1 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4 * function.c (match_asm_constraints_1): Don't do the optimization
5 if input isn't a REG, SUBREG, MEM or constant.
7 2018-03-22 Tom de Vries <tom@codesourcery.com>
9 PR tree-optimization/84956
10 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
17 DECL_INITIAL (decl) to decl at the end.
18 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
21 2018-03-21 Joseph Myers <joseph@codesourcery.com>
23 * doc/extend.texi (__builtin_tgmath): Document when complex
24 integer types are treated as _Complex _Float64.
26 2018-03-21 Tom de Vries <tom@codesourcery.com>
28 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
30 2018-03-21 Jakub Jelinek <jakub@redhat.com>
32 PR tree-optimization/84960
33 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
34 if it is ENTRY block, move them into single succ of ENTRY in that case.
36 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
38 PR tree-optimization/84811
39 * poly-int.h (poly_span_traits): Remove the T3 parameter and
40 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
41 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
42 (known_subrange_p): Update accordingly. Cast each value involved
43 in the size comparison, rather than casting the result of the
46 2018-03-21 Jakub Jelinek <jakub@redhat.com>
48 PR tree-optimization/84982
49 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
50 by flipping the least significant bit rather than all bits from
51 bitpos to bitpos + bitsize - 1.
53 2018-03-21 Nathan Sidwell <nathan@acm.org>
55 * doc/extend.texi (Deprecated Features): Remove mention of
56 long-deleted deprecations.
58 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
61 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
62 * configure: Regenerate.
64 2018-03-21 Tom de Vries <tom@codesourcery.com>
66 PR tree-optimization/83126
67 * tree-parloops.c (num_phis): New function.
68 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
70 2018-03-21 Nathan Sidwell <nathan@acm.org>
72 * doc/extend.texi (Deprecated Features): Update deprecated flags,
73 mention anon-struct/union members and trailing attributes.
75 2018-03-21 Bin Cheng <bin.cheng@arm.com>
77 PR tree-optimization/84969
78 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
79 builtin memset partitions if they set different rhs values.
81 2018-03-21 Jakub Jelinek <jakub@redhat.com>
83 PR rtl-optimization/84989
84 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
85 VEC_DUPLICATE with scalar result mode.
87 2018-03-21 Martin Liska <mliska@suse.cz>
90 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
91 not intended return statement.
93 2018-03-21 Martin Liska <mliska@suse.cz>
96 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
97 (chkp_find_bound_slots_1): Limit number of iterations.
99 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
102 * Minor grammar fixes for x86 options.
104 2018-03-20 Jakub Jelinek <jakub@redhat.com>
107 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
108 holding REG_CFA_RESTORE notes, instead turn them into a USE.
110 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
113 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
114 (altivec_lvx_<mode>_1op): Likewise.
115 (altivec_stvx_<mode>_2op): Likewise.
116 (altivec_stvx_<mode>_1op): Likewise.
117 (altivec_lvx_<VM2:mode>): New define_expand.
118 (altivec_stvx_<VM2:mode>): Likewise.
119 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
120 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
121 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
122 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
123 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
124 (rs6000_gen_lvx): Likewise.
125 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
126 (altivec_expand_stv_builtin): Likewise.
127 (altivec_expand_builtin): Likewise.
128 * config/rs6000/vector.md: Likewise.
130 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
133 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
136 2018-03-20 Richard Biener <rguenther@suse.de>
139 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
140 sign-conversions as zero, fall back to standard scalar_stmt
143 2018-03-20 Martin Liska <mliska@suse.cz>
146 * predict.c (rebuild_frequencies): Handle case when we have
147 PROFILE_ABSENT, but flag_guess_branch_prob is false.
149 2018-03-20 Jakub Jelinek <jakub@redhat.com>
152 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
153 flag_section_anchors.
154 * varasm.c (use_blocks_for_decl_p): Remove hack for
158 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
160 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
161 be created, use lowpart_subreg of operands[0] rather than operands[0]
163 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
164 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
165 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
166 and n constraint instead of aarch64_shift_imm_di and Usd.
167 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
168 (*aarch64_<optab>_reg_minus<mode>3): ... this.
170 2018-03-20 Sudakshina Das <sudi.das@arm.com>
173 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
174 to favor GPR over NEON registers.
175 (<shift>di3_neon): Likewise.
177 2018-03-20 Tom de Vries <tom@codesourcery.com>
180 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
181 (nvptx_process_pars): Emit bar.sync asap and alap.
183 2018-03-20 Tom de Vries <tom@codesourcery.com>
186 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
187 seen_label if seen_label is already set.
189 2018-03-20 Jakub Jelinek <jakub@redhat.com>
192 * config/i386/i386.c (fold_builtin_cpu): For features above 31
193 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
194 Use 1U instead of 1. Formatting fixes.
197 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
198 instead of TREE_TYPE (s1) for the return value.
200 2018-03-19 Jakub Jelinek <jakub@redhat.com>
202 PR tree-optimization/84946
203 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
204 bitsize + bitsize in poly_uint64 rather than poly_int64.
207 * dwarf2asm.c: Include fold-const.c.
208 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
209 of decl rather than decl itself.
211 PR rtl-optimization/84643
212 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
214 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
217 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
218 calling assemble_variable.
220 2018-03-19 Sudakshina Das <sudi.das@arm.com>
223 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
224 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
226 2018-03-19 Jim Wilson <jimw@sifive.com>
229 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
230 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
231 (riscv_first_stack_step): Likewise.
232 (riscv_option_override): Use STACK_BOUNDARY instead of
234 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
236 (BIGGEST_ALIGNMENT): Set to 128.
237 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
238 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
241 2018-03-19 Richard Biener <rguenther@suse.de>
243 PR tree-optimization/84933
244 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
245 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
247 2018-03-19 Richard Biener <rguenther@suse.de>
249 PR tree-optimization/84859
250 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
251 (cond_if_else_store_replacement): Perform sinking operation on
252 single-store BBs regardless of MAX_STORES_TO_SINK setting.
253 Generalize what a BB with a single eligible store is.
255 2018-03-19 Richard Biener <rguenther@suse.de>
257 PR tree-optimization/84929
258 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
259 chrec_is_positive against non-chrec arg.
261 2018-03-19 Tamar Christina <tamar.christina@arm.com>
264 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
266 2018-03-18 Martin Liska <mliska@suse.cz>
268 PR rtl-optimization/84635
269 * regrename.c (build_def_use): Use matches_mode only when
272 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
274 PR tree-optimization/84913
275 * tree-vect-loop.c (vectorizable_reduction): Don't try to
276 vectorize chains of COND_EXPRs.
278 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
280 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
282 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
284 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
286 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
288 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
290 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
291 Kito Cheng <kito.cheng@gmail.com>
293 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
294 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
295 (nds32_adjust_reg_alloc_order): New function.
296 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
298 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
300 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
301 nds32_print_operand, nds32_print_operand_address): Use
302 HOST_WIDE_INT_PRINT_DEC instead.
304 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
306 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
308 2018-03-17 Jakub Jelinek <jakub@redhat.com>
311 * config/i386/i386.c (initial_ix86_tune_features,
312 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
314 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
315 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
316 rather than 1u << ix86_tune. Formatting fix.
317 (ix86_option_override_internal): Change ix86_arch_mask from
318 unsigned int to unsigned HOST_WIDE_INT, initialize to
319 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
320 (ix86_function_specific_restore): Likewise.
322 2018-03-16 Jakub Jelinek <jakub@redhat.com>
325 * postreload.c (reload_combine_recognize_pattern): Perform
326 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
327 truncate_int_for_mode the result for the destination's mode.
330 * hsa-gen.c (mem_type_for_type): Fix comment typo.
331 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
333 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
336 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
339 * lra-assigns.c (lra_split_hard_reg_for): Don't use
340 regno_allocno_class_array and sorted_pseudos.
341 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
342 insns where regno is used.
344 2018-03-16 Martin Liska <mliska@suse.cz>
347 * multiple_target.c (create_dispatcher_calls): Redirect
348 reference in the symbol table.
350 2018-03-16 Martin Liska <mliska@suse.cz>
353 * multiple_target.c (create_dispatcher_calls): Redirect also
356 2018-03-16 Jakub Jelinek <jakub@redhat.com>
360 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
361 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
362 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
364 2018-03-16 Julia Koval <julia.koval@intel.com>
366 * doc/invoke.texi (Skylake Server): Add CLWB.
367 Cannonlake): Remove CLWB.
369 2018-03-16 Jakub Jelinek <jakub@redhat.com>
371 PR tree-optimization/84841
372 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
374 (FLOAT_ONE_CONST_TYPE): Define.
375 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
376 (sort_by_operand_rank): Put entries with higher constant_type last
377 rather than first to match comments.
379 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
381 * config/nios2/nios2.md (movsi_internal): Fix thinko in
384 2018-03-15 Jakub Jelinek <jakub@redhat.com>
387 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
388 check and use address of target always.
390 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
393 * config/i386/i386.c (indirect_thunk_needed): Update comments.
394 (indirect_thunk_bnd_needed): Likewise.
395 (indirect_thunks_used): Likewise.
396 (indirect_thunks_bnd_used): Likewise.
397 (indirect_return_needed): New.
398 (indirect_return_bnd_needed): Likewise.
399 (output_indirect_thunk_function): Add a bool argument for
401 (output_indirect_thunk_function): Don't generate alias for
402 function return thunk.
403 (ix86_code_end): Call output_indirect_thunk_function to generate
404 function return thunks.
405 (ix86_output_function_return): Set indirect_return_bnd_needed
406 and indirect_return_needed instead of indirect_thunk_bnd_needed
407 and indirect_thunk_needed.
409 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
411 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
412 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
413 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
415 2018-03-15 David Malcolm <dmalcolm@redhat.com>
416 Paul Hua <paul.hua.gm@gmail.com>
419 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
421 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
423 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
424 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
427 2018-03-15 Tamar Christina <tamar.christina@arm.com>
430 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
431 instead of GET_MODE_SIZE when comparing Units.
433 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
436 * varasm.c (hash_section): Return an unchangeble hash value
437 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
438 Return !aarch64_can_use_per_function_literal_pools_p ().
440 2018-03-15 Jakub Jelinek <jakub@redhat.com>
443 * optabs.c (emit_conditional_move): Pass address of cmode's copy
444 rather than address of cmode as last argument to prepare_cmp_insn.
446 2018-03-15 Julia Koval <julia.koval@intel.com>
448 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
449 F_AVX512VNNI, F_AVX512BITALG): New.
451 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
454 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
455 insn for floating-point loads and stores.
457 2018-03-14 Carl Love <cel@us.ibm.com>
459 * config/rs6000/rs6000-c.c: Add macro definitions for
460 ALTIVEC_BUILTIN_VEC_PERMXOR.
461 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
462 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
463 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
464 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
466 * config/doc/extend.texi: Add prototypes for vec_permxor.
468 2018-03-14 David Malcolm <dmalcolm@redhat.com>
471 * diagnostic-show-locus.c (class layout_point): Convert m_line
472 from int to linenum_type.
473 (line_span::comparator): Use linenum "compare" function when
474 comparing line numbers.
475 (test_line_span): New function.
476 (layout_range::contains_point): Convert param "row" from int to
478 (layout_range::intersects_line_p): Likewise.
479 (layout::will_show_line_p): Likewise.
480 (layout::print_source_line): Likewise.
481 (layout::should_print_annotation_line_p): Likewise.
482 (layout::print_annotation_line): Likewise.
483 (layout::print_leading_fixits): Likewise.
484 (layout::annotation_line_showed_range_p): Likewise.
485 (struct line_corrections): Likewise for field m_row.
486 (line_corrections::line_corrections): Likewise for param "row".
487 (layout::print_trailing_fixits): Likewise.
488 (layout::get_state_at_point): Likewise.
489 (layout::get_x_bound_for_row): Likewise.
490 (layout::print_line): Likewise.
491 (diagnostic_show_locus): Likewise for locals "last_line" and
493 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
494 * input.c (selftest::test_linenum_comparisons): New function.
495 (selftest::input_c_tests): Call it.
496 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
497 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
498 * selftest.h (ASSERT_GT): New macro.
499 (ASSERT_GT_AT): New macro.
500 (ASSERT_LT): New macro.
501 (ASSERT_LT_AT): New macro.
503 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
505 PR rtl-optimization/84780
506 * combine.c (distribute_links): Don't make a link based on pc_rtx.
508 2018-03-14 Martin Liska <mliska@suse.cz>
510 * tree.c (record_node_allocation_statistics): Use
512 (get_stats_node_kind): New function extracted from
513 record_node_allocation_statistics.
514 (free_node): Use get_stats_node_kind.
516 2018-03-14 Richard Biener <rguenther@suse.de>
518 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
519 that the value-set of ANTIC_IN doesn't grow.
522 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
524 (BB_VISITED_WITH_VISITED_SUCCS): New define.
525 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
527 2018-03-14 Julia Koval <julia.koval@intel.com>
529 * config.gcc (icelake-client, icelake-server): New.
531 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
532 (initial_ix86_arch_features): Ditto.
533 (PTA_SKYLAKE): Add SGX.
534 (PTA_ICELAKE): Remove.
535 (PTA_ICELAKE_CLIENT): New.
536 (PTA_ICELAKE_SERVER): New.
537 (ix86_option_override_internal): Split up icelake on icelake client and
539 (get_builtin_code_for_version): Ditto.
540 (fold_builtin_cpu): Ditto.
541 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
542 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
543 * config/i386/i386.h (processor_type): Ditto.
544 * doc/invoke.texi: Ditto.
546 2018-03-14 Jakub Jelinek <jakub@redhat.com>
549 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
550 INTEGER_CST offset, add it together with bitpos / 8 and
551 sign extend based on POINTER_SIZE.
555 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
558 * config/i386/constraints.md (Yc): New register constraint.
559 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
560 Use Yc constraint for alternative 2 of operand 0. Remove
561 preferred_for_speed attribute.
563 2018-03-14 Richard Biener <rguenther@suse.de>
565 PR tree-optimization/84830
566 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
567 with the old one to avoid oscillations.
569 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
572 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
574 (assign_by_spills): Return a flag of reload assignment failure.
575 Do not process the reload assignment failures. Do not spill other
576 reload pseudos if they has the same reg class. Update n if
578 (lra_assign): Add a return arg. Set up from the result of
579 assign_by_spills call.
580 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
581 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
582 usage_insns if it is not NULL.
583 (spill_hard_reg_in_range): New function.
584 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
585 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
587 (lra_assign): Change prototype.
588 * lra.c (lra): Add code to deal with fails by splitting hard reg
591 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
593 * config/riscv/riscv.opt (mrelax): New option.
594 * config/riscv/riscv.c (riscv_file_start): Emit ".option
595 "norelax" when riscv_mrelax is disabled.
596 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
598 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
601 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
602 reassociation for int modes.
604 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
606 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
607 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
609 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
610 * config/aarch64/aarch64-sve.md
611 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
612 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
613 (*extend<mode><Vwide>2): Rename to...
614 (aarch64_sve_extend<mode><Vwide>2): ...this.
615 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
616 renaming the old pattern to...
617 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
619 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
620 define_expand, renaming the old pattern to...
621 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
622 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
623 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
624 account when deciding which SVE instruction the optab should use.
625 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
627 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
629 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
630 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
631 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
632 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
633 (tlsdesc_small_<mode>): Turn a define_expand and use
634 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
635 (tlsdesc_small_advsimd_<mode>): ...this.
636 (tlsdesc_small_sve_<mode>): New pattern.
638 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
640 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
641 (UNSPEC_UMUL_HIGHPART): New constants.
642 (MUL_HIGHPART): New int iteraor.
643 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
644 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
646 (*<su>mul<mode>3_highpart): New define_insn.
648 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
651 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
654 2018-03-13 Martin Liska <mliska@suse.cz>
657 * (sem_item_optimizer::sem_item_optimizer): Initialize new
659 (sem_item_optimizer::~sem_item_optimizer): Release it.
660 (sem_item_optimizer::merge_classes): Register variable aliases.
661 (sem_item_optimizer::fixup_pt_set): New function.
662 (sem_item_optimizer::fixup_points_to_sets): Likewise.
663 * ipa-icf.h: Declare new variables and functions.
665 2018-03-13 Jakub Jelinek <jakub@redhat.com>
668 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
669 integer_pow2p@2 and test integer_pow2p in condition.
670 (A < 0 ? C : 0): Similarly for @1.
673 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
674 characters starting at p contain '\0' character, don't look beyond
678 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
679 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
682 * reg-stack.c (change_stack): Change update_end var from int to
683 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
684 also call set_block_for_insn on the newly added insns and rescan.
687 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
691 * tree.c (stabilize_reference_1): Return save_expr (e) for
692 STATEMENT_LIST even if it doesn't have side-effects.
694 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
696 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
698 2018-03-12 Renlin Li <renlin.li@arm.com>
700 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
701 aarch64_output_scalar_simd_mov_immediate.
703 2018-03-12 Martin Sebor <msebor@redhat.com>
705 PR tree-optimization/83456
706 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
707 for perfectly overlapping calls to memcpy.
708 (gimple_fold_builtin_memory_chk): Same.
709 (gimple_fold_builtin_strcpy): Handle no-warning.
710 (gimple_fold_builtin_stxcpy_chk): Same.
711 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
713 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
715 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
716 parameter. Use it for SFmode.
717 (rs6000_function_arg_advance_1): Adjust.
718 (rs6000_function_arg): Adjust.
719 (rs6000_gimplify_va_arg): Pass false for that new parameter.
721 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
723 PR rtl-optimization/84169
724 PR rtl-optimization/84780
725 * combine.c (can_combine_p): Check for a 2-insn combination whether
726 the destination register is used between the two insns, too.
728 2018-03-12 Richard Biener <rguenther@suse.de>
730 PR tree-optimization/84803
731 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
732 for refs DR analysis didn't process.
734 2018-03-12 Richard Biener <rguenther@suse.de>
736 PR tree-optimization/84777
737 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
738 force-vectorize loops ignore whether we are optimizing for size.
740 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
742 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
743 (TARGET_MD_ASM_ADJUST): Define.
745 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
746 Kito Cheng <kito.cheng@gmail.com>
747 Chung-Ju Wu <jasonwucj@gmail.com>
749 * config/nds32/nds32.c (nds32_compute_stack_frame,
750 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
751 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
752 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
753 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
754 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
755 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
756 * config/nds32/nds32.md (prologue, epilogue): Use macro
757 NDS32_V3PUSH_AVAILABLE_P to do checking.
759 2018-03-11 Jakub Jelinek <jakub@redhat.com>
762 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
763 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
764 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
765 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
766 addition of most attributes on !orig_type_die or the attribute not
767 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
769 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
770 Chung-Ju Wu <jasonwucj@gmail.com>
772 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
774 * config/nds32/nds32.opt (mvh): New option.
776 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
777 Chung-Ju Wu <jasonwucj@gmail.com>
779 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
781 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
782 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
785 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
786 Chung-Ju Wu <jasonwucj@gmail.com>
788 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
790 * config/nds32/nds32-multiple.md (strlensi): New pattern.
791 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
793 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
794 Kito Cheng <kito.cheng@gmail.com>
795 Chung-Ju Wu <jasonwucj@gmail.com>
797 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
798 UNSPEC_FFMISM and UNSPEC_FLMISM.
799 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
800 for ffb, ffmism and flmism.
801 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
802 (unspec_ffmism): Ditto.
803 (unspec_flmism): Ditto.
804 (nds32_expand_builtin_impl): Check if string extension is available.
805 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
806 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
808 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
811 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
814 * lra-assigns.c (assign_by_spills): Return a flag of reload
815 assignment failure. Do not process the reload assignment
816 failures. Do not spill other reload pseudos if they has the same
818 (lra_assign): Add a return arg. Set up from the result of
819 assign_by_spills call.
820 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
821 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
822 usage_insns if it is not NULL.
823 (spill_hard_reg_in_range): New function.
824 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
825 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
827 (lra_assign): Change prototype.
828 * lra.c (lra): Add code to deal with fails by splitting hard reg
831 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
834 * config/i386/i386.opt: Replace Enforcment with Enforcement.
836 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
839 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
840 (dw_val_node): Add val_symbolic_view.
841 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
842 (symview_upper_bound): New.
843 (new_line_info_table): Initialize symviews_since_reset.
844 (dwarf2out_source_line): Count symviews_since_reset and set
846 (dw_val_equal_p): Handle symview.
847 (add_AT_symview): New.
848 (print_dw_val): Handle symview.
849 (attr_checksum, attr_checksum_ordered): Likewise.
850 (same_dw_val_p, size_of_die): Likewise.
851 (value_format, output_die): Likewise.
852 (add_high_low_attributes): Use add_AT_symview for entry_view.
853 (dwarf2out_finish): Reset symview_upper_bound, clear
856 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
859 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
860 Add strict argument and use it.
861 (rs6000_split_multireg_move): Update for new strict argument.
862 (mem_operand_gpr): Disallow all non-offsettable addresses.
863 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
865 2018-03-09 Jakub Jelinek <jakub@redhat.com>
868 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
869 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
870 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
873 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
874 decl, use remap_type if we want to use the type.
876 2018-03-09 Martin Sebor <msebor@redhat.com>
878 PR tree-optimization/84526
879 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
881 (builtin_access::generic_overlap): Be prepared to handle non-array
884 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
886 PR rtl-optimization/84682
887 * lra-constraints.c (process_address_1): Check is_address flag
888 for address constraints.
889 (process_alt_operands): Likewise.
890 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
891 preprocess_constraints.
892 * recog.h (preprocess_constraints): Add oploc parameter.
894 * recog.c (preprocess_constraints): Test address_operand for
895 CT_ADDRESS constraints.
897 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
900 * lra-assigns.c (assign_by_spills): Return a flag of reload
901 assignment failure. Do not process the reload assignment
902 failures. Do not spill other reload pseudos if they has the same
904 (lra_assign): Add a return arg. Set up from the result of
905 assign_by_spills call.
906 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
907 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
908 usage_insns if it is not NULL.
909 (spill_hard_reg_in_range): New function.
910 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
911 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
913 (lra_assign): Change prototype.
914 * lra.c (lra): Add code to deal with fails by splitting hard reg
917 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
920 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
921 Accept complain bool parameter. Only emit errors if it is true.
922 (arm_parse_cpu_option_name): Likewise.
923 (arm_target_thumb_only): Adjust callers of the above.
924 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
925 prototype to take a default true bool parameter.
926 (arm_parse_arch_option_name): Likewise.
928 2018-03-09 David Malcolm <dmalcolm@redhat.com>
929 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
933 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
934 * configure: Regenerate.
935 * configure.ac ("linker --version-script option"): New.
936 ("linker soname option"): New.
938 2018-03-09 Richard Biener <rguenther@suse.de>
940 PR tree-optimization/84775
941 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
942 immediate uses of predicate stmts and mark them modified.
945 PR tree-optimization/84178
946 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
948 (version_loop_for_if_conversion): Delay update_ssa call.
949 (tree_if_conversion): Delay update_ssa until after predicate
952 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
955 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
956 when the function accesses prior frames.
958 2018-03-08 Jakub Jelinek <jakub@redhat.com>
961 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
962 gen_llsym, otherwise call maybe_gen_llsym.
965 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
966 has ',' character inside of it.
968 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
971 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
972 as clobbering CC_REGNUM.
974 2018-03-08 Richard Biener <rguenther@suse.de>
977 * tree-scalar-evolution.c: Include tree-into-ssa.h.
978 (follow_copies_to_constant): Do not follow SSA names registered
981 2018-03-08 Richard Biener <rguenther@suse.de>
983 PR tree-optimization/84178
984 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
986 (version_loop_for_if_conversion): Delay update_ssa call.
987 (tree_if_conversion): Delay update_ssa until after predicate
990 2018-03-08 David Malcolm <dmalcolm@redhat.com>
992 PR tree-optimization/84178
993 * tree-if-conv.c (release_bb_predicate): Remove the
994 the assertion that the stmts have NULL use_ops.
995 Discard the statements, asserting that they haven't
996 yet been added to a BB.
998 2018-03-08 Richard Biener <rguenther@suse.de>
1000 PR tree-optimization/84746
1001 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1002 (phi_translate): Pass in destination ANTIC_OUT set.
1003 (phi_translate_1): Likewise. For a simplified result lookup
1004 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1005 (phi_translate_set): Adjust.
1006 (do_pre_regular_insertion): Likewise.
1007 (do_pre_partial_partial_insertion): Likewise.
1009 2018-03-08 Martin Liska <mliska@suse.cz>
1011 PR gcov-profile/84735
1012 * doc/gcov.texi: Document usage of profile files.
1013 * gcov-io.h: Document changes in the format.
1015 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1019 * dwarf2out.c (struct dw_line_info_table): Update comments for
1021 (FORCE_RESET_NEXT_VIEW): New.
1022 (FORCE_RESETTING_VIEW_P): New.
1023 (RESETTING_VIEW_P): Check for -1 too.
1024 (ZERO_VIEW_P): Likewise.
1025 (new_line_info_table): Force-reset next view.
1026 (dwarf2out_begin_function): Likewise.
1027 (dwarf2out_source_line): Simplify zero_view_p initialization.
1028 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1029 view directly. Omit view when omitting .loc at line 0.
1031 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1033 PR tree-optimization/84740
1034 * tree-switch-conversion.c (process_switch): Call build_constructors
1035 only if info.phi_count is non-zero.
1037 PR tree-optimization/84739
1038 * tree-tailcall.c (find_tail_calls): Check call arguments against
1039 DECL_ARGUMENTS (current_function_decl) rather than
1040 DECL_ARGUMENTS (func) when checking for tail recursion.
1042 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1044 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1045 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1046 Volker Reichelt's entry and add entries for people that perform
1047 GCC fuzzy testing and report numerous bugs.
1049 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1052 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1053 readonly data in sdata, if that is disabled.
1054 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1055 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1056 -mreadonly-in-sdata option.
1058 2018-03-07 Martin Sebor <msebor@redhat.com>
1060 PR tree-optimization/84468
1061 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1062 basic block when looking for nul assignment.
1064 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1067 * except.h (output_function_exception_table): Adjust prototype.
1068 * except.c (output_function_exception_table): Remove FNNAME parameter
1069 and add SECTION parameter. Ouput one part of the table at a time.
1070 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1071 the first part of the exception table and emit unwind directives.
1072 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1073 (i386_pe_seh_cold_init): Likewise.
1074 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1075 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1076 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1077 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1078 * config/i386/winnt.c: Include except.h.
1079 (struct seh_frame_state): Add reg_offset, after_prologue and
1080 in_cold_section fields.
1081 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1082 (i386_pe_seh_cold_init): New function.
1083 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1084 to seh->in_cold_section.
1085 (seh_emit_push): Record the offset of the push.
1086 (seh_emit_save): Record the offet of the save.
1087 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1088 Test seh->after_prologue to disregard the epilogue.
1089 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1090 (i386_pe_end_cold_function): New function.
1092 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1095 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1096 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1099 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1101 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1102 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1105 * multiple_target.c: Include tree-inline.h and intl.h.
1106 (expand_target_clones): Diagnose and fail if node->definition and
1107 !tree_versionable_function_p (node->decl).
1109 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1111 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1113 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1114 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1115 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1117 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1120 * combine.c (try_combine): Use reg_or_subregno instead of handling
1121 just paradoxical SUBREGs and REGs.
1123 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1125 * config/arc/arc.c (arc_finalize_pic): Remove function.
1126 (arc_must_save_register): We use single base PIC register, remove
1127 checks to save/restore the PIC register.
1128 (arc_expand_prologue): Likewise.
1129 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1131 (arc_verify_short): Likewise.
1132 (arc_attr_type): Likewise.
1133 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1134 (walk_stores): Likewise.
1135 (arc_address_cost): Make it static.
1136 (arc_verify_short): Likewise.
1137 (branch_dest): Likewise.
1138 (arc_attr_type): Likewise.
1139 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1140 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1141 (arc_final_prescan_insn): Remove inserting the nops due to
1142 hardware hazards. It is done in reorg step.
1143 (insn_length_variant_t): Remove.
1144 (insn_length_parameters_t): Likewise.
1145 (arc_insn_length_parameters): Likewise.
1146 (arc_get_insn_variants): Likewise.
1147 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1149 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1152 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1155 PR tree-optimization/84687
1156 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1158 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1160 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1162 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1163 Rename to ppc_speculation_barrier.
1164 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1165 __builtin_ppc_speculation_barrier.
1167 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1170 * combine.c (combine_simplify_rtx): Don't try to simplify if
1171 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1174 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1176 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1177 to 32 bytes when compiling for POWER9.
1179 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1182 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1183 regparm >= 3 with no arg reg available also for calls with
1184 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1187 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1189 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1191 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1194 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1196 2018-03-05 Richard Biener <rguenther@suse.de>
1198 PR tree-optimization/84486
1199 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1200 When inserting a __builtin_assume_aligned call set the LHS
1201 SSA name alignment info accordingly.
1203 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1205 PR tree-optimization/84114
1206 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1207 Avoid reassociation of FLOAT_MODE addition.
1209 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1211 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1212 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1213 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1214 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1215 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1216 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1217 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1219 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1220 __builtin_ia32_wbinvd): New builtins.
1221 (SPECIAL_ARGS2): New.
1222 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1223 (SPECIAL_ARGS2): New.
1224 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1225 (ix86_valid_target_attribute_inner_p): Ditto.
1226 (ix86_init_mmx_sse_builtins): Add special_args2.
1227 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1228 TARGET_WBNOINVD_P): New.
1229 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1230 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1231 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1232 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1233 * config/i386/pconfigintrin.h: New file.
1234 * config/i386/wbnoinvdintrin.h: Ditto.
1235 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1236 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1238 2018-03-05 Richard Biener <rguenther@suse.de>
1240 PR tree-optimization/84670
1241 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1243 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1244 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1245 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1246 doesn't grow if all successors (recursively) were visited at least
1249 2018-03-05 Richard Biener <rguenther@suse.de>
1251 PR tree-optimization/84650
1252 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1253 if executed in the loop pipeline.
1255 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1257 * doc/configfiles.texi (Configuration Files): Move info about
1258 conditionalizing $target-protos.h to...
1259 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1260 differs from $target-protos.h.
1262 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1263 Chung-Ju Wu <jasonwucj@gmail.com>
1265 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1266 * config/nds32/nds32-multiple.md (setmemsi): Define.
1267 * config/nds32/nds32-memory-manipulation.c
1268 (nds32_gen_dup_4_byte_to_word_value): New.
1269 (emit_setmem_word_loop): New.
1270 (emit_setmem_byte_loop): New.
1271 (nds32_expand_setmem_loop): New.
1272 (nds32_expand_setmem_loop_v3m): New.
1273 (nds32_expand_setmem_unroll): New.
1274 (nds32_expand_setmem): New.
1276 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1277 Chung-Ju Wu <jasonwucj@gmail.com>
1279 * config/nds32/nds32-memory-manipulation.c
1280 (nds32_emit_load_store): New.
1281 (nds32_emit_post_inc_load_store): New.
1282 (nds32_emit_mem_move): New.
1283 (nds32_emit_mem_move_block): New.
1284 (nds32_expand_movmemsi_loop_unknown_size): New.
1285 (nds32_expand_movmemsi_loop_known_size): New.
1286 (nds32_expand_movmemsi_loop): New.
1287 (nds32_expand_movmemsi_unroll): New.
1288 (nds32_expand_movmemqi): Rename ...
1289 (nds32_expand_movmemsi): ... to this.
1290 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1291 (movmemsi): ... to this.
1292 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1293 (nds32_expand_movmemsi): ... to this.
1295 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1296 Monk Chiang <sh.chiang04@gmail.com>
1297 Chung-Ju Wu <jasonwucj@gmail.com>
1299 * config/nds32/nds32-protos.h
1300 (nds32_expand_load_multiple): New arguments.
