1 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/84960
4 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5 if it is ENTRY block, move them into single succ of ENTRY in that case.
7 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9 PR tree-optimization/84811
10 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
12 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
13 (known_subrange_p): Update accordingly. Cast each value involved
14 in the size comparison, rather than casting the result of the
17 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19 PR tree-optimization/84982
20 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
21 by flipping the least significant bit rather than all bits from
22 bitpos to bitpos + bitsize - 1.
24 2018-03-21 Nathan Sidwell <nathan@acm.org>
26 * doc/extend.texi (Deprecated Features): Remove mention of
27 long-deleted deprecations.
29 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
32 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
33 * configure: Regenerate.
35 2018-03-21 Tom de Vries <tom@codesourcery.com>
37 PR tree-optimization/83126
38 * tree-parloops.c (num_phis): New function.
39 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
41 2018-03-21 Nathan Sidwell <nathan@acm.org>
43 * doc/extend.texi (Deprecated Features): Update deprecated flags,
44 mention anon-struct/union members and trailing attributes.
46 2018-03-21 Bin Cheng <bin.cheng@arm.com>
48 PR tree-optimization/84969
49 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
50 builtin memset partitions if they set different rhs values.
52 2018-03-21 Jakub Jelinek <jakub@redhat.com>
54 PR rtl-optimization/84989
55 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
56 VEC_DUPLICATE with scalar result mode.
58 2018-03-21 Martin Liska <mliska@suse.cz>
61 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
62 not intended return statement.
64 2018-03-21 Martin Liska <mliska@suse.cz>
67 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
68 (chkp_find_bound_slots_1): Limit number of iterations.
70 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
73 * Minor grammar fixes for x86 options.
75 2018-03-20 Jakub Jelinek <jakub@redhat.com>
78 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
79 holding REG_CFA_RESTORE notes, instead turn them into a USE.
81 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
84 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
85 (altivec_lvx_<mode>_1op): Likewise.
86 (altivec_stvx_<mode>_2op): Likewise.
87 (altivec_stvx_<mode>_1op): Likewise.
88 (altivec_lvx_<VM2:mode>): New define_expand.
89 (altivec_stvx_<VM2:mode>): Likewise.
90 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
91 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
92 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
93 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
94 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
95 (rs6000_gen_lvx): Likewise.
96 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
97 (altivec_expand_stv_builtin): Likewise.
98 (altivec_expand_builtin): Likewise.
99 * config/rs6000/vector.md: Likewise.
101 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
104 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
107 2018-03-20 Richard Biener <rguenther@suse.de>
110 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
111 sign-conversions as zero, fall back to standard scalar_stmt
114 2018-03-20 Martin Liska <mliska@suse.cz>
117 * predict.c (rebuild_frequencies): Handle case when we have
118 PROFILE_ABSENT, but flag_guess_branch_prob is false.
120 2018-03-20 Jakub Jelinek <jakub@redhat.com>
123 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
124 flag_section_anchors.
125 * varasm.c (use_blocks_for_decl_p): Remove hack for
129 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
131 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
132 be created, use lowpart_subreg of operands[0] rather than operands[0]
134 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
135 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
136 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
137 and n constraint instead of aarch64_shift_imm_di and Usd.
138 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
139 (*aarch64_<optab>_reg_minus<mode>3): ... this.
141 2018-03-20 Sudakshina Das <sudi.das@arm.com>
144 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
145 to favor GPR over NEON registers.
146 (<shift>di3_neon): Likewise.
148 2018-03-20 Tom de Vries <tom@codesourcery.com>
151 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
152 (nvptx_process_pars): Emit bar.sync asap and alap.
154 2018-03-20 Tom de Vries <tom@codesourcery.com>
157 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
158 seen_label if seen_label is already set.
160 2018-03-20 Jakub Jelinek <jakub@redhat.com>
163 * config/i386/i386.c (fold_builtin_cpu): For features above 31
164 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
165 Use 1U instead of 1. Formatting fixes.
168 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
169 instead of TREE_TYPE (s1) for the return value.
171 2018-03-19 Jakub Jelinek <jakub@redhat.com>
173 PR tree-optimization/84946
174 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
175 bitsize + bitsize in poly_uint64 rather than poly_int64.
178 * dwarf2asm.c: Include fold-const.c.
179 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
180 of decl rather than decl itself.
182 PR rtl-optimization/84643
183 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
185 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
188 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
189 calling assemble_variable.
191 2018-03-19 Sudakshina Das <sudi.das@arm.com>
194 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
195 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
197 2018-03-19 Jim Wilson <jimw@sifive.com>
200 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
201 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
202 (riscv_first_stack_step): Likewise.
203 (riscv_option_override): Use STACK_BOUNDARY instead of
205 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
207 (BIGGEST_ALIGNMENT): Set to 128.
208 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
209 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
212 2018-03-19 Richard Biener <rguenther@suse.de>
214 PR tree-optimization/84933
215 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
216 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
218 2018-03-19 Richard Biener <rguenther@suse.de>
220 PR tree-optimization/84859
221 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
222 (cond_if_else_store_replacement): Perform sinking operation on
223 single-store BBs regardless of MAX_STORES_TO_SINK setting.
224 Generalize what a BB with a single eligible store is.
226 2018-03-19 Richard Biener <rguenther@suse.de>
228 PR tree-optimization/84929
229 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
230 chrec_is_positive against non-chrec arg.
232 2018-03-19 Tamar Christina <tamar.christina@arm.com>
235 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
237 2018-03-18 Martin Liska <mliska@suse.cz>
239 PR rtl-optimization/84635
240 * regrename.c (build_def_use): Use matches_mode only when
243 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
245 PR tree-optimization/84913
246 * tree-vect-loop.c (vectorizable_reduction): Don't try to
247 vectorize chains of COND_EXPRs.
249 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
251 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
253 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
255 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
257 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
259 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
261 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
262 Kito Cheng <kito.cheng@gmail.com>
264 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
265 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
266 (nds32_adjust_reg_alloc_order): New function.
267 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
269 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
271 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
272 nds32_print_operand, nds32_print_operand_address): Use
273 HOST_WIDE_INT_PRINT_DEC instead.
275 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
277 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
279 2018-03-17 Jakub Jelinek <jakub@redhat.com>
282 * config/i386/i386.c (initial_ix86_tune_features,
283 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
285 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
286 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
287 rather than 1u << ix86_tune. Formatting fix.
288 (ix86_option_override_internal): Change ix86_arch_mask from
289 unsigned int to unsigned HOST_WIDE_INT, initialize to
290 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
291 (ix86_function_specific_restore): Likewise.
293 2018-03-16 Jakub Jelinek <jakub@redhat.com>
296 * postreload.c (reload_combine_recognize_pattern): Perform
297 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
298 truncate_int_for_mode the result for the destination's mode.
301 * hsa-gen.c (mem_type_for_type): Fix comment typo.
302 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
304 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
307 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
310 * lra-assigns.c (lra_split_hard_reg_for): Don't use
311 regno_allocno_class_array and sorted_pseudos.
312 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
313 insns where regno is used.
315 2018-03-16 Martin Liska <mliska@suse.cz>
318 * multiple_target.c (create_dispatcher_calls): Redirect
319 reference in the symbol table.
321 2018-03-16 Martin Liska <mliska@suse.cz>
324 * multiple_target.c (create_dispatcher_calls): Redirect also
327 2018-03-16 Jakub Jelinek <jakub@redhat.com>
331 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
332 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
333 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
335 2018-03-16 Julia Koval <julia.koval@intel.com>
337 * doc/invoke.texi (Skylake Server): Add CLWB.
338 Cannonlake): Remove CLWB.
340 2018-03-16 Jakub Jelinek <jakub@redhat.com>
342 PR tree-optimization/84841
343 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
345 (FLOAT_ONE_CONST_TYPE): Define.
346 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
347 (sort_by_operand_rank): Put entries with higher constant_type last
348 rather than first to match comments.
350 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
352 * config/nios2/nios2.md (movsi_internal): Fix thinko in
355 2018-03-15 Jakub Jelinek <jakub@redhat.com>
358 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
359 check and use address of target always.
361 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
364 * config/i386/i386.c (indirect_thunk_needed): Update comments.
365 (indirect_thunk_bnd_needed): Likewise.
366 (indirect_thunks_used): Likewise.
367 (indirect_thunks_bnd_used): Likewise.
368 (indirect_return_needed): New.
369 (indirect_return_bnd_needed): Likewise.
370 (output_indirect_thunk_function): Add a bool argument for
372 (output_indirect_thunk_function): Don't generate alias for
373 function return thunk.
374 (ix86_code_end): Call output_indirect_thunk_function to generate
375 function return thunks.
376 (ix86_output_function_return): Set indirect_return_bnd_needed
377 and indirect_return_needed instead of indirect_thunk_bnd_needed
378 and indirect_thunk_needed.
380 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
382 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
383 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
384 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
386 2018-03-15 David Malcolm <dmalcolm@redhat.com>
387 Paul Hua <paul.hua.gm@gmail.com>
390 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
392 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
394 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
395 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
398 2018-03-15 Tamar Christina <tamar.christina@arm.com>
401 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
402 instead of GET_MODE_SIZE when comparing Units.
404 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
407 * varasm.c (hash_section): Return an unchangeble hash value
408 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
409 Return !aarch64_can_use_per_function_literal_pools_p ().
411 2018-03-15 Jakub Jelinek <jakub@redhat.com>
414 * optabs.c (emit_conditional_move): Pass address of cmode's copy
415 rather than address of cmode as last argument to prepare_cmp_insn.
417 2018-03-15 Julia Koval <julia.koval@intel.com>
419 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
420 F_AVX512VNNI, F_AVX512BITALG): New.
422 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
425 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
426 insn for floating-point loads and stores.
428 2018-03-14 Carl Love <cel@us.ibm.com>
430 * config/rs6000/rs6000-c.c: Add macro definitions for
431 ALTIVEC_BUILTIN_VEC_PERMXOR.
432 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
433 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
434 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
435 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
437 * config/doc/extend.texi: Add prototypes for vec_permxor.
439 2018-03-14 David Malcolm <dmalcolm@redhat.com>
442 * diagnostic-show-locus.c (class layout_point): Convert m_line
443 from int to linenum_type.
444 (line_span::comparator): Use linenum "compare" function when
445 comparing line numbers.
446 (test_line_span): New function.
447 (layout_range::contains_point): Convert param "row" from int to
449 (layout_range::intersects_line_p): Likewise.
450 (layout::will_show_line_p): Likewise.
451 (layout::print_source_line): Likewise.
452 (layout::should_print_annotation_line_p): Likewise.
453 (layout::print_annotation_line): Likewise.
454 (layout::print_leading_fixits): Likewise.
455 (layout::annotation_line_showed_range_p): Likewise.
456 (struct line_corrections): Likewise for field m_row.
457 (line_corrections::line_corrections): Likewise for param "row".
458 (layout::print_trailing_fixits): Likewise.
459 (layout::get_state_at_point): Likewise.
