6964 (aarch64_fmlsl_lowv2sf): Ditto.
6965 (aarch64_fmlalq_lowv4sf): Ditto.
6966 (aarch64_fmlslq_lowv4sf): Ditto.
6967 (aarch64_fmlal_highv2sf): Ditto.
6968 (aarch64_fmlsl_highv2sf): Ditto.
6969 (aarch64_fmlalq_highv4sf): Ditto.
6970 (aarch64_fmlslq_highv4sf): Ditto.
6971 (aarch64_fmlal_lane_lowv2sf): Ditto.
6972 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6973 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6974 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6975 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6976 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6977 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6978 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6979 (aarch64_fmlal_lane_highv2sf): Ditto.
6980 (aarch64_fmlsl_lane_highv2sf): Ditto.
6981 (aarch64_fmlal_laneq_highv2sf): Ditto.
6982 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6983 (aarch64_fmlalq_lane_highv4sf): Ditto.
6984 (aarch64_fmlsl_lane_highv4sf): Ditto.
6985 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6986 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6987 * config/aarch64/aarch64-simd.md:
6988 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6989 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6990 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6991 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6992 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6993 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6994 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6995 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6996 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6997 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6998 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6999 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7000 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7001 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7002 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7003 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7004 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7005 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7006 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7007 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7008 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7009 (vfmlsl_low_u32): Ditto.
7010 (vfmlalq_low_u32): Ditto.
7011 (vfmlslq_low_u32): Ditto.
7012 (vfmlal_high_u32): Ditto.
7013 (vfmlsl_high_u32): Ditto.
7014 (vfmlalq_high_u32): Ditto.
7015 (vfmlslq_high_u32): Ditto.
7016 (vfmlal_lane_low_u32): Ditto.
7017 (vfmlsl_lane_low_u32): Ditto.
7018 (vfmlal_laneq_low_u32): Ditto.
7019 (vfmlsl_laneq_low_u32): Ditto.
7020 (vfmlalq_lane_low_u32): Ditto.
7021 (vfmlslq_lane_low_u32): Ditto.
7022 (vfmlalq_laneq_low_u32): Ditto.
7023 (vfmlslq_laneq_low_u32): Ditto.
7024 (vfmlal_lane_high_u32): Ditto.
7025 (vfmlsl_lane_high_u32): Ditto.
7026 (vfmlal_laneq_high_u32): Ditto.
7027 (vfmlsl_laneq_high_u32): Ditto.
7028 (vfmlalq_lane_high_u32): Ditto.
7029 (vfmlslq_lane_high_u32): Ditto.
7030 (vfmlalq_laneq_high_u32): Ditto.
7031 (vfmlslq_laneq_high_u32): Ditto.
7032 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7033 (AARCH64_FL_FOR_ARCH8_4): New.
7034 (AARCH64_ISA_F16FML): New ISA flag.
7035 (TARGET_F16FML): New feature flag for fp16fml.
7036 (doc/invoke.texi): Document new fp16fml option.
7038 2018-01-10 Michael Collison <michael.collison@arm.com>
7040 * config/aarch64/aarch64-builtins.c:
7041 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7042 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7043 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7044 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7045 (AARCH64_ISA_SHA3): New ISA flag.
7046 (TARGET_SHA3): New feature flag for sha3.
7047 * config/aarch64/iterators.md (sha512_op): New int attribute.
7048 (CRYPTO_SHA512): New int iterator.
7049 (UNSPEC_SHA512H): New unspec.
7050 (UNSPEC_SHA512H2): Ditto.
7051 (UNSPEC_SHA512SU0): Ditto.
7052 (UNSPEC_SHA512SU1): Ditto.
7053 * config/aarch64/aarch64-simd-builtins.def
7054 (aarch64_crypto_sha512hqv2di): New builtin.
7055 (aarch64_crypto_sha512h2qv2di): Ditto.
7056 (aarch64_crypto_sha512su0qv2di): Ditto.
7057 (aarch64_crypto_sha512su1qv2di): Ditto.
7058 (aarch64_eor3qv8hi): Ditto.
7059 (aarch64_rax1qv2di): Ditto.
7060 (aarch64_xarqv2di): Ditto.
7061 (aarch64_bcaxqv8hi): Ditto.
7062 * config/aarch64/aarch64-simd.md:
7063 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7064 (aarch64_crypto_sha512su0qv2di): Ditto.
7065 (aarch64_crypto_sha512su1qv2di): Ditto.
7066 (aarch64_eor3qv8hi): Ditto.
7067 (aarch64_rax1qv2di): Ditto.
7068 (aarch64_xarqv2di): Ditto.
7069 (aarch64_bcaxqv8hi): Ditto.
7070 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7071 (vsha512h2q_u64): Ditto.
7072 (vsha512su0q_u64): Ditto.
7073 (vsha512su1q_u64): Ditto.
7074 (veor3q_u16): Ditto.
7075 (vrax1q_u64): Ditto.
7077 (vbcaxq_u16): Ditto.
7078 * config/arm/types.md (crypto_sha512): New type attribute.
7079 (crypto_sha3): Ditto.
7080 (doc/invoke.texi): Document new sha3 option.
7082 2018-01-10 Michael Collison <michael.collison@arm.com>
7084 * config/aarch64/aarch64-builtins.c:
7085 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7086 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7087 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7088 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7089 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7090 (AARCH64_ISA_SM4): New ISA flag.
7091 (TARGET_SM4): New feature flag for sm4.
7092 * config/aarch64/aarch64-simd-builtins.def
7093 (aarch64_sm3ss1qv4si): Ditto.
7094 (aarch64_sm3tt1aq4si): Ditto.
7095 (aarch64_sm3tt1bq4si): Ditto.
7096 (aarch64_sm3tt2aq4si): Ditto.
7097 (aarch64_sm3tt2bq4si): Ditto.
7098 (aarch64_sm3partw1qv4si): Ditto.
7099 (aarch64_sm3partw2qv4si): Ditto.
7100 (aarch64_sm4eqv4si): Ditto.
7101 (aarch64_sm4ekeyqv4si): Ditto.
7102 * config/aarch64/aarch64-simd.md:
7103 (aarch64_sm3ss1qv4si): Ditto.
7104 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7105 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7106 (aarch64_sm4eqv4si): Ditto.
7107 (aarch64_sm4ekeyqv4si): Ditto.
7108 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7109 (sm3part_op): Ditto.
7110 (CRYPTO_SM3TT): Ditto.
7111 (CRYPTO_SM3PART): Ditto.
7112 (UNSPEC_SM3SS1): New unspec.
7113 (UNSPEC_SM3TT1A): Ditto.
7114 (UNSPEC_SM3TT1B): Ditto.
7115 (UNSPEC_SM3TT2A): Ditto.
7116 (UNSPEC_SM3TT2B): Ditto.
7117 (UNSPEC_SM3PARTW1): Ditto.
7118 (UNSPEC_SM3PARTW2): Ditto.
7119 (UNSPEC_SM4E): Ditto.
7120 (UNSPEC_SM4EKEY): Ditto.
7121 * config/aarch64/constraints.md (Ui2): New constraint.
7122 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7123 * config/arm/types.md (crypto_sm3): New type attribute.
7124 (crypto_sm4): Ditto.
7125 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7126 (vsm3tt1aq_u32): Ditto.
7127 (vsm3tt1bq_u32): Ditto.
7128 (vsm3tt2aq_u32): Ditto.
7129 (vsm3tt2bq_u32): Ditto.
7130 (vsm3partw1q_u32): Ditto.
7131 (vsm3partw2q_u32): Ditto.
7132 (vsm4eq_u32): Ditto.
7133 (vsm4ekeyq_u32): Ditto.
7134 (doc/invoke.texi): Document new sm4 option.
7136 2018-01-10 Michael Collison <michael.collison@arm.com>
7138 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7139 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7140 (AARCH64_FL_FOR_ARCH8_4): New.
7141 (AARCH64_FL_V8_4): New flag.
7142 (doc/invoke.texi): Document new armv8.4-a option.
7144 2018-01-10 Michael Collison <michael.collison@arm.com>
7146 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7147 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7148 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7149 * config/aarch64/aarch64-option-extension.def: Add
7150 AARCH64_OPT_EXTENSION of 'sha2'.
7151 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7152 (crypto): Disable sha2 and aes if crypto disabled.
7153 (crypto): Enable aes and sha2 if enabled.
7154 (simd): Disable sha2 and aes if simd disabled.
7155 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7157 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7158 (TARGET_SHA2): New feature flag for sha2.
7159 (TARGET_AES): New feature flag for aes.
7160 * config/aarch64/aarch64-simd.md:
7161 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7162 conditional on TARGET_AES.
7163 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7164 (aarch64_crypto_sha1hsi): Make pattern conditional
7166 (aarch64_crypto_sha1hv4si): Ditto.
7167 (aarch64_be_crypto_sha1hv4si): Ditto.
7168 (aarch64_crypto_sha1su1v4si): Ditto.
7169 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7170 (aarch64_crypto_sha1su0v4si): Ditto.
7171 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7172 (aarch64_crypto_sha256su0v4si): Ditto.
7173 (aarch64_crypto_sha256su1v4si): Ditto.
7174 (doc/invoke.texi): Document new aes and sha2 options.