1301 (nds32_expand_store_multiple): Ditto.
1302 (nds32_valid_multiple_load_store): Rename ...
1303 (nds32_valid_multiple_load_store_p): ... to this.
1304 * config/nds32/nds32-memory-manipulation.c
1305 (nds32_expand_load_multiple): Refine implementation.
1306 (nds32_expand_store_multiple): Ditto.
1307 * config/nds32/nds32-multiple.md
1308 (load_multiple): Update nds32_expand_load_multiple interface.
1309 (store_multiple): Update nds32_expand_store_multiple interface.
1310 * config/nds32/nds32-predicates.c
1311 (nds32_valid_multiple_load_store): Rename ...
1312 (nds32_valid_multiple_load_store_p): ... to this and refine
1314 * config/nds32/predicates.md
1315 (nds32_load_multiple_and_update_address_operation): New predicate.
1316 (nds32_store_multiple_and_update_address_operation): New predicate.
1318 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1319 Chung-Ju Wu <jasonwucj@gmail.com>
1321 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1322 (combo): New attribute.
1323 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1325 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1327 * config/nds32/nds32.opt: Change -mcmodel= default value.
1329 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1330 Monk Chiang <sh.chiang04@gmail.com>
1331 Chung-Ju Wu <jasonwucj@gmail.com>
1333 * config/nds32/constants.md (unspec_element): New enum.
1334 * config/nds32/constraints.md (Umw): New constraint.
1335 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1336 * config/nds32/nds32-intrinsic.md: Likewise.
1337 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1338 (nds32_valid_smw_lwm_base_p): New.
1339 (nds32_output_smw_single_word): New.
1340 (nds32_output_lmw_single_word): New.
1341 (nds32_expand_unaligned_load): New.
1342 (nds32_expand_unaligned_store): New.
1343 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1344 (nds32_output_smw_single_word): Declare.
1345 (nds32_output_lmw_single_word): Declare.
1346 (nds32_expand_unaligned_load): Declare.
1347 (nds32_expand_unaligned_store): Declare.
1348 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1349 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1350 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1351 NDS32_BUILTIN_UASTORE_DW.
1352 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1355 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1356 Kito Cheng <kito.cheng@gmail.com>
1357 Chung-Ju Wu <jasonwucj@gmail.com>
1359 * config/nds32/nds32-intrinsic.c
1360 (nds32_expand_builtin_null_ftype_reg): Delete.
1361 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1362 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1363 (nds32_read_argument): New.
1364 (nds32_legitimize_target): Ditto.
1365 (nds32_legitimize_argument): Ditto.
1366 (nds32_check_constant_argument): Ditto.
1367 (nds32_expand_unop_builtin): Ditto.
1368 (nds32_expand_unopimm_builtin): Ditto.
1369 (nds32_expand_binop_builtin): Ditto.
1370 (nds32_builtin_decl_impl): Ditto.
1371 (builtin_description): Ditto.
1372 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1373 (nds32_init_builtins_impl): Ditto.
1374 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1375 (nds32_builtin_decl): New.
1376 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1377 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1379 2018-03-02 Jeff Law <law@redhat.com>
1381 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1382 (redundant_insn, fill_simple_delay_slots): Likewise.
1383 (fill_slots_from_thread): Likewise.
1384 * resource.c (mark_referenced_resources): Likewise.
1385 (mark_set_resources, find_dead_or_set_registers): Likewise.
1387 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1389 * substring-locations.h (format_warning_va): Formatting fix for
1391 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1393 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1394 * substring-locations.c: Include intl.h.
1395 (format_warning_va): Turned into small wrapper around
1396 format_warning_n_va, renamed to ...
1397 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1398 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1400 (format_warning_at_substring_n): New function.
1401 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1402 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1403 format_warning_at_substring with just a shorter name instead of
1404 const function pointer.
1405 (fmtwarn_n): New function.
1406 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1407 appropriate, get rid of all the fmtstr temporaries, move conditionals
1408 with G_() wrapped string literals directly into fmtwarn arguments,
1409 cast dir.len to (int), formatting fixes.
1411 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1413 * doc/invoke.texi: Remove "Cilk Plus" references.
1415 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1416 Richard Biener <rguenther@suse.de>
1419 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1420 for error or warning attributes if CALL_FROM_THUNK_P is set.
1423 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1426 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1427 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1430 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1431 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1433 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1434 instead of -1U in last predictors element's probability member.
1436 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1439 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1440 arguments if they are comparable.
1442 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1444 PR tree-optimization/84634
1445 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1446 masks and masked_loop_p with a single loop_masks, making sure it's
1447 null for bb vectorization.
1449 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1451 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1452 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1453 loop->force_vectorize to check whether there is no alias.
1455 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1458 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1460 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1462 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1463 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1464 instead of a loop around prev_real_insn.
1465 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1469 * config/i386/i386.c (ix86_print_operand): Use conditional
1470 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1473 2018-03-02 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/84427
1476 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1477 (bitmap_set_subtract_values): Rewrite to handle multiple
1480 (prune_clobbered_mems): Likewise.
1481 (phi_translate): Take edge instead of pred/phiblock.
1482 (phi_translate_1): Likewise.
1483 (phi_translate_set): Likewise. Insert all translated
1484 exprs for a value into the set, keeping possibly multiple
1485 expressions per value.
1486 (compute_antic_aux): Adjust for phi_translate changes.
1487 When intersecting union the expressions and prune those
1488 not in the final value set, keeping possibly multiple
1489 expressions per value. Do not use value-insertion
1490 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1491 all expressions. Add verification that the value-sets
1492 only shrink during iteration.
1493 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1494 (do_pre_regular_insertion): Likewise.
1495 (do_pre_partial_partial_insertion): Likewise.
1497 2018-03-02 Richard Biener <rguenther@suse.de>
1500 * config/darwin.c (saved_debug_info_level): New static global.
1501 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1502 (darwin_asm_lto_end): Restore debug info generation settings.
1504 2018-03-01 Martin Liska <mliska@suse.cz>
1507 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1510 2018-03-01 Richard Biener <rguenther@suse.de>
1513 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1514 type annotation with LTO when debug was disabled at compile-time.
1516 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1518 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1520 (mips_final_postscan_insn): Likewise.
1522 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1524 PR rtl-optimization/84528
1525 * alias.c (init_alias_target): Add commentary.
1526 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1527 a unique base value if the frame pointer is not eliminated
1528 to the stack pointer.
1530 2018-03-01 Tom de Vries <tom@codesourcery.com>
1532 PR rtl-optimization/83327
1533 * lra-int.h (hard_regs_spilled_into): Declare.
1534 * lra.c (hard_regs_spilled_into): Define.
1535 (init_reg_info): Init hard_regs_spilled_into.
1536 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1537 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1538 (process_bb_lives): Handle hard_regs_spilled_into.
1539 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1540 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1542 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1544 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1545 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1546 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1547 * config/rs6000/aix72.h: New file.
1549 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1551 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1552 instead of warning_at with conditional singular and plural messages
1556 * stor-layout.c (update_alignment_for_field): For
1557 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1558 && !DECL_PACKED (field), do the alignment update, just use
1559 only desired_align instead of MAX (type_align, desired_align)
1561 (place_field): Don't do known_align < desired_align handling
1562 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1563 is non-NULL, instead do it after rli->prev_field handling and
1564 only if not within a bitfield word. For DECL_PACKED (field)
1565 use type_align of BITS_PER_UNIT.
1567 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1569 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1570 superfluous parentheses and trailing spaces.
1572 2018-02-28 Richard Biener <rguenther@suse.de>
1574 PR tree-optimization/84584
1575 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1576 SCoPs with fake exit edge.
1578 2018-02-28 Martin Liska <mliska@suse.cz>
1581 * timevar.c (timer::print): Fix format to properly print 100%
1584 2018-02-28 Richard Biener <rguenther@suse.de>
1587 * genmatch.c (capture_info::walk_match): Do not mark
1588 captured expressions without operands as expr_p given
1589 they act more like predicates and should be subject to
1590 "lost tail" side-effect preserving.
1592 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1594 PR rtl-optimization/81611
1595 * auto-inc-dec.c (attempt_change): Move dead note from
1596 mem_insn if it's the next use of regno
1597 (find_address): Take address use of reg holding
1598 non-incremented value. Add parm to limit search to the named
1600 (merge_in_block): Attempt to use a mem insn that is the next
1601 use of the original regno.
1603 2018-02-27 Martin Sebor <msebor@redhat.com>
1606 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1607 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1609 2018-02-27 Martin Sebor <msebor@redhat.com>
1611 PR translation/84207
1612 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1613 n argument to unsigned HOST_WIDE_INT.
1614 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1615 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1616 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1617 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1619 2018-02-27 Richard Biener <rguenther@suse.de>
1621 PR tree-optimization/84512
1622 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1623 Do not use the estimate returned from record_stmt_cost for
1624 the scalar iteration cost but sum properly using add_stmt_cost.
1626 2018-02-27 Richard Biener <rguenther@suse.de>
1628 PR tree-optimization/84466
1629 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1630 Adjust last change to less strictly validate use operands.
1632 2018-02-27 Martin Liska <mliska@suse.cz>
1634 PR gcov-profile/84548
1635 * gcov.c (process_file): Allow partial overlap and consider it
1636 also as group functions.
1637 (output_lines): Properly calculate range of lines for a group.
1639 2018-02-27 Martin Liska <mliska@suse.cz>
1641 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1642 'ggc' suffixes. Change first column width.
1643 (timer::print): Fix formatting of the column.
1645 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1647 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1648 preserve inline entry blocks for the sake of debug inline
1649 entry point markers alone.
1650 (remove_unused_locals): Suggest in comments a better place to
1651 force the preservation of inline entry blocks that are
1652 otherwise unused, but do not preserve them.
1654 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1656 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1658 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1661 * config/i386/constraints.md (Bs): Replace
1662 ix86_indirect_branch_register with
1663 TARGET_INDIRECT_BRANCH_REGISTER.
1665 * config/i386/i386.md (indirect_jump): Likewise.
1666 (tablejump): Likewise.
1667 (*sibcall_memory): Likewise.
1668 (*sibcall_value_memory): Likewise.
1669 Peepholes of indirect call and jump via memory: Likewise.
1670 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1671 (*sibcall_value_GOT_32): Likewise.
1672 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1673 (GOT_memory_operand): Likewise.
1674 (call_insn_operand): Likewise.
1675 (sibcall_insn_operand): Likewise.
1676 (GOT32_symbol_operand): Likewise.
1677 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1679 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1681 PR rtl-optimization/83496
1682 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1683 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1684 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1685 redundant insn, if any.
1686 (relax_delay_slots): Likewise.
1687 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1689 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1691 PR tree-optimization/83965
1692 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1693 that grouped statements are part of a reduction chain. Return
1694 true if the statement is not marked as a reduction itself but
1696 (vect_recog_dot_prod_pattern): Don't check whether the statement
1697 is part of a group here.
1698 (vect_recog_sad_pattern): Likewise.
1699 (vect_recog_widen_sum_pattern): Likewise.
1701 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1704 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1706 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1709 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1711 (ix86_output_indirect_function_return): New prototype.
1712 (ix86_split_simple_return_pop_internal): Likewise.
1713 * config/i386/i386.c (indirect_return_via_cx): New.
1714 (indirect_return_via_cx_bnd): Likewise.
1715 (indirect_thunk_name): Handle return va CX_REG.
1716 (output_indirect_thunk_function): Create alias for
1717 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1718 (ix86_output_indirect_jmp): Remove the bool argument.
1719 (ix86_output_indirect_function_return): New function.
1720 (ix86_split_simple_return_pop_internal): Likewise.
1721 * config/i386/i386.md (*indirect_jump): Don't pass false
1722 to ix86_output_indirect_jmp.
1723 (*tablejump_1): Likewise.
1724 (simple_return_pop_internal): Change it to define_insn_and_split.
1725 Call ix86_split_simple_return_pop_internal to split it for
1727 (simple_return_indirect_internal): Call
1728 ix86_output_indirect_function_return instead of
1729 ix86_output_indirect_jmp.
1731 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1734 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1735 memset and value initialization afterwards.
1737 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1739 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1741 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1744 * common/config/aarch64/aarch64-common.c
1745 (aarch_option_optimization_table[]): Switch
1746 off fomit-frame-pointer
1748 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1749 Chung-Ju Wu <jasonwucj@gmail.com>
1751 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1753 (store_multiple): Ditto.
1755 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1757 * config.gcc: Add --with-cpu support for nds32 target.
1758 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1759 * config/nds32/nds32.opt: Add -mcpu= option.
1761 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1763 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1764 isel=yes): Warn for these deprecated options.
1766 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1768 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1769 ISA_2_5_MASKS_EMBEDDED.
1771 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1773 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1774 p->max as pointers rather than using iterative_hash_expr.
1776 2018-02-23 Carl Love <cel@us.ibm.com>
1778 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1779 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1781 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1782 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1783 P8V_BUILTIN_VEC_VUNSIGNED2.
1785 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1788 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1789 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1791 * lra-constraints.c (curr_insn_transform): Set up
1792 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1794 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1795 * lra-eliminations.c (spill_pseudos): Ditto.
1796 (process_insn_for_elimination): Ditto.
1797 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1798 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1799 LRA_NON_CLOBBERED_ALT.
1801 2018-02-22 Martin Sebor <msebor@redhat.com>
1803 PR tree-optimization/84480
1804 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1805 to maybe_diag_stxncpy_trunc. Call it.
1806 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1807 from gimple_fold_builtin_strcpy. Print inlining stack.
1808 (handle_builtin_stxncpy): Print inlining stack.
1809 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1811 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1814 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1815 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1816 and -fcheck-pointer-bounds are used together.
1817 (indirect_thunk_prefix): New enum.
1818 (indirect_thunk_need_prefix): New function.
1819 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1820 "_nt" instead of "_bnd" for NOTRACK prefix.
1821 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1822 (output_indirect_thunk_function): Likewise.
1824 (ix86_code_end): Update output_indirect_thunk_function calls.
1825 (ix86_output_indirect_branch_via_reg): Replace
1826 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1827 (ix86_output_indirect_branch_via_push): Likewise.
1828 (ix86_output_function_return): Likewise.
1829 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1830 incompatible with -fcf-protection=branch and
1831 -fcheck-pointer-bounds.
1833 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1836 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1837 Change gcc_assert call to output_operand_lossage.
1839 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1841 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1843 2018-02-22 DJ Delorie <dj@redhat.com>
1844 Sebastian Perta <sebastian.perta@renesas.com>
1845 Oleg Endo <olegendo@gcc.gnu.org>
1847 * config/rx/rx.c (rx_rtx_costs): New function.
1848 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1850 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1852 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1854 2018-02-22 Martin Liska <mliska@suse.cz>
1857 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1858 Add "native" as a possible value.
1860 2018-02-22 Martin Liska <mliska@suse.cz>
1863 * config/i386/i386.c (ix86_option_override_internal):
1864 Add "native" as a possible value for -march and -mtune.
1866 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1869 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1870 to all type variants.
1872 PR tree-optimization/84503
1873 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1874 width as info->bitpos + info->bitsize - start.
1875 (merged_store_group::merge_overlapping): Simplify width computation.
1876 (check_no_overlap): New function.
1877 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1878 start + width and last_order of the group, fail if check_no_overlap
1880 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1881 to group if check_no_overlap fails.
1883 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1885 * config/rs6000/altivec.md: Delete contraint arguments to
1886 define_expand, define_split, and define_peephole2, and in
1887 define_insn_and_split if always unused.
1888 * config/rs6000/darwin.md: Ditto.
1889 * config/rs6000/dfp.md: Ditto.
1890 * config/rs6000/rs6000.md: Ditto.
1891 * config/rs6000/sync.md: Ditto.
1892 * config/rs6000/vector.md: Ditto.
1893 * config/rs6000/vsx.md: Ditto.
1895 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1897 * config/rs6000/altivec.md: Write output control strings as braced
1898 blocks instead of double-quoted strings.
1899 * config/rs6000/darwin.md: Ditto.
1900 * config/rs6000/rs6000.md: Ditto.
1901 * config/rs6000/vector.md: Ditto.
1902 * config/rs6000/vsx.md: Ditto.
1904 2018-02-21 Jason Merrill <jason@redhat.com>
1906 PR c++/84314 - ICE with templates and fastcall attribute.
1907 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1909 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1911 * ipa-cp.c (determine_versionability): Fix comment typos.
1913 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1916 * ipa-cp.c (determine_versionability): Do not version functions caling
1919 2018-02-21 Martin Liska <mliska@suse.cz>
1922 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1923 Add "native" as a possible value.
1924 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1925 the macro when native cpu detection is available.
1927 2018-02-21 Martin Liska <mliska@suse.cz>
1930 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1931 Add "native" as a possible value.
1932 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1933 when native cpu detection is available.
1935 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1936 Martin Sebor <msebor@redhat.com>
1938 PR tree-optimization/84478
1939 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1941 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1942 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1943 support which is conservatively correct, for 2 only stay conservative
1944 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1945 argument to the 2 argument get_range_strlen, adjust 6 arg
1946 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1948 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1949 (gimple_fold_builtin_strlen): Pass true as last argument to
1952 2018-02-20 Martin Sebor <msebor@redhat.com>
1955 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1956 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1957 (builtin_memref::builtin_memref): Factor out parts into
1958 set_base_and_offset and call it.
1960 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1963 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1964 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1965 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1966 search at the associated MODE_INT.
1968 2018-02-20 Jeff Law <law@redhat.com>
1971 PR tree-optimization/81592
1973 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1974 for range data rather than using global data.
1975 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1976 range data rather than using global data.
1977 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1978 pass it to children as needed.
1979 (struct directive::fmtresult): Similarly.
1980 (struct directive::set_width): Similarly.
1981 (struct directive::set_precision): Similarly.
1982 (format_integer, format_directive, parse_directive): Similarly.
1983 (format_none): Accept unnamed vr_values parameter.
1984 (format_percent, format_floating, format_character): Similarly.
1985 (format_string, format_plain): Similarly.
1986 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1987 the EVRP range analyzer for range data rather than using global data.
1988 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1989 gimple-ssa-evrp-analyze.h
1990 (class sprintf_dom_walker): Add after_dom_children member function.
1991 Add evrp_range_analyzer member.
1992 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1993 range analyzer as needed.
1994 (sprintf_dom_walker::after_dom_children): New member function.
1995 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1997 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1998 (evrp_range_analyzer::pop_to_marker): Likewise.
2000 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2002 PR tree-optimization/84419
2003 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2004 with the required type if its current type is compatible but
2007 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2010 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2011 after vectorization.
2013 2018-02-20 Martin Liska <mliska@suse.cz>
2016 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2017 possible values if we don't have a hint.
2019 2018-02-20 Martin Liska <mliska@suse.cz>
2023 * final.c (shorten_branches): Build align_tab array with one
2025 * opts.c (finish_options): Add alignment option limit check.
2026 (MAX_CODE_ALIGN): Likewise.
2027 (MAX_CODE_ALIGN_VALUE): Likewise.
2028 * doc/invoke.texi: Document maximum allowed option value for
2029 all -falign-* options.
2031 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2034 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2035 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2036 * var-tracking.c (emit_note_insn_var_location): Remove all references
2037 to NOTE_INSN_CALL_ARG_LOCATION.
2038 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2039 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2040 Use copy_rtx_if_shared.
2041 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2042 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2043 (dwarf2out_var_location): Remove handling of
2044 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2046 * final.c (final_scan_insn): Remove all references to
2047 NOTE_INSN_CALL_ARG_LOCATION.
2048 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2049 before dumping final insns.
2050 * except.c (emit_note_eh_region_end): Remove all references to
2051 NOTE_INSN_CALL_ARG_LOCATION.
2052 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2053 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2054 * config/arc/arc.c (hwloop_optimize): Likewise.
2055 * config/arm/arm.c (create_fix_barrier): Likewise.
2056 * config/s390/s390.c (s390_chunkify_start): Likewise.
2057 * config/sh/sh.c (find_barrier): Likewise.
2058 * config/i386/i386.c (rest_of_insert_endbranch,
2059 ix86_seh_fixup_eh_fallthru): Likewise.
2060 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2061 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2062 * config/frv/frv.c (frv_function_prologue): Likewise.
2063 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2065 (note_outside_basic_block_p): Remove all references to
2066 NOTE_INSN_CALL_ARG_LOCATION.
2067 * gengtype.c (adjust_field_rtx_def): Likewise.
2068 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2070 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2071 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2074 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2077 PR tree-optimization/84452
2078 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2079 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2082 2018-02-19 Martin Liska <mliska@suse.cz>
2085 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2087 2018-02-19 Martin Liska <mliska@suse.cz>
2088 Richard Sandiford <richard.sandiford@linaro.org>
2090 PR tree-optimization/82491
2091 * gimple-fold.c (get_base_constructor): Make earlier bail out
2094 2018-02-19 Carl Love <cel@us.ibm.com>
2096 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2097 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2099 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2100 P8V_BUILTIN_VEC_NEG.
2102 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2104 * config/rl78/rl78.md (movdf): New define expand.
2106 2018-02-19 Martin Liska <mliska@suse.cz>
2109 * doc/invoke.texi: Fix typo.
2110 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2112 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2114 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2115 handle rs6000_single_float and rs6000_double_float specially for
2118 2018-02-16 Jeff Law <law@redhat.com>
2120 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2123 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2126 * ipa-inline.c (inline_small_functions): Fix a typo.
2128 2018-02-16 Nathan Sidwell <nathan@acm.org>
2130 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2132 2018-02-16 Carl Love <cel@us.ibm.com>
2134 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2135 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2136 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2137 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2138 expansion to P8V_BUILTIN_VEC_FLOAT2.
2140 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2142 PR rtl-optimization/70023
2143 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2144 src_regno into account.
2146 2018-02-16 Carl Love <cel@us.ibm.com>
2148 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2149 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2150 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2151 * config/rs6000/rs6000.c: Remove case statements for
2152 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2153 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2154 and P9V_BUILTIN_VEC_VINSERT4B.
2155 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2156 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2157 * config/rs6000/vsx.md:
2158 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2161 2018-02-16 Carl Love <cel@us.ibm.com>
2163 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2165 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2167 * config/rs6000/rs6000-c.c: Add the definitions for
2168 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2169 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2170 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2171 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2172 definition for insert4b and define insn *insert3b_internal.
2173 * doc/extend.texi: Add documentation for vec_extract4b.
2175 2018-02-16 Nathan Sidwell <nathan@acm.org>
2177 * doc/extend.texi (Backwards Compatibility): Mention friend
2178 injection. Note for-scope is deprecated.
2179 * doc/invoke.texi (-ffriend-injection): Deprecate.
2181 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2183 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2184 that moved to I2, also allow destinations that are a paradoxical
2185 subreg (instead of a normal reg).
2187 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2190 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2193 2018-02-16 Richard Biener <rguenther@suse.de>
2195 PR tree-optimization/84037
2196 PR tree-optimization/84016
2198 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2199 Adjust vec_construct for the fact we need additional higher latency
2200 128bit inserts for AVX256 and AVX512 vector builds.
2201 (ix86_add_stmt_cost): Scale vector construction cost for
2204 2018-02-16 Richard Biener <rguenther@suse.de>
2206 PR tree-optimization/84417
2207 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2208 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2209 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2211 2018-02-16 Martin Liska <mliska@suse.cz>
2214 * internal-fn.def (ASAN_CHECK): Set proper flags.
2215 (ASAN_MARK): Likewise.
2217 2018-02-16 Julia Koval <julia.koval@intel.com>
2219 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2220 from PTA_CANNONLAKE.
2222 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2225 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2226 Use ++iter rather than iter++ for std::list iterators.
2227 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2228 defer deleting them until all nodes in the forest are processed. Do
2229 free even leaf nodes. Change to_process into auto_vec.
2232 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2233 * vec.h (vec_default_construct): Use memset instead of placement new
2234 if BROKEN_VALUE_INITIALIZATION is defined.
2235 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2236 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2239 PR rtl-optimization/83723
2240 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2241 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2242 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2244 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2246 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2248 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2250 PR rtl-optimization/81443
2251 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2252 from inner REGs to paradoxical SUBREGs.
2254 2018-02-16 Richard Biener <rguenther@suse.de>
2256 PR tree-optimization/84399
2257 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2258 For operands we can analyze at their definition make sure we can
2259 analyze them at each use as well.
2261 2018-02-16 Richard Biener <rguenther@suse.de>
2263 PR tree-optimization/84190
2264 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2265 volatile accesses if the decl isn't volatile.
2267 2018-02-15 Jason Merrill <jason@redhat.com>
2269 PR c++/84314 - ICE with templates and fastcall attribute.
2270 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2271 TYPE_CANONICAL on an existing type.
2273 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2275 PR tree-optimization/84383
2276 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2277 dstoff nor call operand_equal_p if dstbase is NULL.
2279 PR tree-optimization/84334
2280 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2281 also a CONSTANT_CLASS_P, punt.
2283 2018-02-14 Jim Wilson <jimw@sifive.com>
2285 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2286 first SMALL_OPERAND check. New local min_second_step. Move assert
2287 to where locals are set. Add TARGET_RVC support.
2288 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2290 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2292 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2294 2018-02-14 Martin Sebor <msebor@redhat.com>
2296 PR tree-optimization/83698
2297 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2298 arrays constrain the offset range to their bounds.
2299 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2300 (builtin_access::overlap): Avoid setting the size of overlap if it's
2302 (maybe_diag_overlap): Also consider arrays when deciding what values
2303 of offsets to include in diagnostics.
2305 2018-02-14 Martin Sebor <msebor@redhat.com>
2308 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2309 that correspond to the kind of a declaration.
2311 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2314 * config/pa/pa.md: Load address of PIC label using the linkage table
2315 if the label is nonlocal.
2317 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2319 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2320 warning message if user requests -maltivec=be.
2321 * doc/invoke.texi: Document deprecation of -maltivec=be.
2323 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2326 * config/rs6000/rs6000-c.c: Update definitions for
2327 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2328 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2330 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2333 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2334 add _get_ssp intrinsics. Remove argument from
2335 __builtin_ia32_rdssp[d|q].
2336 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2337 * config/i386/i386-builtin.def: Remove argument from
2338 __builtin_ia32_rdssp[d|q].
2339 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2340 ix86_expand_special_args_builtin for _rdssp[d|q].
2341 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2342 Clear register before usage.
2343 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2344 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2346 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2348 PR tree-optimization/84357
2349 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2350 operand 1 of an ARRAY_REF too.
2352 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2355 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2356 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2358 (set_of_reg): New struct.
2359 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2360 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2361 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2363 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2364 Split into bitclr, bitset, bitinvert patterns if appropriate.
2365 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2366 use rx_fuse_in_memory_bitop.
2367 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2368 to named insn, correct maximum insn length.
2370 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2373 * machmode.def: Define a complex mode for PARTIAL_INT.
2374 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2376 * doc/rtl.texi: Document CSPImode.
2377 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2379 (msp430_hard_regno_nregs_with_padding): Likewise.
2381 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2384 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2386 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2388 PR rtl-optimization/84169
2389 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2390 we generated a parallel as new i3 and we split that to new i2 and i3
2391 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2392 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2393 those to i2, not i1. Partially rewrite this scan code.
2395 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2398 * stor-layout.c (place_field): For variable length fields, adjust
2399 offset_align afterwards not just based on the field's alignment,
2400 but also on the size.
2403 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2404 of exps and logs in the use_exp2 case.
2406 2018-02-13 Jeff Law <law@redhat.com>
2408 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2411 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2414 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2418 * common.opt (gas-loc-support, gas-locview-support): New.
2419 (ginline-points, ginternal-reset-location-views): New.
2420 * doc/invoke.texi: Document them. Use @itemx where intended.
2421 (gvariable-location-views): Adjust.
2422 * target.def (reset_location_view): New.
2423 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2424 (TARGET_RESET_LOCATION_VIEW): New.
2425 * doc/tm.texi: Rebuilt.
2426 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2427 (dwarf2out_default_as_locview_support): New.
2428 (output_asm_line_debug_info): Use option variables.
2429 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2430 (output_loc_list): Likewise.
2431 (add_high_low_attributes): Check option variables.
2432 Don't output entry view attribute in strict mode.
2433 (gen_inlined_subroutine_die): Check option variables.
2434 (dwarf2out_inline_entry): Likewise.
2435 (init_sections_and_labels): Likewise.
2436 (dwarf2out_early_finish): Likewise.
2437 (maybe_reset_location_view): New, from...
2438 (dwarf2out_var_location): ... here. Call it.
2439 * debug.h (dwarf2out_default_as_loc_support): Declare.
2440 (dwarf2out_default_as_locview_support): Declare.
2441 * hooks.c (hook_int_rtx_insn_0): New.
2442 * hooks.h (hook_int_rtx_insn_0): Declare.
2443 * toplev.c (process_options): Take -gas-loc-support and
2444 -gas-locview-support from dwarf2out. Enable
2445 -gvariable-location-views by default only with locview
2446 assembler support. Enable -ginternal-reset-location-views by
2447 default only if the target defines the corresponding hook.
2448 Enable -ginline-points by default if location views are
2449 enabled; force it disabled if statement frontiers are
2451 * tree-inline.c (expand_call_inline): Check option variables.
2452 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2454 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2456 PR tree-optimization/84321
2457 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2458 handling. Also check whether the anti-range contains any values
2459 that satisfy the mask; switch to a VR_RANGE if not.
2461 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2464 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2466 2018-02-13 Martin Jambor <mjambor@suse.cz>
2469 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2470 of call statements, also set location of a load to a temporary.
2472 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2474 * config/rl78/rl78.c (add_vector_labels): New function.
2475 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2476 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2477 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2478 which checks that no arguments are passed.
2479 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2480 * doc/extend.texi: Documentation for the new attribute.
2482 2018-02-13 Andreas Schwab <schwab@suse.de>
2484 * config/riscv/linux.h (CPP_SPEC): Define.
2486 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2489 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2490 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2491 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2492 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2493 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2494 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2495 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2496 temporarily for AES and PCLMUL builtins.
2498 PR tree-optimization/84339
2499 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2500 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2504 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2505 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2506 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2508 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2510 * omp-simd-clone.h: New file.
2511 * omp-simd-clone.c: Include omp-simd-clone.h.
2512 (expand_simd_clones): No longer static.
2513 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2514 cgraph.h and omp-simd-clone.h.
2515 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2516 (vect_recog_widen_shift_pattern): Formatting fix.
2517 (vect_pattern_recog_1): Don't check optab for calls.
2520 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2521 operands[2] into a REG before using gen_lowpart on it.
2523 2018-02-12 Jeff Law <law@redhat.com>
2526 * config/sh/sh.c (find_barrier): Consider a sibling call
2529 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2530 successfully back substituting a reg.
2532 2018-02-12 Richard Biener <rguenther@suse.de>
2534 PR tree-optimization/84037
2535 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2536 parameter, move visited init to caller.
2537 (vect_slp_analyze_operations): Separate cost from validity
2538 check, initialize visited once for all instances.
2539 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2541 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2543 (vect_model_promotion_demotion_cost): Likewise.
2544 (vectorizable_bswap): Guard cost modeling with !slp_node
2545 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2547 (vectorizable_call): Likewise.
2548 (vectorizable_conversion): Likewise.
2549 (vectorizable_assignment): Likewise.
2550 (vectorizable_shift): Likewise.
2551 (vectorizable_operation): Likewise.
2552 (vectorizable_store): Likewise.
2553 (vectorizable_load): Likewise.
2554 (vectorizable_condition): Likewise.
2555 (vectorizable_comparison): Likewise.
2557 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2560 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2561 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2562 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2564 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2567 * params.def (inline-min-speedup): Increase from 8 to 15.