460 (layout::get_x_bound_for_row): Likewise.
461 (layout::print_line): Likewise.
462 (diagnostic_show_locus): Likewise for locals "last_line" and
464 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
465 * input.c (selftest::test_linenum_comparisons): New function.
466 (selftest::input_c_tests): Call it.
467 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
468 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
469 * selftest.h (ASSERT_GT): New macro.
470 (ASSERT_GT_AT): New macro.
471 (ASSERT_LT): New macro.
472 (ASSERT_LT_AT): New macro.
474 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
476 PR rtl-optimization/84780
477 * combine.c (distribute_links): Don't make a link based on pc_rtx.
479 2018-03-14 Martin Liska <mliska@suse.cz>
481 * tree.c (record_node_allocation_statistics): Use
483 (get_stats_node_kind): New function extracted from
484 record_node_allocation_statistics.
485 (free_node): Use get_stats_node_kind.
487 2018-03-14 Richard Biener <rguenther@suse.de>
489 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
490 that the value-set of ANTIC_IN doesn't grow.
493 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
495 (BB_VISITED_WITH_VISITED_SUCCS): New define.
496 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
498 2018-03-14 Julia Koval <julia.koval@intel.com>
500 * config.gcc (icelake-client, icelake-server): New.
502 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
503 (initial_ix86_arch_features): Ditto.
504 (PTA_SKYLAKE): Add SGX.
505 (PTA_ICELAKE): Remove.
506 (PTA_ICELAKE_CLIENT): New.
507 (PTA_ICELAKE_SERVER): New.
508 (ix86_option_override_internal): Split up icelake on icelake client and
510 (get_builtin_code_for_version): Ditto.
511 (fold_builtin_cpu): Ditto.
512 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
513 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
514 * config/i386/i386.h (processor_type): Ditto.
515 * doc/invoke.texi: Ditto.
517 2018-03-14 Jakub Jelinek <jakub@redhat.com>
520 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
521 INTEGER_CST offset, add it together with bitpos / 8 and
522 sign extend based on POINTER_SIZE.
526 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
529 * config/i386/constraints.md (Yc): New register constraint.
530 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
531 Use Yc constraint for alternative 2 of operand 0. Remove
532 preferred_for_speed attribute.
534 2018-03-14 Richard Biener <rguenther@suse.de>
536 PR tree-optimization/84830
537 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
538 with the old one to avoid oscillations.
540 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
543 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
545 (assign_by_spills): Return a flag of reload assignment failure.
546 Do not process the reload assignment failures. Do not spill other
547 reload pseudos if they has the same reg class. Update n if
549 (lra_assign): Add a return arg. Set up from the result of
550 assign_by_spills call.
551 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
552 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
553 usage_insns if it is not NULL.
554 (spill_hard_reg_in_range): New function.
555 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
556 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
558 (lra_assign): Change prototype.
559 * lra.c (lra): Add code to deal with fails by splitting hard reg
562 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
564 * config/riscv/riscv.opt (mrelax): New option.
565 * config/riscv/riscv.c (riscv_file_start): Emit ".option
566 "norelax" when riscv_mrelax is disabled.
567 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
569 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
572 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
573 reassociation for int modes.
575 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
577 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
578 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
580 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
581 * config/aarch64/aarch64-sve.md
582 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
583 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
584 (*extend<mode><Vwide>2): Rename to...
585 (aarch64_sve_extend<mode><Vwide>2): ...this.
586 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
587 renaming the old pattern to...
588 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
590 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
591 define_expand, renaming the old pattern to...
592 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
593 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
594 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
595 account when deciding which SVE instruction the optab should use.
596 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
598 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
600 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
601 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
602 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
603 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
604 (tlsdesc_small_<mode>): Turn a define_expand and use
605 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
606 (tlsdesc_small_advsimd_<mode>): ...this.
607 (tlsdesc_small_sve_<mode>): New pattern.
609 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
611 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
612 (UNSPEC_UMUL_HIGHPART): New constants.
613 (MUL_HIGHPART): New int iteraor.
614 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
615 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
617 (*<su>mul<mode>3_highpart): New define_insn.
619 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
622 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
625 2018-03-13 Martin Liska <mliska@suse.cz>
628 * (sem_item_optimizer::sem_item_optimizer): Initialize new
630 (sem_item_optimizer::~sem_item_optimizer): Release it.
631 (sem_item_optimizer::merge_classes): Register variable aliases.
632 (sem_item_optimizer::fixup_pt_set): New function.
633 (sem_item_optimizer::fixup_points_to_sets): Likewise.
634 * ipa-icf.h: Declare new variables and functions.
636 2018-03-13 Jakub Jelinek <jakub@redhat.com>
639 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
640 integer_pow2p@2 and test integer_pow2p in condition.
641 (A < 0 ? C : 0): Similarly for @1.
644 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
645 characters starting at p contain '\0' character, don't look beyond
649 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
650 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
653 * reg-stack.c (change_stack): Change update_end var from int to
654 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
655 also call set_block_for_insn on the newly added insns and rescan.
658 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
662 * tree.c (stabilize_reference_1): Return save_expr (e) for
663 STATEMENT_LIST even if it doesn't have side-effects.
665 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
667 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
669 2018-03-12 Renlin Li <renlin.li@arm.com>
671 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
672 aarch64_output_scalar_simd_mov_immediate.
674 2018-03-12 Martin Sebor <msebor@redhat.com>
676 PR tree-optimization/83456
677 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
678 for perfectly overlapping calls to memcpy.
679 (gimple_fold_builtin_memory_chk): Same.
680 (gimple_fold_builtin_strcpy): Handle no-warning.
681 (gimple_fold_builtin_stxcpy_chk): Same.
682 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
684 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
686 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
687 parameter. Use it for SFmode.
688 (rs6000_function_arg_advance_1): Adjust.
689 (rs6000_function_arg): Adjust.
690 (rs6000_gimplify_va_arg): Pass false for that new parameter.
692 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
694 PR rtl-optimization/84169
695 PR rtl-optimization/84780
696 * combine.c (can_combine_p): Check for a 2-insn combination whether
697 the destination register is used between the two insns, too.
699 2018-03-12 Richard Biener <rguenther@suse.de>
701 PR tree-optimization/84803
702 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
703 for refs DR analysis didn't process.
705 2018-03-12 Richard Biener <rguenther@suse.de>
707 PR tree-optimization/84777
708 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
709 force-vectorize loops ignore whether we are optimizing for size.
711 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
713 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
714 (TARGET_MD_ASM_ADJUST): Define.
716 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
717 Kito Cheng <kito.cheng@gmail.com>
718 Chung-Ju Wu <jasonwucj@gmail.com>
720 * config/nds32/nds32.c (nds32_compute_stack_frame,
721 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
722 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
723 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
724 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
725 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
726 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
727 * config/nds32/nds32.md (prologue, epilogue): Use macro
728 NDS32_V3PUSH_AVAILABLE_P to do checking.
730 2018-03-11 Jakub Jelinek <jakub@redhat.com>
733 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
734 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
735 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
736 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
737 addition of most attributes on !orig_type_die or the attribute not
738 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
740 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
741 Chung-Ju Wu <jasonwucj@gmail.com>
743 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
745 * config/nds32/nds32.opt (mvh): New option.
747 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
748 Chung-Ju Wu <jasonwucj@gmail.com>
750 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
752 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
753 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
756 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
757 Chung-Ju Wu <jasonwucj@gmail.com>
759 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
761 * config/nds32/nds32-multiple.md (strlensi): New pattern.
762 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
764 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
765 Kito Cheng <kito.cheng@gmail.com>
766 Chung-Ju Wu <jasonwucj@gmail.com>
768 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
769 UNSPEC_FFMISM and UNSPEC_FLMISM.
770 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
771 for ffb, ffmism and flmism.
772 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
773 (unspec_ffmism): Ditto.
774 (unspec_flmism): Ditto.
775 (nds32_expand_builtin_impl): Check if string extension is available.
776 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
777 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
779 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
782 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
785 * lra-assigns.c (assign_by_spills): Return a flag of reload
786 assignment failure. Do not process the reload assignment
787 failures. Do not spill other reload pseudos if they has the same
789 (lra_assign): Add a return arg. Set up from the result of
790 assign_by_spills call.
791 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
792 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
793 usage_insns if it is not NULL.
794 (spill_hard_reg_in_range): New function.
795 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
796 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
798 (lra_assign): Change prototype.
799 * lra.c (lra): Add code to deal with fails by splitting hard reg
802 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
805 * config/i386/i386.opt: Replace Enforcment with Enforcement.
807 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
810 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
811 (dw_val_node): Add val_symbolic_view.
812 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
813 (symview_upper_bound): New.
814 (new_line_info_table): Initialize symviews_since_reset.
815 (dwarf2out_source_line): Count symviews_since_reset and set
817 (dw_val_equal_p): Handle symview.
818 (add_AT_symview): New.
819 (print_dw_val): Handle symview.
820 (attr_checksum, attr_checksum_ordered): Likewise.
821 (same_dw_val_p, size_of_die): Likewise.
822 (value_format, output_die): Likewise.
823 (add_high_low_attributes): Use add_AT_symview for entry_view.
824 (dwarf2out_finish): Reset symview_upper_bound, clear
827 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
830 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
831 Add strict argument and use it.
832 (rs6000_split_multireg_move): Update for new strict argument.
833 (mem_operand_gpr): Disallow all non-offsettable addresses.
834 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
836 2018-03-09 Jakub Jelinek <jakub@redhat.com>
839 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
840 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
841 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
844 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
845 decl, use remap_type if we want to use the type.
847 2018-03-09 Martin Sebor <msebor@redhat.com>
849 PR tree-optimization/84526
850 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
852 (builtin_access::generic_overlap): Be prepared to handle non-array
855 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
857 PR rtl-optimization/84682
858 * lra-constraints.c (process_address_1): Check is_address flag
859 for address constraints.
860 (process_alt_operands): Likewise.
861 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
862 preprocess_constraints.
863 * recog.h (preprocess_constraints): Add oploc parameter.
865 * recog.c (preprocess_constraints): Test address_operand for
866 CT_ADDRESS constraints.
868 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
871 * lra-assigns.c (assign_by_spills): Return a flag of reload
872 assignment failure. Do not process the reload assignment
873 failures. Do not spill other reload pseudos if they has the same
875 (lra_assign): Add a return arg. Set up from the result of
876 assign_by_spills call.
877 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
878 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
879 usage_insns if it is not NULL.
880 (spill_hard_reg_in_range): New function.
881 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
882 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
884 (lra_assign): Change prototype.
885 * lra.c (lra): Add code to deal with fails by splitting hard reg
888 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
891 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
892 Accept complain bool parameter. Only emit errors if it is true.
893 (arm_parse_cpu_option_name): Likewise.
894 (arm_target_thumb_only): Adjust callers of the above.
895 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
896 prototype to take a default true bool parameter.
897 (arm_parse_arch_option_name): Likewise.