7176 2018-01-10 Martin Sebor <msebor@redhat.com>
7178 PR tree-optimization/83781
7179 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7182 2018-01-11 Martin Sebor <msebor@gmail.com>
7183 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7185 PR tree-optimization/83501
7186 PR tree-optimization/81703
7188 * tree-ssa-strlen.c (get_string_cst): Rename...
7189 (get_string_len): ...to this. Handle global constants.
7190 (handle_char_store): Adjust.
7192 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7193 Jim Wilson <jimw@sifive.com>
7195 * config/riscv/riscv-protos.h (riscv_output_return): New.
7196 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7197 (riscv_attribute_table, riscv_output_return),
7198 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7199 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7200 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7202 (riscv_expand_prologue): Add early return for naked function.
7203 (riscv_expand_epilogue): Likewise.
7204 (riscv_function_ok_for_sibcall): Return false for naked function.
7205 (riscv_set_current_function): New.
7206 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7207 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7208 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7209 * doc/extend.texi (RISC-V Function Attributes): New.
7211 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7213 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7214 check for 128-bit long double before checking TCmode.
7215 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7216 128-bit long doubles before checking TFmode or TCmode.
7217 (FLOAT128_IBM_P): Likewise.
7219 2018-01-10 Martin Sebor <msebor@redhat.com>
7221 PR tree-optimization/83671
7222 * builtins.c (c_strlen): Unconditionally return zero for the empty
7224 Use -Warray-bounds for warnings.
7225 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7226 for non-constant array indices with COMPONENT_REF, arrays of
7227 arrays, and pointers to arrays.
7228 (gimple_fold_builtin_strlen): Determine and set length range for
7229 non-constant character arrays.
7231 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7234 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7237 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7239 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7241 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7244 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7245 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7246 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7247 indexed_or_indirect_operand predicate.
7248 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7249 (*vsx_le_perm_load_v8hi): Likewise.
7250 (*vsx_le_perm_load_v16qi): Likewise.
7251 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7252 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7253 (*vsx_le_perm_store_v8hi): Likewise.
7254 (*vsx_le_perm_store_v16qi): Likewise.
7255 (eight unnamed splitters): Likewise.
7257 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7259 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7260 * config/rs6000/emmintrin.h: Likewise.
7261 * config/rs6000/mmintrin.h: Likewise.
7262 * config/rs6000/xmmintrin.h: Likewise.
7264 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7267 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7269 * tree.c (tree_nop_conversion): Return true for location wrapper
7271 (maybe_wrap_with_location): New function.
7272 (selftest::check_strip_nops): New function.
7273 (selftest::test_location_wrappers): New function.
7274 (selftest::tree_c_tests): Call it.
7275 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7276 (maybe_wrap_with_location): New decl.
7277 (EXPR_LOCATION_WRAPPER_P): New macro.
7278 (location_wrapper_p): New inline function.
7279 (tree_strip_any_location_wrapper): New inline function.
7281 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7284 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7285 stack_realign_offset for the largest alignment of stack slot
7287 (ix86_find_max_used_stack_alignment): New function.
7288 (ix86_finalize_stack_frame_flags): Use it. Set
7289 max_used_stack_alignment if we don't realign stack.
7290 * config/i386/i386.h (machine_function): Add
7291 max_used_stack_alignment.
7293 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7295 * config/arm/arm.opt (-mbranch-cost): New option.
7296 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7299 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7302 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7303 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7305 2018-01-10 Richard Biener <rguenther@suse.de>
7308 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7309 early out so it also covers the case where we have a non-NULL
7312 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7314 PR tree-optimization/83753
7315 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7316 for non-strided grouped accesses if the number of elements is 1.
7318 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7321 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7322 * i386.h (TARGET_USE_GATHER): Define.
7323 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7325 2018-01-10 Martin Liska <mliska@suse.cz>
7328 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7329 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7331 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7332 CLEANUP_NO_PARTITIONING is not set.
7334 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7336 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7337 for vectors, as a partial revert of r254296.
7338 * rtl.h (const_vec_p): Delete.
7339 (const_vec_duplicate_p): Don't test for vector CONSTs.
7340 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7341 * expmed.c (make_tree): Likewise.
7344 * common.md (E, F): Use CONSTANT_P instead of checking for
7346 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7347 checking for CONST_VECTOR.
7349 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7352 * predict.c (force_edge_cold): Handle in more sane way edges
7355 2018-01-09 Carl Love <cel@us.ibm.com>
7357 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7359 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7360 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7361 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7362 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7363 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7364 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7365 * config/rs6000/rs6000-protos.h: Add extern defition for
7366 rs6000_generate_float2_double_code.
7367 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7369 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7370 (float2_v2df): Add define_expand.
7372 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7375 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7376 op_mode in the force_to_mode call.
7378 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7380 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7381 instead of checking each element individually.
7382 (aarch64_evpc_uzp): Likewise.
7383 (aarch64_evpc_zip): Likewise.
7384 (aarch64_evpc_ext): Likewise.
7385 (aarch64_evpc_rev): Likewise.
7386 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7387 instead of checking each element individually. Return true without
7389 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7390 whether all selected elements come from the same input, instead of
7391 checking each element individually. Remove calls to gen_rtx_REG,
7392 start_sequence and end_sequence and instead assert that no rtl is
7395 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7397 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7398 order of HIGH and CONST checks.
7400 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7402 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7403 if the destination isn't an SSA_NAME.
7405 2018-01-09 Richard Biener <rguenther@suse.de>
7407 PR tree-optimization/83668
7408 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7410 (canonicalize_loop_form): ... here, renamed from ...
7411 (canonicalize_loop_closed_ssa_form): ... this and amended to
7412 swap successor edges for loop exit blocks to make us use
7413 the RPO order we need for initial schedule generation.
7415 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7417 PR tree-optimization/64811
7418 * match.pd: When optimizing comparisons with Inf, avoid
7419 introducing or losing exceptions from comparisons with NaN.
7421 2018-01-09 Martin Liska <mliska@suse.cz>
7424 * asan.c (shadow_mem_size): Add gcc_assert.
7426 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7428 Don't save registers in main().
7431 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7432 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7433 * config/avr/avr.c (avr_set_current_function): Don't error if
7434 naked, OS_task or OS_main are specified at the same time.
7435 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7437 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7439 * common/config/avr/avr-common.c (avr_option_optimization_table):
7440 Switch on -mmain-is-OS_task for optimizing compilations.
7442 2018-01-09 Richard Biener <rguenther@suse.de>
7444 PR tree-optimization/83572
7445 * graphite.c: Include cfganal.h.
7446 (graphite_transform_loops): Connect infinite loops to exit
7447 and remove fake edges at the end.
7449 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7451 * ipa-inline.c (edge_badness): Revert accidental checkin.
7453 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7456 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7457 symbols; not inline clones.
7459 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7462 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7463 hard registers. Formatting fixes.
7465 PR preprocessor/83722
7466 * gcc.c (try_generate_repro): Pass
7467 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7468 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7471 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7472 Kito Cheng <kito.cheng@gmail.com>
7474 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7475 (riscv_leaf_function_p): Delete.
7476 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7478 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7480 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7482 (do_ifelse): New function.
7483 (do_isel): New function.
7484 (do_sub3): New function.
7485 (do_add3): New function.
7486 (do_load_mask_compare): New function.
7487 (do_overlap_load_compare): New function.
7488 (expand_compare_loop): New function.
7489 (expand_block_compare): Call expand_compare_loop() when appropriate.
7490 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7492 (-mblock-compare-inline-loop-limit): New option.
7494 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7497 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7498 Reverse order of second and third operands in first alternative.
7499 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7500 of first and second elements in UNSPEC_VPERMR vector.
7501 (altivec_expand_vec_perm_le): Likewise.
7503 2018-01-08 Jeff Law <law@redhat.com>
7505 PR rtl-optimizatin/81308
7506 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7507 (process_switch): If group_case_labels makes a change, then set
7509 (pass_convert_switch::execute): If a switch is converted, then
7510 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7512 PR rtl-optimization/81308
7513 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7516 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7518 PR target/83663 - Revert r255946
7519 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7520 generation for cases where splatting a value is not useful.
7521 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7522 across a vec_duplicate and a paradoxical subreg forming a vector
7523 mode to a vec_concat.
7525 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7527 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7528 -march=armv8.3-a variants.
7529 * config/arm/t-multilib: Likewise.
7530 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7532 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7534 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7536 (cceq_ior_compare_complement): Give it a name so I can use it, and
7537 change boolean_or_operator predicate to boolean_operator so it can
7538 be used to generate a crand.
7539 (eqne): New code iterator.
7540 (bd/bd_neg): New code_attrs.
7541 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7542 a single define_insn.
7543 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7544 decrement (bdnzt/bdnzf/bdzt/bdzf).
7545 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7546 with the new names of the branch decrement patterns, and added the
7547 names of the branch decrement conditional patterns.
7549 2018-01-08 Richard Biener <rguenther@suse.de>
7551 PR tree-optimization/83563
7552 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7555 2018-01-08 Richard Biener <rguenther@suse.de>
7558 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7560 2018-01-08 Richard Biener <rguenther@suse.de>
7562 PR tree-optimization/83685
7563 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7564 references to abnormals.
7566 2018-01-08 Richard Biener <rguenther@suse.de>
7569 * dwarf2out.c (output_indirect_strings): Handle empty
7570 skeleton_debug_str_hash.