2568 (max-inline-insns-auto): Decrease from 40 to 30.
2569 * ipa-split.c (consider_split): Add some buffer for function to
2570 be considered inlining candidate.
2571 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2574 2018-02-12 Richard Biener <rguenther@suse.de>
2576 PR tree-optimization/84037
2577 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2578 matched stmts if we cannot swap the non-matched ones.
2580 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2582 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2583 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2584 _mm_maskz_scalef_round_ss): New intrinsics.
2585 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2586 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2587 __builtin_ia32_scalefss_round): Remove.
2588 (__builtin_ia32_scalefsd_mask_round,
2589 __builtin_ia32_scalefss_mask_round): New intrinsics.
2590 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2591 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2592 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2593 "<round_constraint>")): Changed to ...
2594 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2595 "<round_scalar_constraint>")): ... this.
2596 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2597 %0, %1, %2<round_op3>}"): Changed to ...
2598 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2599 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2600 %2<round_scalar_mask_op3>}"): ... this.
2601 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2603 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2605 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2606 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2607 (_mm_maskz_sqrt_round_ss): New intrinsics.
2608 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2609 (__builtin_ia32_sqrtsd_mask_round)
2610 (__builtin_ia32_sqrtss_mask_round): New builtins.
2611 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2612 (__builtin_ia32_sqrtss_round): Remove.
2613 (__builtin_ia32_sqrtsd_mask_round)
2614 (__builtin_ia32_sqrtss_mask_round): New builtins.
2615 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2616 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2617 ((match_operand:VF_128 1 "vector_operand"
2618 "xBm,<round_constraint>")): Changed to ...
2619 ((match_operand:VF_128 1 "vector_operand"
2620 "xBm,<round_scalar_constraint>")): ... this.
2621 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2622 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2623 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2624 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2625 %<iptr>1<round_scalar_mask_op3>}): ... this.
2626 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2627 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2629 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2632 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2633 Cast vec_cmpeq result to correct type.
2634 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2635 Cast vec_cmpgt result to correct type.
2637 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2639 * final.c (final_scan_insn_1): Renamed from...
2640 (final_scan_insn): ... this. New wrapper, to recover
2641 seen from the outermost call in recursive ones.
2642 * config/sparc/sparc.c (output_return): Drop seen from call.
2643 (output_sibcall): Likewise.
2644 * config/visium/visium.c (output_branch): Likewise.
2646 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2648 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2651 2018-02-10 Alan Modra <amodra@gmail.com>
2654 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2655 Specify LR as an input.
2657 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2660 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2661 remove_member_access_dummy_vars): New functions.
2662 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2663 lower_omp_1, execute_lower_omp): Use them.
2665 PR rtl-optimization/84308
2666 * shrink-wrap.c (spread_components): Release todo vector.
2668 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2670 PR rtl-optimization/57193
2671 * ira-color.c (struct allocno_color_data): Add member
2672 conflict_allocno_hard_prefs.
2673 (update_conflict_allocno_hard_prefs): New.
2674 (bucket_allocno_compare_func): Add a preference based on
2675 conflict_allocno_hard_prefs.
2676 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2677 (color_allocnos): Remove a dead code. Initiate
2678 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2680 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2683 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2684 constraint from =wa to wa. Avoid a subreg on the output operand,
2685 instead use a pseudo and subreg it in a move.
2686 (p9_xxbrd_<mode>): Changed to ...
2687 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2688 (p9_xxbrd_v2df): New expander.
2689 (p9_xxbrw_<mode>): Changed to ...
2690 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2691 (p9_xxbrw_v4sf): New expander.
2693 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2695 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2697 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2700 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2701 multiply in 32-bit mode.
2702 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2703 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2706 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2708 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2709 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2710 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2711 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2713 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2716 * dwarf2out.c (is_trivial_indirect_ref): New function.
2717 (dwarf2out_late_global_decl): Do not generate a location
2718 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2719 and that are not defined in the current unit.
2721 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2723 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2724 instead of a libcall for UNORDERED.
2726 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2729 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2730 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2732 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2735 * config/s390/s390.c (s390_set_current_function): Invoke
2736 s390_indirect_branch_settings also if fndecl didn't change.
2738 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2740 * config/rs6000/rs6000.md (blockage): Set length to zero.
2742 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2744 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2746 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2749 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2750 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2754 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2755 PARALLEL incoming that failed vt_get_decl_and_offset check.
2758 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2759 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2761 (get_variable_section): For decls in named .bss* sections pass true as
2762 second argument to bss_initializer_p.
2764 2018-02-09 Marek Polacek <polacek@redhat.com>
2765 Jakub Jelinek <jakub@redhat.com>
2768 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2769 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2770 Sync some changes from cxx_fold_indirect_ref.
2772 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2774 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2776 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2777 (BLOCK_INLINE_ENTRY_LABEL): New.
2778 (dwarf2out_var_location): Disregard inline entry markers.
2779 (inline_entry_data): New struct.
2780 (inline_entry_data_hasher): New hashtable type.
2781 (inline_entry_data_hasher::hash): New.
2782 (inline_entry_data_hasher::equal): New.
2783 (inline_entry_data_table): New variable.
2784 (add_high_low_attributes): Add DW_AT_entry_pc and
2785 DW_AT_GNU_entry_view attributes if a pending entry is found
2786 in inline_entry_data_table. Add old entry_pc attribute only
2787 if debug nonbinding markers are disabled.
2788 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2789 markers are enabled.
2790 (block_within_block_p, dwarf2out_inline_entry): New.
2791 (dwarf2out_finish): Check that no entries remained in
2792 inline_entry_data_table.
2793 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2794 (final_scan_insn, notice_source_line): Likewise.
2795 (rest_of_clean_state): Skip inline entry markers.
2796 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2798 * gimple.c (gimple_build_debug_inline_entry): New.
2799 * gimple.h (enum gimple_debug_subcode): Add
2800 GIMPLE_DEBUG_INLINE_ENTRY.
2801 (gimple_build_debug_inline_entry): Declare.
2802 (gimple_debug_inline_entry_p): New.
2803 (gimple_debug_nonbind_marker_p): Adjust.
2804 * insn-notes.def (INLINE_ENTRY): New.
2805 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2806 inline entry marker notes.
2807 (print_insn): Likewise.
2808 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2809 (INSN_DEBUG_MARKER_KIND): Likewise.
2810 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2811 * tree-inline.c (expand_call_inline): Build and insert
2812 debug_inline_entry stmt.
2813 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2814 inline entry blocks early, if nonbind markers are enabled.
2815 (dump_scope_block): Dump fragment info.
2816 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2817 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2818 (gimple_build_debug_inline_entry): New.
2819 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2820 Enable/disable inline entry points too.
2821 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2822 (DEBUG_INSN): Describe inline entry markers.
2824 * common.opt (gvariable-location-views): New.
2825 (gvariable-location-views=incompat5): New.
2826 * config.in: Rebuilt.
2827 * configure: Rebuilt.
2828 * configure.ac: Test assembler for view support.
2829 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2830 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2831 * dwarf2out.c (var_loc_view): New typedef.
2832 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2833 (dwarf2out_locviews_in_attribute): New.
2834 (dwarf2out_locviews_in_loclist): New.
2835 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2836 (enum dw_line_info_opcode): Add LI_adv_address.
2837 (struct dw_line_info_table): Add view.
2838 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2839 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2840 (zero_view_p): New variable.
2841 (ZERO_VIEW_P): New macro.
2842 (output_asm_line_debug_info): New.
2843 (struct var_loc_node): Add view.
2844 (add_AT_view_list, AT_loc_list): New.
2845 (add_var_loc_to_decl): Add view param. Test it against last.
2846 (new_loc_list): Add view params. Record them.
2847 (AT_loc_list_ptr): Handle loc and view lists.
2848 (view_list_to_loc_list_val_node): New.
2849 (print_dw_val): Handle dw_val_class_view_list.
2850 (size_of_die): Likewise.
2851 (value_format): Likewise.
2852 (loc_list_has_views): New.
2853 (gen_llsym): Set vl_symbol too.
2854 (maybe_gen_llsym, skip_loc_list_entry): New.
2855 (dwarf2out_maybe_output_loclist_view_pair): New.
2856 (output_loc_list): Output view list or entries too.
2857 (output_view_list_offset): New.
2858 (output_die): Handle dw_val_class_view_list.
2859 (output_dwarf_version): New.
2860 (output_compilation_unit_header): Use it.
2861 (output_skeleton_debug_sections): Likewise.
2862 (output_rnglists, output_line_info): Likewise.
2863 (output_pubnames, output_aranges): Update version comments.
2864 (output_one_line_info_table): Output view numbers in asm comments.
2865 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2866 Call maybe_gen_llsym.
2867 (loc_list_from_tree_1): Adjust.
2868 (add_AT_location_description): Create view list attribute if
2869 needed, check it's absent otherwise.
2870 (convert_cfa_to_fb_loc_list): Adjust.
2871 (maybe_emit_file): Call output_asm_line_debug_info for test.
2872 (dwarf2out_var_location): Reset views as needed. Precompute
2873 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2874 attribute. Set view.
2875 (new_line_info_table): Reset next view.
2876 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2877 (dwarf2out_source_line): Likewise. Output view resets and labels to
2878 the assembler, or select appropriate line info opcodes.
2879 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2880 (optimize_string_length): Catch it. Adjust.
2881 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2882 dw_val_class_view_list, and remove it if no longer needed.
2883 (hash_loc_list): Hash view numbers.
2884 (loc_list_hasher::equal): Compare them.
2885 (optimize_location_lists): Check whether a view list symbol is
2886 needed, and whether the locview attribute is present, and
2887 whether they match. Remove the locview attribute if no longer
2889 (index_location_lists): Call skip_loc_list_entry for test.
2890 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2891 Use output_dwarf_version.
2892 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2893 (struct dw_val_node): Add val_view_list.
2894 * final.c (SEEN_NEXT_VIEW): New.
2895 (set_next_view_needed): New.
2896 (clear_next_view_needed): New.
2897 (maybe_output_next_view): New.
2898 (final_start_function): Rename to...
2899 (final_start_function_1): ... this. Take pointer to FIRST,
2900 add SEEN parameter. Emit param bindings in the initial view.
2901 (final_start_function): Reintroduce SEEN-less interface.
2902 (final): Rename to...
2903 (final_1): ... this. Take SEEN parameter. Output final pending
2904 next view at the end.
2905 (final): Reintroduce seen-less interface.
2906 (final_scan_insn): Output pending next view before switching
2907 sections or ending a block. Mark the next view as needed when
2908 outputting variable locations. Notify debug backend of section
2909 changes, and of location view changes.
2910 (rest_of_handle_final): Adjust.
2911 * toplev.c (process_options): Autodetect value for debug variable
2912 location views option. Warn on incompat5 without -gdwarf-5.
2913 * doc/invoke.texi (gvariable-location-views): New.
2914 (gvariable-location-views=incompat5): New.
2915 (gno-variable-location-views): New.
2917 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2919 PR tree-optimization/84136
2920 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2921 that the result of find_edge is non-NULL.
2923 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2926 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2927 storing integer register in SImode. Fix cost of 256 and 512
2928 byte aligned SSE register store.
2930 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2932 * config/i386/i386.c (ix86_multiplication_cost): Fix
2933 multiplication cost for TARGET_AVX512DQ.
2935 2018-02-08 Marek Polacek <polacek@redhat.com>
2937 PR tree-optimization/84238
2938 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2941 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2943 PR tree-optimization/84265
2944 * tree-vect-stmts.c (vectorizable_store): Don't treat
2945 VMAT_CONTIGUOUS accesses as grouped.
2946 (vectorizable_load): Likewise.
2948 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2950 PR tree-optimization/81635
2951 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2952 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2953 (test_round_for_mask): New functions.
2954 (wide_int_cc_tests): Call test_round_for_mask.
2955 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2956 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2957 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2958 range returned by get_range_info.
2960 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2963 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2964 * symtab.c: Include builtins.h
2965 (symtab_node::output_to_lto_symbol_table_p): Move here
2966 from lto-streamer-out.c:output_symbol_p.
2967 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2968 (output_symbol_p): Move all logic to symtab.c
2969 (produce_symtab): Update.
2971 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2973 * config/s390/s390-opts.h (enum indirect_branch): Define.
2974 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2975 (s390_indirect_branch_via_thunk)
2976 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2977 (enum s390_indirect_branch_type): Define.
2978 * config/s390/s390.c (struct s390_frame_layout, struct
2979 machine_function): Remove.
2980 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2981 (indirect_branch_table_label_no, indirect_branch_table_name):
2983 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2984 (enum s390_indirect_branch_option): Define.
2985 (s390_return_addr_from_memory): New function.
2986 (s390_handle_string_attribute): New function.
2987 (s390_attribute_table): Add new attribute handler.
2988 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2989 (s390_indirect_branch_via_thunk): New function.
2990 (s390_indirect_branch_via_inline_thunk): New function.
2991 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2992 sibling call optimization for non z10 compiles.
2993 (s390_emit_call): Force indirect branch target to be a single
2994 register. Add r1 clobber for non-z10 compiles.
2995 (s390_emit_epilogue): Emit return jump via return_use expander.
2996 (s390_reorg): Handle JUMP_INSNs as execute targets.
2997 (s390_option_override_internal): Perform validity checks for the
2998 new command line options.
2999 (s390_indirect_branch_attrvalue): New function.
3000 (s390_indirect_branch_settings): New function.
3001 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3002 (s390_output_indirect_thunk_function): New function.
3003 (s390_code_end): Implement target hook.
3004 (s390_case_values_threshold): Implement target hook.
3005 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3007 * config/s390/s390.h (struct s390_frame_layout)
3008 (struct machine_function): Move here from s390.c.
3009 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3010 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3011 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3012 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3013 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3014 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3015 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3016 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3017 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3018 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3019 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3020 (mnemonic attribute): Add values which aren't recognized
3022 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3023 pattern for branch conversion. Fix mnemonic attribute.
3024 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3025 indirect branch via thunk if requested.
3026 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3027 ("*indirect_jump"): Disable for branch conversion using out of
3029 ("indirect_jump_via_thunk<mode>_z10")
3030 ("indirect_jump_via_thunk<mode>")
3031 ("indirect_jump_via_inlinethunk<mode>_z10")
3032 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3033 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3034 ("casesi_jump_via_inlinethunk<mode>_z10")
3035 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3036 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3037 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3038 ("*indirect2_jump"): Disable for branch conversion.
3039 ("casesi_jump"): Turn into expander and expand patterns for branch
3041 ("return_use"): New expander.
3042 ("*return"): Emit return via thunk and rename it to ...
3043 ("*return<mode>"): ... this one.
3044 * config/s390/s390.opt: Add new options and and enum for the
3047 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3049 * lra-constraints.c (match_reload): Unconditionally use
3050 gen_lowpart_SUBREG, rather than selecting between that
3051 and equivalent gen_rtx_SUBREG code.
3053 2018-02-08 Richard Biener <rguenther@suse.de>
3055 PR tree-optimization/84233
3056 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3057 changed flag instead of boguously re-using phi_inserted.
3059 2018-02-08 Martin Jambor <mjambor@suse.cz>
3061 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3062 static local variables.
3064 2018-02-08 Richard Biener <rguenther@suse.de>
3066 PR tree-optimization/84278
3067 * tree-vect-stmts.c (vectorizable_store): When looking for
3068 smaller vector types to perform grouped strided loads/stores
3069 make sure the mode is supported by the target.
3070 (vectorizable_load): Likewise.
3072 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3074 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3075 Increase LDP/STP opportunities by adding adjacent callee-saves.
3077 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3079 PR rtl-optimization/84068
3080 PR rtl-optimization/83459
3081 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3083 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3085 PR tree-optimization/84224
3086 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3087 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3090 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3093 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3094 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3095 count, remove one USE.
3097 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3099 * doc/install.texi (Configuration): Document the
3100 --with-long-double-format={ibm,ieee} PowerPC configuration
3104 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3105 Convert from define_expand to be define_insn_and_split. Rework
3106 float/double/_Float128 conversions to QI/HI/SImode to work with
3107 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3108 conversions to QI/HImode types did a store and then a load to
3109 truncate the value. For conversions to VSX registers, don't split
3110 the insn, instead emit the code directly. Use the code iterator
3111 any_fix to combine signed and unsigned conversions.
3112 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3113 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3114 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3115 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3116 (fix_<mode>di2_hw): Likewise.
3117 (fixuns_<mode>di2_hw): Likewise.
3118 (fix_<mode>si2_hw): Likewise.
3119 (fixuns_<mode>si2_hw): Likewise.
3120 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3121 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3122 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3123 fix<uns>_trunc<SFDF:mode>si2_p8.
3124 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3126 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3127 (fix<uns>_<mode>_mem): Likewise.
3128 (fctiw<u>z_<mode>_mem): Likewise.
3129 (fix<uns>_<mode>_mem): Likewise.
3130 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3131 the register allocator from doing a direct move to the GPRs to do
3132 a store, and instead use the ISA 3.0 store byte/half-word from
3133 vector register instruction. For IEEE 128-bit floating point,
3134 also optimize stores of 32-bit ints.
3135 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3137 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3139 * genextract.c (push_pathstr_operand): New function to support
3141 (walk_rtx): Call push_pathstr_operand.
3142 (print_path): Support [a-zA-Z].
3144 2018-02-07 Richard Biener <rguenther@suse.de>
3146 PR tree-optimization/84037
3147 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3148 (cse_and_gimplify_to_preheader): Declare.
3149 (vect_get_place_in_interleaving_chain): Likewise.
3150 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3152 (_loop_vec_info::~_loop_vec_info): Delete it.
3153 (cse_and_gimplify_to_preheader): New function.
3154 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3155 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3156 (vectorizable_load): Likewise. For grouped stores always base
3157 the IV on the first element.
3158 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3159 condition before gimplifying.
3161 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3163 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3164 *DIV_EXPR and *MOD_EXPR.
3166 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3169 * config/i386/i386.c (ix86_option_override_internal): Mask out
3170 the CF_SET bit when checking -fcf-protection.
3172 2018-02-07 Tom de Vries <tom@codesourcery.com>
3175 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3178 2018-02-07 Richard Biener <rguenther@suse.de>
3180 PR tree-optimization/84204
3181 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3184 PR tree-optimization/84205
3185 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3186 special-case isl_ast_op_zdiv_r.
3188 PR tree-optimization/84223
3189 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3190 Only add conditions from within the region.
3191 (gather_bbs::after_dom_children): Adjust.
3193 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3196 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3197 * config/avr/avr.md: Only post-reload split REG-REG moves if
3198 either register is GENERAL_REG_P.
3200 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3202 PR tree-optimization/84235
3203 * tree-ssa-scopedtables.c
3204 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3205 if the subtraction is performed in floating point type where NaNs are
3206 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3207 build 1. Formatting fix.
3209 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3212 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3213 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3214 and skip it regardless of bb boundaries. Use CALL_P macro,
3215 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3216 unnecessarily, formatting fix.
3218 2018-02-06 Michael Collison <michael.collison@arm.com>
3220 * config/arm/thumb2.md:
3221 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3222 (*thumb_mov_notscc): Ditto.
3224 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3227 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3228 unsigned_fix, not "s".
3230 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3232 * configure.ac (gcc_fn_eh_frame_ro): New function.
3233 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3234 correct .eh_frame permissions.
3235 * configure: Regenerate.
3237 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3239 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3242 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3244 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3245 Display warning message for -mno-speculate-indirect-jumps.
3247 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3249 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3251 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3253 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3255 PR tree-optimization/84225
3256 * tree-eh.c (find_trapping_overflow): Only call
3257 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3259 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3262 * config/i386/i386.c: Reimplement the check of possible options
3263 -mibt/-mshstk conbination. Change error messages.
3264 * doc/invoke.texi: Fix a typo: remove extra '='.
3266 2018-02-06 Marek Polacek <polacek@redhat.com>
3268 PR tree-optimization/84228
3269 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3271 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3274 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3275 emitted arch directives.
3276 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3277 __ARM_FEATURE_COPROC before changing architectures.
3279 2018-02-06 Richard Biener <rguenther@suse.de>
3281 * config/i386/i386.c (print_reg): Fix typo.
3282 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3284 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3286 * configure: Regenerate.
3288 2018-02-05 Martin Sebor <msebor@redhat.com>
3290 PR tree-optimization/83369
3291 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3294 2018-02-05 Martin Liska <mliska@suse.cz>
3296 * doc/invoke.texi: Cherry-pick upstream r323995.
3298 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3300 * ira.c (ira_init_register_move_cost): Adjust comment.
3302 2018-02-05 Martin Liska <mliska@suse.cz>
3304 PR gcov-profile/84137
3305 * doc/gcov.texi: Fix typo in documentation.
3307 2018-02-05 Martin Liska <mliska@suse.cz>
3309 PR gcov-profile/83879
3310 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3311 using dlopen functionality.
3313 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3315 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3316 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3317 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3318 _mm_maskz_range_round_ss): New intrinsics.
3319 (__builtin_ia32_rangesd128_round)
3320 (__builtin_ia32_rangess128_round): Remove.
3321 (__builtin_ia32_rangesd128_mask_round,
3322 __builtin_ia32_rangess128_mask_round): New builtins.
3323 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3324 __builtin_ia32_rangess128_round): Remove.
3325 (__builtin_ia32_rangesd128_mask_round,
3326 __builtin_ia32_rangess128_mask_round): New builtins.
3327 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3328 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3329 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3330 "<round_saeonly_constraint>")): Changed to ...
3331 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3332 "<round_saeonly_scalar_constraint>")): ... this.
3333 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3334 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3335 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3336 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3337 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3339 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3341 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3343 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3344 Remove all values except native, 8540 and 8548.
3346 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3348 * config/i386/i386.c (ix86_output_function_return): Pass
3349 INVALID_REGNUM, instead of -1, as invalid register number to
3350 indirect_thunk_name and output_indirect_thunk.
3352 2018-02-02 Julia Koval <julia.koval@intel.com>
3354 * config.gcc: Add -march=icelake.
3355 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3356 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3357 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3358 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3359 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3360 (processor_target_table): Add icelake.
3361 (ix86_option_override_internal): Handle new PTAs.
3362 (get_builtin_code_for_version): Handle icelake.
3363 (M_INTEL_COREI7_ICELAKE): New.
3364 (fold_builtin_cpu): Handle icelake.
3365 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3366 * doc/invoke.texi: Add -march=icelake.
3368 2018-02-02 Julia Koval <julia.koval@intel.com>
3370 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3371 to wide_int_bitmask.
3372 * wide-int-bitmask.h: New.
3374 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3377 * config/i386/i386.md: Replace Pmode with word_mode in
3378 builtin_setjmp_setup and builtin_longjmp to support x32.
3380 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3384 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3386 (rs6000_supported_cpu_names): New static variable.
3387 (linux_cpu_translation_table): Likewise.
3388 (elf_platform) <cpu>: Define new static variable and use it.
3389 Translate kernel AT_PLATFORM name to canonical name if needed.
3390 Error if platform name is unknown.
3392 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3395 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3397 2018-02-01 Jeff Law <law@redhat.com>
3400 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3401 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3402 the scratch if RELEASE_VIA_POP is false.
3403 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3404 If we have to save a temporary register, decrement SIZE appropriately.
3405 Pass new arguments to release_scratch_register_on_entry.
3406 (ix86_adjust_stack_and_probe): Likewise.
3407 (ix86_emit_probe_stack_range): Pass new arguments to
3408 release_scratch_register_on_entry.
3410 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3412 PR rtl-optimization/84157
3413 * combine.c (change_zero_ext): Use REG_P predicate in
3414 front of HARD_REGISTER_P predicate.
3416 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3418 * config/avr/avr.c (avr_option_override): Move disabling of
3419 -fdelete-null-pointer-checks to...
3420 * common/config/avr/avr-common.c (avr_option_optimization_table):
3423 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3425 PR tree-optimization/81635
3426 * tree-data-ref.c (split_constant_offset_1): For types that
3427 wrap on overflow, try to use range info to prove that wrapping
3430 2018-02-01 Renlin Li <renlin.li@arm.com>
3433 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3435 (aarch64_register_move_cost): Likewise.
3436 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3438 (REG_CLASS_NAMES): Likewise.
3439 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3440 TAILCALL_ADDR_REGS. Remove IP registers.
3441 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3443 2018-02-01 Richard Biener <rguenther@suse.de>
3445 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3446 for specifying RPO order and allow NULL for that.
3447 * domwalk.c (dom_walker::dom_walker): Likewise.
3448 (dom_walker::walk): Handle NULL RPO order.
3449 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3451 (rewrite_update_dom_walker): Likewise.
3452 (mark_def_dom_walker): Likewise.
3454 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3456 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3457 (aarch64_maybe_expand_sve_subreg_move): Declare.
3458 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3459 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3461 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3462 that are semantically a reverse operation.
3463 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3464 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3465 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3467 (aarch64_can_change_mode_class): For big-endian, forbid changes
3468 between two SVE modes if they have different element sizes.
3470 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3472 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3473 the TImode handling for big-endian targets.
3475 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3477 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3478 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3480 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3481 Remove BSWAP handing for big-endian targets and use the form of
3482 LD1RQ appropariate for the mode.
3484 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3486 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3487 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3490 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3493 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3494 check for operands that need to go through aarch64_sve_reload_be.
3496 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3498 PR tree-optimization/81661
3499 PR tree-optimization/84117
3500 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3501 * tree-eh.c: Include gimplify.h.
3502 (find_trapping_overflow, replace_trapping_overflow,
3503 rewrite_to_non_trapping_overflow): New functions.
3504 * tree-vect-loop.c: Include tree-eh.h.
3505 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3506 * tree-data-ref.c: Include tree-eh.h.
3507 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3509 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3511 PR rtl-optimization/84123
3512 * combine.c (change_zero_ext): Check if hard register satisfies
3513 can_change_dest_mode before calling gen_lowpart_SUBREG.
3515 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3518 * ira.c (ira_init_register_move_cost): Remove assert.
3520 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3522 PR rtl-optimization/84071
3523 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3524 * doc/tm.texi: Regenerate.
3526 2018-01-31 Richard Biener <rguenther@suse.de>
3528 PR tree-optimization/84132
3529 * tree-data-ref.c (analyze_miv_subscript): Properly
3530 check whether evolution_function_is_affine_multivariate_p
3531 before calling gcd_of_steps_may_divide_p.
3533 2018-01-31 Julia Koval <julia.koval@intel.com>
3536 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3537 * config/i386/i386.md (rdpid_rex64) New.
3538 (rdpid): Make 32bit only.
3540 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3543 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3544 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3546 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3549 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3551 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3553 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3555 PR rtl-optimization/84071
3556 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3557 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3559 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3561 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3562 (arc_attribute_table): Add 'aux' attribute.
3563 (arc_in_small_data_p): Consider aux like variables.
3564 (arc_is_aux_reg_p): New function.
3565 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3566 (arc_get_aux_arg): New function.
3567 (prepare_move_operands): Handle aux-register access.
3568 (arc_handle_aux_attribute): New function.
3569 * doc/extend.texi (ARC Variable attributes): Add subsection.
3571 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3573 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3574 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3575 (arc_attribute_table): Add 'uncached' attribute.
3576 (arc_print_operand): Print '.di' flag for uncached memory
3578 (arc_in_small_data_p): Do not consider for small data the uncached
3580 (arc_is_uncached_mem_p): New function.
3581 * config/arc/predicates.md (compact_store_memory_operand): Check
3582 for uncached memory accesses.
3583 (nonvol_nonimm_operand): Likewise.
3584 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3586 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3589 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3590 falign-loops=): Add Optimization flag.
3592 2018-01-30 Jeff Law <law@redhat.com>
3595 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3596 INT_REGISTERS_SAVED. Check it prior to calling
3597 get_scratch_register_on_entry.
3598 (ix86_adjust_stack_and_probe): Similarly.
3599 (ix86_emit_probe_stack_range): Similarly.
3600 (ix86_expand_prologue): Corresponding changes.
3602 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3605 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3606 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3608 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3611 * lra-constraints.c (curr_insn_transform): Process AND in the
3614 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3616 PR rtl-optimization/83986
3617 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3618 dependence against last_pending_memory_flush in addition to
3621 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3623 PR tree-optimization/81611
3624 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3627 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3630 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3633 2018-01-30 Richard Biener <rguenther@suse.de>
3634 Jakub Jelinek <jakub@redhat.com>
3636 PR tree-optimization/84111
3637 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3638 inner loops added during recursion, as they don't have up-to-date
3641 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3644 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3645 (can_inline_edge_by_limits_p): ... here.
3646 (can_early_inline_edge_p, check_callers,
3647 update_caller_keys, update_callee_keys, recursive_inlining,
3648 add_new_edges_to_heap, speculation_useful_p,
3649 inline_small_functions,
3650 inline_small_functions, flatten_function,
3651 inline_to_all_callers_1): Update.
3653 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3655 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3658 2018-01-30 Richard Biener <rguenther@suse.de>
3660 PR tree-optimization/83008
3661 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3662 invariant and constant vector uses in stmts when they need
3665 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3668 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3669 * configure: Regenerate.
3671 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3673 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3675 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3676 Use gen_rtx_REG rather than gen_lowpart.
3678 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3680 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3681 rather than 0 when creating partial subregs.
3683 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3685 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3688 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3691 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3692 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3693 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3694 flags. This restores the settings used before the 2017-07-24.
3695 Turning off pre increment/decrement/modify allows IVOPTS to
3696 optimize DF/SF loops where the index is an int.
3698 2018-01-29 Richard Biener <rguenther@suse.de>
3699 Kelvin Nilsen <kelvin@gcc.gnu.org>
3702 * tree-vect-stmts.c (vectorizable_call): Don't call
3703 targetm.vectorize_builtin_md_vectorized_function if callee is
3706 2018-01-22 Carl Love <cel@us.ibm.com>
3708 * doc/extend.tex: Fix typo in second arg in
3709 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3711 2018-01-29 Richard Biener <rguenther@suse.de>
3713 PR tree-optimization/84086
3714 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3715 (flush_ssaname_freelist): When SSA names were released reset
3716 the SCEV hash table.
3718 2018-01-29 Richard Biener <rguenther@suse.de>
3720 PR tree-optimization/84057
3721 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3722 removed paths when removing edges.
3724 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3726 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3727 -mfunction-return=@var{choice}.
3729 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3732 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3734 (layout::print_source_line): Likewise.
3735 (test_get_line_width_without_trailing_whitespace): Add test cases.
3737 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3740 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3743 2018-01-26 Jim Wilson <jimw@sifive.com>
3745 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3747 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3750 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3752 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3753 and CMP + SUB-immediate -> SUBS.
3755 2018-01-26 Martin Sebor <msebor@redhat.com>
3757 PR tree-optimization/83896
3758 * tree-ssa-strlen.c (get_string_len): Rename...
3759 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3760 Avoid assuming length is constant.
3761 (handle_char_store): Use HOST_WIDE_INT for string length.
3763 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3766 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3767 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3769 2018-01-26 Richard Biener <rguenther@suse.de>
3771 PR rtl-optimization/84003
3772 * dse.c (record_store): Only record redundant stores when
3773 the earlier store aliases at least all accesses the later one does.
3775 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3777 PR rtl-optimization/83985
3778 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3779 REG_CFA_RESTORE insns.
3780 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3781 REG_CFA_RESTORE insns here.
3784 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3785 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3787 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3789 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3790 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3791 (arc_init): Likewise.
3792 (arc_override_options): Likewise.
3793 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3795 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3797 * config/arc/arc.h (TARGET_DBNZ): Define.
3798 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3799 properly set the tune attribute.
3800 (dbnz): Use TARGET_DBNZ guard.
3801 * config/arc/arc.opt (mtune): Add core3 option.