899 2018-03-09 David Malcolm <dmalcolm@redhat.com>
900 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
904 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
905 * configure: Regenerate.
906 * configure.ac ("linker --version-script option"): New.
907 ("linker soname option"): New.
909 2018-03-09 Richard Biener <rguenther@suse.de>
911 PR tree-optimization/84775
912 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
913 immediate uses of predicate stmts and mark them modified.
916 PR tree-optimization/84178
917 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
919 (version_loop_for_if_conversion): Delay update_ssa call.
920 (tree_if_conversion): Delay update_ssa until after predicate
923 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
926 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
927 when the function accesses prior frames.
929 2018-03-08 Jakub Jelinek <jakub@redhat.com>
932 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
933 gen_llsym, otherwise call maybe_gen_llsym.
936 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
937 has ',' character inside of it.
939 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
942 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
943 as clobbering CC_REGNUM.
945 2018-03-08 Richard Biener <rguenther@suse.de>
948 * tree-scalar-evolution.c: Include tree-into-ssa.h.
949 (follow_copies_to_constant): Do not follow SSA names registered
952 2018-03-08 Richard Biener <rguenther@suse.de>
954 PR tree-optimization/84178
955 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
957 (version_loop_for_if_conversion): Delay update_ssa call.
958 (tree_if_conversion): Delay update_ssa until after predicate
961 2018-03-08 David Malcolm <dmalcolm@redhat.com>
963 PR tree-optimization/84178
964 * tree-if-conv.c (release_bb_predicate): Remove the
965 the assertion that the stmts have NULL use_ops.
966 Discard the statements, asserting that they haven't
967 yet been added to a BB.
969 2018-03-08 Richard Biener <rguenther@suse.de>
971 PR tree-optimization/84746
972 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
973 (phi_translate): Pass in destination ANTIC_OUT set.
974 (phi_translate_1): Likewise. For a simplified result lookup
975 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
976 (phi_translate_set): Adjust.
977 (do_pre_regular_insertion): Likewise.
978 (do_pre_partial_partial_insertion): Likewise.
980 2018-03-08 Martin Liska <mliska@suse.cz>
982 PR gcov-profile/84735
983 * doc/gcov.texi: Document usage of profile files.
984 * gcov-io.h: Document changes in the format.
986 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
990 * dwarf2out.c (struct dw_line_info_table): Update comments for
992 (FORCE_RESET_NEXT_VIEW): New.
993 (FORCE_RESETTING_VIEW_P): New.
994 (RESETTING_VIEW_P): Check for -1 too.
995 (ZERO_VIEW_P): Likewise.
996 (new_line_info_table): Force-reset next view.
997 (dwarf2out_begin_function): Likewise.
998 (dwarf2out_source_line): Simplify zero_view_p initialization.
999 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1000 view directly. Omit view when omitting .loc at line 0.
1002 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1004 PR tree-optimization/84740
1005 * tree-switch-conversion.c (process_switch): Call build_constructors
1006 only if info.phi_count is non-zero.
1008 PR tree-optimization/84739
1009 * tree-tailcall.c (find_tail_calls): Check call arguments against
1010 DECL_ARGUMENTS (current_function_decl) rather than
1011 DECL_ARGUMENTS (func) when checking for tail recursion.
1013 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1015 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1016 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1017 Volker Reichelt's entry and add entries for people that perform
1018 GCC fuzzy testing and report numerous bugs.
1020 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1023 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1024 readonly data in sdata, if that is disabled.
1025 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1026 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1027 -mreadonly-in-sdata option.
1029 2018-03-07 Martin Sebor <msebor@redhat.com>
1031 PR tree-optimization/84468
1032 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1033 basic block when looking for nul assignment.
1035 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1038 * except.h (output_function_exception_table): Adjust prototype.
1039 * except.c (output_function_exception_table): Remove FNNAME parameter
1040 and add SECTION parameter. Ouput one part of the table at a time.
1041 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1042 the first part of the exception table and emit unwind directives.
1043 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1044 (i386_pe_seh_cold_init): Likewise.
1045 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1046 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1047 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1048 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1049 * config/i386/winnt.c: Include except.h.
1050 (struct seh_frame_state): Add reg_offset, after_prologue and
1051 in_cold_section fields.
1052 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1053 (i386_pe_seh_cold_init): New function.
1054 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1055 to seh->in_cold_section.
1056 (seh_emit_push): Record the offset of the push.
1057 (seh_emit_save): Record the offet of the save.
1058 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1059 Test seh->after_prologue to disregard the epilogue.
1060 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1061 (i386_pe_end_cold_function): New function.
1063 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1066 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1067 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1070 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1072 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1073 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1076 * multiple_target.c: Include tree-inline.h and intl.h.
1077 (expand_target_clones): Diagnose and fail if node->definition and
1078 !tree_versionable_function_p (node->decl).
1080 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1082 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1084 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1085 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1086 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1088 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1091 * combine.c (try_combine): Use reg_or_subregno instead of handling
1092 just paradoxical SUBREGs and REGs.
1094 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1096 * config/arc/arc.c (arc_finalize_pic): Remove function.
1097 (arc_must_save_register): We use single base PIC register, remove
1098 checks to save/restore the PIC register.
1099 (arc_expand_prologue): Likewise.
1100 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1102 (arc_verify_short): Likewise.
1103 (arc_attr_type): Likewise.
1104 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1105 (walk_stores): Likewise.
1106 (arc_address_cost): Make it static.
1107 (arc_verify_short): Likewise.
1108 (branch_dest): Likewise.
1109 (arc_attr_type): Likewise.
1110 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1111 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1112 (arc_final_prescan_insn): Remove inserting the nops due to
1113 hardware hazards. It is done in reorg step.
1114 (insn_length_variant_t): Remove.
1115 (insn_length_parameters_t): Likewise.
1116 (arc_insn_length_parameters): Likewise.
1117 (arc_get_insn_variants): Likewise.
1118 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1120 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1123 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1126 PR tree-optimization/84687
1127 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1129 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1131 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1133 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1134 Rename to ppc_speculation_barrier.
1135 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1136 __builtin_ppc_speculation_barrier.
1138 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1141 * combine.c (combine_simplify_rtx): Don't try to simplify if
1142 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1145 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1147 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1148 to 32 bytes when compiling for POWER9.
1150 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1153 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1154 regparm >= 3 with no arg reg available also for calls with
1155 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1158 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1160 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1162 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1165 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1167 2018-03-05 Richard Biener <rguenther@suse.de>
1169 PR tree-optimization/84486
1170 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1171 When inserting a __builtin_assume_aligned call set the LHS
1172 SSA name alignment info accordingly.
1174 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1176 PR tree-optimization/84114
1177 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1178 Avoid reassociation of FLOAT_MODE addition.
1180 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1182 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1183 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1184 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1185 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1186 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1187 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1188 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1190 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1191 __builtin_ia32_wbinvd): New builtins.
1192 (SPECIAL_ARGS2): New.
1193 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1194 (SPECIAL_ARGS2): New.
1195 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1196 (ix86_valid_target_attribute_inner_p): Ditto.
1197 (ix86_init_mmx_sse_builtins): Add special_args2.
1198 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1199 TARGET_WBNOINVD_P): New.
1200 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1201 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1202 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1203 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1204 * config/i386/pconfigintrin.h: New file.
1205 * config/i386/wbnoinvdintrin.h: Ditto.
1206 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1207 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1209 2018-03-05 Richard Biener <rguenther@suse.de>
1211 PR tree-optimization/84670
1212 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1214 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1215 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1216 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1217 doesn't grow if all successors (recursively) were visited at least
1220 2018-03-05 Richard Biener <rguenther@suse.de>
1222 PR tree-optimization/84650
1223 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1224 if executed in the loop pipeline.
1226 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1228 * doc/configfiles.texi (Configuration Files): Move info about
1229 conditionalizing $target-protos.h to...
1230 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1231 differs from $target-protos.h.
1233 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1234 Chung-Ju Wu <jasonwucj@gmail.com>
1236 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1237 * config/nds32/nds32-multiple.md (setmemsi): Define.
1238 * config/nds32/nds32-memory-manipulation.c
1239 (nds32_gen_dup_4_byte_to_word_value): New.
1240 (emit_setmem_word_loop): New.
1241 (emit_setmem_byte_loop): New.
1242 (nds32_expand_setmem_loop): New.
1243 (nds32_expand_setmem_loop_v3m): New.
1244 (nds32_expand_setmem_unroll): New.
1245 (nds32_expand_setmem): New.
1247 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1248 Chung-Ju Wu <jasonwucj@gmail.com>
1250 * config/nds32/nds32-memory-manipulation.c
1251 (nds32_emit_load_store): New.
1252 (nds32_emit_post_inc_load_store): New.
1253 (nds32_emit_mem_move): New.
1254 (nds32_emit_mem_move_block): New.
1255 (nds32_expand_movmemsi_loop_unknown_size): New.
1256 (nds32_expand_movmemsi_loop_known_size): New.
1257 (nds32_expand_movmemsi_loop): New.
1258 (nds32_expand_movmemsi_unroll): New.
1259 (nds32_expand_movmemqi): Rename ...
1260 (nds32_expand_movmemsi): ... to this.
1261 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1262 (movmemsi): ... to this.
1263 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1264 (nds32_expand_movmemsi): ... to this.
1266 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1267 Monk Chiang <sh.chiang04@gmail.com>
1268 Chung-Ju Wu <jasonwucj@gmail.com>
1270 * config/nds32/nds32-protos.h
1271 (nds32_expand_load_multiple): New arguments.
1272 (nds32_expand_store_multiple): Ditto.
1273 (nds32_valid_multiple_load_store): Rename ...
1274 (nds32_valid_multiple_load_store_p): ... to this.
1275 * config/nds32/nds32-memory-manipulation.c
1276 (nds32_expand_load_multiple): Refine implementation.
1277 (nds32_expand_store_multiple): Ditto.
1278 * config/nds32/nds32-multiple.md
1279 (load_multiple): Update nds32_expand_load_multiple interface.
1280 (store_multiple): Update nds32_expand_store_multiple interface.
1281 * config/nds32/nds32-predicates.c
1282 (nds32_valid_multiple_load_store): Rename ...
1283 (nds32_valid_multiple_load_store_p): ... to this and refine
1285 * config/nds32/predicates.md
1286 (nds32_load_multiple_and_update_address_operation): New predicate.
1287 (nds32_store_multiple_and_update_address_operation): New predicate.
1289 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1290 Chung-Ju Wu <jasonwucj@gmail.com>
1292 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1293 (combo): New attribute.
1294 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1296 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1298 * config/nds32/nds32.opt: Change -mcmodel= default value.
1300 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1301 Monk Chiang <sh.chiang04@gmail.com>
1302 Chung-Ju Wu <jasonwucj@gmail.com>
1304 * config/nds32/constants.md (unspec_element): New enum.
1305 * config/nds32/constraints.md (Umw): New constraint.
1306 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1307 * config/nds32/nds32-intrinsic.md: Likewise.