7571 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7573 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7575 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7576 (emit_store_direct): Likewise.
7577 (arc_trampoline_adjust_address): Likewise.
7578 (arc_asm_trampoline_template): New function.
7579 (arc_initialize_trampoline): Use asm_trampoline_template.
7580 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7581 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7582 * config/arc/arc.md (flush_icache): Delete pattern.
7584 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7586 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7587 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7590 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7593 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7594 by not USED_FOR_TARGET.
7595 (make_pass_resolve_sw_modes): Likewise.
7597 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7599 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7602 2018-01-08 Richard Biener <rguenther@suse.de>
7605 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7607 2018-01-08 Richard Biener <rguenther@suse.de>
7610 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7612 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7615 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7616 basic blocks with a small number of successors.
7617 (convert_control_dep_chain_into_preds): Improve handling of
7619 (dump_predicates): Split apart into...
7620 (dump_pred_chain): ...here...
7621 (dump_pred_info): ...and here.
7622 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7623 (can_chain_union_be_invalidated_p): Improve check for invalidation
7625 (uninit_uses_cannot_happen): Avoid unnecessary if
7626 convert_control_dep_chain_into_preds yielded nothing.
7628 2018-01-06 Martin Sebor <msebor@redhat.com>
7630 PR tree-optimization/83640
7631 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7632 subtracting negative offset from size.
7633 (builtin_access::overlap): Adjust offset bounds of the access to fall
7634 within the size of the object if possible.
7636 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7638 PR rtl-optimization/83699
7639 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7640 extract_bit_field_as_subreg to cases in which the extracted
7641 value is also a vector.
7643 * lra-constraints.c (process_alt_operands): Test for the equivalence
7644 substitutions when detecting a possible reload cycle.
7646 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7649 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7650 by default if flag_selective_schedling{,2}. Formatting fixes.
7652 PR rtl-optimization/83682
7653 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7654 if it has non-VECTOR_MODE element mode.
7655 (vec_duplicate_p): Likewise.
7658 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7659 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7661 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7664 * config/i386/i386-builtin.def
7665 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7666 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7667 Require also OPTION_MASK_ISA_AVX512F in addition to
7668 OPTION_MASK_ISA_GFNI.
7669 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7670 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7671 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7672 to OPTION_MASK_ISA_GFNI.
7673 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7674 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7675 OPTION_MASK_ISA_AVX512BW.
7676 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7677 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7678 addition to OPTION_MASK_ISA_GFNI.
7679 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7680 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7681 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7682 to OPTION_MASK_ISA_GFNI.
7683 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7684 a requirement for all ISAs rather than any of them with a few
7686 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7688 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7689 bitmasks to be enabled with 3 exceptions, instead of requiring any
7690 enabled ISA with lots of exceptions.
7691 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7692 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7693 Change avx512bw in isa attribute to avx512f.
7694 * config/i386/sgxintrin.h: Add license boilerplate.
7695 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7696 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7697 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7698 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7700 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7701 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7702 temporarily sse2 rather than sse if not enabled already.
7705 * config/i386/sse.md (VI248_VLBW): Rename to ...
7706 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7707 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7708 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7709 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7710 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7711 mode iterator instead of VI248_VLBW.
7713 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7715 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7716 (record_modified): Skip clobbers; add debug output.
7717 (param_change_prob): Use sreal frequencies.
7719 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7721 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7722 punt for user-aligned variables.
7724 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7726 * tree-chrec.c (chrec_contains_symbols): Return true for
7729 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7732 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7733 of (x|y) == x for BICS pattern.
7735 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7737 PR tree-optimization/83605
7738 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7739 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7742 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7744 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7745 * config/epiphany/rtems.h: New file.
7747 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7748 Uros Bizjak <ubizjak@gmail.com>
7751 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7752 QIreg_operand instead of register_operand predicate.
7753 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7754 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7755 comments instead of -fmitigate[-_]rop.
7757 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7760 * cgraphunit.c (symbol_table::compile): Switch to text_section
7761 before calling assembly_start debug hook.
7762 * run-rtl-passes.c (run_rtl_passes): Likewise.
7765 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7767 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7768 range_int_cst_p rather than !symbolic_range_p before calling
7769 extract_range_from_multiplicative_op_1.
7771 2018-01-04 Jeff Law <law@redhat.com>
7773 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7774 redundant test in assertion.
7776 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7778 * doc/rtl.texi: Document machine_mode wrapper classes.
7780 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7782 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7785 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7787 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7788 the VEC_PERM_EXPR fold to fail.
7790 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7793 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7794 to switched_sections.
7796 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7799 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7802 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7805 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7806 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7808 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7811 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7812 is BLKmode and bitpos not zero or mode change is needed.
7814 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7817 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7820 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7823 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7824 instead of MULT rtx. Update all corresponding splitters.
7826 (*ssub<modesuffix>): Ditto.
7828 (*cmp_sadd_di): Update split patterns.
7829 (*cmp_sadd_si): Ditto.
7830 (*cmp_sadd_sidi): Ditto.
7831 (*cmp_ssub_di): Ditto.
7832 (*cmp_ssub_si): Ditto.
7833 (*cmp_ssub_sidi): Ditto.
7834 * config/alpha/predicates.md (const23_operand): New predicate.
7835 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7836 Look for ASHIFT, not MULT inner operand.
7837 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7839 2018-01-04 Martin Liska <mliska@suse.cz>
7841 PR gcov-profile/83669
7842 * gcov.c (output_intermediate_file): Add version to intermediate
7844 * doc/gcov.texi: Document new field 'version' in intermediate
7845 file format. Fix location of '-k' option of gcov command.
7847 2018-01-04 Martin Liska <mliska@suse.cz>
7850 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7852 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7854 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7856 2018-01-03 Martin Sebor <msebor@redhat.com>
7858 PR tree-optimization/83655
7859 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7860 checking calls with invalid arguments.
7862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7864 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7865 (vectorizable_mask_load_store): Delete.
7866 (vectorizable_call): Return false for masked loads and stores.
7867 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7868 instead of gimple_assign_rhs1.
7869 (vectorizable_load): Handle IFN_MASK_LOAD.
7870 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7874 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7876 (vectorizable_mask_load_store): ...here.
7877 (vectorizable_load): ...and here.
7879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7881 * tree-vect-stmts.c (vect_build_all_ones_mask)
7882 (vect_build_zero_merge_argument): New functions, split out from...
7883 (vectorizable_load): ...here.
7885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7887 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7889 (vectorizable_mask_load_store): ...here.
7890 (vectorizable_store): ...and here.
7892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7894 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7896 (vectorizable_mask_load_store): ...here.
7898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7900 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7901 (vect_model_store_cost): Take a vec_load_store_type instead of a
7903 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7904 (vect_model_store_cost): Take a vec_load_store_type instead of a
7906 (vectorizable_mask_load_store): Update accordingly.
7907 (vectorizable_store): Likewise.
7908 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7912 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7913 IFN_MASK_LOAD calls here rather than...
7914 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7917 Alan Hayward <alan.hayward@arm.com>
7918 David Sherwood <david.sherwood@arm.com>
7920 * expmed.c (extract_bit_field_1): For vector extracts,
7921 fall back to extract_bit_field_as_subreg if vec_extract
7924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7925 Alan Hayward <alan.hayward@arm.com>
7926 David Sherwood <david.sherwood@arm.com>
7928 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7929 they are variable or constant sized.
7930 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7931 slots for constant-sized data.
7933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7934 Alan Hayward <alan.hayward@arm.com>
7935 David Sherwood <david.sherwood@arm.com>
7937 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7938 handling COND_EXPRs with boolean comparisons, try to find a better
7939 basis for the mask type than the boolean itself.
7941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7943 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7944 is calculated and how it can be overridden.
7945 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7946 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7948 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7952 Alan Hayward <alan.hayward@arm.com>
7953 David Sherwood <david.sherwood@arm.com>
7955 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7956 Remove the mode argument.
7957 (aarch64_simd_valid_immediate): Remove the mode and inverse
7959 * config/aarch64/iterators.md (bitsize): New iterator.
7960 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7961 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7962 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7963 aarch64_simd_valid_immediate.
7964 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7965 (aarch64_reg_or_bic_imm): Likewise.
7966 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7967 with an insn_type enum and msl with a modifier_type enum.
7968 Replace element_width with a scalar_mode. Change the shift
7969 to unsigned int. Add constructors for scalar_float_mode and
7970 scalar_int_mode elements.
7971 (aarch64_vect_float_const_representable_p): Delete.
7972 (aarch64_can_const_movi_rtx_p)
7973 (aarch64_simd_scalar_immediate_valid_for_move)
7974 (aarch64_simd_make_constant): Update call to
7975 aarch64_simd_valid_immediate.
7976 (aarch64_advsimd_valid_immediate_hs): New function.
7977 (aarch64_advsimd_valid_immediate): Likewise.
7978 (aarch64_simd_valid_immediate): Remove mode and inverse
7979 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7980 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7981 and aarch64_float_const_representable_p on the result.
7982 (aarch64_output_simd_mov_immediate): Remove mode argument.
7983 Update call to aarch64_simd_valid_immediate and use of
7984 simd_immediate_info.
7985 (aarch64_output_scalar_simd_mov_immediate): Update call
7988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7989 Alan Hayward <alan.hayward@arm.com>
7990 David Sherwood <david.sherwood@arm.com>
7992 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7993 (mode_nunits): Likewise CONST_MODE_NUNITS.