3803 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3805 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3806 recognize new pic like addresses.
3807 (arc_delegitimize_address): Clean up.
3809 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3811 * config/arc/arc-arches.def: Option mrf16 valid for all
3813 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3814 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3815 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3816 * config/arc/arc-tables.opt: Regenerate.
3817 * config/arc/arc.c (arc_conditional_register_usage): Handle
3818 reduced register file case.
3819 (arc_file_start): Set must have build attributes.
3820 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3822 * config/arc/arc.opt (mrf16): Add new option.
3823 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3824 * config/arc/genmultilib.awk: Handle new mrf16 option.
3825 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3826 * config/arc/t-multilib: Regenerate.
3827 * doc/invoke.texi (ARC Options): Document mrf16 option.
3829 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3831 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3832 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3833 (arc_attribute_table): Add 'secure_call' attribute.
3834 (arc_print_operand): Print secure call operand.
3835 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3837 (arc_is_secure_call_p): New function. * config/arc/arc.md
3838 (call_i): Add support for sjli instruction.
3839 (call_value_i): Likewise.
3840 * config/arc/constraints.md (Csc): New constraint.
3842 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3843 John Eric Martin <John.Martin@emmicro-us.com>
3845 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3846 * config/arc/arc.c (_arc_jli_section): New struct.
3847 (arc_jli_section): New type.
3848 (rc_jli_sections): New static variable.
3849 (arc_handle_jli_attribute): New function.
3850 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3851 (arc_file_end): New function.
3852 (TARGET_ASM_FILE_END): Define.
3853 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3854 (arc_add_jli_section): New function.
3855 (jli_call_scan): Likewise.
3856 (arc_reorg): Call jli_call_scan.
3857 (arc_output_addsi): Remove 'S' from printing asm operand.
3858 (arc_is_jli_call_p): New function.
3859 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3861 (movhi_insn): Likewise.
3862 (movsi_insn): Likewise.
3863 (movsi_set_cc_insn): Likewise.
3864 (loadqi_update): Likewise.
3865 (load_zeroextendqisi_update): Likewise.
3866 (load_signextendqisi_update): Likewise.
3867 (loadhi_update): Likewise.
3868 (load_zeroextendhisi_update): Likewise.
3869 (load_signextendhisi_update): Likewise.
3870 (loadsi_update): Likewise.
3871 (loadsf_update): Likewise.
3872 (movsicc_insn): Likewise.
3873 (bset_insn): Likewise.
3874 (bxor_insn): Likewise.
3875 (bclr_insn): Likewise.
3876 (bmsk_insn): Likewise.
3877 (bicsi3_insn): Likewise.
3878 (cmpsi_cc_c_insn): Likewise.
3879 (movsi_ne): Likewise.
3880 (movsi_cond_exec): Likewise.
3881 (clrsbsi2): Likewise.
3892 (call_i): Remove 'S' asm letter, add jli instruction.
3893 (call_value_i): Likewise.
3894 * config/arc/arc.op (mjli-always): New option.
3895 * config/arc/constraints.md (Cji): New constraint.
3896 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3898 (subsf3_fpx): Likewise.
3899 (mulsf3_fpx): Likewise.
3900 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3902 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3904 * doc/invoke.texi (ARC): Document mjli-always option.
3906 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3908 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3909 avoid addition with 0 and use incw and decw where possible.
3911 2018-01-26 Richard Biener <rguenther@suse.de>
3913 PR tree-optimization/81082
3914 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3915 association if it requires casting to unsigned.
3916 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3917 from fold_plusminus_mult_expr to catch important cases late when
3918 range info is available.
3920 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3922 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3923 * configure.ac (hidden_linkonce): New test.
3924 * configure: Regenerate.
3925 * config.in: Regenerate.
3927 2018-01-26 Julia Koval <julia.koval@intel.com>
3929 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3930 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3931 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3932 _mm_mask_bitshuffle_epi64_mask): Fix type.
3933 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3934 USI_FTYPE_V4DI_V4DI_USI): Remove.
3935 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3936 __builtin_ia32_vpshufbitqmb256_mask,
3937 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3938 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3939 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3941 2018-01-26 Alan Modra <amodra@gmail.com>
3944 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3945 UNSPEC_VBPERMQ. Sort other unspecs.
3947 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3949 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3951 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3954 * predict.c (drop_profile): Do not push/pop cfun; update also
3956 (handle_missing_profiles): Fix logic looking for zero profiles.
3958 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3961 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3962 on functions with #pragma omp declare simd or functions with simd
3964 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3965 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3966 Remove trailing \n from warning_at calls.
3968 2018-01-25 Tom de Vries <tom@codesourcery.com>
3971 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3972 for neutered workers.
3974 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3977 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3978 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3980 2018-01-24 Jeff Law <law@redhat.com>
3983 * i386.c (get_probe_interval): Move to earlier point.
3984 (ix86_compute_frame_layout): If -fstack-clash-protection and
3985 the frame is larger than the probe interval, then use pushes
3986 to save registers rather than reg->mem moves.
3987 (ix86_expand_prologue): Remove conditional for int_registers_saved
3990 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3993 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3994 min/max for never referenced object.
3996 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3999 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4001 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4002 attributes from DECL_ATTRIBUTES (decl) without affecting
4003 DECL_ATTRIBUTES (current_function_decl).
4004 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4005 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4007 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4009 PR tree-optimization/83979
4010 * fold-const.c (fold_comparison): Use constant_boolean_node
4011 instead of boolean_{true,false}_node.
4013 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4015 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4018 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4020 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4021 Simplify the clause that sets the length attribute.
4022 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4023 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4024 clause that sets the length attribute.
4025 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4027 2018-01-24 Tom de Vries <tom@codesourcery.com>
4030 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4031 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4032 Add strict parameter.
4033 (prevent_branch_around_nothing): Insert dummy insn between branch to
4034 label and label with no ptx insn inbetween.
4035 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4037 2018-01-24 Tom de Vries <tom@codesourcery.com>
4040 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4041 for neutered threads in warp.
4042 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4044 2018-01-24 Richard Biener <rguenther@suse.de>
4046 PR tree-optimization/83176
4047 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4050 2018-01-24 Richard Biener <rguenther@suse.de>
4052 PR tree-optimization/82819
4053 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4054 code generating pluses that are no-ops in the target precision.
4056 2018-01-24 Richard Biener <rguenther@suse.de>
4059 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4061 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4063 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4064 to merge probabilities.
4065 * predict.c (probably_never_executed): Also mark as cold functions
4066 with global 0 profile and guessed local profile.
4067 * profile-count.c (profile_probability::combine_with_count): New
4069 * profile-count.h (profile_probability::operator*,
4070 profile_probability::operator*=, profile_probability::operator/,
4071 profile_probability::operator/=): Reduce precision to adjusted
4072 and set value to guessed on contradictory divisions.
4073 (profile_probability::combine_with_freq): Remove.
4074 (profile_probability::combine_wiht_count): Declare.
4075 (profile_count::force_nonzero):: Set to adjusted.
4076 (profile_count::probability_in):: Set quality to adjusted.
4077 * tree-ssa-tail-merge.c (replace_block_by): Use
4080 2018-01-23 Andrew Waterman <andrew@sifive.com>
4081 Jim Wilson <jimw@sifive.com>
4083 * config/riscv/riscv.c (riscv_stack_boundary): New.
4084 (riscv_option_override): Set riscv_stack_boundary. Handle
4085 riscv_preferred_stack_boundary_arg.
4086 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4087 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4088 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4089 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4090 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4091 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4093 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4096 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4097 of struct ix86_frame.
4098 (ix86_expand_epilogue): Likewise. Add a local variable for
4099 the reg_save_offset field in struct ix86_frame.
4101 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4103 PR tree-optimization/82604
4104 * tree-loop-distribution.c (enum partition_kind): New enum item
4105 PKIND_PARTIAL_MEMSET.
4106 (partition_builtin_p): Support above new enum item.
4107 (generate_code_for_partition): Ditto.
4108 (compute_access_range): Differentiate cases that equality can be
4109 proven at all loops, the innermost loops or no loops.
4110 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4111 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4112 (finalize_partitions, distribute_loop): Don't fuse partition of
4113 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4114 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4117 2018-01-23 Martin Liska <mliska@suse.cz>
4119 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4120 order to ignore the predictor.
4121 (PRED_POLYMORPHIC_CALL): Likewise.
4122 (PRED_RECURSIVE_CALL): Likewise.
4124 2018-01-23 Martin Liska <mliska@suse.cz>
4126 * tree-profile.c (tree_profiling): Print function header to
4127 aware reader which function we are working on.
4128 * value-prof.c (gimple_find_values_to_profile): Do not print
4129 not interesting value histograms.
4131 2018-01-23 Martin Liska <mliska@suse.cz>
4133 * profile-count.h (enum profile_quality): Add
4134 profile_uninitialized as the first value. Do not number values
4135 as they are zero based.
4136 (profile_count::verify): Update sanity check.
4137 (profile_probability::verify): Likewise.
4139 2018-01-23 Nathan Sidwell <nathan@acm.org>
4141 * doc/invoke.texi (ffor-scope): Deprecate.
4143 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4145 PR tree-optimization/83510
4146 * domwalk.c (set_all_edges_as_executable): New function.
4147 (dom_walker::dom_walker): Convert bool param
4148 "skip_unreachable_blocks" to enum reachability. Move setup of
4149 edge flags to set_all_edges_as_executable and only do it when
4150 reachability is REACHABLE_BLOCKS.
4151 * domwalk.h (enum dom_walker::reachability): New enum.
4152 (dom_walker::dom_walker): Convert bool param
4153 "skip_unreachable_blocks" to enum reachability.
4154 (set_all_edges_as_executable): New decl.
4155 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4156 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4158 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4159 but converting true to REACHABLE_BLOCKS.
4160 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4162 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4163 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4164 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4166 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4167 if check_all_array_refs will be called.
4169 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4171 * tree.c (selftest::test_location_wrappers): Add more test
4174 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4176 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4177 (selftest::test_bit_in_range): Likewise.
4179 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4182 * doc/sourcebuild.texi (vect_float): Say that the selector
4183 only describes the situation when -funsafe-math-optimizations is on.
4184 (vect_float_strict): Document.
4186 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4188 PR tree-optimization/83965
4189 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4190 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4191 instead of checking only for a reduction.
4192 (vect_recog_widen_sum_pattern): Likewise.
4194 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4196 * predict.c (probably_never_executed): Only use precise profile info.
4197 (compute_function_frequency): Skip after inlining hack since we now
4198 have quality checking.
4200 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4202 * profile-count.h (profile_probability::very_unlikely,
4203 profile_probability::unlikely, profile_probability::even): Set
4204 precision to guessed.
4206 2018-01-23 Richard Biener <rguenther@suse.de>
4208 PR tree-optimization/83963
4209 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4210 Properly terminate dominator walk when crossing the exit edge not
4211 when visiting its source block.
4213 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4216 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4217 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4219 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4221 PR tree-optimization/83957
4222 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4223 semicolon after for body surrounded by braces.
4225 PR tree-optimization/83081
4226 * profile-count.h (profile_probability::split): New method.
4227 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4228 Use profile_probability::split.
4229 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4230 when splitting a single conditional jump into 2.
4232 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4234 PR tree-optimization/69452
4235 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4238 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4240 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4241 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4242 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4244 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4246 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4247 * config/rl78/rl78.md (movdi): New define_expand.
4248 * config/rl78/rl78.c (rl78_split_movdi): New function.
4250 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4253 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4255 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4256 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4257 128-bit to produce an UNSPEC move to get the double word with the
4258 signbit and then a shift directly to do signbit.
4259 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4260 implementation with a new version that just does either a direct
4261 move or a regular move. Move memory interface to separate insns.
4262 Move insns so they are next to the expander.
4263 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4264 with signbit move. Split big and little endian case.
4265 (signbit<mode>2_dm_mem_le): Likewise.
4266 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4267 (signbit<mode>2_dm2): Likewise.
4269 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4271 * config/rl78/rl78.md (anddi3): New define_expand.
4273 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4275 * config/rl78/rl78.md (umindi3): New define_expand.
4277 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4279 * config/rl78/rl78.md (smindi3): New define_expand.
4281 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4283 * config/rl78/rl78.md (smaxdi3): New define_expand.
4285 2018-01-22 Carl Love <cel@us.ibm.com>
4287 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4288 LVX_V1TI): Add macro expansion.
4289 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4290 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4291 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4292 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4293 Change check to determine if the instruction is a byte reversing
4294 entry. Fix typo in comment.
4295 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4296 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4297 Add def_builtin calls for new builtins.
4298 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4299 Add define_insn expansion.
4301 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4303 * config/rl78/rl78.md (umaxdi3): New define_expand.
4305 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4307 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4308 for non-QImode registers.
4310 2018-01-22 Richard Biener <rguenther@suse.de>
4312 PR tree-optimization/83963
4313 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4314 including the loop exit block.
4315 (scop_detection::merge_sese): Likewise.
4316 (scop_detection::add_scop): Do it here instead.
4318 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4320 * doc/sourcebuild.texi (arm_softfloat): Document.
4322 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4325 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4326 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4327 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4329 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4330 David Edelsohn <dje.gcc@gmail.com>
4333 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4334 Change "crset eq" to "crset 2".
4335 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4336 (*call_indirect_aix<mode>_nospec): Likewise.
4337 (*call_value_indirect_aix<mode>_nospec): Likewise.
4338 (*call_indirect_elfv2<mode>_nospec): Likewise.
4339 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4340 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4341 change assembly output from . to $.
4342 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4343 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4344 (*tablejump<mode>_internal1_nospec): Likewise.
4346 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4349 * config/sh/sh_optimize_sett_clrt.cc:
4350 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4352 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4354 PR tree-optimization/83940
4355 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4356 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4357 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4358 use it to pass back the definition type.
4359 (vect_check_store_rhs): Likewise rhs_dt_out.
4360 (vect_build_gather_load_calls): Add a mask_dt argument and use
4361 it instead of a call to vect_is_simple_use.
4362 (vectorizable_store): Update calls to vect_check_load_store_mask
4363 and vect_check_store_rhs. Use the dt returned by the latter instead
4364 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4365 instead of calls to vect_is_simple_use. Pass the scalar rather
4366 than the vector operand to vect_is_simple_use when handling
4367 second and subsequent copies of an rhs value.
4368 (vectorizable_load): Update calls to vect_check_load_store_mask
4369 and vect_build_gather_load_calls. Use the cached mask_dt and
4370 gs_info.offset_dt instead of calls to vect_is_simple_use.
4372 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4375 * tree-emutls.c: Include gimplify.h.
4376 (lower_emutls_2): New function.
4377 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4378 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4379 it before further processing.
4382 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4383 UINTVAL (trueop1) instead of INTVAL (op1).
4385 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4389 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4390 INCOMING_FRAME_SP_OFFSET if not defined.
4391 (scan_trace): Add ENTRY argument. If true and
4392 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4393 emit a note to adjust the CFA offset.
4394 (create_cfi_notes): Adjust scan_trace callers.
4395 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4396 INCOMING_FRAME_SP_OFFSET in the CIE.
4397 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4398 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4400 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4401 * doc/tm.texi: Regenerated.
4403 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4405 PR rtl-optimization/83147
4406 * lra-constraints.c (remove_inheritance_pseudos): Use
4407 lra_substitute_pseudo_within_insn.
4409 2018-01-19 Tom de Vries <tom@codesourcery.com>
4410 Cesar Philippidis <cesar@codesourcery.com>
4413 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4415 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4418 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4419 spaces for function labels.
4421 2018-01-19 Martin Liska <mliska@suse.cz>
4423 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4424 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4425 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4426 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4427 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4428 (PRED_CONST_RETURN): Change from 69 to 65.
4429 (PRED_NULL_RETURN): Change from 91 to 71.
4430 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4431 (PRED_LOOP_GUARD): Change from 66 to 73.
4433 2018-01-19 Martin Liska <mliska@suse.cz>
4435 * predict.c (predict_insn_def): Add new assert.
4436 (struct branch_predictor): Change type to signed integer.
4437 (test_prediction_value_range): Amend test to cover
4439 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4440 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4441 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4442 (PRED_LOOP_IV_COMPARE): Likewise.
4443 * predict.h (PROB_UNINITIALIZED): Define new constant.
4445 2018-01-19 Martin Liska <mliska@suse.cz>
4447 * predict.c (dump_prediction): Add new format for
4448 analyze_brprob.py script which is enabled with -details
4450 * profile-count.h (precise_p): New function.
4452 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4454 PR tree-optimization/83922
4455 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4456 there are no statements that need masking.
4457 (vect_active_double_reduction_p): New function.
4458 (vect_analyze_loop_operations): Use it when handling phis that
4459 are not in the loop header.
4461 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4463 PR tree-optimization/83914
4464 * tree-vect-loop.c (vectorizable_induction): Don't convert
4465 init_expr or apply the peeling adjustment for inductions
4466 that are nested within the vectorized loop.
4468 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4470 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4473 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4477 * function.h (gimplify_parameters): Add gimple_seq * argument.
4478 * function.c: Include gimple.h and options.h.
4479 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4480 for the added local temporaries if needed.
4481 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4482 if there are any parameter cleanups, wrap whole body into a
4483 try/finally with the cleanups.
4485 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4488 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4489 Use GET_MODE_CLASS for scalar floating point.
4491 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4495 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4496 Fix call of call_cgraph_insertion_hooks.
4498 2018-01-18 Martin Sebor <msebor@redhat.com>
4500 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4502 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4505 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4508 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4511 * common.opt: (-ffile-prefix-map): New option.
4512 * opts.c (common_handle_option): Defer it.
4513 * opts-global.c (handle_common_deferred_options): Handle it.
4514 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4515 * file-prefix-map.h: New file.
4516 (remap_debug_filename, add_debug_prefix_map): ...here.
4517 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4518 * final.c (debug_prefix_map, add_debug_prefix_map
4519 remap_debug_filename): Move to...
4520 * file-prefix-map.c: New file.
4521 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4522 generalize, get rid of alloca(), use strrchr() instead of strchr().
4523 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4524 Implement in terms of add_prefix_map().
4525 (remap_macro_filename, remap_debug_filename): Implement in term of
4527 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4528 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4529 * dbxout.c: Include file-prefix-map.h.
4530 * varasm.c: Likewise.
4531 * vmsdbgout.c: Likewise.
4532 * xcoffout.c: Likewise.
4533 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4534 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4535 * doc/invoke.texi (-ffile-prefix-map): Document.
4536 (-fdebug-prefix-map): Update description.
4538 2018-01-18 Martin Liska <mliska@suse.cz>
4540 * config/i386/i386.c (indirect_thunk_name): Document that also
4542 (output_indirect_thunk): Document why both instructions
4543 (pause and lfence) are generated.
4545 2018-01-18 Richard Biener <rguenther@suse.de>
4547 PR tree-optimization/83887
4548 * graphite-scop-detection.c
4549 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4550 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4551 (scop_detection::merge_sese): Re-implement with a flood-fill
4552 algorithm that properly finds a SESE region if it exists.
4554 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4557 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4558 pointer_diff optimizations use view_convert instead of convert.
4560 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4562 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4563 Generate different code for -mno-speculate-indirect-jumps.
4564 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4565 (*call_indirect_aix<mode>): Disable for
4566 -mno-speculate-indirect-jumps.
4567 (*call_indirect_aix<mode>_nospec): New define_insn.
4568 (*call_value_indirect_aix<mode>): Disable for
4569 -mno-speculate-indirect-jumps.
4570 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4571 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4572 -mno-speculate-indirect-jumps.
4573 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4575 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4577 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4578 long double type, set the flags for noting the default long double
4579 type, even if we don't pass or return a long double type.
4581 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4584 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4587 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4589 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4590 support for merge[hl].
4591 (fold_mergehl_helper): New helper function.
4592 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4593 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4594 (altivec_vmrglw_direct): Add xxmrglw insn.
4596 2018-01-17 Andrew Waterman <andrew@sifive.com>
4598 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4599 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4601 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4604 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4605 call the lto_location_cache before reading the
4606 DECL_SOURCE_LOCATION of the types.
4608 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4609 Richard Sandiford <richard.sandiford@linaro.org>
4611 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4612 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4613 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4614 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4615 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4617 * config/aarch64/constraints.md (aarch64_movti_operand):
4619 * config/aarch64/predicates.md (Uti): Add new constraint.
4621 2018-01-17 Carl Love <cel@us.ibm.com>
4623 * config/rs6000/vsx.md (define_expand xl_len_r,
4624 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4625 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4627 (define_expand, define_insn): Move the shift left from the
4628 define_insn to the define_expand for lxvl and stxvl instructions.
4629 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4630 and XL_LEN_R definitions to PURE.
4632 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4634 * config/i386/i386.c (indirect_thunk_name): Declare regno
4635 as unsigned int. Compare regno with INVALID_REGNUM.
4636 (output_indirect_thunk): Ditto.
4637 (output_indirect_thunk_function): Ditto.
4638 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4639 in the call to output_indirect_thunk_function.
4641 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4644 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4645 rather than the size of inner_type to determine the stack slot size
4646 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4648 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4651 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4654 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4656 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4657 endian Linux systems to optionally enable multilibs for selecting
4658 the long double type if the user configured an explicit type.
4659 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4660 have no long double multilibs if not defined.
4661 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4662 warn if the user used -mabi={ieee,ibm}longdouble and we built
4663 multilibs for long double.
4664 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4665 appropriate multilib option.
4666 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4668 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4669 for building long double multilibs.
4670 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4672 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4674 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4677 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4679 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4682 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4685 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4688 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4690 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4691 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4693 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4695 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4696 different rtl trees depending on TARGET_64BIT.
4697 (rs6000_gen_lvx): Likewise.
4699 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4701 * config/visium/visium.md (nop): Tweak comment.
4702 (hazard_nop): Likewise.
4704 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4706 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4707 -mspeculate-indirect-jumps.
4708 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4709 for -mno-speculate-indirect-jumps.
4710 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4711 (*call_value_indirect_elfv2<mode>): Disable for
4712 -mno-speculate-indirect-jumps.
4713 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4714 (indirect_jump): Emit different RTL for
4715 -mno-speculate-indirect-jumps.
4716 (*indirect_jump<mode>): Disable for
4717 -mno-speculate-indirect-jumps.
4718 (*indirect_jump<mode>_nospec): New define_insn.
4719 (tablejump): Emit different RTL for
4720 -mno-speculate-indirect-jumps.
4721 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4722 (tablejumpsi_nospec): New define_expand.
4723 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4724 (tablejumpdi_nospec): New define_expand.
4725 (*tablejump<mode>_internal1): Disable for
4726 -mno-speculate-indirect-jumps.
4727 (*tablejump<mode>_internal1_nospec): New define_insn.
4728 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4731 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4733 * caller-save.c (insert_save): Drop unnecessary parameter. All
4736 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4737 Richard Biener <rguenth@suse.de>
4740 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4741 return early, inline manually is_gimple_sizepos. Make sure if we
4742 call gimplify_expr we don't end up with a gimple constant.
4743 * tree.c (variably_modified_type_p): Don't return true for
4744 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4745 * gimplify.h (is_gimple_sizepos): Remove.
4747 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4749 PR tree-optimization/83857
4750 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4751 vectorizable_live_operation for pure SLP statements.
4752 (vectorizable_live_operation): Handle PHIs.
4754 2018-01-16 Richard Biener <rguenther@suse.de>
4756 PR tree-optimization/83867
4757 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4758 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4760 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4763 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4764 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4765 If off is not INTEGER_CST, issue a may not be aligned warning
4766 rather than isn't aligned. Use isn%'t rather than isn't.
4767 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4769 <case MULT_EXPR>: Improve the case when bottom and one of the
4770 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4771 operand, in that case check if the other operand is multiple of
4772 bottom divided by the INTEGER_CST operand.
4774 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4777 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4778 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4779 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4780 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4781 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4782 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4783 (pa_function_arg_size): New function.
4785 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4787 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4788 in a separate statement.
4790 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4792 PR tree-optimization/83847
4793 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4794 group gathers and scatters.
4796 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4798 PR rtl-optimization/86620
4799 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4801 PR rtl-optimization/83213
4802 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4803 to last if both are JUMP_INSNs.
4805 PR tree-optimization/83843
4806 * gimple-ssa-store-merging.c
4807 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4808 store_immediate_info for bswap/nop orig_stores.
4810 2018-01-15 Andrew Waterman <andrew@sifive.com>
4812 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4814 <UDIV>: Increase cost if !TARGET_DIV.
4816 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4818 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4819 (define_attr "cr_logical_3op"): New.
4820 (cceq_ior_compare): Adjust.
4821 (cceq_ior_compare_complement): Adjust.
4822 (*cceq_rev_compare): Adjust.
4823 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4824 (is_cracked_insn): Adjust.
4825 (insn_must_be_first_in_group): Adjust.
4826 * config/rs6000/40x.md: Adjust.
4827 * config/rs6000/440.md: Adjust.
4828 * config/rs6000/476.md: Adjust.
4829 * config/rs6000/601.md: Adjust.
4830 * config/rs6000/603.md: Adjust.
4831 * config/rs6000/6xx.md: Adjust.
4832 * config/rs6000/7450.md: Adjust.
4833 * config/rs6000/7xx.md: Adjust.
4834 * config/rs6000/8540.md: Adjust.
4835 * config/rs6000/cell.md: Adjust.
4836 * config/rs6000/e300c2c3.md: Adjust.
4837 * config/rs6000/e500mc.md: Adjust.
4838 * config/rs6000/e500mc64.md: Adjust.
4839 * config/rs6000/e5500.md: Adjust.
4840 * config/rs6000/e6500.md: Adjust.
4841 * config/rs6000/mpc.md: Adjust.
4842 * config/rs6000/power4.md: Adjust.
4843 * config/rs6000/power5.md: Adjust.
4844 * config/rs6000/power6.md: Adjust.
4845 * config/rs6000/power7.md: Adjust.
4846 * config/rs6000/power8.md: Adjust.
4847 * config/rs6000/power9.md: Adjust.
4848 * config/rs6000/rs64.md: Adjust.
4849 * config/rs6000/titan.md: Adjust.
4851 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4853 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4854 ix86_indirect_branch_register logic.
4856 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4858 * config/i386/constraints.md (Bs): Update
4859 ix86_indirect_branch_register check. Don't check
4860 ix86_indirect_branch_register with GOT_memory_operand.
4862 * config/i386/predicates.md (GOT_memory_operand): Don't check
4863 ix86_indirect_branch_register here.
4864 (GOT32_symbol_operand): Likewise.
4866 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4868 * config/i386/predicates.md (constant_call_address_operand):
4869 Rewrite ix86_indirect_branch_register logic.
4870 (sibcall_insn_operand): Likewise.
4872 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4874 * config/i386/constraints.md (Bs): Replace
4875 ix86_indirect_branch_thunk_register with
4876 ix86_indirect_branch_register.
4878 * config/i386/i386.md (indirect_jump): Likewise.
4879 (tablejump): Likewise.
4880 (*sibcall_memory): Likewise.
4881 (*sibcall_value_memory): Likewise.
4882 Peepholes of indirect call and jump via memory: Likewise.
4883 * config/i386/i386.opt: Likewise.
4884 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4885 (GOT_memory_operand): Likewise.
4886 (call_insn_operand): Likewise.
4887 (sibcall_insn_operand): Likewise.
4888 (GOT32_symbol_operand): Likewise.
4890 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4893 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4894 type rather than type addr's type points to.
4895 (expand_omp_atomic_mutex): Likewise.
4896 (expand_omp_atomic): Likewise.
4898 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4901 * config/i386/i386.c (output_indirect_thunk_function): Use
4902 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4903 for __x86_return_thunk.
4905 2018-01-15 Richard Biener <rguenther@suse.de>
4908 * expmed.c (extract_bit_field_1): Fix typo.
4910 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4913 * config/arm/iterators.md (VF): New mode iterator.
4914 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4915 Remove integer-related logic from pattern.
4916 (neon_vabd<mode>_3): Likewise.
4918 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4921 * common.opt (fstrict-overflow): No longer an alias.
4922 (fwrapv-pointer): New option.
4923 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4924 also for pointer types based on flag_wrapv_pointer.
4925 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4926 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4927 opts->x_flag_wrapv got set.
4928 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4929 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4930 POINTER_TYPE_OVERFLOW_UNDEFINED.
4931 * match.pd: Likewise in address comparison pattern.
4932 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4934 2018-01-15 Richard Biener <rguenther@suse.de>
4937 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4938 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4939 Reset type names to their identifier if their TYPE_DECL doesn't
4940 have linkage (and thus is used for ODR and devirt).
4941 (save_debug_info_for_decl): Remove.
4942 (save_debug_info_for_type): Likewise.
4943 (add_tree_to_fld_list): Adjust.
4944 * tree-pretty-print.c (dump_generic_node): Make dumping of
4945 type names more robust.
4947 2018-01-15 Richard Biener <rguenther@suse.de>
4949 * BASE-VER: Bump to 8.0.1.
4951 2018-01-14 Martin Sebor <msebor@redhat.com>
4954 * builtins.c (check_access): Avoid warning when the no-warning bit
4957 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4959 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4960 * ira-color (allocno_hard_regs_compare): Likewise.
4962 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4965 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4966 Use .pushsection/.popsection.
4968 2018-01-14 Martin Sebor <msebor@redhat.com>
4971 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4973 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4975 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4976 entry from extra_headers.
4977 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4978 extra_headers, make the list bitwise identical to the i?86-*-* one.
4980 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4982 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4983 -mcmodel=large with -mindirect-branch=thunk,
4984 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4985 -mfunction-return=thunk-extern.
4986 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4987 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4988 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4990 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4992 * config/i386/i386.c (print_reg): Print the name of the full
4993 integer register without '%'.
4994 (ix86_print_operand): Handle 'V'.
4995 * doc/extend.texi: Document 'V' modifier.
4997 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4999 * config/i386/constraints.md (Bs): Disallow memory operand for
5000 -mindirect-branch-register.
5002 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5003 (GOT_memory_operand): Likewise.
5004 (call_insn_operand): Likewise.
5005 (sibcall_insn_operand): Likewise.
5006 (GOT32_symbol_operand): Likewise.
5007 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5008 for -mindirect-branch-register.
5009 (tablejump): Likewise.
5010 (*sibcall_memory): Likewise.
5011 (*sibcall_value_memory): Likewise.
5012 Disallow peepholes of indirect call and jump via memory for
5013 -mindirect-branch-register.
5014 (*call_pop): Replace m with Bw.
5015 (*call_value_pop): Likewise.
5016 (*sibcall_pop_memory): Replace m with Bs.
5017 * config/i386/i386.opt (mindirect-branch-register): New option.
5018 * doc/invoke.texi: Document -mindirect-branch-register option.
5020 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5022 * config/i386/i386-protos.h (ix86_output_function_return): New.
5023 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5024 set function_return_type.
5025 (indirect_thunk_name): Add ret_p to indicate thunk for function
5027 (output_indirect_thunk_function): Pass false to
5028 indirect_thunk_name.
5029 (ix86_output_indirect_branch_via_reg): Likewise.
5030 (ix86_output_indirect_branch_via_push): Likewise.
5031 (output_indirect_thunk_function): Create alias for function
5032 return thunk if regno < 0.
5033 (ix86_output_function_return): New function.
5034 (ix86_handle_fndecl_attribute): Handle function_return.
5035 (ix86_attribute_table): Add function_return.
5036 * config/i386/i386.h (machine_function): Add
5037 function_return_type.
5038 * config/i386/i386.md (simple_return_internal): Use
5039 ix86_output_function_return.
5040 (simple_return_internal_long): Likewise.
5041 * config/i386/i386.opt (mfunction-return=): New option.
5042 (indirect_branch): Mention -mfunction-return=.
5043 * doc/extend.texi: Document function_return function attribute.