1308 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1309 (nds32_valid_smw_lwm_base_p): New.
1310 (nds32_output_smw_single_word): New.
1311 (nds32_output_lmw_single_word): New.
1312 (nds32_expand_unaligned_load): New.
1313 (nds32_expand_unaligned_store): New.
1314 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1315 (nds32_output_smw_single_word): Declare.
1316 (nds32_output_lmw_single_word): Declare.
1317 (nds32_expand_unaligned_load): Declare.
1318 (nds32_expand_unaligned_store): Declare.
1319 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1320 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1321 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1322 NDS32_BUILTIN_UASTORE_DW.
1323 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1326 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1327 Kito Cheng <kito.cheng@gmail.com>
1328 Chung-Ju Wu <jasonwucj@gmail.com>
1330 * config/nds32/nds32-intrinsic.c
1331 (nds32_expand_builtin_null_ftype_reg): Delete.
1332 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1333 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1334 (nds32_read_argument): New.
1335 (nds32_legitimize_target): Ditto.
1336 (nds32_legitimize_argument): Ditto.
1337 (nds32_check_constant_argument): Ditto.
1338 (nds32_expand_unop_builtin): Ditto.
1339 (nds32_expand_unopimm_builtin): Ditto.
1340 (nds32_expand_binop_builtin): Ditto.
1341 (nds32_builtin_decl_impl): Ditto.
1342 (builtin_description): Ditto.
1343 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1344 (nds32_init_builtins_impl): Ditto.
1345 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1346 (nds32_builtin_decl): New.
1347 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1348 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1350 2018-03-02 Jeff Law <law@redhat.com>
1352 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1353 (redundant_insn, fill_simple_delay_slots): Likewise.
1354 (fill_slots_from_thread): Likewise.
1355 * resource.c (mark_referenced_resources): Likewise.
1356 (mark_set_resources, find_dead_or_set_registers): Likewise.
1358 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1360 * substring-locations.h (format_warning_va): Formatting fix for
1362 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1364 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1365 * substring-locations.c: Include intl.h.
1366 (format_warning_va): Turned into small wrapper around
1367 format_warning_n_va, renamed to ...
1368 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1369 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1371 (format_warning_at_substring_n): New function.
1372 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1373 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1374 format_warning_at_substring with just a shorter name instead of
1375 const function pointer.
1376 (fmtwarn_n): New function.
1377 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1378 appropriate, get rid of all the fmtstr temporaries, move conditionals
1379 with G_() wrapped string literals directly into fmtwarn arguments,
1380 cast dir.len to (int), formatting fixes.
1382 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1384 * doc/invoke.texi: Remove "Cilk Plus" references.
1386 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1387 Richard Biener <rguenther@suse.de>
1390 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1391 for error or warning attributes if CALL_FROM_THUNK_P is set.
1394 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1397 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1398 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1401 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1402 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1404 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1405 instead of -1U in last predictors element's probability member.
1407 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1410 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1411 arguments if they are comparable.
1413 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1415 PR tree-optimization/84634
1416 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1417 masks and masked_loop_p with a single loop_masks, making sure it's
1418 null for bb vectorization.
1420 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1422 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1423 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1424 loop->force_vectorize to check whether there is no alias.
1426 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1429 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1431 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1433 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1434 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1435 instead of a loop around prev_real_insn.
1436 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1440 * config/i386/i386.c (ix86_print_operand): Use conditional
1441 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1444 2018-03-02 Richard Biener <rguenther@suse.de>
1446 PR tree-optimization/84427
1447 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1448 (bitmap_set_subtract_values): Rewrite to handle multiple
1451 (prune_clobbered_mems): Likewise.
1452 (phi_translate): Take edge instead of pred/phiblock.
1453 (phi_translate_1): Likewise.
1454 (phi_translate_set): Likewise. Insert all translated
1455 exprs for a value into the set, keeping possibly multiple
1456 expressions per value.
1457 (compute_antic_aux): Adjust for phi_translate changes.
1458 When intersecting union the expressions and prune those
1459 not in the final value set, keeping possibly multiple
1460 expressions per value. Do not use value-insertion
1461 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1462 all expressions. Add verification that the value-sets
1463 only shrink during iteration.
1464 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1465 (do_pre_regular_insertion): Likewise.
1466 (do_pre_partial_partial_insertion): Likewise.
1468 2018-03-02 Richard Biener <rguenther@suse.de>
1471 * config/darwin.c (saved_debug_info_level): New static global.
1472 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1473 (darwin_asm_lto_end): Restore debug info generation settings.
1475 2018-03-01 Martin Liska <mliska@suse.cz>
1478 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1481 2018-03-01 Richard Biener <rguenther@suse.de>
1484 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1485 type annotation with LTO when debug was disabled at compile-time.
1487 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1489 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1491 (mips_final_postscan_insn): Likewise.
1493 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1495 PR rtl-optimization/84528
1496 * alias.c (init_alias_target): Add commentary.
1497 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1498 a unique base value if the frame pointer is not eliminated
1499 to the stack pointer.
1501 2018-03-01 Tom de Vries <tom@codesourcery.com>
1503 PR rtl-optimization/83327
1504 * lra-int.h (hard_regs_spilled_into): Declare.
1505 * lra.c (hard_regs_spilled_into): Define.
1506 (init_reg_info): Init hard_regs_spilled_into.
1507 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1508 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1509 (process_bb_lives): Handle hard_regs_spilled_into.
1510 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1511 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1513 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1515 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1516 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1517 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1518 * config/rs6000/aix72.h: New file.
1520 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1522 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1523 instead of warning_at with conditional singular and plural messages
1527 * stor-layout.c (update_alignment_for_field): For
1528 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1529 && !DECL_PACKED (field), do the alignment update, just use
1530 only desired_align instead of MAX (type_align, desired_align)
1532 (place_field): Don't do known_align < desired_align handling
1533 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1534 is non-NULL, instead do it after rli->prev_field handling and
1535 only if not within a bitfield word. For DECL_PACKED (field)
1536 use type_align of BITS_PER_UNIT.
1538 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1540 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1541 superfluous parentheses and trailing spaces.
1543 2018-02-28 Richard Biener <rguenther@suse.de>
1545 PR tree-optimization/84584
1546 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1547 SCoPs with fake exit edge.
1549 2018-02-28 Martin Liska <mliska@suse.cz>
1552 * timevar.c (timer::print): Fix format to properly print 100%
1555 2018-02-28 Richard Biener <rguenther@suse.de>
1558 * genmatch.c (capture_info::walk_match): Do not mark
1559 captured expressions without operands as expr_p given
1560 they act more like predicates and should be subject to
1561 "lost tail" side-effect preserving.
1563 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1565 PR rtl-optimization/81611
1566 * auto-inc-dec.c (attempt_change): Move dead note from
1567 mem_insn if it's the next use of regno
1568 (find_address): Take address use of reg holding
1569 non-incremented value. Add parm to limit search to the named
1571 (merge_in_block): Attempt to use a mem insn that is the next
1572 use of the original regno.
1574 2018-02-27 Martin Sebor <msebor@redhat.com>
1577 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1578 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1580 2018-02-27 Martin Sebor <msebor@redhat.com>
1582 PR translation/84207
1583 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1584 n argument to unsigned HOST_WIDE_INT.
1585 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1586 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1587 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1588 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1590 2018-02-27 Richard Biener <rguenther@suse.de>
1592 PR tree-optimization/84512
1593 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1594 Do not use the estimate returned from record_stmt_cost for
1595 the scalar iteration cost but sum properly using add_stmt_cost.
1597 2018-02-27 Richard Biener <rguenther@suse.de>
1599 PR tree-optimization/84466
1600 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1601 Adjust last change to less strictly validate use operands.
1603 2018-02-27 Martin Liska <mliska@suse.cz>
1605 PR gcov-profile/84548
1606 * gcov.c (process_file): Allow partial overlap and consider it
1607 also as group functions.
1608 (output_lines): Properly calculate range of lines for a group.
1610 2018-02-27 Martin Liska <mliska@suse.cz>
1612 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1613 'ggc' suffixes. Change first column width.
1614 (timer::print): Fix formatting of the column.
1616 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1618 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1619 preserve inline entry blocks for the sake of debug inline
1620 entry point markers alone.
1621 (remove_unused_locals): Suggest in comments a better place to
1622 force the preservation of inline entry blocks that are
1623 otherwise unused, but do not preserve them.
1625 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1627 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1629 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1632 * config/i386/constraints.md (Bs): Replace
1633 ix86_indirect_branch_register with
1634 TARGET_INDIRECT_BRANCH_REGISTER.
1636 * config/i386/i386.md (indirect_jump): Likewise.
1637 (tablejump): Likewise.
1638 (*sibcall_memory): Likewise.
1639 (*sibcall_value_memory): Likewise.
1640 Peepholes of indirect call and jump via memory: Likewise.
1641 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1642 (*sibcall_value_GOT_32): Likewise.
1643 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1644 (GOT_memory_operand): Likewise.
1645 (call_insn_operand): Likewise.
1646 (sibcall_insn_operand): Likewise.
1647 (GOT32_symbol_operand): Likewise.
1648 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1650 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1652 PR rtl-optimization/83496
1653 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1654 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1655 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1656 redundant insn, if any.
1657 (relax_delay_slots): Likewise.
1658 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1660 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1662 PR tree-optimization/83965
1663 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1664 that grouped statements are part of a reduction chain. Return
1665 true if the statement is not marked as a reduction itself but
1667 (vect_recog_dot_prod_pattern): Don't check whether the statement
1668 is part of a group here.
1669 (vect_recog_sad_pattern): Likewise.
1670 (vect_recog_widen_sum_pattern): Likewise.
1672 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1675 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1677 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1680 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1682 (ix86_output_indirect_function_return): New prototype.
1683 (ix86_split_simple_return_pop_internal): Likewise.
1684 * config/i386/i386.c (indirect_return_via_cx): New.
1685 (indirect_return_via_cx_bnd): Likewise.
1686 (indirect_thunk_name): Handle return va CX_REG.
1687 (output_indirect_thunk_function): Create alias for
1688 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1689 (ix86_output_indirect_jmp): Remove the bool argument.
1690 (ix86_output_indirect_function_return): New function.
1691 (ix86_split_simple_return_pop_internal): Likewise.
1692 * config/i386/i386.md (*indirect_jump): Don't pass false
1693 to ix86_output_indirect_jmp.
1694 (*tablejump_1): Likewise.
1695 (simple_return_pop_internal): Change it to define_insn_and_split.
1696 Call ix86_split_simple_return_pop_internal to split it for
1698 (simple_return_indirect_internal): Call
1699 ix86_output_indirect_function_return instead of
1700 ix86_output_indirect_jmp.
1702 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1705 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1706 memset and value initialization afterwards.
1708 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1710 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1712 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1715 * common/config/aarch64/aarch64-common.c
1716 (aarch_option_optimization_table[]): Switch
1717 off fomit-frame-pointer
1719 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1720 Chung-Ju Wu <jasonwucj@gmail.com>
1722 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1724 (store_multiple): Ditto.