7994 * machmode.def (ADJUST_NUNITS): Document.
7995 * genmodes.c (mode_data::need_nunits_adj): New field.
7996 (blank_mode): Update accordingly.
7997 (adj_nunits): New variable.
7998 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8000 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8001 listed in adj_nunits.
8002 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8003 listed in adj_nunits. Don't emit case statements for such modes.
8004 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8005 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8006 nothing if adj_nunits is nonnull.
8007 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8008 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8009 (emit_mode_fbit): Update use of print_maybe_const_decl.
8010 (emit_move_size): Likewise. Treat the array as non-const
8012 (emit_mode_adjustments): Handle adj_nunits.
8014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8016 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8017 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8018 (VECTOR_MODES): Use it.
8019 (make_vector_modes): Take the prefix as an argument.
8021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8022 Alan Hayward <alan.hayward@arm.com>
8023 David Sherwood <david.sherwood@arm.com>
8025 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8026 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8027 for MODE_VECTOR_BOOL.
8028 * machmode.def (VECTOR_BOOL_MODE): Document.
8029 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8030 (make_vector_bool_mode): New function.
8031 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8033 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8034 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8036 * stor-layout.c (int_mode_for_mode): Likewise.
8037 * tree.c (build_vector_type_for_mode): Likewise.
8038 * varasm.c (output_constant_pool_2): Likewise.
8039 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8040 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8041 for MODE_VECTOR_BOOL.
8042 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8043 of mode class checks.
8044 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8045 instead of a list of mode class checks.
8046 (expand_vector_scalar_condition): Likewise.
8047 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8050 Alan Hayward <alan.hayward@arm.com>
8051 David Sherwood <david.sherwood@arm.com>
8053 * machmode.h (mode_size): Change from unsigned short to
8055 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8056 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8057 or if measurement_type is not polynomial.
8058 (fixed_size_mode::includes_p): Check for constant-sized modes.
8059 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8060 return a poly_uint16 rather than an unsigned short.
8061 (emit_mode_size): Change the type of mode_size from unsigned short
8062 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8063 (emit_mode_adjustments): Cope with polynomial vector sizes.
8064 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8066 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8068 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8069 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8070 * caller-save.c (setup_save_areas): Likewise.
8071 (replace_reg_with_saved_mem): Likewise.
8072 * calls.c (emit_library_call_value_1): Likewise.
8073 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8074 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8075 (gen_lowpart_for_combine): Likewise.
8076 * convert.c (convert_to_integer_1): Likewise.
8077 * cse.c (equiv_constant, cse_insn): Likewise.
8078 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8079 (cselib_subst_to_values): Likewise.
8080 * dce.c (word_dce_process_block): Likewise.
8081 * df-problems.c (df_word_lr_mark_ref): Likewise.
8082 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8083 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8084 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8085 (rtl_for_decl_location): Likewise.
8086 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8087 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8088 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8089 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8090 (expand_expr_real_1): Likewise.
8091 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8092 (pad_below): Likewise.
8093 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8094 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8095 * ira.c (get_subreg_tracking_sizes): Likewise.
8096 * ira-build.c (ira_create_allocno_objects): Likewise.
8097 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8098 (ira_sort_regnos_for_alter_reg): Likewise.
8099 * ira-costs.c (record_operand_costs): Likewise.
8100 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8101 (resolve_simple_move): Likewise.
8102 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8103 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8104 (lra_constraints): Likewise.
8105 (CONST_POOL_OK_P): Reject variable-sized modes.
8106 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8107 (add_pseudo_to_slot, lra_spill): Likewise.
8108 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8109 * optabs-query.c (get_best_extraction_insn): Likewise.
8110 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8111 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8112 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8113 * recog.c (offsettable_address_addr_space_p): Likewise.
8114 * regcprop.c (maybe_mode_change): Likewise.
8115 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8116 * regrename.c (build_def_use): Likewise.
8117 * regstat.c (dump_reg_info): Likewise.
8118 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8119 (find_reloads, find_reloads_subreg_address): Likewise.
8120 * reload1.c (eliminate_regs_1): Likewise.
8121 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8122 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8123 (simplify_binary_operation_1, simplify_subreg): Likewise.
8124 * targhooks.c (default_function_arg_padding): Likewise.
8125 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8126 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8127 (verify_gimple_assign_ternary): Likewise.
8128 * tree-inline.c (estimate_move_cost): Likewise.
8129 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8130 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8131 (get_address_cost_ainc): Likewise.
8132 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8133 (vect_supportable_dr_alignment): Likewise.
8134 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8135 (vectorizable_reduction): Likewise.
8136 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8137 (vectorizable_operation, vectorizable_load): Likewise.
8138 * tree.c (build_same_sized_truth_vector_type): Likewise.
8139 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8140 * var-tracking.c (emit_note_insn_var_location): Likewise.
8141 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8142 (ADDR_VEC_ALIGN): Likewise.
8144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8145 Alan Hayward <alan.hayward@arm.com>
8146 David Sherwood <david.sherwood@arm.com>
8148 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8150 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8151 or if measurement_type is polynomial.
8152 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8153 * combine.c (make_extraction): Likewise.
8154 * dse.c (find_shift_sequence): Likewise.
8155 * dwarf2out.c (mem_loc_descriptor): Likewise.
8156 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8157 (extract_bit_field, extract_low_bits): Likewise.
8158 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8159 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8160 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8161 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8162 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8163 * reload.c (find_reloads): Likewise.
8164 * reload1.c (alter_reg): Likewise.
8165 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8166 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8167 * tree-if-conv.c (predicate_mem_writes): Likewise.
8168 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8169 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8170 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8171 * valtrack.c (dead_debug_insert_temp): Likewise.
8172 * varasm.c (mergeable_constant_section): Likewise.
8173 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8176 Alan Hayward <alan.hayward@arm.com>
8177 David Sherwood <david.sherwood@arm.com>
8179 * expr.c (expand_assignment): Cope with polynomial mode sizes
8180 when assigning to a CONCAT.
8182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8183 Alan Hayward <alan.hayward@arm.com>
8184 David Sherwood <david.sherwood@arm.com>
8186 * machmode.h (mode_precision): Change from unsigned short to
8188 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8190 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8191 or if measurement_type is not polynomial.
8192 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8193 in which the mode is already known to be a scalar_int_mode.
8194 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8195 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8197 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8198 for GET_MODE_PRECISION.
8199 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8200 for GET_MODE_PRECISION.
8201 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8203 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8204 (expand_field_assignment, make_extraction): Likewise.
8205 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8206 (get_last_value): Likewise.
8207 * convert.c (convert_to_integer_1): Likewise.
8208 * cse.c (cse_insn): Likewise.
8209 * expr.c (expand_expr_real_1): Likewise.
8210 * lra-constraints.c (simplify_operand_subreg): Likewise.
8211 * optabs-query.c (can_atomic_load_p): Likewise.
8212 * optabs.c (expand_atomic_load): Likewise.
8213 (expand_atomic_store): Likewise.
8214 * ree.c (combine_reaching_defs): Likewise.
8215 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8216 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8217 * tree.h (type_has_mode_precision_p): Likewise.
8218 * ubsan.c (instrument_si_overflow): Likewise.
8220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8221 Alan Hayward <alan.hayward@arm.com>
8222 David Sherwood <david.sherwood@arm.com>
8224 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8225 polynomial numbers of units.
8226 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8227 (valid_vector_subparts_p): New function.
8228 (build_vector_type): Remove temporary shim and take the number
8229 of units as a poly_uint64 rather than an int.
8230 (build_opaque_vector_type): Take the number of units as a
8231 poly_uint64 rather than an int.
8232 * tree.c (build_vector_from_ctor): Handle polynomial
8233 TYPE_VECTOR_SUBPARTS.
8234 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8235 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8236 (build_vector_from_val): If the number of units is variable,
8237 use build_vec_duplicate_cst for constant operands and
8238 VEC_DUPLICATE_EXPR otherwise.
8239 (make_vector_type): Remove temporary is_constant ().
8240 (build_vector_type, build_opaque_vector_type): Take the number of
8241 units as a poly_uint64 rather than an int.
8242 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8244 * cfgexpand.c (expand_debug_expr): Likewise.
8245 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8246 (store_constructor, expand_expr_real_1): Likewise.
8247 (const_scalar_mask_from_tree): Likewise.
8248 * fold-const-call.c (fold_const_reduction): Likewise.
8249 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8250 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8251 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8252 (fold_relational_const): Likewise.
8253 (native_interpret_vector): Likewise. Change the size from an
8254 int to an unsigned int.
8255 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8256 TYPE_VECTOR_SUBPARTS.
8257 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8258 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8259 duplicating a non-constant operand into a variable-length vector.
8260 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8261 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8262 * ipa-icf.c (sem_variable::equals): Likewise.
8263 * match.pd: Likewise.
8264 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8265 * print-tree.c (print_node): Likewise.
8266 * stor-layout.c (layout_type): Likewise.
8267 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8268 * tree-cfg.c (verify_gimple_comparison): Likewise.
8269 (verify_gimple_assign_binary): Likewise.
8270 (verify_gimple_assign_ternary): Likewise.
8271 (verify_gimple_assign_single): Likewise.