5044 * doc/invoke.texi: Document -mfunction-return= option.
5046 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5048 * config/i386/i386-opts.h (indirect_branch): New.
5049 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5050 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5051 with local indirect jump when converting indirect call and jump.
5052 (ix86_set_indirect_branch_type): New.
5053 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5054 (indirectlabelno): New.
5055 (indirect_thunk_needed): Likewise.
5056 (indirect_thunk_bnd_needed): Likewise.
5057 (indirect_thunks_used): Likewise.
5058 (indirect_thunks_bnd_used): Likewise.
5059 (INDIRECT_LABEL): Likewise.
5060 (indirect_thunk_name): Likewise.
5061 (output_indirect_thunk): Likewise.
5062 (output_indirect_thunk_function): Likewise.
5063 (ix86_output_indirect_branch_via_reg): Likewise.
5064 (ix86_output_indirect_branch_via_push): Likewise.
5065 (ix86_output_indirect_branch): Likewise.
5066 (ix86_output_indirect_jmp): Likewise.
5067 (ix86_code_end): Call output_indirect_thunk_function if needed.
5068 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5070 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5071 (ix86_attribute_table): Add indirect_branch.
5072 * config/i386/i386.h (machine_function): Add indirect_branch_type
5073 and has_local_indirect_jump.
5074 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5076 (tablejump): Likewise.
5077 (*indirect_jump): Use ix86_output_indirect_jmp.
5078 (*tablejump_1): Likewise.
5079 (simple_return_indirect_internal): Likewise.
5080 * config/i386/i386.opt (mindirect-branch=): New option.
5081 (indirect_branch): New.
5084 (thunk-inline): Likewise.
5085 (thunk-extern): Likewise.
5086 * doc/extend.texi: Document indirect_branch function attribute.
5087 * doc/invoke.texi: Document -mindirect-branch= option.
5089 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5092 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5094 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5096 * ipa-inline.c (want_inline_small_function_p): Return false if
5097 inlining has already failed with CIF_FINAL_ERROR.
5098 (update_caller_keys): Call want_inline_small_function_p before
5100 (update_callee_keys): Likewise.
5102 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5104 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5106 (rs6000_quadword_masked_address_p): Likewise.
5107 (quad_aligned_load_p): Likewise.
5108 (quad_aligned_store_p): Likewise.
5109 (const_load_sequence_p): Add comment to describe the outer-most loop.
5110 (mimic_memory_attributes_and_flags): New function.
5111 (rs6000_gen_stvx): Likewise.
5112 (replace_swapped_aligned_store): Likewise.
5113 (rs6000_gen_lvx): Likewise.
5114 (replace_swapped_aligned_load): Likewise.
5115 (replace_swapped_load_constant): Capitalize argument name in
5116 comment describing this function.
5117 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5118 and stores that access quad-word aligned addresses and replace
5119 with stvx or lvx instructions when appropriate.
5120 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5121 New function prototype.
5122 (rs6000_quadword_masked_address_p): Likewise.
5123 (rs6000_gen_lvx): Likewise.
5124 (rs6000_gen_stvx): Likewise.
5125 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5126 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5127 when memory address is aligned.
5128 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5129 this split to select lvx instruction when memory address is aligned.
5130 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5131 instruction when memory address is aligned.
5132 (*vsx_le_perm_load_v16qi): Likewise.
5133 (four unnamed splitters): Modify to select the stvx instruction
5134 when memory is aligned.
5136 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5138 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5139 which appears in the queue multiple times.
5141 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5142 Alan Hayward <alan.hayward@arm.com>
5143 David Sherwood <david.sherwood@arm.com>
5145 * tree-vectorizer.h (vec_lower_bound): New structure.
5146 (_loop_vec_info): Add check_nonzero and lower_bounds.
5147 (LOOP_VINFO_CHECK_NONZERO): New macro.
5148 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5149 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5150 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5151 fields. Make seg_len the distance travelled, not including the
5153 (dr_direction_indicator): Declare.
5154 (dr_zero_step_indicator): Likewise.
5155 (dr_known_forward_stride_p): Likewise.
5156 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5158 (runtime_alias_check_p): Allow runtime alias checks with
5160 (operator ==): Compare access_size and align.
5161 (prune_runtime_alias_test_list): Rework for new distinction between
5162 the access_size and seg_len.
5163 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5165 (get_segment_min_max): New function.
5166 (create_intersect_range_checks): Use it.
5167 (dr_step_indicator): New function.
5168 (dr_direction_indicator): Likewise.
5169 (dr_zero_step_indicator): Likewise.
5170 (dr_known_forward_stride_p): Likewise.
5171 * tree-loop-distribution.c (data_ref_segment_size): Return
5172 DR_STEP * (niters - 1).
5173 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5175 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5176 (vect_preserves_scalar_order_p): New function, split out from...
5177 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5178 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5179 (vect_vfa_access_size): New function.
5180 (vect_vfa_align): Likewise.
5181 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5182 (dump_lower_bound): New function.
5183 (vect_check_lower_bound): Likewise.
5184 (vect_small_gap_p): Likewise.
5185 (vectorizable_with_step_bound_p): Likewise.
5186 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5187 depencies if the vectorization factor is 1. Convert the checks
5188 for nonzero steps into checks on the bounds of DR_STEP. Try using
5189 a bunds check for variable steps if the minimum required step is
5190 relatively small. Update calls to the dr_with_seg_len
5191 constructor and to vect_compile_time_alias.
5192 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5194 (vect_loop_versioning): Call it.
5195 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5197 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5200 Alan Hayward <alan.hayward@arm.com>
5201 David Sherwood <david.sherwood@arm.com>
5203 * doc/sourcebuild.texi (vect_scatter_store): Document.
5204 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5206 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5208 * genopinit.c (main): Add supports_vec_scatter_store and
5209 supports_vec_scatter_store_cached to target_optabs.
5210 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5211 IFN_MASK_SCATTER_STORE.
5212 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5214 * internal-fn.h (internal_store_fn_p): Declare.
5215 (internal_fn_stored_value_index): Likewise.
5216 * internal-fn.c (scatter_store_direct): New macro.
5217 (expand_scatter_store_optab_fn): New function.
5218 (direct_scatter_store_optab_supported_p): New macro.
5219 (internal_store_fn_p): New function.
5220 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5221 IFN_MASK_SCATTER_STORE.
5222 (internal_fn_mask_index): Likewise.
5223 (internal_fn_stored_value_index): New function.
5224 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5226 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5227 * optabs-query.c (supports_vec_scatter_store_p): New function.
5228 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5229 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5230 true for scatter stores.
5231 (vect_gather_scatter_fn_p): Handle scatter stores too.
5232 (vect_check_gather_scatter): Consider using scatter stores if
5233 supports_vec_scatter_store_p.
5234 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5236 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5237 internal_fn_stored_value_index.
5238 (check_load_store_masking): Handle scatter stores too.
5239 (vect_get_store_rhs): Make public.
5240 (vectorizable_call): Use internal_store_fn_p.
5241 (vectorizable_store): Handle scatter store internal functions.
5242 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5243 when deciding whether the end of the group has been reached.
5244 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5245 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5246 (mask_scatter_store<mode>): New insns.
5248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5249 Alan Hayward <alan.hayward@arm.com>
5250 David Sherwood <david.sherwood@arm.com>
5252 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5253 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5254 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5256 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5257 Use vect_truncate_gather_scatter_offset if we can't treat the
5258 operation as a normal gather load or scatter store.
5259 (get_group_load_store_type): Take the gather_scatter_info
5260 as argument. Try using a gather load or scatter store for
5261 single-element groups.
5262 (get_load_store_type): Update calls to get_group_load_store_type
5263 and vect_use_strided_gather_scatters_p.
5265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5266 Alan Hayward <alan.hayward@arm.com>
5267 David Sherwood <david.sherwood@arm.com>
5269 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5270 optional tree argument.
5271 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5273 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5274 but continue to use the current value as a fallback.
5275 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5276 to compare the updates.
5277 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5278 (get_load_store_type): Use it when handling a strided access.
5279 (vect_get_strided_load_store_ops): New function.
5280 (vect_get_data_ptr_increment): Likewise.
5281 (vectorizable_load): Handle strided gather loads. Always pass
5282 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5284 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5285 Alan Hayward <alan.hayward@arm.com>
5286 David Sherwood <david.sherwood@arm.com>
5288 * doc/md.texi (gather_load@var{m}): Document.
5289 (mask_gather_load@var{m}): Likewise.
5290 * genopinit.c (main): Add supports_vec_gather_load and
5291 supports_vec_gather_load_cached to target_optabs.
5292 * optabs-tree.c (init_tree_optimization_optabs): Use
5293 ggc_cleared_alloc to allocate target_optabs.
5294 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5295 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5297 * internal-fn.h (internal_load_fn_p): Declare.
5298 (internal_gather_scatter_fn_p): Likewise.
5299 (internal_fn_mask_index): Likewise.
5300 (internal_gather_scatter_fn_supported_p): Likewise.
5301 * internal-fn.c (gather_load_direct): New macro.
5302 (expand_gather_load_optab_fn): New function.
5303 (direct_gather_load_optab_supported_p): New macro.
5304 (direct_internal_fn_optab): New function.
5305 (internal_load_fn_p): Likewise.
5306 (internal_gather_scatter_fn_p): Likewise.
5307 (internal_fn_mask_index): Likewise.
5308 (internal_gather_scatter_fn_supported_p): Likewise.
5309 * optabs-query.c (supports_at_least_one_mode_p): New function.
5310 (supports_vec_gather_load_p): Likewise.
5311 * optabs-query.h (supports_vec_gather_load_p): Declare.
5312 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5313 and memory_type field.
5314 (NUM_PATTERNS): Bump to 15.
5315 * tree-vect-data-refs.c: Include internal-fn.h.
5316 (vect_gather_scatter_fn_p): New function.
5317 (vect_describe_gather_scatter_call): Likewise.
5318 (vect_check_gather_scatter): Try using internal functions for
5319 gather loads. Recognize existing calls to a gather load function.
5320 (vect_analyze_data_refs): Consider using gather loads if
5321 supports_vec_gather_load_p.
5322 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5323 (vect_get_gather_scatter_offset_type): Likewise.
5324 (vect_convert_mask_for_vectype): Likewise.
5325 (vect_add_conversion_to_patterm): Likewise.
5326 (vect_try_gather_scatter_pattern): Likewise.
5327 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5328 (vect_vect_recog_func_ptrs): Add it.
5329 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5330 internal_fn_mask_index and internal_gather_scatter_fn_p.
5331 (check_load_store_masking): Take the gather_scatter_info as an
5332 argument and handle gather loads.
5333 (vect_get_gather_scatter_ops): New function.
5334 (vectorizable_call): Check internal_load_fn_p.
5335 (vectorizable_load): Likewise. Handle gather load internal
5337 (vectorizable_store): Update call to check_load_store_masking.
5338 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5339 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5340 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5341 (aarch64_gather_scale_operand_d): New predicates.
5342 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5343 (mask_gather_load<mode>): New insns.
5345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5346 Alan Hayward <alan.hayward@arm.com>
5347 David Sherwood <david.sherwood@arm.com>
5349 * optabs.def (fold_left_plus_optab): New optab.
5350 * doc/md.texi (fold_left_plus_@var{m}): Document.
5351 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5352 * internal-fn.c (fold_left_direct): Define.
5353 (expand_fold_left_optab_fn): Likewise.
5354 (direct_fold_left_optab_supported_p): Likewise.
5355 * fold-const-call.c (fold_const_fold_left): New function.
5356 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5357 * tree-parloops.c (valid_reduction_p): New function.
5358 (gather_scalar_reductions): Use it.
5359 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5360 (vect_finish_replace_stmt): Declare.
5361 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5362 (needs_fold_left_reduction_p): New function, split out from...
5363 (vect_is_simple_reduction): ...here. Accept reductions that
5364 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5365 (vect_force_simple_reduction): Also store the reduction type in
5366 the assignment's STMT_VINFO_REDUC_TYPE.
5367 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5368 (merge_with_identity): New function.
5369 (vect_expand_fold_left): Likewise.
5370 (vectorize_fold_left_reduction): Likewise.
5371 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5372 scalar phi in place for it. Check for target support and reject
5373 cases that would reassociate the operation. Defer the transform
5374 phase to vectorize_fold_left_reduction.
5375 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5376 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5377 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5381 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5382 call to ifc_temp_var.
5384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5385 Alan Hayward <alan.hayward@arm.com>
5386 David Sherwood <david.sherwood@arm.com>
5388 * target.def (legitimize_address_displacement): Take the original
5389 offset as a poly_int.
5390 * targhooks.h (default_legitimize_address_displacement): Update
5392 * targhooks.c (default_legitimize_address_displacement): Likewise.
5393 * doc/tm.texi: Regenerate.
5394 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5395 as an argument, moving assert of ad->disp == ad->disp_term to...
5396 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5397 Try calling targetm.legitimize_address_displacement before expanding
5398 the address rather than afterwards, and adjust for the new interface.
5399 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5400 Match the new hook interface. Handle SVE addresses.
5401 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5406 * Makefile.in (OBJS): Add early-remat.o.
5407 * target.def (select_early_remat_modes): New hook.
5408 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5409 * doc/tm.texi: Regenerate.
5410 * targhooks.h (default_select_early_remat_modes): Declare.
5411 * targhooks.c (default_select_early_remat_modes): New function.
5412 * timevar.def (TV_EARLY_REMAT): New timevar.
5413 * passes.def (pass_early_remat): New pass.
5414 * tree-pass.h (make_pass_early_remat): Declare.
5415 * early-remat.c: New file.
5416 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5418 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5421 Alan Hayward <alan.hayward@arm.com>
5422 David Sherwood <david.sherwood@arm.com>
5424 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5425 vfm1 with a bound_epilog parameter.
5426 (vect_do_peeling): Update calls accordingly, and move the prologue
5427 call earlier in the function. Treat the base bound_epilog as 0 for
5428 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5429 this base when peeling for gaps.
5430 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5431 with fully-masked loops.
5432 (vect_estimate_min_profitable_iters): Handle the single peeled
5433 iteration in that case.
5435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5436 Alan Hayward <alan.hayward@arm.com>
5437 David Sherwood <david.sherwood@arm.com>
5439 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5440 single-element interleaving even if the size is not a power of 2.
5441 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5442 accesses for single-element interleaving if the group size is
5445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5446 Alan Hayward <alan.hayward@arm.com>
5447 David Sherwood <david.sherwood@arm.com>
5449 * doc/md.texi (fold_extract_last_@var{m}): Document.
5450 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5451 * optabs.def (fold_extract_last_optab): New optab.
5452 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5453 * internal-fn.c (fold_extract_direct): New macro.
5454 (expand_fold_extract_optab_fn): Likewise.
5455 (direct_fold_extract_optab_supported_p): Likewise.
5456 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5457 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5458 EXTRACT_LAST_REDUCTION.
5459 (get_initial_def_for_reduction): Do not create an initial vector
5460 for EXTRACT_LAST_REDUCTION reductions.
5461 (vectorizable_reduction): Leave the scalar phi in place for
5462 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5463 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5464 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5465 transform phase to vectorizable_condition.
5466 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5468 (vect_finish_stmt_generation): ...here.
5469 (vect_finish_replace_stmt): New function.
5470 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5471 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5473 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5476 Alan Hayward <alan.hayward@arm.com>
5477 David Sherwood <david.sherwood@arm.com>
5479 * doc/md.texi (extract_last_@var{m}): Document.
5480 * optabs.def (extract_last_optab): New optab.
5481 * internal-fn.def (EXTRACT_LAST): New internal function.
5482 * internal-fn.c (cond_unary_direct): New macro.
5483 (expand_cond_unary_optab_fn): Likewise.
5484 (direct_cond_unary_optab_supported_p): Likewise.
5485 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5486 loops using EXTRACT_LAST.
5487 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5488 (extract_last_<mode>): ...this optab.
5489 (vec_extract<mode><Vel>): Update accordingly.
5491 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5492 Alan Hayward <alan.hayward@arm.com>
5493 David Sherwood <david.sherwood@arm.com>
5495 * target.def (empty_mask_is_expensive): New hook.
5496 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5497 * doc/tm.texi: Regenerate.
5498 * targhooks.h (default_empty_mask_is_expensive): Declare.
5499 * targhooks.c (default_empty_mask_is_expensive): New function.
5500 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5501 if the target says that empty masks are expensive.
5502 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5504 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5507 Alan Hayward <alan.hayward@arm.com>
5508 David Sherwood <david.sherwood@arm.com>
5510 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5511 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5512 (vect_use_loop_mask_for_alignment_p): New function.
5513 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5514 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5515 niters_skip argument. Make sure that the first niters_skip elements
5516 of the first iteration are inactive.
5517 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5518 Update call to vect_set_loop_masks_directly.
5519 (get_misalign_in_elems): New function, split out from...
5520 (vect_gen_prolog_loop_niters): ...here.
5521 (vect_update_init_of_dr): Take a code argument that specifies whether
5522 the adjustment should be added or subtracted.
5523 (vect_update_init_of_drs): Likewise.
5524 (vect_prepare_for_masked_peels): New function.
5525 (vect_do_peeling): Skip prologue peeling if we're using a mask
5526 instead. Update call to vect_update_inits_of_drs.
5527 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5529 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5530 alignment. Do not include the number of peeled iterations in
5531 the minimum threshold in that case.
5532 (vectorizable_induction): Adjust the start value down by
5533 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5534 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5535 Take the number of skipped iterations into account when calculating
5537 * tree-vect-stmts.c (vect_gen_while_not): New function.
5539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5540 Alan Hayward <alan.hayward@arm.com>
5541 David Sherwood <david.sherwood@arm.com>
5543 * doc/sourcebuild.texi (vect_fully_masked): Document.
5544 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5546 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5548 (vect_analyze_loop_2): ...here. Don't check the vectorization
5549 factor against the number of loop iterations if the loop is
5552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5553 Alan Hayward <alan.hayward@arm.com>
5554 David Sherwood <david.sherwood@arm.com>
5556 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5557 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5558 (dump_groups): Update accordingly.
5559 (iv_use::mem_type): New member variable.
5560 (address_p): New function.
5561 (record_use): Add a mem_type argument and initialize the new
5563 (record_group_use): Add a mem_type argument. Use address_p.
5564 Remove obsolete null checks of base_object. Update call to record_use.
5565 (find_interesting_uses_op): Update call to record_group_use.
5566 (find_interesting_uses_cond): Likewise.
5567 (find_interesting_uses_address): Likewise.
5568 (get_mem_type_for_internal_fn): New function.
5569 (find_address_like_use): Likewise.
5570 (find_interesting_uses_stmt): Try find_address_like_use before
5571 calling find_interesting_uses_op.
5572 (addr_offset_valid_p): Use the iv mem_type field as the type
5573 of the addressed memory.
5574 (add_autoinc_candidates): Likewise.
5575 (get_address_cost): Likewise.
5576 (split_small_address_groups_p): Use address_p.
5577 (split_address_groups): Likewise.
5578 (add_iv_candidate_for_use): Likewise.
5579 (autoinc_possible_for_pair): Likewise.
5580 (rewrite_groups): Likewise.
5581 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5582 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5583 (get_alias_ptr_type_for_ptr_address): New function.
5584 (rewrite_use_address): Rewrite address uses in calls that were
5585 identified by find_address_like_use.
5587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5588 Alan Hayward <alan.hayward@arm.com>
5589 David Sherwood <david.sherwood@arm.com>
5591 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5593 * gimple-expr.h (is_gimple_addressable: Likewise.
5594 * gimple-expr.c (is_gimple_address): Likewise.
5595 * internal-fn.c (expand_call_mem_ref): New function.
5596 (expand_mask_load_optab_fn): Use it.
5597 (expand_mask_store_optab_fn): Likewise.
5599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5600 Alan Hayward <alan.hayward@arm.com>
5601 David Sherwood <david.sherwood@arm.com>
5603 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5604 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5605 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5606 (cond_umax@var{mode}): Document.
5607 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5608 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5609 (cond_umin_optab, cond_umax_optab): New optabs.
5610 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5611 (COND_IOR, COND_XOR): New internal functions.
5612 * internal-fn.h (get_conditional_internal_fn): Declare.
5613 * internal-fn.c (cond_binary_direct): New macro.
5614 (expand_cond_binary_optab_fn): Likewise.
5615 (direct_cond_binary_optab_supported_p): Likewise.
5616 (get_conditional_internal_fn): New function.
5617 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5618 Cope with reduction statements that are vectorized as calls rather
5620 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5621 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5622 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5623 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5624 (UNSPEC_COND_EOR): New unspecs.
5625 (optab): Add mappings for them.
5626 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5627 (sve_int_op, sve_fp_op): New int attributes.
5629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5630 Alan Hayward <alan.hayward@arm.com>
5631 David Sherwood <david.sherwood@arm.com>
5633 * optabs.def (while_ult_optab): New optab.
5634 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5635 * internal-fn.def (WHILE_ULT): New internal function.
5636 * internal-fn.h (direct_internal_fn_supported_p): New override
5637 that takes two types as argument.
5638 * internal-fn.c (while_direct): New macro.
5639 (expand_while_optab_fn): New function.
5640 (convert_optab_supported_p): Likewise.
5641 (direct_while_optab_supported_p): New macro.
5642 * wide-int.h (wi::udiv_ceil): New function.
5643 * tree-vectorizer.h (rgroup_masks): New structure.
5644 (vec_loop_masks): New typedef.
5645 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5647 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5648 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5649 (vect_max_vf): New function.
5650 (slpeel_make_loop_iterate_ntimes): Delete.
5651 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5652 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5653 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5654 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5655 internal-fn.h, stor-layout.h and optabs-query.h.
5656 (vect_set_loop_mask): New function.
5657 (add_preheader_seq): Likewise.
5658 (add_header_seq): Likewise.
5659 (interleave_supported_p): Likewise.
5660 (vect_maybe_permute_loop_masks): Likewise.
5661 (vect_set_loop_masks_directly): Likewise.
5662 (vect_set_loop_condition_masked): Likewise.
5663 (vect_set_loop_condition_unmasked): New function, split out from
5664 slpeel_make_loop_iterate_ntimes.
5665 (slpeel_make_loop_iterate_ntimes): Rename to..
5666 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5667 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5668 (vect_do_peeling): Update call accordingly.
5669 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5671 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5672 mask_compare_type, can_fully_mask_p and fully_masked_p.
5673 (release_vec_loop_masks): New function.
5674 (_loop_vec_info): Use it to free the loop masks.
5675 (can_produce_all_loop_masks_p): New function.
5676 (vect_get_max_nscalars_per_iter): Likewise.
5677 (vect_verify_full_masking): Likewise.
5678 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5679 retries, and free the mask rgroups before retrying. Check loop-wide
5680 reasons for disallowing fully-masked loops. Make the final decision
5681 about whether use a fully-masked loop or not.
5682 (vect_estimate_min_profitable_iters): Do not assume that peeling
5683 for the number of iterations will be needed for fully-masked loops.
5684 (vectorizable_reduction): Disable fully-masked loops.
5685 (vectorizable_live_operation): Likewise.
5686 (vect_halve_mask_nunits): New function.
5687 (vect_double_mask_nunits): Likewise.
5688 (vect_record_loop_mask): Likewise.
5689 (vect_get_loop_mask): Likewise.
5690 (vect_transform_loop): Handle the case in which the final loop
5691 iteration might handle a partial vector. Call vect_set_loop_condition
5692 instead of slpeel_make_loop_iterate_ntimes.
5693 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5694 (check_load_store_masking): New function.
5695 (prepare_load_store_mask): Likewise.
5696 (vectorizable_store): Handle fully-masked loops.
5697 (vectorizable_load): Likewise.
5698 (supportable_widening_operation): Use vect_halve_mask_nunits for
5700 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5701 (vect_gen_while): New function.
5702 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5703 (aarch64_uqdec<mode>): New insn.
5705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5706 Alan Hayward <alan.hayward@arm.com>
5707 David Sherwood <david.sherwood@arm.com>
5709 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5710 (reduc_xor_scal_optab): New optabs.
5711 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5712 (reduc_xor_scal_@var{m}): Document.
5713 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5714 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5716 * fold-const-call.c (fold_const_call): Handle them.
5717 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5718 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5719 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5720 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5721 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5722 (UNSPEC_XORV): New unspecs.
5723 (optab): Add entries for them.
5724 (BITWISEV): New int iterator.
5725 (bit_reduc_op): New int attributes.
5727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5728 Alan Hayward <alan.hayward@arm.com>
5729 David Sherwood <david.sherwood@arm.com>
5731 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5732 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5733 * optabs.def (vec_shl_insert_optab): New optab.
5734 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5735 (duplicate_and_interleave): Likewise.
5736 * tree-vect-loop.c: Include internal-fn.h.
5737 (neutral_op_for_slp_reduction): New function, split out from
5738 get_initial_defs_for_reduction.
5739 (get_initial_def_for_reduction): Handle option 2 for variable-length
5740 vectors by loading the neutral value into a vector and then shifting
5741 the initial value into element 0.
5742 (get_initial_defs_for_reduction): Replace the code argument with
5743 the neutral value calculated by neutral_op_for_slp_reduction.
5744 Use gimple_build_vector for constant-length vectors.
5745 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5746 but the first group_size elements have a neutral value.
5747 Use duplicate_and_interleave otherwise.
5748 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5749 Update call to get_initial_defs_for_reduction. Handle SLP
5750 reductions for variable-length vectors by creating one vector
5751 result for each scalar result, with the elements associated
5752 with other scalar results stubbed out with the neutral value.
5753 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5754 Require IFN_VEC_SHL_INSERT for double reductions on
5755 variable-length vectors, or SLP reductions that have
5756 a neutral value. Require can_duplicate_and_interleave_p
5757 support for variable-length unchained SLP reductions if there
5758 is no neutral value, such as for MIN/MAX reductions. Also require
5759 the number of vector elements to be a multiple of the number of
5760 SLP statements when doing variable-length unchained SLP reductions.
5761 Update call to vect_create_epilog_for_reduction.
5762 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5763 and remove initial values.
5764 (duplicate_and_interleave): Make public.
5765 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5766 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5769 Alan Hayward <alan.hayward@arm.com>
5770 David Sherwood <david.sherwood@arm.com>
5772 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5773 (can_duplicate_and_interleave_p): New function.
5774 (vect_get_and_check_slp_defs): Take the vector of statements
5775 rather than just the current one. Remove excess parentheses.
5776 Restriction rejectinon of vect_constant_def and vect_external_def
5777 for variable-length vectors to boolean types, or types for which
5778 can_duplicate_and_interleave_p is false.
5779 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5780 (duplicate_and_interleave): New function.
5781 (vect_get_constant_vectors): Use gimple_build_vector for
5782 constant-length vectors and suitable variable-length constant
5783 vectors. Use duplicate_and_interleave for other variable-length
5784 vectors. Don't defer the update when inserting new statements.
5786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5787 Alan Hayward <alan.hayward@arm.com>
5788 David Sherwood <david.sherwood@arm.com>
5790 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5791 min_profitable_iters doesn't go negative.
5793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5794 Alan Hayward <alan.hayward@arm.com>
5795 David Sherwood <david.sherwood@arm.com>
5797 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5798 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5799 * optabs.def (vec_mask_load_lanes_optab): New optab.
5800 (vec_mask_store_lanes_optab): Likewise.
5801 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5802 (MASK_STORE_LANES): Likewise.
5803 * internal-fn.c (mask_load_lanes_direct): New macro.
5804 (mask_store_lanes_direct): Likewise.
5805 (expand_mask_load_optab_fn): Handle masked operations.
5806 (expand_mask_load_lanes_optab_fn): New macro.
5807 (expand_mask_store_optab_fn): Handle masked operations.
5808 (expand_mask_store_lanes_optab_fn): New macro.
5809 (direct_mask_load_lanes_optab_supported_p): Likewise.
5810 (direct_mask_store_lanes_optab_supported_p): Likewise.
5811 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5813 (vect_load_lanes_supported): Likewise.
5814 * tree-vect-data-refs.c (strip_conversion): New function.
5815 (can_group_stmts_p): Likewise.
5816 (vect_analyze_data_ref_accesses): Use it instead of checking
5817 for a pair of assignments.
5818 (vect_store_lanes_supported): Take a masked_p parameter.
5819 (vect_load_lanes_supported): Likewise.
5820 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5821 vect_store_lanes_supported and vect_load_lanes_supported.
5822 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5823 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5824 parameter. Don't allow gaps for masked accesses.
5825 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5826 and vect_load_lanes_supported.
5827 (get_load_store_type): Take a masked_p parameter and update
5828 call to get_group_load_store_type.
5829 (vectorizable_store): Update call to get_load_store_type.
5830 Handle IFN_MASK_STORE_LANES.
5831 (vectorizable_load): Update call to get_load_store_type.
5832 Handle IFN_MASK_LOAD_LANES.
5834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5835 Alan Hayward <alan.hayward@arm.com>
5836 David Sherwood <david.sherwood@arm.com>
5838 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5840 * config/aarch64/aarch64-protos.h
5841 (aarch64_sve_struct_memory_operand_p): Declare.
5842 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5843 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5844 (VPRED, vpred): Handle SVE structure modes.
5845 * config/aarch64/constraints.md (Utx): New constraint.
5846 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5847 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5848 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5849 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5850 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5851 structure modes. Split into pieces after RA.
5852 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5853 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5855 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5856 SVE structure modes.
5857 (aarch64_classify_address): Likewise.
5858 (sizetochar): Move earlier in file.
5859 (aarch64_print_operand): Handle SVE register lists.
5860 (aarch64_array_mode): New function.
5861 (aarch64_sve_struct_memory_operand_p): Likewise.
5862 (TARGET_ARRAY_MODE): Redefine.
5864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5865 Alan Hayward <alan.hayward@arm.com>
5866 David Sherwood <david.sherwood@arm.com>
5868 * target.def (array_mode): New target hook.
5869 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5870 * doc/tm.texi: Regenerate.
5871 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5872 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5873 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5875 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5879 Alan Hayward <alan.hayward@arm.com>
5880 David Sherwood <david.sherwood@arm.com>
5882 * fold-const.c (fold_binary_loc): Check the argument types
5883 rather than the result type when testing for a vector operation.
5885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5887 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5888 * doc/tm.texi: Regenerate.
5890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5891 Alan Hayward <alan.hayward@arm.com>
5892 David Sherwood <david.sherwood@arm.com>
5894 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5895 (sve): Document new AArch64 extension.
5896 * doc/md.texi (w): Extend the description of the AArch64
5897 constraint to include SVE vectors.
5898 (Upl, Upa): Document new AArch64 predicate constraints.
5899 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5901 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5902 (msve-vector-bits=): New option.
5903 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5904 SVE when these are disabled.
5905 (sve): New extension.
5906 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5907 modes. Adjust their number of units based on aarch64_sve_vg.
5908 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5909 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5910 aarch64_addr_query_type.
5911 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5912 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5913 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5914 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5915 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5916 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5917 (aarch64_simd_imm_zero_p): Delete.
5918 (aarch64_check_zero_based_sve_index_immediate): Declare.
5919 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5920 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5921 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5922 (aarch64_sve_float_mul_immediate_p): Likewise.
5923 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5925 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5926 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5927 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5928 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5929 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5930 (aarch64_regmode_natural_size): Likewise.
5931 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5932 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5934 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5935 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5936 for VG and the SVE predicate registers.
5937 (V_ALIASES): Add a "z"-prefixed alias.
5938 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5939 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5940 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5941 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5942 (REG_CLASS_NAMES): Add entries for them.
5943 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5944 and the predicate registers.
5945 (aarch64_sve_vg): Declare.
5946 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5947 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5948 (REGMODE_NATURAL_SIZE): Define.