1726 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1728 * config.gcc: Add --with-cpu support for nds32 target.
1729 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1730 * config/nds32/nds32.opt: Add -mcpu= option.
1732 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1734 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1735 isel=yes): Warn for these deprecated options.
1737 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1739 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1740 ISA_2_5_MASKS_EMBEDDED.
1742 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1744 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1745 p->max as pointers rather than using iterative_hash_expr.
1747 2018-02-23 Carl Love <cel@us.ibm.com>
1749 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1750 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1752 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1753 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1754 P8V_BUILTIN_VEC_VUNSIGNED2.
1756 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1759 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1760 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1762 * lra-constraints.c (curr_insn_transform): Set up
1763 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1765 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1766 * lra-eliminations.c (spill_pseudos): Ditto.
1767 (process_insn_for_elimination): Ditto.
1768 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1769 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1770 LRA_NON_CLOBBERED_ALT.
1772 2018-02-22 Martin Sebor <msebor@redhat.com>
1774 PR tree-optimization/84480
1775 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1776 to maybe_diag_stxncpy_trunc. Call it.
1777 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1778 from gimple_fold_builtin_strcpy. Print inlining stack.
1779 (handle_builtin_stxncpy): Print inlining stack.
1780 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1782 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1785 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1786 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1787 and -fcheck-pointer-bounds are used together.
1788 (indirect_thunk_prefix): New enum.
1789 (indirect_thunk_need_prefix): New function.
1790 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1791 "_nt" instead of "_bnd" for NOTRACK prefix.
1792 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1793 (output_indirect_thunk_function): Likewise.
1795 (ix86_code_end): Update output_indirect_thunk_function calls.
1796 (ix86_output_indirect_branch_via_reg): Replace
1797 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1798 (ix86_output_indirect_branch_via_push): Likewise.
1799 (ix86_output_function_return): Likewise.
1800 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1801 incompatible with -fcf-protection=branch and
1802 -fcheck-pointer-bounds.
1804 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1807 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1808 Change gcc_assert call to output_operand_lossage.
1810 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1812 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1814 2018-02-22 DJ Delorie <dj@redhat.com>
1815 Sebastian Perta <sebastian.perta@renesas.com>
1816 Oleg Endo <olegendo@gcc.gnu.org>
1818 * config/rx/rx.c (rx_rtx_costs): New function.
1819 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1821 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1823 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1825 2018-02-22 Martin Liska <mliska@suse.cz>
1828 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1829 Add "native" as a possible value.
1831 2018-02-22 Martin Liska <mliska@suse.cz>
1834 * config/i386/i386.c (ix86_option_override_internal):
1835 Add "native" as a possible value for -march and -mtune.
1837 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1840 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1841 to all type variants.
1843 PR tree-optimization/84503
1844 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1845 width as info->bitpos + info->bitsize - start.
1846 (merged_store_group::merge_overlapping): Simplify width computation.
1847 (check_no_overlap): New function.
1848 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1849 start + width and last_order of the group, fail if check_no_overlap
1851 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1852 to group if check_no_overlap fails.
1854 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1856 * config/rs6000/altivec.md: Delete contraint arguments to
1857 define_expand, define_split, and define_peephole2, and in
1858 define_insn_and_split if always unused.
1859 * config/rs6000/darwin.md: Ditto.
1860 * config/rs6000/dfp.md: Ditto.
1861 * config/rs6000/rs6000.md: Ditto.
1862 * config/rs6000/sync.md: Ditto.
1863 * config/rs6000/vector.md: Ditto.
1864 * config/rs6000/vsx.md: Ditto.
1866 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1868 * config/rs6000/altivec.md: Write output control strings as braced
1869 blocks instead of double-quoted strings.
1870 * config/rs6000/darwin.md: Ditto.
1871 * config/rs6000/rs6000.md: Ditto.
1872 * config/rs6000/vector.md: Ditto.
1873 * config/rs6000/vsx.md: Ditto.
1875 2018-02-21 Jason Merrill <jason@redhat.com>
1877 PR c++/84314 - ICE with templates and fastcall attribute.
1878 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1880 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1882 * ipa-cp.c (determine_versionability): Fix comment typos.
1884 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1887 * ipa-cp.c (determine_versionability): Do not version functions caling
1890 2018-02-21 Martin Liska <mliska@suse.cz>
1893 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1894 Add "native" as a possible value.
1895 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1896 the macro when native cpu detection is available.
1898 2018-02-21 Martin Liska <mliska@suse.cz>
1901 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1902 Add "native" as a possible value.
1903 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1904 when native cpu detection is available.
1906 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1907 Martin Sebor <msebor@redhat.com>
1909 PR tree-optimization/84478
1910 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1912 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1913 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1914 support which is conservatively correct, for 2 only stay conservative
1915 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1916 argument to the 2 argument get_range_strlen, adjust 6 arg
1917 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1919 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1920 (gimple_fold_builtin_strlen): Pass true as last argument to
1923 2018-02-20 Martin Sebor <msebor@redhat.com>
1926 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1927 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1928 (builtin_memref::builtin_memref): Factor out parts into
1929 set_base_and_offset and call it.
1931 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1934 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1935 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1936 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1937 search at the associated MODE_INT.
1939 2018-02-20 Jeff Law <law@redhat.com>
1942 PR tree-optimization/81592
1944 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1945 for range data rather than using global data.
1946 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1947 range data rather than using global data.
1948 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1949 pass it to children as needed.
1950 (struct directive::fmtresult): Similarly.
1951 (struct directive::set_width): Similarly.
1952 (struct directive::set_precision): Similarly.
1953 (format_integer, format_directive, parse_directive): Similarly.
1954 (format_none): Accept unnamed vr_values parameter.
1955 (format_percent, format_floating, format_character): Similarly.
1956 (format_string, format_plain): Similarly.
1957 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1958 the EVRP range analyzer for range data rather than using global data.
1959 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1960 gimple-ssa-evrp-analyze.h
1961 (class sprintf_dom_walker): Add after_dom_children member function.
1962 Add evrp_range_analyzer member.
1963 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1964 range analyzer as needed.
1965 (sprintf_dom_walker::after_dom_children): New member function.
1966 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1968 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1969 (evrp_range_analyzer::pop_to_marker): Likewise.
1971 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1973 PR tree-optimization/84419
1974 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1975 with the required type if its current type is compatible but
1978 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1981 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1982 after vectorization.
1984 2018-02-20 Martin Liska <mliska@suse.cz>
1987 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1988 possible values if we don't have a hint.
1990 2018-02-20 Martin Liska <mliska@suse.cz>
1994 * final.c (shorten_branches): Build align_tab array with one
1996 * opts.c (finish_options): Add alignment option limit check.
1997 (MAX_CODE_ALIGN): Likewise.
1998 (MAX_CODE_ALIGN_VALUE): Likewise.
1999 * doc/invoke.texi: Document maximum allowed option value for
2000 all -falign-* options.
2002 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2005 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2006 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2007 * var-tracking.c (emit_note_insn_var_location): Remove all references
2008 to NOTE_INSN_CALL_ARG_LOCATION.
2009 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2010 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2011 Use copy_rtx_if_shared.
2012 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2013 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2014 (dwarf2out_var_location): Remove handling of
2015 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2017 * final.c (final_scan_insn): Remove all references to
2018 NOTE_INSN_CALL_ARG_LOCATION.
2019 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2020 before dumping final insns.
2021 * except.c (emit_note_eh_region_end): Remove all references to
2022 NOTE_INSN_CALL_ARG_LOCATION.
2023 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2024 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2025 * config/arc/arc.c (hwloop_optimize): Likewise.
2026 * config/arm/arm.c (create_fix_barrier): Likewise.
2027 * config/s390/s390.c (s390_chunkify_start): Likewise.
2028 * config/sh/sh.c (find_barrier): Likewise.
2029 * config/i386/i386.c (rest_of_insert_endbranch,
2030 ix86_seh_fixup_eh_fallthru): Likewise.
2031 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2032 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2033 * config/frv/frv.c (frv_function_prologue): Likewise.
2034 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2036 (note_outside_basic_block_p): Remove all references to
2037 NOTE_INSN_CALL_ARG_LOCATION.
2038 * gengtype.c (adjust_field_rtx_def): Likewise.
2039 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2041 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2042 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2045 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2048 PR tree-optimization/84452
2049 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2050 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2053 2018-02-19 Martin Liska <mliska@suse.cz>
2056 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2058 2018-02-19 Martin Liska <mliska@suse.cz>
2059 Richard Sandiford <richard.sandiford@linaro.org>
2061 PR tree-optimization/82491
2062 * gimple-fold.c (get_base_constructor): Make earlier bail out
2065 2018-02-19 Carl Love <cel@us.ibm.com>
2067 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2068 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2070 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2071 P8V_BUILTIN_VEC_NEG.
2073 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2075 * config/rl78/rl78.md (movdf): New define expand.
2077 2018-02-19 Martin Liska <mliska@suse.cz>
2080 * doc/invoke.texi: Fix typo.
2081 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2083 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2085 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2086 handle rs6000_single_float and rs6000_double_float specially for
2089 2018-02-16 Jeff Law <law@redhat.com>
2091 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2094 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2097 * ipa-inline.c (inline_small_functions): Fix a typo.
2099 2018-02-16 Nathan Sidwell <nathan@acm.org>
2101 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2103 2018-02-16 Carl Love <cel@us.ibm.com>
2105 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2106 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2107 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2108 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2109 expansion to P8V_BUILTIN_VEC_FLOAT2.
2111 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2113 PR rtl-optimization/70023
2114 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2115 src_regno into account.
2117 2018-02-16 Carl Love <cel@us.ibm.com>
2119 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2120 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2121 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2122 * config/rs6000/rs6000.c: Remove case statements for
2123 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2124 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2125 and P9V_BUILTIN_VEC_VINSERT4B.
2126 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2127 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2128 * config/rs6000/vsx.md:
2129 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2132 2018-02-16 Carl Love <cel@us.ibm.com>
2134 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2136 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2138 * config/rs6000/rs6000-c.c: Add the definitions for
2139 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2140 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2141 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2142 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2143 definition for insert4b and define insn *insert3b_internal.
2144 * doc/extend.texi: Add documentation for vec_extract4b.
2146 2018-02-16 Nathan Sidwell <nathan@acm.org>
2148 * doc/extend.texi (Backwards Compatibility): Mention friend
2149 injection. Note for-scope is deprecated.
2150 * doc/invoke.texi (-ffriend-injection): Deprecate.
2152 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2154 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2155 that moved to I2, also allow destinations that are a paradoxical
2156 subreg (instead of a normal reg).
2158 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2161 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2164 2018-02-16 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/84037
2167 PR tree-optimization/84016
2169 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2170 Adjust vec_construct for the fact we need additional higher latency
2171 128bit inserts for AVX256 and AVX512 vector builds.