8272 * tree-pretty-print.c (dump_generic_node): Likewise.
8273 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8274 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8275 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8276 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8277 (vect_shift_permute_load_chain): Likewise.
8278 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8279 (expand_vector_condition, optimize_vector_constructor): Likewise.
8280 (lower_vec_perm, get_compute_type): Likewise.
8281 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8282 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8283 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8284 (vect_recog_mask_conversion_pattern): Likewise.
8285 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8286 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8287 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8288 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8289 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8290 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8291 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8292 (supportable_widening_operation): Likewise.
8293 (supportable_narrowing_operation): Likewise.
8294 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8296 * varasm.c (output_constant): Likewise.
8298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8299 Alan Hayward <alan.hayward@arm.com>
8300 David Sherwood <david.sherwood@arm.com>
8302 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8303 so that both the length == 3 and length != 3 cases set up their
8304 own permute vectors. Add comments explaining why we know the
8305 number of elements is constant.
8306 (vect_permute_load_chain): Likewise.
8308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8309 Alan Hayward <alan.hayward@arm.com>
8310 David Sherwood <david.sherwood@arm.com>
8312 * machmode.h (mode_nunits): Change from unsigned char to
8314 (ONLY_FIXED_SIZE_MODES): New macro.
8315 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8316 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8317 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8319 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8320 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8321 or if measurement_type is not polynomial.
8322 * genmodes.c (ZERO_COEFFS): New macro.
8323 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8325 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8326 Use ZERO_COEFFS when emitting initializers.
8327 * data-streamer.h (bp_pack_poly_value): New function.
8328 (bp_unpack_poly_value): Likewise.
8329 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8330 for GET_MODE_NUNITS.
8331 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8332 for GET_MODE_NUNITS.
8333 * tree.c (make_vector_type): Remove temporary shim and make
8334 the real function take the number of units as a poly_uint64
8336 (build_vector_type_for_mode): Handle polynomial nunits.
8337 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8338 * emit-rtl.c (const_vec_series_p_1): Likewise.
8339 (gen_rtx_CONST_VECTOR): Likewise.
8340 * fold-const.c (test_vec_duplicate_folding): Likewise.
8341 * genrecog.c (validate_pattern): Likewise.
8342 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8343 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8344 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8345 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8346 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8347 * rtlanal.c (subreg_get_info): Likewise.
8348 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8349 (vect_grouped_load_supported): Likewise.
8350 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8351 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8352 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8353 (simplify_const_unary_operation, simplify_binary_operation_1)
8354 (simplify_const_binary_operation, simplify_ternary_operation)
8355 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8356 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8357 instead of CONST_VECTOR_NUNITS.
8358 * varasm.c (output_constant_pool_2): Likewise.
8359 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8360 explicit-encoded elements in the XVEC for variable-length vectors.
8362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8364 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8367 Alan Hayward <alan.hayward@arm.com>
8368 David Sherwood <david.sherwood@arm.com>
8370 * coretypes.h (fixed_size_mode): Declare.
8371 (fixed_size_mode_pod): New typedef.
8372 * builtins.h (target_builtins::x_apply_args_mode)
8373 (target_builtins::x_apply_result_mode): Change type to
8374 fixed_size_mode_pod.
8375 * builtins.c (apply_args_size, apply_result_size, result_vector)
8376 (expand_builtin_apply_args_1, expand_builtin_apply)
8377 (expand_builtin_return): Update accordingly.
8379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8381 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8382 * cselib.c (cselib_hash_rtx): Likewise.
8383 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8384 CONST_VECTOR encoding.
8386 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8387 Jeff Law <law@redhat.com>
8390 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8391 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8392 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8393 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8396 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8397 explicitly probe *sp in a noreturn function if there were any callee
8398 register saves or frame pointer is needed.
8400 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8403 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8404 BLKmode for ternary, binary or unary expressions.
8407 * var-tracking.c (delete_vta_debug_insn): New inline function.
8408 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8409 insns from get_insns () to NULL instead of each bb separately.
8410 Use delete_vta_debug_insn. No longer static.
8411 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8412 delete_vta_debug_insns callers.
8413 * rtl.h (delete_vta_debug_insns): Declare.
8414 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8415 instead of variable_tracking_main.
8417 2018-01-03 Martin Sebor <msebor@redhat.com>
8419 PR tree-optimization/83603
8420 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8421 arguments past the endof the argument list in functions declared
8422 without a prototype.
8423 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8424 Avoid checking when arguments are null.
8426 2018-01-03 Martin Sebor <msebor@redhat.com>
8429 * doc/extend.texi (attribute const): Fix a typo.
8430 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8431 issuing -Wsuggest-attribute for void functions.
8433 2018-01-03 Martin Sebor <msebor@redhat.com>
8435 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8436 offset_int::from instead of wide_int::to_shwi.
8437 (maybe_diag_overlap): Remove assertion.
8438 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8439 * gimple-ssa-sprintf.c (format_directive): Same.
8440 (parse_directive): Same.
8441 (sprintf_dom_walker::compute_format_length): Same.
8442 (try_substitute_return_value): Same.
8444 2018-01-03 Jeff Law <law@redhat.com>
8447 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8448 non-constant residual for zero at runtime and avoid probing in
8449 that case. Reorganize code for trailing problem to mirror handling
8452 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8454 PR tree-optimization/83501
8455 * tree-ssa-strlen.c (get_string_cst): New.
8456 (handle_char_store): Call get_string_cst.
8458 2018-01-03 Martin Liska <mliska@suse.cz>
8460 PR tree-optimization/83593
8461 * tree-ssa-strlen.c: Include tree-cfg.h.
8462 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8463 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8464 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8466 (strlen_dom_walker::before_dom_children): Call
8467 gimple_purge_dead_eh_edges. Dump tranformation with details
8469 (strlen_dom_walker::before_dom_children): Update call by adding
8470 new argument cleanup_eh.
8471 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8473 2018-01-03 Martin Liska <mliska@suse.cz>
8476 * cif-code.def (VARIADIC_THUNK): New enum value.
8477 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8480 2018-01-03 Jan Beulich <jbeulich@suse.com>
8482 * sse.md (mov<mode>_internal): Tighten condition for when to use
8483 vmovdqu<ssescalarsize> for TI and OI modes.
8485 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8487 Update copyright years.
8489 2018-01-03 Martin Liska <mliska@suse.cz>
8492 * ipa-visibility.c (function_and_variable_visibility): Skip
8493 functions with noipa attribure.
8495 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8497 * gcc.c (process_command): Update copyright notice dates.
8498 * gcov-dump.c (print_version): Ditto.
8499 * gcov.c (print_version): Ditto.
8500 * gcov-tool.c (print_version): Ditto.
8501 * gengtype.c (create_file): Ditto.
8502 * doc/cpp.texi: Bump @copying's copyright year.
8503 * doc/cppinternals.texi: Ditto.
8504 * doc/gcc.texi: Ditto.
8505 * doc/gccint.texi: Ditto.
8506 * doc/gcov.texi: Ditto.
8507 * doc/install.texi: Ditto.
8508 * doc/invoke.texi: Ditto.
8510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8512 * vector-builder.h (vector_builder::m_full_nelts): Change from
8513 unsigned int to poly_uint64.
8514 (vector_builder::full_nelts): Update prototype accordingly.
8515 (vector_builder::new_vector): Likewise.
8516 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8517 (vector_builder::operator ==): Likewise.
8518 (vector_builder::finalize): Likewise.
8519 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8520 Take the number of elements as a poly_uint64 rather than an
8522 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8523 from unsigned int to poly_uint64.
8524 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8525 (vec_perm_indices::new_vector): Likewise.
8526 (vec_perm_indices::length): Likewise.
8527 (vec_perm_indices::nelts_per_input): Likewise.
8528 (vec_perm_indices::input_nelts): Likewise.
8529 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8530 number of elements per input as a poly_uint64 rather than an
8531 unsigned int. Use the original encoding for variable-length
8532 vectors, rather than clamping each individual element.
8533 For the second and subsequent elements in each pattern,
8534 clamp the step and base before clamping their sum.
8535 (vec_perm_indices::series_p): Handle polynomial element counts.
8536 (vec_perm_indices::all_in_range_p): Likewise.
8537 (vec_perm_indices_to_tree): Likewise.
8538 (vec_perm_indices_to_rtx): Likewise.
8539 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8540 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8541 (tree_vector_builder::new_binary_operation): Handle polynomial
8542 element counts. Return false if we need to know the number
8543 of elements at compile time.
8544 * fold-const.c (fold_vec_perm): Punt if the number of elements
8545 isn't known at compile time.
8547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8549 * vec-perm-indices.h (vec_perm_builder): Change element type
8550 from HOST_WIDE_INT to poly_int64.
8551 (vec_perm_indices::element_type): Update accordingly.
8552 (vec_perm_indices::clamp): Handle polynomial element_types.
8553 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8554 (vec_perm_indices::all_in_range_p): Likewise.
8555 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8557 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8558 polynomial vec_perm_indices element types.
8559 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8560 * fold-const.c (fold_vec_perm): Likewise.
8561 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8562 * tree-vect-generic.c (lower_vec_perm): Likewise.
8563 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8564 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8565 element type to HOST_WIDE_INT.
8567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8568 Alan Hayward <alan.hayward@arm.com>
8569 David Sherwood <david.sherwood@arm.com>
8571 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8572 rather than an int. Use plus_constant.