5949 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5951 * config/aarch64/aarch64.c: Include cfgrtl.h.
5952 (simd_immediate_info): Add a constructor for series vectors,
5953 and an associated step field.
5954 (aarch64_sve_vg): New variable.
5955 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5956 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5957 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5958 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5959 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5960 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5961 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5962 (aarch64_get_mask_mode): New functions.
5963 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5964 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5965 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5966 predicate modes and predicate registers. Explicitly restrict
5967 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5968 to store a vector mode if it is recognized by
5969 aarch64_classify_vector_mode.
5970 (aarch64_regmode_natural_size): New function.
5971 (aarch64_hard_regno_caller_save_mode): Return the original mode
5973 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5974 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5975 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5976 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5978 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5979 does not overlap dest if the function is frame-related. Handle
5981 (aarch64_split_add_offset): New function.
5982 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5983 them aarch64_add_offset.
5984 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5985 and update call to aarch64_sub_sp.
5986 (aarch64_add_cfa_expression): New function.
5987 (aarch64_expand_prologue): Pass extra temporary registers to the
5988 functions above. Handle the case in which we need to emit new
5989 DW_CFA_expressions for registers that were originally saved
5990 relative to the stack pointer, but now have to be expressed
5991 relative to the frame pointer.
5992 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5994 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5995 IP0 and IP1 values for SVE frames.
5996 (aarch64_expand_vec_series): New function.
5997 (aarch64_expand_sve_widened_duplicate): Likewise.
5998 (aarch64_expand_sve_const_vector): Likewise.
5999 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6000 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6001 into the register, rather than emitting a SET directly.
6002 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6003 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6004 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6005 (offset_9bit_signed_scaled_p): New functions.
6006 (aarch64_replicate_bitmask_imm): New function.
6007 (aarch64_bitmask_imm): Use it.
6008 (aarch64_cannot_force_const_mem): Reject expressions involving
6009 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6010 (aarch64_classify_index): Handle SVE indices, by requiring
6011 a plain register index with a scale that matches the element size.
6012 (aarch64_classify_address): Handle SVE addresses. Assert that
6013 the mode of the address is VOIDmode or an integer mode.
6014 Update call to aarch64_classify_symbol.
6015 (aarch64_classify_symbolic_expression): Update call to
6016 aarch64_classify_symbol.
6017 (aarch64_const_vec_all_in_range_p): New function.
6018 (aarch64_print_vector_float_operand): Likewise.
6019 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6020 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6021 and the FP immediates 1.0 and 0.5.
6022 (aarch64_print_address_internal): Handle SVE addresses.
6023 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6024 (aarch64_regno_regclass): Handle predicate registers.
6025 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6027 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6028 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6029 (aarch64_convert_sve_vector_bits): New function.
6030 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6031 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6033 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6034 Handle SVE vector and predicate modes. Accept VL-based constants
6035 that need only one temporary register, and VL offsets that require
6036 no temporary registers.
6037 (aarch64_conditional_register_usage): Mark the predicate registers
6038 as fixed if SVE isn't available.
6039 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6040 Return true for SVE vector and predicate modes.
6041 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6042 rather than an unsigned int. Handle SVE modes.
6043 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6045 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6047 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6048 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6049 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6050 (aarch64_sve_float_mul_immediate_p): New functions.
6051 (aarch64_sve_valid_immediate): New function.
6052 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6053 Explicitly reject structure modes. Check for INDEX constants.
6054 Handle PTRUE and PFALSE constants.
6055 (aarch64_check_zero_based_sve_index_immediate): New function.
6056 (aarch64_simd_imm_zero_p): Delete.
6057 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6058 vector modes. Accept constants in the range of CNT[BHWD].
6059 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6060 ask for an Advanced SIMD mode.
6061 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6062 (aarch64_simd_vector_alignment): Handle SVE predicates.
6063 (aarch64_vectorize_preferred_vector_alignment): New function.
6064 (aarch64_simd_vector_alignment_reachable): Use it instead of
6066 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6067 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6069 (MAX_VECT_LEN): Delete.
6070 (expand_vec_perm_d): Add a vec_flags field.
6071 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6072 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6073 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6075 (aarch64_evpc_rev): Rename to...
6076 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6077 (aarch64_evpc_rev_global): New function.
6078 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6079 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6081 (aarch64_evpc_sve_tbl): New function.
6082 (aarch64_expand_vec_perm_const_1): Update after rename of
6083 aarch64_evpc_rev. Handle SVE permutes too, trying
6084 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6085 than aarch64_evpc_tbl.
6086 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6087 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6088 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6089 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6090 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6091 (aarch64_expand_sve_vcond): New functions.
6092 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6093 of aarch64_vector_mode_p.
6094 (aarch64_dwarf_poly_indeterminate_value): New function.
6095 (aarch64_compute_pressure_classes): Likewise.
6096 (aarch64_can_change_mode_class): Likewise.
6097 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6098 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6099 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6100 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6101 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6102 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6103 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6104 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6106 (Dn, Dl, Dr): Accept const as well as const_vector.
6107 (Dz): Likewise. Compare against CONST0_RTX.
6108 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6109 of "vector" where appropriate.
6110 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6111 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6112 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6113 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6114 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6115 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6116 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6117 (v_int_equiv): Extend to SVE modes.
6118 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6120 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6121 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6122 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6123 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6124 (SVE_COND_FP_CMP): New int iterators.
6125 (perm_hilo): Handle the new unpack unspecs.
6126 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6128 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6129 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6130 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6131 (aarch64_equality_operator, aarch64_constant_vector_operand)
6132 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6133 (aarch64_sve_nonimmediate_operand): Likewise.
6134 (aarch64_sve_general_operand): Likewise.
6135 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6136 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6137 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6138 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6139 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6140 (aarch64_sve_float_arith_immediate): Likewise.
6141 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6142 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6143 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6144 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6145 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6146 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6147 (aarch64_sve_float_arith_operand): Likewise.
6148 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6149 (aarch64_sve_float_mul_operand): Likewise.
6150 (aarch64_sve_vec_perm_operand): Likewise.
6151 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6152 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6153 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6154 as well as const_vector.
6155 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6156 in file. Use CONST0_RTX and CONSTM1_RTX.
6157 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6158 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6159 Use aarch64_simd_imm_zero.
6160 * config/aarch64/aarch64-sve.md: New file.
6161 * config/aarch64/aarch64.md: Include it.
6162 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6163 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6164 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6165 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6166 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6167 (sve): New attribute.
6168 (enabled): Disable instructions with the sve attribute unless
6170 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6171 aarch64_expand_mov_immediate.
6172 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6173 CNT[BHSD] immediates.
6174 (movti): Split CONST_POLY_INT moves into two halves.
6175 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6176 Split additions that need a temporary here if the destination
6177 is the stack pointer.
6178 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6179 (*add<mode>3_poly_1): New instruction.
6180 (set_clobber_cc): New expander.
6182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6184 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6185 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6186 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6187 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6188 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6189 Change innermode from fixed_mode_size to machine_mode.
6190 (simplify_subreg): Update call accordingly. Handle a constant-sized
6191 subreg of a variable-length CONST_VECTOR.
6193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6194 Alan Hayward <alan.hayward@arm.com>
6195 David Sherwood <david.sherwood@arm.com>
6197 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6198 (add_offset_to_base): New function, split out from...
6199 (create_mem_ref): ...here. When handling a scale other than 1,
6200 check first whether the address is valid without the offset.
6201 Add it into the base if so, leaving the index and scale as-is.
6203 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6206 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6207 fold_for_warn before checking if arg2 is INTEGER_CST.
6209 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6211 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6212 (store_multiple_operation): Delete.
6213 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6214 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6215 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6216 guarded by TARGET_STRING.
6217 (rs6000_output_load_multiple): Delete.
6218 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6219 OPTION_MASK_STRING / TARGET_STRING handling.
6220 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6221 (const rs6000_opt_masks) <"string">: Change mask to 0.
6222 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6223 (MASK_STRING): Delete.
6224 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6226 (load_multiple): Delete.
6233 (store_multiple): Delete.
6240 (movmemsi_8reg): Delete.
6241 (corresponding unnamed define_insn): Delete.
6242 (movmemsi_6reg): Delete.
6243 (corresponding unnamed define_insn): Delete.
6244 (movmemsi_4reg): Delete.
6245 (corresponding unnamed define_insn): Delete.
6246 (movmemsi_2reg): Delete.
6247 (corresponding unnamed define_insn): Delete.
6248 (movmemsi_1reg): Delete.
6249 (corresponding unnamed define_insn): Delete.
6250 * config/rs6000/rs6000.opt (mno-string): New.
6251 (mstring): Replace by deprecation warning stub.
6252 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6254 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6256 * regrename.c (regrename_do_replace): If replacing the same
6257 reg multiple times, try to reuse last created gen_raw_REG.
6260 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6261 main to workaround a bug in GDB.
6263 2018-01-12 Tom de Vries <tom@codesourcery.com>
6266 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6268 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6270 PR rtl-optimization/80481
6271 * ira-color.c (get_cap_member): New function.
6272 (allocnos_conflict_by_live_ranges_p): Use it.
6273 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6274 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6276 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6279 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6280 (*saddl_se_1): Ditto.
6282 (*ssubl_se_1): Ditto.
6284 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6286 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6287 rather than wi::to_widest for DR_INITs.
6288 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6289 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6290 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6292 (vect_analyze_group_access_1): Note that here.
6294 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6296 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6297 polynomial type sizes.
6299 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6301 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6302 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6303 (gimple_add_tmp_var): Likewise.
6305 2018-01-12 Martin Liska <mliska@suse.cz>
6307 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6308 (gimple_alloc_sizes): Likewise.
6309 (dump_gimple_statistics): Use PRIu64 in printf format.
6310 * gimple.h: Change uint64_t to int.
6312 2018-01-12 Martin Liska <mliska@suse.cz>
6314 * tree-core.h: Use uint64_t instead of int.
6315 * tree.c (tree_node_counts): Likewise.
6316 (tree_node_sizes): Likewise.
6317 (dump_tree_statistics): Use PRIu64 in printf format.
6319 2018-01-12 Martin Liska <mliska@suse.cz>
6321 * Makefile.in: As qsort_chk is implemented in vec.c, add
6322 vec.o to linkage of gencfn-macros.
6323 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6324 passing the info to record_node_allocation_statistics.
6325 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6327 * ggc-common.c (struct ggc_usage): Add operator== and use
6328 it in operator< and compare function.
6329 * mem-stats.h (struct mem_usage): Likewise.
6330 * vec.c (struct vec_usage): Remove operator< and compare
6331 function. Can be simply inherited.
6333 2018-01-12 Martin Jambor <mjambor@suse.cz>
6336 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6337 * tree-ssa-math-opts.c: Include domwalk.h.
6338 (convert_mult_to_fma_1): New function.
6339 (fma_transformation_info): New type.
6340 (fma_deferring_state): Likewise.
6341 (cancel_fma_deferring): New function.
6342 (result_of_phi): Likewise.
6343 (last_fma_candidate_feeds_initial_phi): Likewise.
6344 (convert_mult_to_fma): Added deferring logic, split actual
6345 transformation to convert_mult_to_fma_1.
6346 (math_opts_dom_walker): New type.
6347 (math_opts_dom_walker::after_dom_children): New method, body moved
6348 here from pass_optimize_widening_mul::execute, added deferring logic
6350 (pass_optimize_widening_mul::execute): Moved most of code to
6351 math_opts_dom_walker::after_dom_children.
6352 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6353 * config/i386/i386.c (ix86_option_override_internal): Added
6354 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6356 2018-01-12 Richard Biener <rguenther@suse.de>
6359 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6360 inline instance vars.
6362 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6365 * config/rx/rx.c (rx_is_restricted_memory_address):
6368 2018-01-12 Richard Biener <rguenther@suse.de>
6370 PR tree-optimization/80846
6371 * target.def (split_reduction): New target hook.
6372 * targhooks.c (default_split_reduction): New function.
6373 * targhooks.h (default_split_reduction): Declare.
6374 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6375 target requests first reduce vectors by combining low and high
6377 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6378 (get_vectype_for_scalar_type_and_size): Export.
6379 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6380 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6381 * doc/tm.texi: Regenerate.
6382 * config/i386/i386.c (ix86_split_reduction): Implement
6383 TARGET_VECTORIZE_SPLIT_REDUCTION.
6385 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6388 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6389 in PIC mode except for TARGET_VXWORKS_RTP.
6390 * config/sparc/sparc.c: Include cfgrtl.h.
6391 (TARGET_INIT_PIC_REG): Define.
6392 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6393 (sparc_pic_register_p): New predicate.
6394 (sparc_legitimate_address_p): Use it.
6395 (sparc_legitimize_pic_address): Likewise.
6396 (sparc_delegitimize_address): Likewise.
6397 (sparc_mode_dependent_address_p): Likewise.
6398 (gen_load_pcrel_sym): Remove 4th parameter.
6399 (load_got_register): Adjust call to above. Remove obsolete stuff.
6400 (sparc_expand_prologue): Do not call load_got_register here.
6401 (sparc_flat_expand_prologue): Likewise.
6402 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6403 (sparc_use_pseudo_pic_reg): New function.
6404 (sparc_init_pic_reg): Likewise.
6405 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6406 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6408 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6410 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6411 Add item for branch_cost.
6413 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6415 PR rtl-optimization/83565
6416 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6417 not extend the result to a larger mode for rotate operations.
6418 (num_sign_bit_copies1): Likewise.
6420 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6423 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6425 Use values-Xc.o for -pedantic.
6426 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6428 2018-01-12 Martin Liska <mliska@suse.cz>
6431 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6433 (possible_polymorphic_call_targets): Use it.
6434 (ipa_devirt): Likewise.
6436 2018-01-12 Martin Liska <mliska@suse.cz>
6438 * profile-count.h (enum profile_quality): Use 0 as invalid
6439 enum value of profile_quality.
6441 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6443 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6444 -mext-string options.
6446 2018-01-12 Richard Biener <rguenther@suse.de>
6448 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6449 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6450 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6452 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6454 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6456 * configure.ac (--with-long-double-format): Add support for the
6457 configuration option to change the default long double format on
6459 * config.gcc (powerpc*-linux*-*): Likewise.
6460 * configure: Regenerate.
6461 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6462 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6463 used without modification.
6465 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6467 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6468 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6469 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6470 MISC_BUILTIN_SPEC_BARRIER.
6471 (rs6000_init_builtins): Likewise.
6472 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6474 (speculation_barrier): New define_insn.
6475 * doc/extend.texi: Document __builtin_speculation_barrier.
6477 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6480 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6481 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6482 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6484 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6485 integral modes instead of "ss" and "sd".
6486 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6487 vectors with 32-bit and 64-bit elements.
6488 (vecdupssescalarmodesuffix): New mode attribute.
6489 (vec_dup<mode>): Use it.
6491 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6494 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6495 frame if argument is passed on stack.
6497 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6500 * ree.c (combine_reaching_defs): Optimize also
6501 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6502 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6504 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6507 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6509 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6512 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6513 after they are computed.
6515 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6517 PR tree-optimization/83695
6518 * gimple-loop-linterchange.cc
6519 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6520 reset cached scev information after interchange.
6521 (pass_linterchange::execute): Remove call to scev_reset_htab.
6523 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6525 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6526 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6527 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6528 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6529 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6530 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6531 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6532 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6533 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6534 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6535 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6536 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6537 (V_lane_reg): Likewise.
6538 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6540 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6541 (vfmal_lane_low<mode>_intrinsic,
6542 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6543 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6544 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6545 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6546 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6547 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6549 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6551 * config/arm/arm-cpus.in (fp16fml): New feature.
6552 (ALL_SIMD): Add fp16fml.
6553 (armv8.2-a): Add fp16fml as an option.
6554 (armv8.3-a): Likewise.
6555 (armv8.4-a): Add fp16fml as part of fp16.
6556 * config/arm/arm.h (TARGET_FP16FML): Define.
6557 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6559 * config/arm/arm-modes.def (V2HF): Define.
6560 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6561 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6562 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6563 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6564 vfmsl_low, vfmsl_high): New set of builtins.
6565 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6566 (vfml_op): New code attribute.
6567 (VFMLHALVES): New int iterator.
6568 (VFML, VFMLSEL): New mode attributes.
6569 (V_reg): Define mapping for V2HF.
6570 (V_hi, V_lo): New mode attributes.
6571 (VF_constraint): Likewise.
6572 (vfml_half, vfml_half_selector): New int attributes.
6573 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6575 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6576 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6578 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6579 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6580 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6581 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6583 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6584 Document new effective target and option set.
6586 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6588 * config/arm/arm-cpus.in (armv8_4): New feature.
6589 (ARMv8_4a): New fgroup.
6590 (armv8.4-a): New arch.
6591 * config/arm/arm-tables.opt: Regenerate.
6592 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6593 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6594 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6595 Add matching rules for -march=armv8.4-a and extensions.
6596 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6598 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6601 * config/rx/rx.md (BW): New mode attribute.
6602 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6604 2018-01-11 Richard Biener <rguenther@suse.de>
6606 PR tree-optimization/83435
6607 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6608 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6609 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6611 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6612 Alan Hayward <alan.hayward@arm.com>
6613 David Sherwood <david.sherwood@arm.com>
6615 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6617 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6618 (aarch64_print_address_internal): Use it to check for a zero offset.
6620 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6621 Alan Hayward <alan.hayward@arm.com>
6622 David Sherwood <david.sherwood@arm.com>
6624 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6625 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6626 Return a poly_int64 rather than a HOST_WIDE_INT.
6627 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6628 rather than a HOST_WIDE_INT.
6629 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6630 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6631 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6632 final_offset from HOST_WIDE_INT to poly_int64.
6633 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6634 to_constant when getting the number of units in an Advanced SIMD
6636 (aarch64_builtin_vectorized_function): Check for a constant number
6638 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6640 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6641 attribute instead of GET_MODE_NUNITS.
6642 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6643 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6644 GET_MODE_SIZE for fixed-size registers.
6645 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6646 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6647 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6648 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6649 (aarch64_print_operand, aarch64_print_address_internal)
6650 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6651 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6652 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6653 Handle polynomial GET_MODE_SIZE.
6654 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6655 wider than SImode without modification.
6656 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6657 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6658 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6659 passing and returning SVE modes.
6660 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6661 rather than GEN_INT.
6662 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6663 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6664 (aarch64_allocate_and_probe_stack_space): Likewise.
6665 (aarch64_layout_frame): Cope with polynomial offsets.
6666 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6667 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6669 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6670 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6671 poly_int64 rather than a HOST_WIDE_INT.
6672 (aarch64_get_separate_components, aarch64_process_components)
6673 (aarch64_expand_prologue, aarch64_expand_epilogue)
6674 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6675 (aarch64_anchor_offset): New function, split out from...
6676 (aarch64_legitimize_address): ...here.
6677 (aarch64_builtin_vectorization_cost): Handle polynomial
6678 TYPE_VECTOR_SUBPARTS.
6679 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6681 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6682 number of elements from the PARALLEL rather than the mode.
6683 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6684 rather than GET_MODE_BITSIZE.
6685 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6686 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6687 (aarch64_expand_vec_perm_const_1): Handle polynomial
6688 d->perm.length () and d->perm elements.
6689 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6690 Apply to_constant to d->perm elements.
6691 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6692 polynomial CONST_VECTOR_NUNITS.
6693 (aarch64_move_pointer): Take amount as a poly_int64 rather
6695 (aarch64_progress_pointer): Avoid temporary variable.
6696 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6697 the mode attribute instead of GET_MODE.
6699 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6700 Alan Hayward <alan.hayward@arm.com>
6701 David Sherwood <david.sherwood@arm.com>
6703 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6704 x exists before using it.
6705 (aarch64_add_constant_internal): Rename to...
6706 (aarch64_add_offset_1): ...this. Replace regnum with separate
6707 src and dest rtxes. Handle the case in which they're different,
6708 including when the offset is zero. Replace scratchreg with an rtx.
6709 Use 2 additions if there is no spare register into which we can
6710 move a 16-bit constant.
6711 (aarch64_add_constant): Delete.
6712 (aarch64_add_offset): Replace reg with separate src and dest
6713 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6714 Use aarch64_add_offset_1.
6715 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6716 an rtx rather than an int. Take the delta as a poly_int64
6717 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6718 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6719 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6720 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6721 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6723 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6724 aarch64_add_constant.
6726 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6728 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6729 Use scalar_float_mode.
6731 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6733 * config/aarch64/aarch64-simd.md
6734 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6735 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6736 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6737 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6738 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6739 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6740 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6741 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6742 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6743 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6745 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6748 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6749 targ_options->x_arm_arch_string is non NULL.
6751 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6753 * config/aarch64/aarch64.h
6754 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6756 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6759 * expmed.c (emit_store_flag_force): Swap if const op0
6760 and change VOIDmode to mode of op0.
6762 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6764 PR rtl-optimization/83761
6765 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6766 than bytes to mode_for_size.
6768 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6771 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6772 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6775 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6778 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6779 when in layout mode.
6780 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6781 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6784 2018-01-10 Michael Collison <michael.collison@arm.com>
6786 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6787 * config/aarch64/aarch64-option-extension.def: Add
6788 AARCH64_OPT_EXTENSION of 'fp16fml'.
6789 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6790 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6791 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6792 * config/aarch64/constraints.md (Ui7): New constraint.
6793 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6794 (VFMLA_SEL_W): Ditto.
6797 (VFMLA16_LOW): New int iterator.
6798 (VFMLA16_HIGH): Ditto.
6799 (UNSPEC_FMLAL): New unspec.
6800 (UNSPEC_FMLSL): Ditto.
6801 (UNSPEC_FMLAL2): Ditto.
6802 (UNSPEC_FMLSL2): Ditto.
6803 (f16mac): New code attribute.
6804 * config/aarch64/aarch64-simd-builtins.def
6805 (aarch64_fmlal_lowv2sf): Ditto.
6806 (aarch64_fmlsl_lowv2sf): Ditto.
6807 (aarch64_fmlalq_lowv4sf): Ditto.
6808 (aarch64_fmlslq_lowv4sf): Ditto.
6809 (aarch64_fmlal_highv2sf): Ditto.
6810 (aarch64_fmlsl_highv2sf): Ditto.
6811 (aarch64_fmlalq_highv4sf): Ditto.
6812 (aarch64_fmlslq_highv4sf): Ditto.
6813 (aarch64_fmlal_lane_lowv2sf): Ditto.
6814 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6815 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6816 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6817 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6818 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6819 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6820 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6821 (aarch64_fmlal_lane_highv2sf): Ditto.
6822 (aarch64_fmlsl_lane_highv2sf): Ditto.
6823 (aarch64_fmlal_laneq_highv2sf): Ditto.
6824 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6825 (aarch64_fmlalq_lane_highv4sf): Ditto.
6826 (aarch64_fmlsl_lane_highv4sf): Ditto.
6827 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6828 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6829 * config/aarch64/aarch64-simd.md:
6830 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6831 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6832 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6833 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6834 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6835 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6836 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6837 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6838 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6839 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6840 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6841 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6842 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6843 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6844 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6845 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6846 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6847 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6848 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6849 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6850 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6851 (vfmlsl_low_u32): Ditto.
6852 (vfmlalq_low_u32): Ditto.
6853 (vfmlslq_low_u32): Ditto.
6854 (vfmlal_high_u32): Ditto.
6855 (vfmlsl_high_u32): Ditto.
6856 (vfmlalq_high_u32): Ditto.
6857 (vfmlslq_high_u32): Ditto.
6858 (vfmlal_lane_low_u32): Ditto.
6859 (vfmlsl_lane_low_u32): Ditto.
6860 (vfmlal_laneq_low_u32): Ditto.
6861 (vfmlsl_laneq_low_u32): Ditto.
6862 (vfmlalq_lane_low_u32): Ditto.
6863 (vfmlslq_lane_low_u32): Ditto.
6864 (vfmlalq_laneq_low_u32): Ditto.
6865 (vfmlslq_laneq_low_u32): Ditto.
6866 (vfmlal_lane_high_u32): Ditto.
6867 (vfmlsl_lane_high_u32): Ditto.
6868 (vfmlal_laneq_high_u32): Ditto.
6869 (vfmlsl_laneq_high_u32): Ditto.
6870 (vfmlalq_lane_high_u32): Ditto.
6871 (vfmlslq_lane_high_u32): Ditto.
6872 (vfmlalq_laneq_high_u32): Ditto.
6873 (vfmlslq_laneq_high_u32): Ditto.
6874 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6875 (AARCH64_FL_FOR_ARCH8_4): New.
6876 (AARCH64_ISA_F16FML): New ISA flag.
6877 (TARGET_F16FML): New feature flag for fp16fml.
6878 (doc/invoke.texi): Document new fp16fml option.
6880 2018-01-10 Michael Collison <michael.collison@arm.com>
6882 * config/aarch64/aarch64-builtins.c:
6883 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6884 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6885 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6886 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6887 (AARCH64_ISA_SHA3): New ISA flag.
6888 (TARGET_SHA3): New feature flag for sha3.
6889 * config/aarch64/iterators.md (sha512_op): New int attribute.
6890 (CRYPTO_SHA512): New int iterator.
6891 (UNSPEC_SHA512H): New unspec.
6892 (UNSPEC_SHA512H2): Ditto.
6893 (UNSPEC_SHA512SU0): Ditto.
6894 (UNSPEC_SHA512SU1): Ditto.
6895 * config/aarch64/aarch64-simd-builtins.def
6896 (aarch64_crypto_sha512hqv2di): New builtin.
6897 (aarch64_crypto_sha512h2qv2di): Ditto.
6898 (aarch64_crypto_sha512su0qv2di): Ditto.
6899 (aarch64_crypto_sha512su1qv2di): Ditto.
6900 (aarch64_eor3qv8hi): Ditto.
6901 (aarch64_rax1qv2di): Ditto.
6902 (aarch64_xarqv2di): Ditto.
6903 (aarch64_bcaxqv8hi): Ditto.
6904 * config/aarch64/aarch64-simd.md:
6905 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6906 (aarch64_crypto_sha512su0qv2di): Ditto.
6907 (aarch64_crypto_sha512su1qv2di): Ditto.
6908 (aarch64_eor3qv8hi): Ditto.
6909 (aarch64_rax1qv2di): Ditto.
6910 (aarch64_xarqv2di): Ditto.
6911 (aarch64_bcaxqv8hi): Ditto.
6912 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6913 (vsha512h2q_u64): Ditto.
6914 (vsha512su0q_u64): Ditto.
6915 (vsha512su1q_u64): Ditto.
6916 (veor3q_u16): Ditto.
6917 (vrax1q_u64): Ditto.
6919 (vbcaxq_u16): Ditto.
6920 * config/arm/types.md (crypto_sha512): New type attribute.
6921 (crypto_sha3): Ditto.
6922 (doc/invoke.texi): Document new sha3 option.
6924 2018-01-10 Michael Collison <michael.collison@arm.com>
6926 * config/aarch64/aarch64-builtins.c:
6927 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6928 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6929 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6930 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6931 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6932 (AARCH64_ISA_SM4): New ISA flag.
6933 (TARGET_SM4): New feature flag for sm4.
6934 * config/aarch64/aarch64-simd-builtins.def
6935 (aarch64_sm3ss1qv4si): Ditto.
6936 (aarch64_sm3tt1aq4si): Ditto.
6937 (aarch64_sm3tt1bq4si): Ditto.
6938 (aarch64_sm3tt2aq4si): Ditto.
6939 (aarch64_sm3tt2bq4si): Ditto.
6940 (aarch64_sm3partw1qv4si): Ditto.
6941 (aarch64_sm3partw2qv4si): Ditto.
6942 (aarch64_sm4eqv4si): Ditto.
6943 (aarch64_sm4ekeyqv4si): Ditto.
6944 * config/aarch64/aarch64-simd.md:
6945 (aarch64_sm3ss1qv4si): Ditto.
6946 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6947 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6948 (aarch64_sm4eqv4si): Ditto.
6949 (aarch64_sm4ekeyqv4si): Ditto.
6950 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6951 (sm3part_op): Ditto.
6952 (CRYPTO_SM3TT): Ditto.
6953 (CRYPTO_SM3PART): Ditto.
6954 (UNSPEC_SM3SS1): New unspec.
6955 (UNSPEC_SM3TT1A): Ditto.
6956 (UNSPEC_SM3TT1B): Ditto.
6957 (UNSPEC_SM3TT2A): Ditto.
6958 (UNSPEC_SM3TT2B): Ditto.
6959 (UNSPEC_SM3PARTW1): Ditto.
6960 (UNSPEC_SM3PARTW2): Ditto.
6961 (UNSPEC_SM4E): Ditto.
6962 (UNSPEC_SM4EKEY): Ditto.
6963 * config/aarch64/constraints.md (Ui2): New constraint.
6964 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6965 * config/arm/types.md (crypto_sm3): New type attribute.
6966 (crypto_sm4): Ditto.
6967 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6968 (vsm3tt1aq_u32): Ditto.
6969 (vsm3tt1bq_u32): Ditto.
6970 (vsm3tt2aq_u32): Ditto.
6971 (vsm3tt2bq_u32): Ditto.
6972 (vsm3partw1q_u32): Ditto.
6973 (vsm3partw2q_u32): Ditto.
6974 (vsm4eq_u32): Ditto.
6975 (vsm4ekeyq_u32): Ditto.
6976 (doc/invoke.texi): Document new sm4 option.
6978 2018-01-10 Michael Collison <michael.collison@arm.com>
6980 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6981 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6982 (AARCH64_FL_FOR_ARCH8_4): New.
6983 (AARCH64_FL_V8_4): New flag.
6984 (doc/invoke.texi): Document new armv8.4-a option.
6986 2018-01-10 Michael Collison <michael.collison@arm.com>
6988 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6989 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6990 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6991 * config/aarch64/aarch64-option-extension.def: Add
6992 AARCH64_OPT_EXTENSION of 'sha2'.
6993 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6994 (crypto): Disable sha2 and aes if crypto disabled.
6995 (crypto): Enable aes and sha2 if enabled.
6996 (simd): Disable sha2 and aes if simd disabled.
6997 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6999 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7000 (TARGET_SHA2): New feature flag for sha2.
7001 (TARGET_AES): New feature flag for aes.
7002 * config/aarch64/aarch64-simd.md:
7003 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7004 conditional on TARGET_AES.
7005 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7006 (aarch64_crypto_sha1hsi): Make pattern conditional
7008 (aarch64_crypto_sha1hv4si): Ditto.
7009 (aarch64_be_crypto_sha1hv4si): Ditto.
7010 (aarch64_crypto_sha1su1v4si): Ditto.
7011 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7012 (aarch64_crypto_sha1su0v4si): Ditto.
7013 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7014 (aarch64_crypto_sha256su0v4si): Ditto.
7015 (aarch64_crypto_sha256su1v4si): Ditto.
7016 (doc/invoke.texi): Document new aes and sha2 options.
7018 2018-01-10 Martin Sebor <msebor@redhat.com>
7020 PR tree-optimization/83781
7021 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7024 2018-01-11 Martin Sebor <msebor@gmail.com>
7025 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7027 PR tree-optimization/83501
7028 PR tree-optimization/81703
7030 * tree-ssa-strlen.c (get_string_cst): Rename...
7031 (get_string_len): ...to this. Handle global constants.
7032 (handle_char_store): Adjust.
7034 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7035 Jim Wilson <jimw@sifive.com>
7037 * config/riscv/riscv-protos.h (riscv_output_return): New.
7038 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7039 (riscv_attribute_table, riscv_output_return),
7040 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7041 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7042 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7044 (riscv_expand_prologue): Add early return for naked function.
7045 (riscv_expand_epilogue): Likewise.
7046 (riscv_function_ok_for_sibcall): Return false for naked function.
7047 (riscv_set_current_function): New.