2172 (ix86_add_stmt_cost): Scale vector construction cost for
2175 2018-02-16 Richard Biener <rguenther@suse.de>
2177 PR tree-optimization/84417
2178 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2179 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2180 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2182 2018-02-16 Martin Liska <mliska@suse.cz>
2185 * internal-fn.def (ASAN_CHECK): Set proper flags.
2186 (ASAN_MARK): Likewise.
2188 2018-02-16 Julia Koval <julia.koval@intel.com>
2190 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2191 from PTA_CANNONLAKE.
2193 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2196 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2197 Use ++iter rather than iter++ for std::list iterators.
2198 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2199 defer deleting them until all nodes in the forest are processed. Do
2200 free even leaf nodes. Change to_process into auto_vec.
2203 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2204 * vec.h (vec_default_construct): Use memset instead of placement new
2205 if BROKEN_VALUE_INITIALIZATION is defined.
2206 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2207 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2210 PR rtl-optimization/83723
2211 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2212 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2213 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2215 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2217 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2219 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2221 PR rtl-optimization/81443
2222 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2223 from inner REGs to paradoxical SUBREGs.
2225 2018-02-16 Richard Biener <rguenther@suse.de>
2227 PR tree-optimization/84399
2228 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2229 For operands we can analyze at their definition make sure we can
2230 analyze them at each use as well.
2232 2018-02-16 Richard Biener <rguenther@suse.de>
2234 PR tree-optimization/84190
2235 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2236 volatile accesses if the decl isn't volatile.
2238 2018-02-15 Jason Merrill <jason@redhat.com>
2240 PR c++/84314 - ICE with templates and fastcall attribute.
2241 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2242 TYPE_CANONICAL on an existing type.
2244 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2246 PR tree-optimization/84383
2247 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2248 dstoff nor call operand_equal_p if dstbase is NULL.
2250 PR tree-optimization/84334
2251 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2252 also a CONSTANT_CLASS_P, punt.
2254 2018-02-14 Jim Wilson <jimw@sifive.com>
2256 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2257 first SMALL_OPERAND check. New local min_second_step. Move assert
2258 to where locals are set. Add TARGET_RVC support.
2259 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2261 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2263 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2265 2018-02-14 Martin Sebor <msebor@redhat.com>
2267 PR tree-optimization/83698
2268 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2269 arrays constrain the offset range to their bounds.
2270 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2271 (builtin_access::overlap): Avoid setting the size of overlap if it's
2273 (maybe_diag_overlap): Also consider arrays when deciding what values
2274 of offsets to include in diagnostics.
2276 2018-02-14 Martin Sebor <msebor@redhat.com>
2279 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2280 that correspond to the kind of a declaration.
2282 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2285 * config/pa/pa.md: Load address of PIC label using the linkage table
2286 if the label is nonlocal.
2288 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2290 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2291 warning message if user requests -maltivec=be.
2292 * doc/invoke.texi: Document deprecation of -maltivec=be.
2294 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2297 * config/rs6000/rs6000-c.c: Update definitions for
2298 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2299 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2301 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2304 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2305 add _get_ssp intrinsics. Remove argument from
2306 __builtin_ia32_rdssp[d|q].
2307 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2308 * config/i386/i386-builtin.def: Remove argument from
2309 __builtin_ia32_rdssp[d|q].
2310 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2311 ix86_expand_special_args_builtin for _rdssp[d|q].
2312 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2313 Clear register before usage.
2314 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2315 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2317 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2319 PR tree-optimization/84357
2320 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2321 operand 1 of an ARRAY_REF too.
2323 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2326 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2327 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2329 (set_of_reg): New struct.
2330 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2331 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2332 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2334 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2335 Split into bitclr, bitset, bitinvert patterns if appropriate.
2336 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2337 use rx_fuse_in_memory_bitop.
2338 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2339 to named insn, correct maximum insn length.
2341 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2344 * machmode.def: Define a complex mode for PARTIAL_INT.
2345 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2347 * doc/rtl.texi: Document CSPImode.
2348 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2350 (msp430_hard_regno_nregs_with_padding): Likewise.
2352 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2355 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2357 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2359 PR rtl-optimization/84169
2360 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2361 we generated a parallel as new i3 and we split that to new i2 and i3
2362 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2363 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2364 those to i2, not i1. Partially rewrite this scan code.
2366 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2369 * stor-layout.c (place_field): For variable length fields, adjust
2370 offset_align afterwards not just based on the field's alignment,
2371 but also on the size.
2374 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2375 of exps and logs in the use_exp2 case.
2377 2018-02-13 Jeff Law <law@redhat.com>
2379 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2382 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2385 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2389 * common.opt (gas-loc-support, gas-locview-support): New.
2390 (ginline-points, ginternal-reset-location-views): New.
2391 * doc/invoke.texi: Document them. Use @itemx where intended.
2392 (gvariable-location-views): Adjust.
2393 * target.def (reset_location_view): New.
2394 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2395 (TARGET_RESET_LOCATION_VIEW): New.
2396 * doc/tm.texi: Rebuilt.
2397 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2398 (dwarf2out_default_as_locview_support): New.
2399 (output_asm_line_debug_info): Use option variables.
2400 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2401 (output_loc_list): Likewise.
2402 (add_high_low_attributes): Check option variables.
2403 Don't output entry view attribute in strict mode.
2404 (gen_inlined_subroutine_die): Check option variables.
2405 (dwarf2out_inline_entry): Likewise.
2406 (init_sections_and_labels): Likewise.
2407 (dwarf2out_early_finish): Likewise.
2408 (maybe_reset_location_view): New, from...
2409 (dwarf2out_var_location): ... here. Call it.
2410 * debug.h (dwarf2out_default_as_loc_support): Declare.
2411 (dwarf2out_default_as_locview_support): Declare.
2412 * hooks.c (hook_int_rtx_insn_0): New.
2413 * hooks.h (hook_int_rtx_insn_0): Declare.
2414 * toplev.c (process_options): Take -gas-loc-support and
2415 -gas-locview-support from dwarf2out. Enable
2416 -gvariable-location-views by default only with locview
2417 assembler support. Enable -ginternal-reset-location-views by
2418 default only if the target defines the corresponding hook.
2419 Enable -ginline-points by default if location views are
2420 enabled; force it disabled if statement frontiers are
2422 * tree-inline.c (expand_call_inline): Check option variables.
2423 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2425 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2427 PR tree-optimization/84321
2428 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2429 handling. Also check whether the anti-range contains any values
2430 that satisfy the mask; switch to a VR_RANGE if not.
2432 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2435 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2437 2018-02-13 Martin Jambor <mjambor@suse.cz>
2440 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2441 of call statements, also set location of a load to a temporary.
2443 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2445 * config/rl78/rl78.c (add_vector_labels): New function.
2446 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2447 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2448 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2449 which checks that no arguments are passed.
2450 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2451 * doc/extend.texi: Documentation for the new attribute.
2453 2018-02-13 Andreas Schwab <schwab@suse.de>
2455 * config/riscv/linux.h (CPP_SPEC): Define.
2457 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2460 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2461 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2462 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2463 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2464 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2465 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2466 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2467 temporarily for AES and PCLMUL builtins.
2469 PR tree-optimization/84339
2470 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2471 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2475 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2476 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2477 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2479 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2481 * omp-simd-clone.h: New file.
2482 * omp-simd-clone.c: Include omp-simd-clone.h.
2483 (expand_simd_clones): No longer static.
2484 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2485 cgraph.h and omp-simd-clone.h.
2486 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2487 (vect_recog_widen_shift_pattern): Formatting fix.
2488 (vect_pattern_recog_1): Don't check optab for calls.
2491 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2492 operands[2] into a REG before using gen_lowpart on it.
2494 2018-02-12 Jeff Law <law@redhat.com>
2497 * config/sh/sh.c (find_barrier): Consider a sibling call
2500 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2501 successfully back substituting a reg.
2503 2018-02-12 Richard Biener <rguenther@suse.de>
2505 PR tree-optimization/84037
2506 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2507 parameter, move visited init to caller.
2508 (vect_slp_analyze_operations): Separate cost from validity
2509 check, initialize visited once for all instances.
2510 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2512 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2514 (vect_model_promotion_demotion_cost): Likewise.
2515 (vectorizable_bswap): Guard cost modeling with !slp_node
2516 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2518 (vectorizable_call): Likewise.
2519 (vectorizable_conversion): Likewise.
2520 (vectorizable_assignment): Likewise.
2521 (vectorizable_shift): Likewise.
2522 (vectorizable_operation): Likewise.
2523 (vectorizable_store): Likewise.
2524 (vectorizable_load): Likewise.
2525 (vectorizable_condition): Likewise.
2526 (vectorizable_comparison): Likewise.
2528 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2531 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2532 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2533 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2535 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2538 * params.def (inline-min-speedup): Increase from 8 to 15.
2539 (max-inline-insns-auto): Decrease from 40 to 30.
2540 * ipa-split.c (consider_split): Add some buffer for function to
2541 be considered inlining candidate.
2542 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2545 2018-02-12 Richard Biener <rguenther@suse.de>
2547 PR tree-optimization/84037
2548 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2549 matched stmts if we cannot swap the non-matched ones.
2551 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2553 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2554 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2555 _mm_maskz_scalef_round_ss): New intrinsics.
2556 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2557 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2558 __builtin_ia32_scalefss_round): Remove.
2559 (__builtin_ia32_scalefsd_mask_round,
2560 __builtin_ia32_scalefss_mask_round): New intrinsics.
2561 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2562 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2563 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2564 "<round_constraint>")): Changed to ...
2565 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2566 "<round_scalar_constraint>")): ... this.
2567 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2568 %0, %1, %2<round_op3>}"): Changed to ...
2569 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2570 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2571 %2<round_scalar_mask_op3>}"): ... this.
2572 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2574 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2576 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2577 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2578 (_mm_maskz_sqrt_round_ss): New intrinsics.
2579 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2580 (__builtin_ia32_sqrtsd_mask_round)
2581 (__builtin_ia32_sqrtss_mask_round): New builtins.
2582 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2583 (__builtin_ia32_sqrtss_round): Remove.
2584 (__builtin_ia32_sqrtsd_mask_round)
2585 (__builtin_ia32_sqrtss_mask_round): New builtins.
2586 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2587 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2588 ((match_operand:VF_128 1 "vector_operand"
2589 "xBm,<round_constraint>")): Changed to ...
2590 ((match_operand:VF_128 1 "vector_operand"
2591 "xBm,<round_scalar_constraint>")): ... this.
2592 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2593 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2594 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2595 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2596 %<iptr>1<round_scalar_mask_op3>}): ... this.
2597 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2598 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2600 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2603 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2604 Cast vec_cmpeq result to correct type.
2605 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2606 Cast vec_cmpgt result to correct type.
2608 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2610 * final.c (final_scan_insn_1): Renamed from...
2611 (final_scan_insn): ... this. New wrapper, to recover
2612 seen from the outermost call in recursive ones.
2613 * config/sparc/sparc.c (output_return): Drop seen from call.
2614 (output_sibcall): Likewise.