8573 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8574 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8577 Alan Hayward <alan.hayward@arm.com>
8578 David Sherwood <david.sherwood@arm.com>
8580 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8581 a HOST_WIDE_INT to a poly_int64.
8583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8584 Alan Hayward <alan.hayward@arm.com>
8585 David Sherwood <david.sherwood@arm.com>
8587 * calls.c (load_register_parameters): Cope with polynomial
8588 mode sizes. Require a constant size for BLKmode parameters
8589 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8590 forces a parameter to be padded at the lsb end in order to
8591 fill a complete number of words, require the parameter size
8592 to be ordered wrt UNITS_PER_WORD.
8594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8595 Alan Hayward <alan.hayward@arm.com>
8596 David Sherwood <david.sherwood@arm.com>
8598 * reload1.c (spill_stack_slot_width): Change element type
8599 from unsigned int to poly_uint64_pod.
8600 (alter_reg): Treat mode sizes as polynomial.
8602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8603 Alan Hayward <alan.hayward@arm.com>
8604 David Sherwood <david.sherwood@arm.com>
8606 * reload.c (complex_word_subreg_p): New function.
8607 (reload_inner_reg_of_subreg, push_reload): Use it.
8609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8610 Alan Hayward <alan.hayward@arm.com>
8611 David Sherwood <david.sherwood@arm.com>
8613 * lra-constraints.c (process_alt_operands): Reject matched
8614 operands whose sizes aren't ordered.
8615 (match_reload): Refer to this check here.
8617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8618 Alan Hayward <alan.hayward@arm.com>
8619 David Sherwood <david.sherwood@arm.com>
8621 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8622 that the mode size is in the set {1, 2, 4, 8, 16}.
8624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8625 Alan Hayward <alan.hayward@arm.com>
8626 David Sherwood <david.sherwood@arm.com>
8628 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8629 Use plus_constant instead of gen_rtx_PLUS.
8631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8632 Alan Hayward <alan.hayward@arm.com>
8633 David Sherwood <david.sherwood@arm.com>
8635 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8636 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8637 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8638 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8639 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8640 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8641 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8642 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8643 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8644 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8646 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8647 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8648 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8649 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8650 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8651 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8652 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8653 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8654 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8655 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8657 * expr.c (emit_move_resolve_push): Treat the input and result
8658 of PUSH_ROUNDING as a poly_int64.
8659 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8660 (emit_push_insn): Likewise.
8661 * lra-eliminations.c (mark_not_eliminable): Likewise.
8662 * recog.c (push_operand): Likewise.
8663 * reload1.c (elimination_effects): Likewise.
8664 * rtlanal.c (nonzero_bits1): Likewise.
8665 * calls.c (store_one_arg): Likewise. Require the padding to be
8666 known at compile time.
8668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8669 Alan Hayward <alan.hayward@arm.com>
8670 David Sherwood <david.sherwood@arm.com>
8672 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8673 Use plus_constant instead of gen_rtx_PLUS.
8675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8676 Alan Hayward <alan.hayward@arm.com>
8677 David Sherwood <david.sherwood@arm.com>
8679 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8683 Alan Hayward <alan.hayward@arm.com>
8684 David Sherwood <david.sherwood@arm.com>
8686 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8687 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8688 via stack temporaries. Treat the mode size as polynomial too.
8690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8691 Alan Hayward <alan.hayward@arm.com>
8692 David Sherwood <david.sherwood@arm.com>
8694 * expr.c (expand_expr_real_2): When handling conversions involving
8695 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8696 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8697 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8706 Alan Hayward <alan.hayward@arm.com>
8707 David Sherwood <david.sherwood@arm.com>
8709 * combine.c (can_change_dest_mode): Handle polynomial
8710 REGMODE_NATURAL_SIZE.
8711 * expmed.c (store_bit_field_1): Likewise.
8712 * expr.c (store_constructor): Likewise.
8713 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8714 and polynomial REGMODE_NATURAL_SIZE.
8715 (gen_lowpart_common): Likewise.
8716 * reginfo.c (record_subregs_of_mode): Likewise.
8717 * rtlanal.c (read_modify_subreg_p): Likewise.
8719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8720 Alan Hayward <alan.hayward@arm.com>
8721 David Sherwood <david.sherwood@arm.com>
8723 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8724 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
8732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8733 Alan Hayward <alan.hayward@arm.com>
8734 David Sherwood <david.sherwood@arm.com>
8736 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8737 in a POINTER_PLUS_EXPR.
8739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8740 Alan Hayward <alan.hayward@arm.com>
8741 David Sherwood <david.sherwood@arm.com>
8743 * omp-simd-clone.c (simd_clone_subparts): New function.
8744 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8745 (ipa_simd_modify_function_body): Likewise.
8747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8748 Alan Hayward <alan.hayward@arm.com>
8749 David Sherwood <david.sherwood@arm.com>
8751 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8752 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8753 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8754 (expand_vector_condition, vector_element): Likewise.
8755 (subparts_gt): New function.
8756 (get_compute_type): Use subparts_gt.
8757 (count_type_subparts): Delete.
8758 (expand_vector_operations_1): Use subparts_gt instead of
8759 count_type_subparts.
8761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8762 Alan Hayward <alan.hayward@arm.com>
8763 David Sherwood <david.sherwood@arm.com>
8765 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8766 (vect_compile_time_alias): ...this new function. Do the calculation
8767 on poly_ints rather than trees.
8768 (vect_prune_runtime_alias_test_list): Update call accordingly.
8770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8771 Alan Hayward <alan.hayward@arm.com>
8772 David Sherwood <david.sherwood@arm.com>
8774 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8776 (vect_schedule_slp_instance): Likewise.
8778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8779 Alan Hayward <alan.hayward@arm.com>
8780 David Sherwood <david.sherwood@arm.com>
8782 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8783 constant and extern definitions for variable-length vectors.
8784 (vect_get_constant_vectors): Note that the number of units
8785 is known to be constant.
8787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8788 Alan Hayward <alan.hayward@arm.com>
8789 David Sherwood <david.sherwood@arm.com>
8791 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8792 of units as polynomial. Choose between WIDE and NARROW based
8795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8796 Alan Hayward <alan.hayward@arm.com>
8797 David Sherwood <david.sherwood@arm.com>
8799 * tree-vect-stmts.c (simd_clone_subparts): New function.
8800 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8803 Alan Hayward <alan.hayward@arm.com>
8804 David Sherwood <david.sherwood@arm.com>
8806 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8807 vectors as polynomial. Use build_index_vector for
8810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8811 Alan Hayward <alan.hayward@arm.com>
8812 David Sherwood <david.sherwood@arm.com>
8814 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8815 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8816 for variable-length vectors.
8817 (vectorizable_mask_load_store): Treat the number of units as
8818 polynomial, asserting that it is constant if the condition has
8819 already been enforced.
8820 (vectorizable_store, vectorizable_load): Likewise.
8822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8823 Alan Hayward <alan.hayward@arm.com>
8824 David Sherwood <david.sherwood@arm.com>
8826 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8827 of units as polynomial. Punt if we can't tell at compile time
8828 which vector contains the final result.
8830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8831 Alan Hayward <alan.hayward@arm.com>
8832 David Sherwood <david.sherwood@arm.com>
8834 * tree-vect-loop.c (vectorizable_induction): Treat the number
8835 of units as polynomial. Punt on SLP inductions. Use an integer
8836 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8837 cast of such a series for variable-length floating-point
8840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8841 Alan Hayward <alan.hayward@arm.com>
8842 David Sherwood <david.sherwood@arm.com>
8844 * tree.h (build_index_vector): Declare.
8845 * tree.c (build_index_vector): New function.
8846 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8847 of units as polynomial, forcibly converting it to a constant if
8848 vectorizable_reduction has already enforced the condition.
8849 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8850 to create a {1,2,3,...} vector.
8851 (vectorizable_reduction): Treat the number of units as polynomial.
8852 Choose vectype_in based on the largest scalar element size rather
8853 than the smallest number of units. Enforce the restrictions
8856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8857 Alan Hayward <alan.hayward@arm.com>
8858 David Sherwood <david.sherwood@arm.com>
8860 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8861 number of units as polynomial.
8863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8864 Alan Hayward <alan.hayward@arm.com>
8865 David Sherwood <david.sherwood@arm.com>
8867 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8868 * target.def (autovectorize_vector_sizes): Return the vector sizes
8869 by pointer, using vector_sizes rather than a bitmask.
8870 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8871 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8872 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8874 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8875 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8876 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8877 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8878 * omp-general.c (omp_max_vf): Likewise.
8879 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8880 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8881 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8882 * tree-vect-slp.c (vect_slp_bb): Likewise.
8883 * doc/tm.texi: Regenerate.
8884 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8886 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8887 the vector size as a poly_uint64 rather than an unsigned int.
8888 (current_vector_size): Change from an unsigned int to a poly_uint64.
8889 (get_vectype_for_scalar_type): Update accordingly.
8890 * tree.h (build_truth_vector_type): Take the size and number of
8891 units as a poly_uint64 rather than an unsigned int.
8892 (build_vector_type): Add a temporary overload that takes
8893 the number of units as a poly_uint64 rather than an unsigned int.
8894 * tree.c (make_vector_type): Likewise.
8895 (build_truth_vector_type): Take the number of units as a poly_uint64
8896 rather than an unsigned int.