7048 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7049 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7050 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7051 * doc/extend.texi (RISC-V Function Attributes): New.
7053 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7055 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7056 check for 128-bit long double before checking TCmode.
7057 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7058 128-bit long doubles before checking TFmode or TCmode.
7059 (FLOAT128_IBM_P): Likewise.
7061 2018-01-10 Martin Sebor <msebor@redhat.com>
7063 PR tree-optimization/83671
7064 * builtins.c (c_strlen): Unconditionally return zero for the empty
7066 Use -Warray-bounds for warnings.
7067 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7068 for non-constant array indices with COMPONENT_REF, arrays of
7069 arrays, and pointers to arrays.
7070 (gimple_fold_builtin_strlen): Determine and set length range for
7071 non-constant character arrays.
7073 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7076 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7079 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7081 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7083 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7086 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7087 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7088 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7089 indexed_or_indirect_operand predicate.
7090 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7091 (*vsx_le_perm_load_v8hi): Likewise.
7092 (*vsx_le_perm_load_v16qi): Likewise.
7093 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7094 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7095 (*vsx_le_perm_store_v8hi): Likewise.
7096 (*vsx_le_perm_store_v16qi): Likewise.
7097 (eight unnamed splitters): Likewise.
7099 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7101 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7102 * config/rs6000/emmintrin.h: Likewise.
7103 * config/rs6000/mmintrin.h: Likewise.
7104 * config/rs6000/xmmintrin.h: Likewise.
7106 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7109 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7111 * tree.c (tree_nop_conversion): Return true for location wrapper
7113 (maybe_wrap_with_location): New function.
7114 (selftest::check_strip_nops): New function.
7115 (selftest::test_location_wrappers): New function.
7116 (selftest::tree_c_tests): Call it.
7117 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7118 (maybe_wrap_with_location): New decl.
7119 (EXPR_LOCATION_WRAPPER_P): New macro.
7120 (location_wrapper_p): New inline function.
7121 (tree_strip_any_location_wrapper): New inline function.
7123 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7126 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7127 stack_realign_offset for the largest alignment of stack slot
7129 (ix86_find_max_used_stack_alignment): New function.
7130 (ix86_finalize_stack_frame_flags): Use it. Set
7131 max_used_stack_alignment if we don't realign stack.
7132 * config/i386/i386.h (machine_function): Add
7133 max_used_stack_alignment.
7135 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7137 * config/arm/arm.opt (-mbranch-cost): New option.
7138 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7141 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7144 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7145 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7147 2018-01-10 Richard Biener <rguenther@suse.de>
7150 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7151 early out so it also covers the case where we have a non-NULL
7154 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7156 PR tree-optimization/83753
7157 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7158 for non-strided grouped accesses if the number of elements is 1.
7160 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7163 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7164 * i386.h (TARGET_USE_GATHER): Define.
7165 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7167 2018-01-10 Martin Liska <mliska@suse.cz>
7170 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7171 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7173 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7174 CLEANUP_NO_PARTITIONING is not set.
7176 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7178 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7179 for vectors, as a partial revert of r254296.
7180 * rtl.h (const_vec_p): Delete.
7181 (const_vec_duplicate_p): Don't test for vector CONSTs.
7182 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7183 * expmed.c (make_tree): Likewise.
7186 * common.md (E, F): Use CONSTANT_P instead of checking for
7188 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7189 checking for CONST_VECTOR.
7191 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7194 * predict.c (force_edge_cold): Handle in more sane way edges
7197 2018-01-09 Carl Love <cel@us.ibm.com>
7199 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7201 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7202 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7203 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7204 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7205 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7206 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7207 * config/rs6000/rs6000-protos.h: Add extern defition for
7208 rs6000_generate_float2_double_code.
7209 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7211 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7212 (float2_v2df): Add define_expand.
7214 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7217 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7218 op_mode in the force_to_mode call.
7220 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7222 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7223 instead of checking each element individually.
7224 (aarch64_evpc_uzp): Likewise.
7225 (aarch64_evpc_zip): Likewise.
7226 (aarch64_evpc_ext): Likewise.
7227 (aarch64_evpc_rev): Likewise.
7228 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7229 instead of checking each element individually. Return true without
7231 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7232 whether all selected elements come from the same input, instead of
7233 checking each element individually. Remove calls to gen_rtx_REG,
7234 start_sequence and end_sequence and instead assert that no rtl is
7237 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7239 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7240 order of HIGH and CONST checks.
7242 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7244 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7245 if the destination isn't an SSA_NAME.
7247 2018-01-09 Richard Biener <rguenther@suse.de>
7249 PR tree-optimization/83668
7250 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7252 (canonicalize_loop_form): ... here, renamed from ...
7253 (canonicalize_loop_closed_ssa_form): ... this and amended to
7254 swap successor edges for loop exit blocks to make us use
7255 the RPO order we need for initial schedule generation.
7257 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7259 PR tree-optimization/64811
7260 * match.pd: When optimizing comparisons with Inf, avoid
7261 introducing or losing exceptions from comparisons with NaN.
7263 2018-01-09 Martin Liska <mliska@suse.cz>
7266 * asan.c (shadow_mem_size): Add gcc_assert.
7268 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7270 Don't save registers in main().
7273 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7274 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7275 * config/avr/avr.c (avr_set_current_function): Don't error if
7276 naked, OS_task or OS_main are specified at the same time.
7277 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7279 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7281 * common/config/avr/avr-common.c (avr_option_optimization_table):
7282 Switch on -mmain-is-OS_task for optimizing compilations.
7284 2018-01-09 Richard Biener <rguenther@suse.de>
7286 PR tree-optimization/83572
7287 * graphite.c: Include cfganal.h.
7288 (graphite_transform_loops): Connect infinite loops to exit
7289 and remove fake edges at the end.
7291 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7293 * ipa-inline.c (edge_badness): Revert accidental checkin.
7295 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7298 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7299 symbols; not inline clones.
7301 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7304 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7305 hard registers. Formatting fixes.
7307 PR preprocessor/83722
7308 * gcc.c (try_generate_repro): Pass
7309 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7310 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7313 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7314 Kito Cheng <kito.cheng@gmail.com>
7316 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7317 (riscv_leaf_function_p): Delete.
7318 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7320 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7322 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7324 (do_ifelse): New function.
7325 (do_isel): New function.
7326 (do_sub3): New function.
7327 (do_add3): New function.
7328 (do_load_mask_compare): New function.
7329 (do_overlap_load_compare): New function.
7330 (expand_compare_loop): New function.
7331 (expand_block_compare): Call expand_compare_loop() when appropriate.
7332 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7334 (-mblock-compare-inline-loop-limit): New option.
7336 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7339 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7340 Reverse order of second and third operands in first alternative.
7341 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7342 of first and second elements in UNSPEC_VPERMR vector.
7343 (altivec_expand_vec_perm_le): Likewise.
7345 2018-01-08 Jeff Law <law@redhat.com>
7347 PR rtl-optimizatin/81308
7348 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7349 (process_switch): If group_case_labels makes a change, then set
7351 (pass_convert_switch::execute): If a switch is converted, then
7352 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7354 PR rtl-optimization/81308
7355 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7358 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7360 PR target/83663 - Revert r255946
7361 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7362 generation for cases where splatting a value is not useful.
7363 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7364 across a vec_duplicate and a paradoxical subreg forming a vector
7365 mode to a vec_concat.
7367 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7369 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7370 -march=armv8.3-a variants.
7371 * config/arm/t-multilib: Likewise.
7372 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7374 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7376 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7378 (cceq_ior_compare_complement): Give it a name so I can use it, and
7379 change boolean_or_operator predicate to boolean_operator so it can
7380 be used to generate a crand.
7381 (eqne): New code iterator.
7382 (bd/bd_neg): New code_attrs.
7383 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7384 a single define_insn.
7385 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7386 decrement (bdnzt/bdnzf/bdzt/bdzf).
7387 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7388 with the new names of the branch decrement patterns, and added the
7389 names of the branch decrement conditional patterns.
7391 2018-01-08 Richard Biener <rguenther@suse.de>
7393 PR tree-optimization/83563
7394 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7397 2018-01-08 Richard Biener <rguenther@suse.de>
7400 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7402 2018-01-08 Richard Biener <rguenther@suse.de>
7404 PR tree-optimization/83685
7405 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7406 references to abnormals.
7408 2018-01-08 Richard Biener <rguenther@suse.de>
7411 * dwarf2out.c (output_indirect_strings): Handle empty
7412 skeleton_debug_str_hash.
7413 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7415 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7417 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7418 (emit_store_direct): Likewise.
7419 (arc_trampoline_adjust_address): Likewise.
7420 (arc_asm_trampoline_template): New function.
7421 (arc_initialize_trampoline): Use asm_trampoline_template.
7422 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7423 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7424 * config/arc/arc.md (flush_icache): Delete pattern.
7426 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7428 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7429 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7432 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7435 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7436 by not USED_FOR_TARGET.
7437 (make_pass_resolve_sw_modes): Likewise.
7439 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7441 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7444 2018-01-08 Richard Biener <rguenther@suse.de>
7447 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7449 2018-01-08 Richard Biener <rguenther@suse.de>
7452 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7454 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7457 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7458 basic blocks with a small number of successors.
7459 (convert_control_dep_chain_into_preds): Improve handling of
7461 (dump_predicates): Split apart into...
7462 (dump_pred_chain): ...here...
7463 (dump_pred_info): ...and here.
7464 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7465 (can_chain_union_be_invalidated_p): Improve check for invalidation
7467 (uninit_uses_cannot_happen): Avoid unnecessary if
7468 convert_control_dep_chain_into_preds yielded nothing.
7470 2018-01-06 Martin Sebor <msebor@redhat.com>
7472 PR tree-optimization/83640
7473 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7474 subtracting negative offset from size.
7475 (builtin_access::overlap): Adjust offset bounds of the access to fall
7476 within the size of the object if possible.
7478 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7480 PR rtl-optimization/83699
7481 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7482 extract_bit_field_as_subreg to cases in which the extracted
7483 value is also a vector.
7485 * lra-constraints.c (process_alt_operands): Test for the equivalence
7486 substitutions when detecting a possible reload cycle.
7488 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7491 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7492 by default if flag_selective_schedling{,2}. Formatting fixes.
7494 PR rtl-optimization/83682
7495 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7496 if it has non-VECTOR_MODE element mode.
7497 (vec_duplicate_p): Likewise.
7500 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7501 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7503 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7506 * config/i386/i386-builtin.def
7507 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7508 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7509 Require also OPTION_MASK_ISA_AVX512F in addition to
7510 OPTION_MASK_ISA_GFNI.
7511 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7512 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7513 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7514 to OPTION_MASK_ISA_GFNI.
7515 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7516 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7517 OPTION_MASK_ISA_AVX512BW.
7518 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7519 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7520 addition to OPTION_MASK_ISA_GFNI.
7521 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7522 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7523 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7524 to OPTION_MASK_ISA_GFNI.
7525 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7526 a requirement for all ISAs rather than any of them with a few
7528 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7530 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7531 bitmasks to be enabled with 3 exceptions, instead of requiring any
7532 enabled ISA with lots of exceptions.
7533 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7534 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7535 Change avx512bw in isa attribute to avx512f.
7536 * config/i386/sgxintrin.h: Add license boilerplate.
7537 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7538 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7539 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7540 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7542 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7543 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7544 temporarily sse2 rather than sse if not enabled already.
7547 * config/i386/sse.md (VI248_VLBW): Rename to ...
7548 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7549 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7550 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7551 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7552 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7553 mode iterator instead of VI248_VLBW.
7555 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7557 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7558 (record_modified): Skip clobbers; add debug output.
7559 (param_change_prob): Use sreal frequencies.
7561 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7563 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7564 punt for user-aligned variables.
7566 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7568 * tree-chrec.c (chrec_contains_symbols): Return true for
7571 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7574 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7575 of (x|y) == x for BICS pattern.
7577 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7579 PR tree-optimization/83605
7580 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7581 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7584 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7586 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7587 * config/epiphany/rtems.h: New file.
7589 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7590 Uros Bizjak <ubizjak@gmail.com>
7593 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7594 QIreg_operand instead of register_operand predicate.
7595 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7596 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7597 comments instead of -fmitigate[-_]rop.
7599 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7602 * cgraphunit.c (symbol_table::compile): Switch to text_section
7603 before calling assembly_start debug hook.
7604 * run-rtl-passes.c (run_rtl_passes): Likewise.
7607 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7609 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7610 range_int_cst_p rather than !symbolic_range_p before calling
7611 extract_range_from_multiplicative_op_1.
7613 2018-01-04 Jeff Law <law@redhat.com>
7615 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7616 redundant test in assertion.
7618 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7620 * doc/rtl.texi: Document machine_mode wrapper classes.
7622 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7624 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7627 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7629 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7630 the VEC_PERM_EXPR fold to fail.
7632 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7635 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7636 to switched_sections.
7638 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7641 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7644 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7647 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7648 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7650 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7653 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7654 is BLKmode and bitpos not zero or mode change is needed.
7656 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7659 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7662 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7665 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7666 instead of MULT rtx. Update all corresponding splitters.
7668 (*ssub<modesuffix>): Ditto.
7670 (*cmp_sadd_di): Update split patterns.
7671 (*cmp_sadd_si): Ditto.
7672 (*cmp_sadd_sidi): Ditto.
7673 (*cmp_ssub_di): Ditto.
7674 (*cmp_ssub_si): Ditto.
7675 (*cmp_ssub_sidi): Ditto.
7676 * config/alpha/predicates.md (const23_operand): New predicate.
7677 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7678 Look for ASHIFT, not MULT inner operand.
7679 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7681 2018-01-04 Martin Liska <mliska@suse.cz>
7683 PR gcov-profile/83669
7684 * gcov.c (output_intermediate_file): Add version to intermediate
7686 * doc/gcov.texi: Document new field 'version' in intermediate
7687 file format. Fix location of '-k' option of gcov command.
7689 2018-01-04 Martin Liska <mliska@suse.cz>
7692 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7694 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7696 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7698 2018-01-03 Martin Sebor <msebor@redhat.com>
7700 PR tree-optimization/83655
7701 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7702 checking calls with invalid arguments.
7704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7706 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7707 (vectorizable_mask_load_store): Delete.
7708 (vectorizable_call): Return false for masked loads and stores.
7709 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7710 instead of gimple_assign_rhs1.
7711 (vectorizable_load): Handle IFN_MASK_LOAD.
7712 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7716 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7718 (vectorizable_mask_load_store): ...here.
7719 (vectorizable_load): ...and here.
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7723 * tree-vect-stmts.c (vect_build_all_ones_mask)
7724 (vect_build_zero_merge_argument): New functions, split out from...
7725 (vectorizable_load): ...here.
7727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7729 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7731 (vectorizable_mask_load_store): ...here.
7732 (vectorizable_store): ...and here.
7734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7736 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7738 (vectorizable_mask_load_store): ...here.
7740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7742 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7743 (vect_model_store_cost): Take a vec_load_store_type instead of a
7745 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7746 (vect_model_store_cost): Take a vec_load_store_type instead of a
7748 (vectorizable_mask_load_store): Update accordingly.
7749 (vectorizable_store): Likewise.
7750 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7754 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7755 IFN_MASK_LOAD calls here rather than...
7756 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7759 Alan Hayward <alan.hayward@arm.com>
7760 David Sherwood <david.sherwood@arm.com>
7762 * expmed.c (extract_bit_field_1): For vector extracts,
7763 fall back to extract_bit_field_as_subreg if vec_extract
7766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7767 Alan Hayward <alan.hayward@arm.com>
7768 David Sherwood <david.sherwood@arm.com>
7770 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7771 they are variable or constant sized.
7772 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7773 slots for constant-sized data.
7775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7776 Alan Hayward <alan.hayward@arm.com>
7777 David Sherwood <david.sherwood@arm.com>
7779 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7780 handling COND_EXPRs with boolean comparisons, try to find a better
7781 basis for the mask type than the boolean itself.
7783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7785 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7786 is calculated and how it can be overridden.
7787 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7788 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7790 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7794 Alan Hayward <alan.hayward@arm.com>
7795 David Sherwood <david.sherwood@arm.com>
7797 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7798 Remove the mode argument.
7799 (aarch64_simd_valid_immediate): Remove the mode and inverse
7801 * config/aarch64/iterators.md (bitsize): New iterator.
7802 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7803 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7804 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7805 aarch64_simd_valid_immediate.
7806 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7807 (aarch64_reg_or_bic_imm): Likewise.
7808 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7809 with an insn_type enum and msl with a modifier_type enum.
7810 Replace element_width with a scalar_mode. Change the shift
7811 to unsigned int. Add constructors for scalar_float_mode and
7812 scalar_int_mode elements.
7813 (aarch64_vect_float_const_representable_p): Delete.
7814 (aarch64_can_const_movi_rtx_p)
7815 (aarch64_simd_scalar_immediate_valid_for_move)
7816 (aarch64_simd_make_constant): Update call to
7817 aarch64_simd_valid_immediate.
7818 (aarch64_advsimd_valid_immediate_hs): New function.
7819 (aarch64_advsimd_valid_immediate): Likewise.
7820 (aarch64_simd_valid_immediate): Remove mode and inverse
7821 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7822 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7823 and aarch64_float_const_representable_p on the result.
7824 (aarch64_output_simd_mov_immediate): Remove mode argument.
7825 Update call to aarch64_simd_valid_immediate and use of
7826 simd_immediate_info.
7827 (aarch64_output_scalar_simd_mov_immediate): Update call
7830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7831 Alan Hayward <alan.hayward@arm.com>
7832 David Sherwood <david.sherwood@arm.com>
7834 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7835 (mode_nunits): Likewise CONST_MODE_NUNITS.
7836 * machmode.def (ADJUST_NUNITS): Document.
7837 * genmodes.c (mode_data::need_nunits_adj): New field.
7838 (blank_mode): Update accordingly.
7839 (adj_nunits): New variable.
7840 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7842 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7843 listed in adj_nunits.
7844 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7845 listed in adj_nunits. Don't emit case statements for such modes.
7846 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7847 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7848 nothing if adj_nunits is nonnull.
7849 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7850 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7851 (emit_mode_fbit): Update use of print_maybe_const_decl.
7852 (emit_move_size): Likewise. Treat the array as non-const
7854 (emit_mode_adjustments): Handle adj_nunits.
7856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7858 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7859 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7860 (VECTOR_MODES): Use it.
7861 (make_vector_modes): Take the prefix as an argument.
7863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7864 Alan Hayward <alan.hayward@arm.com>
7865 David Sherwood <david.sherwood@arm.com>
7867 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7868 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7869 for MODE_VECTOR_BOOL.
7870 * machmode.def (VECTOR_BOOL_MODE): Document.
7871 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7872 (make_vector_bool_mode): New function.
7873 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7875 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7876 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7878 * stor-layout.c (int_mode_for_mode): Likewise.
7879 * tree.c (build_vector_type_for_mode): Likewise.
7880 * varasm.c (output_constant_pool_2): Likewise.
7881 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7882 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7883 for MODE_VECTOR_BOOL.
7884 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7885 of mode class checks.
7886 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7887 instead of a list of mode class checks.
7888 (expand_vector_scalar_condition): Likewise.
7889 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7892 Alan Hayward <alan.hayward@arm.com>
7893 David Sherwood <david.sherwood@arm.com>
7895 * machmode.h (mode_size): Change from unsigned short to
7897 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7898 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7899 or if measurement_type is not polynomial.
7900 (fixed_size_mode::includes_p): Check for constant-sized modes.
7901 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7902 return a poly_uint16 rather than an unsigned short.
7903 (emit_mode_size): Change the type of mode_size from unsigned short
7904 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7905 (emit_mode_adjustments): Cope with polynomial vector sizes.
7906 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7908 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7910 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7911 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7912 * caller-save.c (setup_save_areas): Likewise.
7913 (replace_reg_with_saved_mem): Likewise.
7914 * calls.c (emit_library_call_value_1): Likewise.
7915 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7916 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7917 (gen_lowpart_for_combine): Likewise.
7918 * convert.c (convert_to_integer_1): Likewise.
7919 * cse.c (equiv_constant, cse_insn): Likewise.
7920 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7921 (cselib_subst_to_values): Likewise.
7922 * dce.c (word_dce_process_block): Likewise.
7923 * df-problems.c (df_word_lr_mark_ref): Likewise.
7924 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7925 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7926 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7927 (rtl_for_decl_location): Likewise.
7928 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7929 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7930 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7931 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7932 (expand_expr_real_1): Likewise.
7933 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7934 (pad_below): Likewise.
7935 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7936 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7937 * ira.c (get_subreg_tracking_sizes): Likewise.
7938 * ira-build.c (ira_create_allocno_objects): Likewise.
7939 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7940 (ira_sort_regnos_for_alter_reg): Likewise.
7941 * ira-costs.c (record_operand_costs): Likewise.
7942 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7943 (resolve_simple_move): Likewise.
7944 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7945 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7946 (lra_constraints): Likewise.
7947 (CONST_POOL_OK_P): Reject variable-sized modes.
7948 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7949 (add_pseudo_to_slot, lra_spill): Likewise.
7950 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7951 * optabs-query.c (get_best_extraction_insn): Likewise.
7952 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7953 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7954 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7955 * recog.c (offsettable_address_addr_space_p): Likewise.
7956 * regcprop.c (maybe_mode_change): Likewise.
7957 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7958 * regrename.c (build_def_use): Likewise.
7959 * regstat.c (dump_reg_info): Likewise.
7960 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7961 (find_reloads, find_reloads_subreg_address): Likewise.
7962 * reload1.c (eliminate_regs_1): Likewise.
7963 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7964 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7965 (simplify_binary_operation_1, simplify_subreg): Likewise.
7966 * targhooks.c (default_function_arg_padding): Likewise.
7967 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7968 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7969 (verify_gimple_assign_ternary): Likewise.
7970 * tree-inline.c (estimate_move_cost): Likewise.
7971 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7972 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7973 (get_address_cost_ainc): Likewise.
7974 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7975 (vect_supportable_dr_alignment): Likewise.
7976 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7977 (vectorizable_reduction): Likewise.
7978 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7979 (vectorizable_operation, vectorizable_load): Likewise.
7980 * tree.c (build_same_sized_truth_vector_type): Likewise.
7981 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7982 * var-tracking.c (emit_note_insn_var_location): Likewise.
7983 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7984 (ADDR_VEC_ALIGN): Likewise.
7986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 Alan Hayward <alan.hayward@arm.com>
7988 David Sherwood <david.sherwood@arm.com>
7990 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7992 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7993 or if measurement_type is polynomial.
7994 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7995 * combine.c (make_extraction): Likewise.
7996 * dse.c (find_shift_sequence): Likewise.
7997 * dwarf2out.c (mem_loc_descriptor): Likewise.
7998 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7999 (extract_bit_field, extract_low_bits): Likewise.
8000 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8001 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8002 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8003 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8004 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8005 * reload.c (find_reloads): Likewise.
8006 * reload1.c (alter_reg): Likewise.
8007 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8008 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8009 * tree-if-conv.c (predicate_mem_writes): Likewise.
8010 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8011 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8012 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8013 * valtrack.c (dead_debug_insert_temp): Likewise.
8014 * varasm.c (mergeable_constant_section): Likewise.
8015 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8018 Alan Hayward <alan.hayward@arm.com>
8019 David Sherwood <david.sherwood@arm.com>
8021 * expr.c (expand_assignment): Cope with polynomial mode sizes
8022 when assigning to a CONCAT.
8024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8025 Alan Hayward <alan.hayward@arm.com>
8026 David Sherwood <david.sherwood@arm.com>
8028 * machmode.h (mode_precision): Change from unsigned short to
8030 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8032 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8033 or if measurement_type is not polynomial.
8034 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8035 in which the mode is already known to be a scalar_int_mode.
8036 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8037 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8039 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8040 for GET_MODE_PRECISION.
8041 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8042 for GET_MODE_PRECISION.
8043 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8045 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8046 (expand_field_assignment, make_extraction): Likewise.
8047 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8048 (get_last_value): Likewise.
8049 * convert.c (convert_to_integer_1): Likewise.
8050 * cse.c (cse_insn): Likewise.
8051 * expr.c (expand_expr_real_1): Likewise.
8052 * lra-constraints.c (simplify_operand_subreg): Likewise.
8053 * optabs-query.c (can_atomic_load_p): Likewise.
8054 * optabs.c (expand_atomic_load): Likewise.
8055 (expand_atomic_store): Likewise.
8056 * ree.c (combine_reaching_defs): Likewise.
8057 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8058 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8059 * tree.h (type_has_mode_precision_p): Likewise.
8060 * ubsan.c (instrument_si_overflow): Likewise.
8062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8063 Alan Hayward <alan.hayward@arm.com>
8064 David Sherwood <david.sherwood@arm.com>
8066 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8067 polynomial numbers of units.
8068 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8069 (valid_vector_subparts_p): New function.
8070 (build_vector_type): Remove temporary shim and take the number
8071 of units as a poly_uint64 rather than an int.
8072 (build_opaque_vector_type): Take the number of units as a
8073 poly_uint64 rather than an int.
8074 * tree.c (build_vector_from_ctor): Handle polynomial
8075 TYPE_VECTOR_SUBPARTS.
8076 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8077 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8078 (build_vector_from_val): If the number of units is variable,
8079 use build_vec_duplicate_cst for constant operands and
8080 VEC_DUPLICATE_EXPR otherwise.
8081 (make_vector_type): Remove temporary is_constant ().
8082 (build_vector_type, build_opaque_vector_type): Take the number of
8083 units as a poly_uint64 rather than an int.
8084 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8086 * cfgexpand.c (expand_debug_expr): Likewise.
8087 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8088 (store_constructor, expand_expr_real_1): Likewise.
8089 (const_scalar_mask_from_tree): Likewise.
8090 * fold-const-call.c (fold_const_reduction): Likewise.
8091 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8092 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8093 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8094 (fold_relational_const): Likewise.
8095 (native_interpret_vector): Likewise. Change the size from an
8096 int to an unsigned int.
8097 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8098 TYPE_VECTOR_SUBPARTS.
8099 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8100 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8101 duplicating a non-constant operand into a variable-length vector.
8102 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8103 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8104 * ipa-icf.c (sem_variable::equals): Likewise.
8105 * match.pd: Likewise.
8106 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8107 * print-tree.c (print_node): Likewise.
8108 * stor-layout.c (layout_type): Likewise.
8109 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8110 * tree-cfg.c (verify_gimple_comparison): Likewise.
8111 (verify_gimple_assign_binary): Likewise.
8112 (verify_gimple_assign_ternary): Likewise.
8113 (verify_gimple_assign_single): Likewise.
8114 * tree-pretty-print.c (dump_generic_node): Likewise.
8115 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8116 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8117 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8118 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8119 (vect_shift_permute_load_chain): Likewise.
8120 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8121 (expand_vector_condition, optimize_vector_constructor): Likewise.
8122 (lower_vec_perm, get_compute_type): Likewise.
8123 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8124 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8125 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8126 (vect_recog_mask_conversion_pattern): Likewise.
8127 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8128 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8129 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8130 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8131 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8132 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8133 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8134 (supportable_widening_operation): Likewise.
8135 (supportable_narrowing_operation): Likewise.
8136 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8138 * varasm.c (output_constant): Likewise.
8140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8141 Alan Hayward <alan.hayward@arm.com>
8142 David Sherwood <david.sherwood@arm.com>
8144 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8145 so that both the length == 3 and length != 3 cases set up their
8146 own permute vectors. Add comments explaining why we know the
8147 number of elements is constant.
8148 (vect_permute_load_chain): Likewise.
8150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8151 Alan Hayward <alan.hayward@arm.com>
8152 David Sherwood <david.sherwood@arm.com>
8154 * machmode.h (mode_nunits): Change from unsigned char to
8156 (ONLY_FIXED_SIZE_MODES): New macro.
8157 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8158 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8159 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8161 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8162 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8163 or if measurement_type is not polynomial.
8164 * genmodes.c (ZERO_COEFFS): New macro.
8165 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8167 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8168 Use ZERO_COEFFS when emitting initializers.
8169 * data-streamer.h (bp_pack_poly_value): New function.
8170 (bp_unpack_poly_value): Likewise.
8171 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8172 for GET_MODE_NUNITS.
8173 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8174 for GET_MODE_NUNITS.
8175 * tree.c (make_vector_type): Remove temporary shim and make
8176 the real function take the number of units as a poly_uint64
8178 (build_vector_type_for_mode): Handle polynomial nunits.
8179 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8180 * emit-rtl.c (const_vec_series_p_1): Likewise.
8181 (gen_rtx_CONST_VECTOR): Likewise.
8182 * fold-const.c (test_vec_duplicate_folding): Likewise.
8183 * genrecog.c (validate_pattern): Likewise.
8184 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8185 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8186 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8187 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8188 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8189 * rtlanal.c (subreg_get_info): Likewise.
8190 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8191 (vect_grouped_load_supported): Likewise.
8192 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8193 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8194 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8195 (simplify_const_unary_operation, simplify_binary_operation_1)
8196 (simplify_const_binary_operation, simplify_ternary_operation)
8197 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8198 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8199 instead of CONST_VECTOR_NUNITS.
8200 * varasm.c (output_constant_pool_2): Likewise.
8201 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8202 explicit-encoded elements in the XVEC for variable-length vectors.
8204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8206 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8209 Alan Hayward <alan.hayward@arm.com>
8210 David Sherwood <david.sherwood@arm.com>
8212 * coretypes.h (fixed_size_mode): Declare.
8213 (fixed_size_mode_pod): New typedef.
8214 * builtins.h (target_builtins::x_apply_args_mode)
8215 (target_builtins::x_apply_result_mode): Change type to
8216 fixed_size_mode_pod.
8217 * builtins.c (apply_args_size, apply_result_size, result_vector)
8218 (expand_builtin_apply_args_1, expand_builtin_apply)
8219 (expand_builtin_return): Update accordingly.
8221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8223 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8224 * cselib.c (cselib_hash_rtx): Likewise.
8225 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8226 CONST_VECTOR encoding.
8228 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8229 Jeff Law <law@redhat.com>
8232 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8233 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8234 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8235 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8238 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8239 explicitly probe *sp in a noreturn function if there were any callee
8240 register saves or frame pointer is needed.
8242 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8245 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8246 BLKmode for ternary, binary or unary expressions.
8249 * var-tracking.c (delete_vta_debug_insn): New inline function.
8250 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8251 insns from get_insns () to NULL instead of each bb separately.
8252 Use delete_vta_debug_insn. No longer static.
8253 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8254 delete_vta_debug_insns callers.
8255 * rtl.h (delete_vta_debug_insns): Declare.
8256 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8257 instead of variable_tracking_main.
8259 2018-01-03 Martin Sebor <msebor@redhat.com>
8261 PR tree-optimization/83603
8262 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8263 arguments past the endof the argument list in functions declared
8264 without a prototype.
8265 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8266 Avoid checking when arguments are null.
8268 2018-01-03 Martin Sebor <msebor@redhat.com>
8271 * doc/extend.texi (attribute const): Fix a typo.
8272 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8273 issuing -Wsuggest-attribute for void functions.
8275 2018-01-03 Martin Sebor <msebor@redhat.com>
8277 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8278 offset_int::from instead of wide_int::to_shwi.
8279 (maybe_diag_overlap): Remove assertion.
8280 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8281 * gimple-ssa-sprintf.c (format_directive): Same.
8282 (parse_directive): Same.
8283 (sprintf_dom_walker::compute_format_length): Same.
8284 (try_substitute_return_value): Same.
8286 2018-01-03 Jeff Law <law@redhat.com>
8289 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8290 non-constant residual for zero at runtime and avoid probing in
8291 that case. Reorganize code for trailing problem to mirror handling
8294 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8296 PR tree-optimization/83501
8297 * tree-ssa-strlen.c (get_string_cst): New.