2615 * config/visium/visium.c (output_branch): Likewise.
2617 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2619 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2622 2018-02-10 Alan Modra <amodra@gmail.com>
2625 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2626 Specify LR as an input.
2628 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2631 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2632 remove_member_access_dummy_vars): New functions.
2633 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2634 lower_omp_1, execute_lower_omp): Use them.
2636 PR rtl-optimization/84308
2637 * shrink-wrap.c (spread_components): Release todo vector.
2639 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2641 PR rtl-optimization/57193
2642 * ira-color.c (struct allocno_color_data): Add member
2643 conflict_allocno_hard_prefs.
2644 (update_conflict_allocno_hard_prefs): New.
2645 (bucket_allocno_compare_func): Add a preference based on
2646 conflict_allocno_hard_prefs.
2647 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2648 (color_allocnos): Remove a dead code. Initiate
2649 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2651 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2654 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2655 constraint from =wa to wa. Avoid a subreg on the output operand,
2656 instead use a pseudo and subreg it in a move.
2657 (p9_xxbrd_<mode>): Changed to ...
2658 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2659 (p9_xxbrd_v2df): New expander.
2660 (p9_xxbrw_<mode>): Changed to ...
2661 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2662 (p9_xxbrw_v4sf): New expander.
2664 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2666 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2668 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2671 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2672 multiply in 32-bit mode.
2673 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2674 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2677 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2679 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2680 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2681 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2682 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2684 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2687 * dwarf2out.c (is_trivial_indirect_ref): New function.
2688 (dwarf2out_late_global_decl): Do not generate a location
2689 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2690 and that are not defined in the current unit.
2692 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2694 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2695 instead of a libcall for UNORDERED.
2697 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2700 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2701 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2703 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2706 * config/s390/s390.c (s390_set_current_function): Invoke
2707 s390_indirect_branch_settings also if fndecl didn't change.
2709 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2711 * config/rs6000/rs6000.md (blockage): Set length to zero.
2713 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2715 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2717 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2720 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2721 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2725 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2726 PARALLEL incoming that failed vt_get_decl_and_offset check.
2729 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2730 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2732 (get_variable_section): For decls in named .bss* sections pass true as
2733 second argument to bss_initializer_p.
2735 2018-02-09 Marek Polacek <polacek@redhat.com>
2736 Jakub Jelinek <jakub@redhat.com>
2739 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2740 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2741 Sync some changes from cxx_fold_indirect_ref.
2743 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2745 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2747 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2748 (BLOCK_INLINE_ENTRY_LABEL): New.
2749 (dwarf2out_var_location): Disregard inline entry markers.
2750 (inline_entry_data): New struct.
2751 (inline_entry_data_hasher): New hashtable type.
2752 (inline_entry_data_hasher::hash): New.
2753 (inline_entry_data_hasher::equal): New.
2754 (inline_entry_data_table): New variable.
2755 (add_high_low_attributes): Add DW_AT_entry_pc and
2756 DW_AT_GNU_entry_view attributes if a pending entry is found
2757 in inline_entry_data_table. Add old entry_pc attribute only
2758 if debug nonbinding markers are disabled.
2759 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2760 markers are enabled.
2761 (block_within_block_p, dwarf2out_inline_entry): New.
2762 (dwarf2out_finish): Check that no entries remained in
2763 inline_entry_data_table.
2764 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2765 (final_scan_insn, notice_source_line): Likewise.
2766 (rest_of_clean_state): Skip inline entry markers.
2767 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2769 * gimple.c (gimple_build_debug_inline_entry): New.
2770 * gimple.h (enum gimple_debug_subcode): Add
2771 GIMPLE_DEBUG_INLINE_ENTRY.
2772 (gimple_build_debug_inline_entry): Declare.
2773 (gimple_debug_inline_entry_p): New.
2774 (gimple_debug_nonbind_marker_p): Adjust.
2775 * insn-notes.def (INLINE_ENTRY): New.
2776 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2777 inline entry marker notes.
2778 (print_insn): Likewise.
2779 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2780 (INSN_DEBUG_MARKER_KIND): Likewise.
2781 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2782 * tree-inline.c (expand_call_inline): Build and insert
2783 debug_inline_entry stmt.
2784 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2785 inline entry blocks early, if nonbind markers are enabled.
2786 (dump_scope_block): Dump fragment info.
2787 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2788 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2789 (gimple_build_debug_inline_entry): New.
2790 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2791 Enable/disable inline entry points too.
2792 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2793 (DEBUG_INSN): Describe inline entry markers.
2795 * common.opt (gvariable-location-views): New.
2796 (gvariable-location-views=incompat5): New.
2797 * config.in: Rebuilt.
2798 * configure: Rebuilt.
2799 * configure.ac: Test assembler for view support.
2800 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2801 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2802 * dwarf2out.c (var_loc_view): New typedef.
2803 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2804 (dwarf2out_locviews_in_attribute): New.
2805 (dwarf2out_locviews_in_loclist): New.
2806 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2807 (enum dw_line_info_opcode): Add LI_adv_address.
2808 (struct dw_line_info_table): Add view.
2809 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2810 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2811 (zero_view_p): New variable.
2812 (ZERO_VIEW_P): New macro.
2813 (output_asm_line_debug_info): New.
2814 (struct var_loc_node): Add view.
2815 (add_AT_view_list, AT_loc_list): New.
2816 (add_var_loc_to_decl): Add view param. Test it against last.
2817 (new_loc_list): Add view params. Record them.
2818 (AT_loc_list_ptr): Handle loc and view lists.
2819 (view_list_to_loc_list_val_node): New.
2820 (print_dw_val): Handle dw_val_class_view_list.
2821 (size_of_die): Likewise.
2822 (value_format): Likewise.
2823 (loc_list_has_views): New.
2824 (gen_llsym): Set vl_symbol too.
2825 (maybe_gen_llsym, skip_loc_list_entry): New.
2826 (dwarf2out_maybe_output_loclist_view_pair): New.
2827 (output_loc_list): Output view list or entries too.
2828 (output_view_list_offset): New.
2829 (output_die): Handle dw_val_class_view_list.
2830 (output_dwarf_version): New.
2831 (output_compilation_unit_header): Use it.
2832 (output_skeleton_debug_sections): Likewise.
2833 (output_rnglists, output_line_info): Likewise.
2834 (output_pubnames, output_aranges): Update version comments.
2835 (output_one_line_info_table): Output view numbers in asm comments.
2836 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2837 Call maybe_gen_llsym.
2838 (loc_list_from_tree_1): Adjust.
2839 (add_AT_location_description): Create view list attribute if
2840 needed, check it's absent otherwise.
2841 (convert_cfa_to_fb_loc_list): Adjust.
2842 (maybe_emit_file): Call output_asm_line_debug_info for test.
2843 (dwarf2out_var_location): Reset views as needed. Precompute
2844 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2845 attribute. Set view.
2846 (new_line_info_table): Reset next view.
2847 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2848 (dwarf2out_source_line): Likewise. Output view resets and labels to
2849 the assembler, or select appropriate line info opcodes.
2850 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2851 (optimize_string_length): Catch it. Adjust.
2852 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2853 dw_val_class_view_list, and remove it if no longer needed.
2854 (hash_loc_list): Hash view numbers.
2855 (loc_list_hasher::equal): Compare them.
2856 (optimize_location_lists): Check whether a view list symbol is
2857 needed, and whether the locview attribute is present, and
2858 whether they match. Remove the locview attribute if no longer
2860 (index_location_lists): Call skip_loc_list_entry for test.
2861 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2862 Use output_dwarf_version.
2863 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2864 (struct dw_val_node): Add val_view_list.
2865 * final.c (SEEN_NEXT_VIEW): New.
2866 (set_next_view_needed): New.
2867 (clear_next_view_needed): New.
2868 (maybe_output_next_view): New.
2869 (final_start_function): Rename to...
2870 (final_start_function_1): ... this. Take pointer to FIRST,
2871 add SEEN parameter. Emit param bindings in the initial view.
2872 (final_start_function): Reintroduce SEEN-less interface.
2873 (final): Rename to...
2874 (final_1): ... this. Take SEEN parameter. Output final pending
2875 next view at the end.
2876 (final): Reintroduce seen-less interface.
2877 (final_scan_insn): Output pending next view before switching
2878 sections or ending a block. Mark the next view as needed when
2879 outputting variable locations. Notify debug backend of section
2880 changes, and of location view changes.
2881 (rest_of_handle_final): Adjust.
2882 * toplev.c (process_options): Autodetect value for debug variable
2883 location views option. Warn on incompat5 without -gdwarf-5.
2884 * doc/invoke.texi (gvariable-location-views): New.
2885 (gvariable-location-views=incompat5): New.
2886 (gno-variable-location-views): New.
2888 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2890 PR tree-optimization/84136
2891 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2892 that the result of find_edge is non-NULL.
2894 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2897 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2898 storing integer register in SImode. Fix cost of 256 and 512
2899 byte aligned SSE register store.
2901 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2903 * config/i386/i386.c (ix86_multiplication_cost): Fix
2904 multiplication cost for TARGET_AVX512DQ.
2906 2018-02-08 Marek Polacek <polacek@redhat.com>
2908 PR tree-optimization/84238
2909 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2912 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2914 PR tree-optimization/84265
2915 * tree-vect-stmts.c (vectorizable_store): Don't treat
2916 VMAT_CONTIGUOUS accesses as grouped.
2917 (vectorizable_load): Likewise.
2919 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2921 PR tree-optimization/81635
2922 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2923 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2924 (test_round_for_mask): New functions.
2925 (wide_int_cc_tests): Call test_round_for_mask.
2926 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2927 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2928 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2929 range returned by get_range_info.
2931 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2934 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2935 * symtab.c: Include builtins.h
2936 (symtab_node::output_to_lto_symbol_table_p): Move here
2937 from lto-streamer-out.c:output_symbol_p.
2938 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2939 (output_symbol_p): Move all logic to symtab.c
2940 (produce_symtab): Update.
2942 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2944 * config/s390/s390-opts.h (enum indirect_branch): Define.
2945 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2946 (s390_indirect_branch_via_thunk)
2947 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2948 (enum s390_indirect_branch_type): Define.
2949 * config/s390/s390.c (struct s390_frame_layout, struct
2950 machine_function): Remove.
2951 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2952 (indirect_branch_table_label_no, indirect_branch_table_name):
2954 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2955 (enum s390_indirect_branch_option): Define.
2956 (s390_return_addr_from_memory): New function.
2957 (s390_handle_string_attribute): New function.
2958 (s390_attribute_table): Add new attribute handler.
2959 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2960 (s390_indirect_branch_via_thunk): New function.
2961 (s390_indirect_branch_via_inline_thunk): New function.
2962 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2963 sibling call optimization for non z10 compiles.
2964 (s390_emit_call): Force indirect branch target to be a single
2965 register. Add r1 clobber for non-z10 compiles.
2966 (s390_emit_epilogue): Emit return jump via return_use expander.
2967 (s390_reorg): Handle JUMP_INSNs as execute targets.