8898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8899 Alan Hayward <alan.hayward@arm.com>
8900 David Sherwood <david.sherwood@arm.com>
8902 * target.def (get_mask_mode): Take the number of units and length
8903 as poly_uint64s rather than unsigned ints.
8904 * targhooks.h (default_get_mask_mode): Update accordingly.
8905 * targhooks.c (default_get_mask_mode): Likewise.
8906 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8907 * doc/tm.texi: Regenerate.
8909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8910 Alan Hayward <alan.hayward@arm.com>
8911 David Sherwood <david.sherwood@arm.com>
8913 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8914 * omp-general.c (omp_max_vf): Likewise.
8915 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8916 (expand_omp_simd): Handle polynomial safelen.
8917 * omp-low.c (omplow_simd_context): Add a default constructor.
8918 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8919 (lower_rec_simd_input_clauses): Update accordingly.
8920 (lower_rec_input_clauses): Likewise.
8922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8923 Alan Hayward <alan.hayward@arm.com>
8924 David Sherwood <david.sherwood@arm.com>
8926 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8927 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8928 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8929 (vect_analyze_slp_cost): Likewise.
8930 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8931 (vect_model_load_cost): Likewise.
8933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8934 Alan Hayward <alan.hayward@arm.com>
8935 David Sherwood <david.sherwood@arm.com>
8937 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8938 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8939 from an unsigned int * to a poly_uint64_pod *.
8940 (calculate_unrolling_factor): New function.
8941 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8944 Alan Hayward <alan.hayward@arm.com>
8945 David Sherwood <david.sherwood@arm.com>
8947 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8948 from an unsigned int to a poly_uint64.
8949 (_loop_vec_info::slp_unrolling_factor): Likewise.
8950 (_loop_vec_info::vectorization_factor): Change from an int
8952 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8953 (vect_get_num_vectors): New function.
8954 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8955 (vect_get_num_copies): Use vect_get_num_vectors.
8956 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8957 to an unsigned int *.
8958 (vect_analyze_data_refs): Change min_vf from an int * to a
8960 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8961 than an unsigned HOST_WIDE_INT.
8962 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8963 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8964 to an unsigned int *.
8965 (vect_analyze_data_ref_dependences): Likewise.
8966 (vect_compute_data_ref_alignment): Handle polynomial vf.
8967 (vect_enhance_data_refs_alignment): Likewise.
8968 (vect_prune_runtime_alias_test_list): Likewise.
8969 (vect_shift_permute_load_chain): Likewise.
8970 (vect_supportable_dr_alignment): Likewise.
8971 (dependence_distance_ge_vf): Take the vectorization factor as a
8972 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8973 (vect_analyze_data_refs): Change min_vf from an int * to a
8975 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8976 vfm1 as a poly_uint64 rather than an int. Make the same change
8977 for the returned bound_scalar.
8978 (vect_gen_vector_loop_niters): Handle polynomial vf.
8979 (vect_do_peeling): Likewise. Update call to
8980 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8981 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8983 * tree-vect-loop.c (vect_determine_vectorization_factor)
8984 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8985 (vect_get_known_peeling_cost): Likewise.
8986 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8987 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8988 (vect_transform_loop): Likewise. Use the lowest possible VF when
8989 updating the upper bounds of the loop.
8990 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8992 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8993 polynomial unroll factors.
8994 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8995 (vect_make_slp_decision): Likewise.
8996 (vect_supported_load_permutation_p): Likewise, and polynomial
8998 (vect_analyze_slp_cost): Handle polynomial vf.
8999 (vect_slp_analyze_node_operations): Likewise.
9000 (vect_slp_analyze_bb_1): Likewise.
9001 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9002 than an unsigned HOST_WIDE_INT.
9003 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9004 (vectorizable_load): Handle polynomial vf.
9005 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9007 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * match.pd: Handle bit operations involving three constants
9014 and try to fold one pair.
9016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9018 * tree-vect-loop-manip.c: Include gimple-fold.h.
9019 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9020 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9021 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9022 Add a path that uses a step of VF instead of 1, but disable it
9024 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9025 and niters_no_overflow parameters. Update calls to
9026 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9027 Create a new SSA name if the latter choses to use a ste other
9028 than zero, and return it via niters_vector_mult_vf_var.
9029 * tree-vect-loop.c (vect_transform_loop): Update calls to
9030 vect_do_peeling, vect_gen_vector_loop_niters and
9031 slpeel_make_loop_iterate_ntimes.
9032 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9033 (vect_gen_vector_loop_niters): Update declarations after above changes.
9035 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9037 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9038 128-bit round to integer instructions.
9039 (ceil<mode>2): Likewise.
9040 (btrunc<mode>2): Likewise.
9041 (round<mode>2): Likewise.
9043 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9045 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9046 unaligned VSX load/store on P8/P9.
9047 (expand_block_clear): Allow the use of unaligned VSX
9048 load/store on P8/P9.
9050 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9052 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9054 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9055 swap associated with both a load and a store.
9057 2018-01-02 Andrew Waterman <andrew@sifive.com>
9059 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9060 * config/riscv/riscv.md (clear_cache): Use it.
9062 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9064 * web.c: Remove out-of-date comment.
9066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9068 * expr.c (fixup_args_size_notes): Check that any existing
9069 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9070 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9071 (emit_single_push_insn): ...here.
9073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9075 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9076 (const_vector_encoded_nelts): New function.
9077 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9078 (const_vector_int_elt, const_vector_elt): Declare.
9079 * emit-rtl.c (const_vector_int_elt_1): New function.
9080 (const_vector_elt): Likewise.
9081 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9082 of CONST_VECTOR_ELT.
9084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9086 * expr.c: Include rtx-vector-builder.h.
9087 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9088 directly on the tree encoding.
9089 (const_vector_from_tree): Likewise.
9090 * optabs.c: Include rtx-vector-builder.h.
9091 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9092 sequence of "u" values.
9093 * vec-perm-indices.c: Include rtx-vector-builder.h.
9094 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9095 directly on the vec_perm_indices encoding.
9097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9099 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9100 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9101 * rtx-vector-builder.h: New file.
9102 * rtx-vector-builder.c: Likewise.
9103 * rtl.h (rtx_def::u2): Add a const_vector field.
9104 (CONST_VECTOR_NPATTERNS): New macro.
9105 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9106 (CONST_VECTOR_DUPLICATE_P): Likewise.
9107 (CONST_VECTOR_STEPPED_P): Likewise.
9108 (CONST_VECTOR_ENCODED_ELT): Likewise.
9109 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9110 (unwrap_const_vec_duplicate): Likewise.
9111 (const_vec_series_p): Check for a non-duplicated vector encoding.
9112 Say that the function only returns true for integer vectors.
9113 * emit-rtl.c: Include rtx-vector-builder.h.
9114 (gen_const_vec_duplicate_1): Delete.
9115 (gen_const_vector): Call gen_const_vec_duplicate instead of
9116 gen_const_vec_duplicate_1.
9117 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9118 (gen_const_vec_duplicate): Use rtx_vector_builder.
9119 (gen_const_vec_series): Likewise.
9120 (gen_rtx_CONST_VECTOR): Likewise.
9121 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9122 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9123 Build a new vector rather than modifying a CONST_VECTOR in-place.
9124 (handle_special_swappables): Update call accordingly.
9125 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9126 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9127 Build a new vector rather than modifying a CONST_VECTOR in-place.
9128 (handle_special_swappables): Update call accordingly.
9130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9132 * simplify-rtx.c (simplify_const_binary_operation): Use
9133 CONST_VECTOR_ELT instead of XVECEXP.
9135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9137 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9138 the selector elements to be different from the data elements
9139 if the selector is a VECTOR_CST.
9140 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9141 ssizetype for the selector.
9143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9145 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9146 before testing each element individually.
9147 * tree-vect-generic.c (lower_vec_perm): Likewise.
9149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9151 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9152 * selftest-run-tests.c (selftest::run_tests): Call it.
9153 * vector-builder.h (vector_builder::operator ==): New function.
9154 (vector_builder::operator !=): Likewise.
9155 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9156 (vec_perm_indices::all_from_input_p): New function.
9157 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9158 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9159 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9160 instead of reading the VECTOR_CST directly. Detect whether both
9161 vector inputs are the same before constructing the vec_perm_indices,
9162 and update the number of inputs argument accordingly. Use the
9163 utility functions added above. Only construct sel2 if we need to.
9165 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9167 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9168 the broadcast of the low byte.
9169 (expand_mult_highpart): Use an explicit encoding for the permutes.
9170 * optabs-query.c (can_mult_highpart_p): Likewise.
9171 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9172 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9173 (vectorizable_bswap): Likewise.
9174 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9175 explicit encoding for the power-of-2 permutes.
9176 (vect_permute_store_chain): Likewise.
9177 (vect_grouped_load_supported): Likewise.
9178 (vect_permute_load_chain): Likewise.
9180 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9182 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9183 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9184 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9185 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9186 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9187 (vect_gen_perm_mask_any): Likewise.
9189 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9191 * int-vector-builder.h: New file.
9192 * vec-perm-indices.h: Include int-vector-builder.h.
9193 (vec_perm_indices): Redefine as an int_vector_builder.
9194 (auto_vec_perm_indices): Delete.
9195 (vec_perm_builder): Redefine as a stand-alone class.