8298 (handle_char_store): Call get_string_cst.
8300 2018-01-03 Martin Liska <mliska@suse.cz>
8302 PR tree-optimization/83593
8303 * tree-ssa-strlen.c: Include tree-cfg.h.
8304 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8305 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8306 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8308 (strlen_dom_walker::before_dom_children): Call
8309 gimple_purge_dead_eh_edges. Dump tranformation with details
8311 (strlen_dom_walker::before_dom_children): Update call by adding
8312 new argument cleanup_eh.
8313 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8315 2018-01-03 Martin Liska <mliska@suse.cz>
8318 * cif-code.def (VARIADIC_THUNK): New enum value.
8319 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8322 2018-01-03 Jan Beulich <jbeulich@suse.com>
8324 * sse.md (mov<mode>_internal): Tighten condition for when to use
8325 vmovdqu<ssescalarsize> for TI and OI modes.
8327 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8329 Update copyright years.
8331 2018-01-03 Martin Liska <mliska@suse.cz>
8334 * ipa-visibility.c (function_and_variable_visibility): Skip
8335 functions with noipa attribure.
8337 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8339 * gcc.c (process_command): Update copyright notice dates.
8340 * gcov-dump.c (print_version): Ditto.
8341 * gcov.c (print_version): Ditto.
8342 * gcov-tool.c (print_version): Ditto.
8343 * gengtype.c (create_file): Ditto.
8344 * doc/cpp.texi: Bump @copying's copyright year.
8345 * doc/cppinternals.texi: Ditto.
8346 * doc/gcc.texi: Ditto.
8347 * doc/gccint.texi: Ditto.
8348 * doc/gcov.texi: Ditto.
8349 * doc/install.texi: Ditto.
8350 * doc/invoke.texi: Ditto.
8352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8354 * vector-builder.h (vector_builder::m_full_nelts): Change from
8355 unsigned int to poly_uint64.
8356 (vector_builder::full_nelts): Update prototype accordingly.
8357 (vector_builder::new_vector): Likewise.
8358 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8359 (vector_builder::operator ==): Likewise.
8360 (vector_builder::finalize): Likewise.
8361 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8362 Take the number of elements as a poly_uint64 rather than an
8364 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8365 from unsigned int to poly_uint64.
8366 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8367 (vec_perm_indices::new_vector): Likewise.
8368 (vec_perm_indices::length): Likewise.
8369 (vec_perm_indices::nelts_per_input): Likewise.
8370 (vec_perm_indices::input_nelts): Likewise.
8371 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8372 number of elements per input as a poly_uint64 rather than an
8373 unsigned int. Use the original encoding for variable-length
8374 vectors, rather than clamping each individual element.
8375 For the second and subsequent elements in each pattern,
8376 clamp the step and base before clamping their sum.
8377 (vec_perm_indices::series_p): Handle polynomial element counts.
8378 (vec_perm_indices::all_in_range_p): Likewise.
8379 (vec_perm_indices_to_tree): Likewise.
8380 (vec_perm_indices_to_rtx): Likewise.
8381 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8382 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8383 (tree_vector_builder::new_binary_operation): Handle polynomial
8384 element counts. Return false if we need to know the number
8385 of elements at compile time.
8386 * fold-const.c (fold_vec_perm): Punt if the number of elements
8387 isn't known at compile time.
8389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8391 * vec-perm-indices.h (vec_perm_builder): Change element type
8392 from HOST_WIDE_INT to poly_int64.
8393 (vec_perm_indices::element_type): Update accordingly.
8394 (vec_perm_indices::clamp): Handle polynomial element_types.
8395 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8396 (vec_perm_indices::all_in_range_p): Likewise.
8397 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8399 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8400 polynomial vec_perm_indices element types.
8401 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8402 * fold-const.c (fold_vec_perm): Likewise.
8403 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8404 * tree-vect-generic.c (lower_vec_perm): Likewise.
8405 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8406 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8407 element type to HOST_WIDE_INT.
8409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8410 Alan Hayward <alan.hayward@arm.com>
8411 David Sherwood <david.sherwood@arm.com>
8413 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8414 rather than an int. Use plus_constant.
8415 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8416 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8419 Alan Hayward <alan.hayward@arm.com>
8420 David Sherwood <david.sherwood@arm.com>
8422 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8423 a HOST_WIDE_INT to a poly_int64.
8425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8426 Alan Hayward <alan.hayward@arm.com>
8427 David Sherwood <david.sherwood@arm.com>
8429 * calls.c (load_register_parameters): Cope with polynomial
8430 mode sizes. Require a constant size for BLKmode parameters
8431 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8432 forces a parameter to be padded at the lsb end in order to
8433 fill a complete number of words, require the parameter size
8434 to be ordered wrt UNITS_PER_WORD.
8436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8437 Alan Hayward <alan.hayward@arm.com>
8438 David Sherwood <david.sherwood@arm.com>
8440 * reload1.c (spill_stack_slot_width): Change element type
8441 from unsigned int to poly_uint64_pod.
8442 (alter_reg): Treat mode sizes as polynomial.
8444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8445 Alan Hayward <alan.hayward@arm.com>
8446 David Sherwood <david.sherwood@arm.com>
8448 * reload.c (complex_word_subreg_p): New function.
8449 (reload_inner_reg_of_subreg, push_reload): Use it.
8451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8452 Alan Hayward <alan.hayward@arm.com>
8453 David Sherwood <david.sherwood@arm.com>
8455 * lra-constraints.c (process_alt_operands): Reject matched
8456 operands whose sizes aren't ordered.
8457 (match_reload): Refer to this check here.
8459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8460 Alan Hayward <alan.hayward@arm.com>
8461 David Sherwood <david.sherwood@arm.com>
8463 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8464 that the mode size is in the set {1, 2, 4, 8, 16}.
8466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8467 Alan Hayward <alan.hayward@arm.com>
8468 David Sherwood <david.sherwood@arm.com>
8470 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8471 Use plus_constant instead of gen_rtx_PLUS.
8473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8474 Alan Hayward <alan.hayward@arm.com>
8475 David Sherwood <david.sherwood@arm.com>
8477 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8478 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8479 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8480 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8481 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8482 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8483 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8484 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8485 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8486 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8488 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8489 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8490 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8491 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8492 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8493 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8494 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8495 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8496 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8497 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8499 * expr.c (emit_move_resolve_push): Treat the input and result
8500 of PUSH_ROUNDING as a poly_int64.
8501 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8502 (emit_push_insn): Likewise.
8503 * lra-eliminations.c (mark_not_eliminable): Likewise.
8504 * recog.c (push_operand): Likewise.
8505 * reload1.c (elimination_effects): Likewise.
8506 * rtlanal.c (nonzero_bits1): Likewise.
8507 * calls.c (store_one_arg): Likewise. Require the padding to be
8508 known at compile time.
8510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8511 Alan Hayward <alan.hayward@arm.com>
8512 David Sherwood <david.sherwood@arm.com>
8514 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8515 Use plus_constant instead of gen_rtx_PLUS.
8517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8518 Alan Hayward <alan.hayward@arm.com>
8519 David Sherwood <david.sherwood@arm.com>
8521 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8525 Alan Hayward <alan.hayward@arm.com>
8526 David Sherwood <david.sherwood@arm.com>
8528 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8529 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8530 via stack temporaries. Treat the mode size as polynomial too.
8532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8533 Alan Hayward <alan.hayward@arm.com>
8534 David Sherwood <david.sherwood@arm.com>
8536 * expr.c (expand_expr_real_2): When handling conversions involving
8537 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8538 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8539 as a poly_uint64 too.
8541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8542 Alan Hayward <alan.hayward@arm.com>
8543 David Sherwood <david.sherwood@arm.com>
8545 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8548 Alan Hayward <alan.hayward@arm.com>
8549 David Sherwood <david.sherwood@arm.com>
8551 * combine.c (can_change_dest_mode): Handle polynomial
8552 REGMODE_NATURAL_SIZE.
8553 * expmed.c (store_bit_field_1): Likewise.
8554 * expr.c (store_constructor): Likewise.
8555 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8556 and polynomial REGMODE_NATURAL_SIZE.
8557 (gen_lowpart_common): Likewise.
8558 * reginfo.c (record_subregs_of_mode): Likewise.
8559 * rtlanal.c (read_modify_subreg_p): Likewise.
8561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8562 Alan Hayward <alan.hayward@arm.com>
8563 David Sherwood <david.sherwood@arm.com>
8565 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8566 numbers of elements.
8568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8569 Alan Hayward <alan.hayward@arm.com>
8570 David Sherwood <david.sherwood@arm.com>
8572 * match.pd: Cope with polynomial numbers of vector elements.
8574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8575 Alan Hayward <alan.hayward@arm.com>
8576 David Sherwood <david.sherwood@arm.com>
8578 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8579 in a POINTER_PLUS_EXPR.
8581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8582 Alan Hayward <alan.hayward@arm.com>
8583 David Sherwood <david.sherwood@arm.com>
8585 * omp-simd-clone.c (simd_clone_subparts): New function.
8586 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8587 (ipa_simd_modify_function_body): Likewise.
8589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8590 Alan Hayward <alan.hayward@arm.com>
8591 David Sherwood <david.sherwood@arm.com>
8593 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8594 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8595 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8596 (expand_vector_condition, vector_element): Likewise.
8597 (subparts_gt): New function.
8598 (get_compute_type): Use subparts_gt.
8599 (count_type_subparts): Delete.
8600 (expand_vector_operations_1): Use subparts_gt instead of
8601 count_type_subparts.
8603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8604 Alan Hayward <alan.hayward@arm.com>
8605 David Sherwood <david.sherwood@arm.com>
8607 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8608 (vect_compile_time_alias): ...this new function. Do the calculation
8609 on poly_ints rather than trees.
8610 (vect_prune_runtime_alias_test_list): Update call accordingly.
8612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8613 Alan Hayward <alan.hayward@arm.com>
8614 David Sherwood <david.sherwood@arm.com>
8616 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8618 (vect_schedule_slp_instance): Likewise.
8620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8621 Alan Hayward <alan.hayward@arm.com>
8622 David Sherwood <david.sherwood@arm.com>
8624 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8625 constant and extern definitions for variable-length vectors.
8626 (vect_get_constant_vectors): Note that the number of units
8627 is known to be constant.
8629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8630 Alan Hayward <alan.hayward@arm.com>
8631 David Sherwood <david.sherwood@arm.com>
8633 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8634 of units as polynomial. Choose between WIDE and NARROW based
8637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8638 Alan Hayward <alan.hayward@arm.com>
8639 David Sherwood <david.sherwood@arm.com>
8641 * tree-vect-stmts.c (simd_clone_subparts): New function.
8642 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8645 Alan Hayward <alan.hayward@arm.com>
8646 David Sherwood <david.sherwood@arm.com>
8648 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8649 vectors as polynomial. Use build_index_vector for
8652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8653 Alan Hayward <alan.hayward@arm.com>
8654 David Sherwood <david.sherwood@arm.com>
8656 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8657 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8658 for variable-length vectors.
8659 (vectorizable_mask_load_store): Treat the number of units as
8660 polynomial, asserting that it is constant if the condition has
8661 already been enforced.
8662 (vectorizable_store, vectorizable_load): Likewise.
8664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8665 Alan Hayward <alan.hayward@arm.com>
8666 David Sherwood <david.sherwood@arm.com>
8668 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8669 of units as polynomial. Punt if we can't tell at compile time
8670 which vector contains the final result.
8672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8673 Alan Hayward <alan.hayward@arm.com>
8674 David Sherwood <david.sherwood@arm.com>
8676 * tree-vect-loop.c (vectorizable_induction): Treat the number
8677 of units as polynomial. Punt on SLP inductions. Use an integer
8678 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8679 cast of such a series for variable-length floating-point
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 * tree.h (build_index_vector): Declare.
8687 * tree.c (build_index_vector): New function.
8688 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8689 of units as polynomial, forcibly converting it to a constant if
8690 vectorizable_reduction has already enforced the condition.
8691 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8692 to create a {1,2,3,...} vector.
8693 (vectorizable_reduction): Treat the number of units as polynomial.
8694 Choose vectype_in based on the largest scalar element size rather
8695 than the smallest number of units. Enforce the restrictions
8698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8699 Alan Hayward <alan.hayward@arm.com>
8700 David Sherwood <david.sherwood@arm.com>
8702 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8703 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8710 * target.def (autovectorize_vector_sizes): Return the vector sizes
8711 by pointer, using vector_sizes rather than a bitmask.
8712 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8713 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8714 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8716 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8717 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8718 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8719 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8720 * omp-general.c (omp_max_vf): Likewise.
8721 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8722 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8723 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8724 * tree-vect-slp.c (vect_slp_bb): Likewise.
8725 * doc/tm.texi: Regenerate.
8726 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8728 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8729 the vector size as a poly_uint64 rather than an unsigned int.
8730 (current_vector_size): Change from an unsigned int to a poly_uint64.
8731 (get_vectype_for_scalar_type): Update accordingly.
8732 * tree.h (build_truth_vector_type): Take the size and number of
8733 units as a poly_uint64 rather than an unsigned int.
8734 (build_vector_type): Add a temporary overload that takes
8735 the number of units as a poly_uint64 rather than an unsigned int.
8736 * tree.c (make_vector_type): Likewise.
8737 (build_truth_vector_type): Take the number of units as a poly_uint64
8738 rather than an unsigned int.
8740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8741 Alan Hayward <alan.hayward@arm.com>
8742 David Sherwood <david.sherwood@arm.com>
8744 * target.def (get_mask_mode): Take the number of units and length
8745 as poly_uint64s rather than unsigned ints.
8746 * targhooks.h (default_get_mask_mode): Update accordingly.
8747 * targhooks.c (default_get_mask_mode): Likewise.
8748 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8749 * doc/tm.texi: Regenerate.
8751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8752 Alan Hayward <alan.hayward@arm.com>
8753 David Sherwood <david.sherwood@arm.com>
8755 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8756 * omp-general.c (omp_max_vf): Likewise.
8757 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8758 (expand_omp_simd): Handle polynomial safelen.
8759 * omp-low.c (omplow_simd_context): Add a default constructor.
8760 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8761 (lower_rec_simd_input_clauses): Update accordingly.
8762 (lower_rec_input_clauses): Likewise.
8764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8765 Alan Hayward <alan.hayward@arm.com>
8766 David Sherwood <david.sherwood@arm.com>
8768 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8769 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8770 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8771 (vect_analyze_slp_cost): Likewise.
8772 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8773 (vect_model_load_cost): Likewise.
8775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8776 Alan Hayward <alan.hayward@arm.com>
8777 David Sherwood <david.sherwood@arm.com>
8779 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8780 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8781 from an unsigned int * to a poly_uint64_pod *.
8782 (calculate_unrolling_factor): New function.
8783 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8786 Alan Hayward <alan.hayward@arm.com>
8787 David Sherwood <david.sherwood@arm.com>
8789 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8790 from an unsigned int to a poly_uint64.
8791 (_loop_vec_info::slp_unrolling_factor): Likewise.
8792 (_loop_vec_info::vectorization_factor): Change from an int
8794 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8795 (vect_get_num_vectors): New function.
8796 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8797 (vect_get_num_copies): Use vect_get_num_vectors.
8798 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8799 to an unsigned int *.
8800 (vect_analyze_data_refs): Change min_vf from an int * to a
8802 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8803 than an unsigned HOST_WIDE_INT.
8804 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8805 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8806 to an unsigned int *.
8807 (vect_analyze_data_ref_dependences): Likewise.
8808 (vect_compute_data_ref_alignment): Handle polynomial vf.
8809 (vect_enhance_data_refs_alignment): Likewise.
8810 (vect_prune_runtime_alias_test_list): Likewise.
8811 (vect_shift_permute_load_chain): Likewise.
8812 (vect_supportable_dr_alignment): Likewise.
8813 (dependence_distance_ge_vf): Take the vectorization factor as a
8814 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8815 (vect_analyze_data_refs): Change min_vf from an int * to a
8817 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8818 vfm1 as a poly_uint64 rather than an int. Make the same change
8819 for the returned bound_scalar.
8820 (vect_gen_vector_loop_niters): Handle polynomial vf.
8821 (vect_do_peeling): Likewise. Update call to
8822 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8823 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8825 * tree-vect-loop.c (vect_determine_vectorization_factor)
8826 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8827 (vect_get_known_peeling_cost): Likewise.
8828 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8829 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8830 (vect_transform_loop): Likewise. Use the lowest possible VF when
8831 updating the upper bounds of the loop.
8832 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8834 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8835 polynomial unroll factors.
8836 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8837 (vect_make_slp_decision): Likewise.
8838 (vect_supported_load_permutation_p): Likewise, and polynomial
8840 (vect_analyze_slp_cost): Handle polynomial vf.
8841 (vect_slp_analyze_node_operations): Likewise.
8842 (vect_slp_analyze_bb_1): Likewise.
8843 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8844 than an unsigned HOST_WIDE_INT.
8845 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8846 (vectorizable_load): Handle polynomial vf.
8847 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8849 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8852 Alan Hayward <alan.hayward@arm.com>
8853 David Sherwood <david.sherwood@arm.com>
8855 * match.pd: Handle bit operations involving three constants
8856 and try to fold one pair.
8858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8860 * tree-vect-loop-manip.c: Include gimple-fold.h.
8861 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8862 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8863 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8864 Add a path that uses a step of VF instead of 1, but disable it
8866 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8867 and niters_no_overflow parameters. Update calls to
8868 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8869 Create a new SSA name if the latter choses to use a ste other
8870 than zero, and return it via niters_vector_mult_vf_var.
8871 * tree-vect-loop.c (vect_transform_loop): Update calls to
8872 vect_do_peeling, vect_gen_vector_loop_niters and
8873 slpeel_make_loop_iterate_ntimes.
8874 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8875 (vect_gen_vector_loop_niters): Update declarations after above changes.
8877 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8879 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8880 128-bit round to integer instructions.
8881 (ceil<mode>2): Likewise.
8882 (btrunc<mode>2): Likewise.
8883 (round<mode>2): Likewise.
8885 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8887 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8888 unaligned VSX load/store on P8/P9.
8889 (expand_block_clear): Allow the use of unaligned VSX
8890 load/store on P8/P9.
8892 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8894 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8896 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8897 swap associated with both a load and a store.
8899 2018-01-02 Andrew Waterman <andrew@sifive.com>
8901 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8902 * config/riscv/riscv.md (clear_cache): Use it.
8904 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8906 * web.c: Remove out-of-date comment.
8908 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8910 * expr.c (fixup_args_size_notes): Check that any existing
8911 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8912 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8913 (emit_single_push_insn): ...here.
8915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8917 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8918 (const_vector_encoded_nelts): New function.
8919 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8920 (const_vector_int_elt, const_vector_elt): Declare.
8921 * emit-rtl.c (const_vector_int_elt_1): New function.
8922 (const_vector_elt): Likewise.
8923 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8924 of CONST_VECTOR_ELT.
8926 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8928 * expr.c: Include rtx-vector-builder.h.
8929 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8930 directly on the tree encoding.
8931 (const_vector_from_tree): Likewise.
8932 * optabs.c: Include rtx-vector-builder.h.
8933 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8934 sequence of "u" values.
8935 * vec-perm-indices.c: Include rtx-vector-builder.h.
8936 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8937 directly on the vec_perm_indices encoding.
8939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8941 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8942 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8943 * rtx-vector-builder.h: New file.
8944 * rtx-vector-builder.c: Likewise.
8945 * rtl.h (rtx_def::u2): Add a const_vector field.
8946 (CONST_VECTOR_NPATTERNS): New macro.
8947 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8948 (CONST_VECTOR_DUPLICATE_P): Likewise.
8949 (CONST_VECTOR_STEPPED_P): Likewise.
8950 (CONST_VECTOR_ENCODED_ELT): Likewise.
8951 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8952 (unwrap_const_vec_duplicate): Likewise.
8953 (const_vec_series_p): Check for a non-duplicated vector encoding.
8954 Say that the function only returns true for integer vectors.
8955 * emit-rtl.c: Include rtx-vector-builder.h.
8956 (gen_const_vec_duplicate_1): Delete.
8957 (gen_const_vector): Call gen_const_vec_duplicate instead of
8958 gen_const_vec_duplicate_1.
8959 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8960 (gen_const_vec_duplicate): Use rtx_vector_builder.
8961 (gen_const_vec_series): Likewise.
8962 (gen_rtx_CONST_VECTOR): Likewise.
8963 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8964 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8965 Build a new vector rather than modifying a CONST_VECTOR in-place.
8966 (handle_special_swappables): Update call accordingly.
8967 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8968 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8969 Build a new vector rather than modifying a CONST_VECTOR in-place.
8970 (handle_special_swappables): Update call accordingly.
8972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8974 * simplify-rtx.c (simplify_const_binary_operation): Use
8975 CONST_VECTOR_ELT instead of XVECEXP.
8977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8979 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8980 the selector elements to be different from the data elements
8981 if the selector is a VECTOR_CST.
8982 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8983 ssizetype for the selector.
8985 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8987 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8988 before testing each element individually.
8989 * tree-vect-generic.c (lower_vec_perm): Likewise.
8991 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8993 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8994 * selftest-run-tests.c (selftest::run_tests): Call it.
8995 * vector-builder.h (vector_builder::operator ==): New function.
8996 (vector_builder::operator !=): Likewise.
8997 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8998 (vec_perm_indices::all_from_input_p): New function.
8999 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9000 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9001 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9002 instead of reading the VECTOR_CST directly. Detect whether both
9003 vector inputs are the same before constructing the vec_perm_indices,
9004 and update the number of inputs argument accordingly. Use the
9005 utility functions added above. Only construct sel2 if we need to.
9007 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9009 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9010 the broadcast of the low byte.
9011 (expand_mult_highpart): Use an explicit encoding for the permutes.
9012 * optabs-query.c (can_mult_highpart_p): Likewise.
9013 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9014 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9015 (vectorizable_bswap): Likewise.
9016 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9017 explicit encoding for the power-of-2 permutes.
9018 (vect_permute_store_chain): Likewise.
9019 (vect_grouped_load_supported): Likewise.
9020 (vect_permute_load_chain): Likewise.
9022 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9024 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9025 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9026 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9027 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9028 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9029 (vect_gen_perm_mask_any): Likewise.
9031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9033 * int-vector-builder.h: New file.
9034 * vec-perm-indices.h: Include int-vector-builder.h.
9035 (vec_perm_indices): Redefine as an int_vector_builder.
9036 (auto_vec_perm_indices): Delete.
9037 (vec_perm_builder): Redefine as a stand-alone class.
9038 (vec_perm_indices::vec_perm_indices): New function.
9039 (vec_perm_indices::clamp): Likewise.
9040 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9041 (vec_perm_indices::new_vector): New function.
9042 (vec_perm_indices::new_expanded_vector): Update for new
9043 vec_perm_indices class.
9044 (vec_perm_indices::rotate_inputs): New function.
9045 (vec_perm_indices::all_in_range_p): Operate directly on the
9046 encoded form, without computing elided elements.
9047 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9048 encoding. Update for new vec_perm_indices class.
9049 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9050 the given vec_perm_builder.
9051 (expand_vec_perm_var): Update vec_perm_builder constructor.
9052 (expand_mult_highpart): Use vec_perm_builder instead of
9053 auto_vec_perm_indices.
9054 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9055 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9056 or double series encoding as appropriate.
9057 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9058 vec_perm_indices instead of auto_vec_perm_indices.
9059 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9060 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9061 (vect_permute_store_chain): Likewise.
9062 (vect_grouped_load_supported): Likewise.
9063 (vect_permute_load_chain): Likewise.
9064 (vect_shift_permute_load_chain): Likewise.
9065 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9066 (vect_transform_slp_perm_load): Likewise.
9067 (vect_schedule_slp_instance): Likewise.
9068 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9069 (vectorizable_mask_load_store): Likewise.
9070 (vectorizable_bswap): Likewise.
9071 (vectorizable_store): Likewise.
9072 (vectorizable_load): Likewise.
9073 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9074 vec_perm_indices instead of auto_vec_perm_indices. Use
9075 tree_to_vec_perm_builder to read the vector from a tree.
9076 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9077 vec_perm_builder instead of a vec_perm_indices.
9078 (have_whole_vector_shift): Use vec_perm_builder and
9079 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9080 truncation to calc_vec_perm_mask_for_shift.
9081 (vect_create_epilog_for_reduction): Likewise.
9082 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9083 from auto_vec_perm_indices to vec_perm_indices.
9084 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9085 instead of changing individual elements.
9086 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9087 the vector in d.perm.
9088 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9089 from auto_vec_perm_indices to vec_perm_indices.
9090 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9091 instead of changing individual elements.
9092 (arm_vectorize_vec_perm_const): Use new_vector to install
9093 the vector in d.perm.
9094 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9095 Update vec_perm_builder constructor.
9096 (rs6000_expand_interleave): Likewise.
9097 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9098 (rs6000_expand_interleave): Likewise.
9100 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9102 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9103 to qimode could truncate the indices.
9104 * optabs.c (expand_vec_perm_var): Likewise.
9106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9108 * Makefile.in (OBJS): Add vec-perm-indices.o.
9109 * vec-perm-indices.h: New file.
9110 * vec-perm-indices.c: Likewise.
9111 * target.h (vec_perm_indices): Replace with a forward class
9113 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9114 * optabs.h: Include vec-perm-indices.h.
9115 (expand_vec_perm): Delete.
9116 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9117 (expand_vec_perm_const): Declare.
9118 * target.def (vec_perm_const_ok): Replace with...
9119 (vec_perm_const): ...this new hook.
9120 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9121 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9122 * doc/tm.texi: Regenerate.
9123 * optabs.def (vec_perm_const): Delete.
9124 * doc/md.texi (vec_perm_const): Likewise.
9125 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9126 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9127 expand_vec_perm for constant permutation vectors. Assert that
9128 the mode of variable permutation vectors is the integer equivalent
9129 of the mode that is being permuted.
9130 * optabs-query.h (selector_fits_mode_p): Declare.
9131 * optabs-query.c: Include vec-perm-indices.h.
9132 (selector_fits_mode_p): New function.
9133 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9134 is defined, instead of checking whether the vec_perm_const_optab
9135 exists. Use targetm.vectorize.vec_perm_const instead of
9136 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9137 fit in the vector mode before using a variable permute.
9138 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9139 vec_perm_indices instead of an rtx.
9140 (expand_vec_perm): Replace with...
9141 (expand_vec_perm_const): ...this new function. Take the selector
9142 as a vec_perm_indices rather than an rtx. Also take the mode of
9143 the selector. Update call to shift_amt_for_vec_perm_mask.
9144 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9145 Use vec_perm_indices::new_expanded_vector to expand the original
9146 selector into bytes. Check whether the indices fit in the vector
9147 mode before using a variable permute.
9148 (expand_vec_perm_var): Make global.
9149 (expand_mult_highpart): Use expand_vec_perm_const.
9150 * fold-const.c: Includes vec-perm-indices.h.
9151 * tree-ssa-forwprop.c: Likewise.
9152 * tree-vect-data-refs.c: Likewise.
9153 * tree-vect-generic.c: Likewise.
9154 * tree-vect-loop.c: Likewise.
9155 * tree-vect-slp.c: Likewise.
9156 * tree-vect-stmts.c: Likewise.
9157 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9159 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9160 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9161 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9162 (aarch64_vectorize_vec_perm_const): ...this new function.
9163 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9164 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9165 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9166 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9167 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9168 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9169 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9171 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9172 check for NEON modes.
9173 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9174 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9175 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9176 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9178 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9179 the old VEC_PERM_CONST conditions.
9180 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9181 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9182 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9183 (ia64_vectorize_vec_perm_const_ok): Merge into...
9184 (ia64_vectorize_vec_perm_const): ...this new function.
9185 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9186 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9187 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9188 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9189 * config/mips/mips.c (mips_expand_vec_perm_const)
9190 (mips_vectorize_vec_perm_const_ok): Merge into...
9191 (mips_vectorize_vec_perm_const): ...this new function.
9192 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9193 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9194 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9195 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9196 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9197 (rs6000_expand_vec_perm_const): Delete.
9198 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9200 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9201 (altivec_expand_vec_perm_const_le): Take each operand individually.
9202 Operate on constant selectors rather than rtxes.
9203 (altivec_expand_vec_perm_const): Likewise. Update call to
9204 altivec_expand_vec_perm_const_le.
9205 (rs6000_expand_vec_perm_const): Delete.
9206 (rs6000_vectorize_vec_perm_const_ok): Delete.
9207 (rs6000_vectorize_vec_perm_const): New function.
9208 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9209 an element count and rtx array.
9210 (rs6000_expand_extract_even): Update call accordingly.
9211 (rs6000_expand_interleave): Likewise.
9212 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9213 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9214 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9215 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9216 (rs6000_expand_vec_perm_const): Delete.
9217 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9218 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9219 (altivec_expand_vec_perm_const_le): Take each operand individually.
9220 Operate on constant selectors rather than rtxes.
9221 (altivec_expand_vec_perm_const): Likewise. Update call to
9222 altivec_expand_vec_perm_const_le.
9223 (rs6000_expand_vec_perm_const): Delete.
9224 (rs6000_vectorize_vec_perm_const_ok): Delete.
9225 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9226 reference to the SPE evmerge intructions.
9227 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9228 an element count and rtx array.
9229 (rs6000_expand_extract_even): Update call accordingly.
9230 (rs6000_expand_interleave): Likewise.
9231 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9232 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9234 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9236 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9238 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9239 vector mode and that that mode matches the mode of the data
9241 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9242 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9243 directly using expand_vec_perm_1 when forcing selectors into
9245 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9247 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9249 * optabs-query.h (can_vec_perm_p): Delete.
9250 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9251 * optabs-query.c (can_vec_perm_p): Split into...
9252 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9253 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9254 particular selector is valid.
9255 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9256 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9257 (vect_grouped_load_supported): Likewise.
9258 (vect_shift_permute_load_chain): Likewise.
9259 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9260 (vect_transform_slp_perm_load): Likewise.
9261 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9262 (vectorizable_bswap): Likewise.
9263 (vect_gen_perm_mask_checked): Likewise.
9264 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9265 implementations of variable permutation vectors into account
9266 when deciding which selector to use.
9267 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9268 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9269 with a false third argument.
9270 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9271 to test whether the constant selector is valid and can_vec_perm_var_p
9272 to test whether a variable selector is valid.
9274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9276 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9277 * optabs-query.c (can_vec_perm_p): Likewise.
9278 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9279 instead of vec_perm_indices.
9280 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9281 (vect_gen_perm_mask_checked): Likewise,
9282 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9283 (vect_gen_perm_mask_checked): Likewise,
9285 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9287 * optabs-query.h (qimode_for_vec_perm): Declare.
9288 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9289 (qimode_for_vec_perm): ...this new function.
9290 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9292 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9294 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9295 does not have a conditional at the top.
9297 2018-01-02 Richard Biener <rguenther@suse.de>
9299 * ipa-inline.c (big_speedup_p): Fix expression.
9301 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9304 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9307 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9311 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9312 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9313 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9314 cond_taken_branch_cost 3->4.
9316 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9318 PR tree-optimization/83581
9319 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9320 TODO_cleanup_cfg if any changes have been made.
9323 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9324 convert_modes if target mode has the right side, but different mode
9328 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9329 last argument when extracting from CONCAT. If either from_real or
9330 from_imag is NULL, use expansion through memory. If result is not
9331 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9332 the parts directly to inner mode, if even that fails, use expansion
9336 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9337 check for bswap in mode rather than HImode and use that in expand_unop
9340 Copyright (C) 2018 Free Software Foundation, Inc.
9342 Copying and distribu