9196 (vec_perm_indices::vec_perm_indices): New function.
9197 (vec_perm_indices::clamp): Likewise.
9198 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9199 (vec_perm_indices::new_vector): New function.
9200 (vec_perm_indices::new_expanded_vector): Update for new
9201 vec_perm_indices class.
9202 (vec_perm_indices::rotate_inputs): New function.
9203 (vec_perm_indices::all_in_range_p): Operate directly on the
9204 encoded form, without computing elided elements.
9205 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9206 encoding. Update for new vec_perm_indices class.
9207 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9208 the given vec_perm_builder.
9209 (expand_vec_perm_var): Update vec_perm_builder constructor.
9210 (expand_mult_highpart): Use vec_perm_builder instead of
9211 auto_vec_perm_indices.
9212 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9213 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9214 or double series encoding as appropriate.
9215 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9216 vec_perm_indices instead of auto_vec_perm_indices.
9217 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9218 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9219 (vect_permute_store_chain): Likewise.
9220 (vect_grouped_load_supported): Likewise.
9221 (vect_permute_load_chain): Likewise.
9222 (vect_shift_permute_load_chain): Likewise.
9223 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9224 (vect_transform_slp_perm_load): Likewise.
9225 (vect_schedule_slp_instance): Likewise.
9226 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9227 (vectorizable_mask_load_store): Likewise.
9228 (vectorizable_bswap): Likewise.
9229 (vectorizable_store): Likewise.
9230 (vectorizable_load): Likewise.
9231 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9232 vec_perm_indices instead of auto_vec_perm_indices. Use
9233 tree_to_vec_perm_builder to read the vector from a tree.
9234 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9235 vec_perm_builder instead of a vec_perm_indices.
9236 (have_whole_vector_shift): Use vec_perm_builder and
9237 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9238 truncation to calc_vec_perm_mask_for_shift.
9239 (vect_create_epilog_for_reduction): Likewise.
9240 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9241 from auto_vec_perm_indices to vec_perm_indices.
9242 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9243 instead of changing individual elements.
9244 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9245 the vector in d.perm.
9246 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9247 from auto_vec_perm_indices to vec_perm_indices.
9248 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9249 instead of changing individual elements.
9250 (arm_vectorize_vec_perm_const): Use new_vector to install
9251 the vector in d.perm.
9252 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9253 Update vec_perm_builder constructor.
9254 (rs6000_expand_interleave): Likewise.
9255 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9256 (rs6000_expand_interleave): Likewise.
9258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9260 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9261 to qimode could truncate the indices.
9262 * optabs.c (expand_vec_perm_var): Likewise.
9264 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9266 * Makefile.in (OBJS): Add vec-perm-indices.o.
9267 * vec-perm-indices.h: New file.
9268 * vec-perm-indices.c: Likewise.
9269 * target.h (vec_perm_indices): Replace with a forward class
9271 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9272 * optabs.h: Include vec-perm-indices.h.
9273 (expand_vec_perm): Delete.
9274 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9275 (expand_vec_perm_const): Declare.
9276 * target.def (vec_perm_const_ok): Replace with...
9277 (vec_perm_const): ...this new hook.
9278 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9279 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9280 * doc/tm.texi: Regenerate.
9281 * optabs.def (vec_perm_const): Delete.
9282 * doc/md.texi (vec_perm_const): Likewise.
9283 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9284 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9285 expand_vec_perm for constant permutation vectors. Assert that
9286 the mode of variable permutation vectors is the integer equivalent
9287 of the mode that is being permuted.
9288 * optabs-query.h (selector_fits_mode_p): Declare.
9289 * optabs-query.c: Include vec-perm-indices.h.
9290 (selector_fits_mode_p): New function.
9291 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9292 is defined, instead of checking whether the vec_perm_const_optab
9293 exists. Use targetm.vectorize.vec_perm_const instead of
9294 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9295 fit in the vector mode before using a variable permute.
9296 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9297 vec_perm_indices instead of an rtx.
9298 (expand_vec_perm): Replace with...
9299 (expand_vec_perm_const): ...this new function. Take the selector
9300 as a vec_perm_indices rather than an rtx. Also take the mode of
9301 the selector. Update call to shift_amt_for_vec_perm_mask.
9302 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9303 Use vec_perm_indices::new_expanded_vector to expand the original
9304 selector into bytes. Check whether the indices fit in the vector
9305 mode before using a variable permute.
9306 (expand_vec_perm_var): Make global.
9307 (expand_mult_highpart): Use expand_vec_perm_const.
9308 * fold-const.c: Includes vec-perm-indices.h.
9309 * tree-ssa-forwprop.c: Likewise.
9310 * tree-vect-data-refs.c: Likewise.
9311 * tree-vect-generic.c: Likewise.
9312 * tree-vect-loop.c: Likewise.
9313 * tree-vect-slp.c: Likewise.
9314 * tree-vect-stmts.c: Likewise.
9315 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9317 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9318 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9319 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9320 (aarch64_vectorize_vec_perm_const): ...this new function.
9321 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9322 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9323 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9324 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9325 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9326 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9327 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9329 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9330 check for NEON modes.
9331 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9332 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9333 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9334 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9336 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9337 the old VEC_PERM_CONST conditions.
9338 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9339 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9340 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9341 (ia64_vectorize_vec_perm_const_ok): Merge into...
9342 (ia64_vectorize_vec_perm_const): ...this new function.
9343 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9344 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9345 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9346 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9347 * config/mips/mips.c (mips_expand_vec_perm_const)
9348 (mips_vectorize_vec_perm_const_ok): Merge into...
9349 (mips_vectorize_vec_perm_const): ...this new function.
9350 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9351 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9352 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9353 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9354 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9355 (rs6000_expand_vec_perm_const): Delete.
9356 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9358 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9359 (altivec_expand_vec_perm_const_le): Take each operand individually.
9360 Operate on constant selectors rather than rtxes.
9361 (altivec_expand_vec_perm_const): Likewise. Update call to
9362 altivec_expand_vec_perm_const_le.
9363 (rs6000_expand_vec_perm_const): Delete.
9364 (rs6000_vectorize_vec_perm_const_ok): Delete.
9365 (rs6000_vectorize_vec_perm_const): New function.
9366 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9367 an element count and rtx array.
9368 (rs6000_expand_extract_even): Update call accordingly.
9369 (rs6000_expand_interleave): Likewise.
9370 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9371 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9372 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9373 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9374 (rs6000_expand_vec_perm_const): Delete.
9375 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9376 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9377 (altivec_expand_vec_perm_const_le): Take each operand individually.
9378 Operate on constant selectors rather than rtxes.
9379 (altivec_expand_vec_perm_const): Likewise. Update call to
9380 altivec_expand_vec_perm_const_le.
9381 (rs6000_expand_vec_perm_const): Delete.
9382 (rs6000_vectorize_vec_perm_const_ok): Delete.
9383 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9384 reference to the SPE evmerge intructions.
9385 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9386 an element count and rtx array.
9387 (rs6000_expand_extract_even): Update call accordingly.
9388 (rs6000_expand_interleave): Likewise.
9389 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9390 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9392 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9396 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9397 vector mode and that that mode matches the mode of the data
9399 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9400 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9401 directly using expand_vec_perm_1 when forcing selectors into
9403 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9405 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9407 * optabs-query.h (can_vec_perm_p): Delete.
9408 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9409 * optabs-query.c (can_vec_perm_p): Split into...
9410 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9411 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9412 particular selector is valid.
9413 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9414 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9415 (vect_grouped_load_supported): Likewise.
9416 (vect_shift_permute_load_chain): Likewise.
9417 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9418 (vect_transform_slp_perm_load): Likewise.
9419 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9420 (vectorizable_bswap): Likewise.
9421 (vect_gen_perm_mask_checked): Likewise.
9422 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9423 implementations of variable permutation vectors into account
9424 when deciding which selector to use.
9425 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9426 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9427 with a false third argument.
9428 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9429 to test whether the constant selector is valid and can_vec_perm_var_p
9430 to test whether a variable selector is valid.
9432 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9434 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9435 * optabs-query.c (can_vec_perm_p): Likewise.
9436 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9437 instead of vec_perm_indices.
9438 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9439 (vect_gen_perm_mask_checked): Likewise,
9440 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9441 (vect_gen_perm_mask_checked): Likewise,
9443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9445 * optabs-query.h (qimode_for_vec_perm): Declare.
9446 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9447 (qimode_for_vec_perm): ...this new function.
9448 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9450 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9452 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9453 does not have a conditional at the top.
9455 2018-01-02 Richard Biener <rguenther@suse.de>
9457 * ipa-inline.c (big_speedup_p): Fix expression.
9459 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9462 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9465 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9469 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9470 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9471 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9472 cond_taken_branch_cost 3->4.
9474 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9476 PR tree-optimization/83581
9477 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9478 TODO_cleanup_cfg if any changes have been made.
9481 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9482 convert_modes if target mode has the right side, but different mode
9486 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9487 last argument when extracting from CONCAT. If either from_real or
9488 from_imag is NULL, use expansion through memory. If result is not
9489 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9490 the parts directly to inner mode, if even that fails, use expansion
9494 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9495 check for bswap in mode rather than HImode and use that in expand_unop
9498 Copyright (C) 2018 Free Software Foundation, Inc.
9500 Copying and distribu