2968 (s390_option_override_internal): Perform validity checks for the
2969 new command line options.
2970 (s390_indirect_branch_attrvalue): New function.
2971 (s390_indirect_branch_settings): New function.
2972 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2973 (s390_output_indirect_thunk_function): New function.
2974 (s390_code_end): Implement target hook.
2975 (s390_case_values_threshold): Implement target hook.
2976 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2978 * config/s390/s390.h (struct s390_frame_layout)
2979 (struct machine_function): Move here from s390.c.
2980 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2981 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2982 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2983 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2984 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2985 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2986 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2987 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2988 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2989 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2990 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2991 (mnemonic attribute): Add values which aren't recognized
2993 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2994 pattern for branch conversion. Fix mnemonic attribute.
2995 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2996 indirect branch via thunk if requested.
2997 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2998 ("*indirect_jump"): Disable for branch conversion using out of
3000 ("indirect_jump_via_thunk<mode>_z10")
3001 ("indirect_jump_via_thunk<mode>")
3002 ("indirect_jump_via_inlinethunk<mode>_z10")
3003 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3004 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3005 ("casesi_jump_via_inlinethunk<mode>_z10")
3006 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3007 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3008 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3009 ("*indirect2_jump"): Disable for branch conversion.
3010 ("casesi_jump"): Turn into expander and expand patterns for branch
3012 ("return_use"): New expander.
3013 ("*return"): Emit return via thunk and rename it to ...
3014 ("*return<mode>"): ... this one.
3015 * config/s390/s390.opt: Add new options and and enum for the
3018 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3020 * lra-constraints.c (match_reload): Unconditionally use
3021 gen_lowpart_SUBREG, rather than selecting between that
3022 and equivalent gen_rtx_SUBREG code.
3024 2018-02-08 Richard Biener <rguenther@suse.de>
3026 PR tree-optimization/84233
3027 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3028 changed flag instead of boguously re-using phi_inserted.
3030 2018-02-08 Martin Jambor <mjambor@suse.cz>
3032 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3033 static local variables.
3035 2018-02-08 Richard Biener <rguenther@suse.de>
3037 PR tree-optimization/84278
3038 * tree-vect-stmts.c (vectorizable_store): When looking for
3039 smaller vector types to perform grouped strided loads/stores
3040 make sure the mode is supported by the target.
3041 (vectorizable_load): Likewise.
3043 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3045 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3046 Increase LDP/STP opportunities by adding adjacent callee-saves.
3048 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3050 PR rtl-optimization/84068
3051 PR rtl-optimization/83459
3052 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3054 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3056 PR tree-optimization/84224
3057 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3058 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3061 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3064 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3065 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3066 count, remove one USE.
3068 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3070 * doc/install.texi (Configuration): Document the
3071 --with-long-double-format={ibm,ieee} PowerPC configuration
3075 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3076 Convert from define_expand to be define_insn_and_split. Rework
3077 float/double/_Float128 conversions to QI/HI/SImode to work with
3078 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3079 conversions to QI/HImode types did a store and then a load to
3080 truncate the value. For conversions to VSX registers, don't split
3081 the insn, instead emit the code directly. Use the code iterator
3082 any_fix to combine signed and unsigned conversions.
3083 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3084 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3085 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3086 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3087 (fix_<mode>di2_hw): Likewise.
3088 (fixuns_<mode>di2_hw): Likewise.
3089 (fix_<mode>si2_hw): Likewise.
3090 (fixuns_<mode>si2_hw): Likewise.
3091 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3092 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3093 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3094 fix<uns>_trunc<SFDF:mode>si2_p8.
3095 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3097 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3098 (fix<uns>_<mode>_mem): Likewise.
3099 (fctiw<u>z_<mode>_mem): Likewise.
3100 (fix<uns>_<mode>_mem): Likewise.
3101 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3102 the register allocator from doing a direct move to the GPRs to do
3103 a store, and instead use the ISA 3.0 store byte/half-word from
3104 vector register instruction. For IEEE 128-bit floating point,
3105 also optimize stores of 32-bit ints.
3106 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3108 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3110 * genextract.c (push_pathstr_operand): New function to support
3112 (walk_rtx): Call push_pathstr_operand.
3113 (print_path): Support [a-zA-Z].
3115 2018-02-07 Richard Biener <rguenther@suse.de>
3117 PR tree-optimization/84037
3118 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3119 (cse_and_gimplify_to_preheader): Declare.
3120 (vect_get_place_in_interleaving_chain): Likewise.
3121 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3123 (_loop_vec_info::~_loop_vec_info): Delete it.
3124 (cse_and_gimplify_to_preheader): New function.
3125 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3126 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3127 (vectorizable_load): Likewise. For grouped stores always base
3128 the IV on the first element.
3129 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3130 condition before gimplifying.
3132 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3134 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3135 *DIV_EXPR and *MOD_EXPR.
3137 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3140 * config/i386/i386.c (ix86_option_override_internal): Mask out
3141 the CF_SET bit when checking -fcf-protection.
3143 2018-02-07 Tom de Vries <tom@codesourcery.com>
3146 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3149 2018-02-07 Richard Biener <rguenther@suse.de>
3151 PR tree-optimization/84204
3152 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3155 PR tree-optimization/84205
3156 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3157 special-case isl_ast_op_zdiv_r.
3159 PR tree-optimization/84223
3160 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3161 Only add conditions from within the region.
3162 (gather_bbs::after_dom_children): Adjust.
3164 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3167 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3168 * config/avr/avr.md: Only post-reload split REG-REG moves if
3169 either register is GENERAL_REG_P.
3171 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3173 PR tree-optimization/84235
3174 * tree-ssa-scopedtables.c
3175 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3176 if the subtraction is performed in floating point type where NaNs are
3177 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3178 build 1. Formatting fix.
3180 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3183 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3184 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3185 and skip it regardless of bb boundaries. Use CALL_P macro,
3186 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3187 unnecessarily, formatting fix.
3189 2018-02-06 Michael Collison <michael.collison@arm.com>
3191 * config/arm/thumb2.md:
3192 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3193 (*thumb_mov_notscc): Ditto.
3195 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3198 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3199 unsigned_fix, not "s".
3201 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3203 * configure.ac (gcc_fn_eh_frame_ro): New function.
3204 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3205 correct .eh_frame permissions.
3206 * configure: Regenerate.
3208 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3210 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3213 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3215 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3216 Display warning message for -mno-speculate-indirect-jumps.
3218 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3220 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3222 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3224 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3226 PR tree-optimization/84225
3227 * tree-eh.c (find_trapping_overflow): Only call
3228 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3230 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3233 * config/i386/i386.c: Reimplement the check of possible options
3234 -mibt/-mshstk conbination. Change error messages.
3235 * doc/invoke.texi: Fix a typo: remove extra '='.
3237 2018-02-06 Marek Polacek <polacek@redhat.com>
3239 PR tree-optimization/84228
3240 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3242 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3245 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3246 emitted arch directives.
3247 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3248 __ARM_FEATURE_COPROC before changing architectures.
3250 2018-02-06 Richard Biener <rguenther@suse.de>
3252 * config/i386/i386.c (print_reg): Fix typo.
3253 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3255 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3257 * configure: Regenerate.
3259 2018-02-05 Martin Sebor <msebor@redhat.com>
3261 PR tree-optimization/83369
3262 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3265 2018-02-05 Martin Liska <mliska@suse.cz>
3267 * doc/invoke.texi: Cherry-pick upstream r323995.
3269 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3271 * ira.c (ira_init_register_move_cost): Adjust comment.
3273 2018-02-05 Martin Liska <mliska@suse.cz>
3275 PR gcov-profile/84137
3276 * doc/gcov.texi: Fix typo in documentation.
3278 2018-02-05 Martin Liska <mliska@suse.cz>
3280 PR gcov-profile/83879
3281 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3282 using dlopen functionality.
3284 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3286 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3287 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3288 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3289 _mm_maskz_range_round_ss): New intrinsics.
3290 (__builtin_ia32_rangesd128_round)
3291 (__builtin_ia32_rangess128_round): Remove.
3292 (__builtin_ia32_rangesd128_mask_round,
3293 __builtin_ia32_rangess128_mask_round): New builtins.
3294 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3295 __builtin_ia32_rangess128_round): Remove.
3296 (__builtin_ia32_rangesd128_mask_round,
3297 __builtin_ia32_rangess128_mask_round): New builtins.
3298 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3299 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3300 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3301 "<round_saeonly_constraint>")): Changed to ...
3302 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3303 "<round_saeonly_scalar_constraint>")): ... this.
3304 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3305 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3306 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3307 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3308 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3310 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3312 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3314 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3315 Remove all values except native, 8540 and 8548.
3317 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3319 * config/i386/i386.c (ix86_output_function_return): Pass
3320 INVALID_REGNUM, instead of -1, as invalid register number to
3321 indirect_thunk_name and output_indirect_thunk.
3323 2018-02-02 Julia Koval <julia.koval@intel.com>
3325 * config.gcc: Add -march=icelake.
3326 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3327 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3328 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3329 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3330 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3331 (processor_target_table): Add icelake.
3332 (ix86_option_override_internal): Handle new PTAs.
3333 (get_builtin_code_for_version): Handle icelake.
3334 (M_INTEL_COREI7_ICELAKE): New.
3335 (fold_builtin_cpu): Handle icelake.
3336 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3337 * doc/invoke.texi: Add -march=icelake.
3339 2018-02-02 Julia Koval <julia.koval@intel.com>
3341 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3342 to wide_int_bitmask.
3343 * wide-int-bitmask.h: New.
3345 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3348 * config/i386/i386.md: Replace Pmode with word_mode in
3349 builtin_setjmp_setup and builtin_longjmp to support x32.
3351 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3355 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3357 (rs6000_supported_cpu_names): New static variable.
3358 (linux_cpu_translation_table): Likewise.
3359 (elf_platform) <cpu>: Define new static variable and use it.
3360 Translate kernel AT_PLATFORM name to canonical name if needed.
3361 Error if platform name is unknown.
3363 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3366 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3368 2018-02-01 Jeff Law <law@redhat.com>
3371 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3372 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3373 the scratch if RELEASE_VIA_POP is false.
3374 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3375 If we have to save a temporary register, decrement SIZE appropriately.
3376 Pass new arguments to release_scratch_register_on_entry.
3377 (ix86_adjust_stack_and_probe): Likewise.
3378 (ix86_emit_probe_stack_range): Pass new arguments to
3379 release_scratch_register_on_entry.
3381 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3383 PR rtl-optimization/84157
3384 * combine.c (change_zero_ext): Use REG_P predicate in
3385 front of HARD_REGISTER_P predicate.
3387 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3389 * config/avr/avr.c (avr_option_override): Move disabling of
3390 -fdelete-null-pointer-checks to...
3391 * common/config/avr/avr-common.c (avr_option_optimization_table):
3394 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3396 PR tree-optimization/81635
3397 * tree-data-ref.c (split_constant_offset_1): For types that
3398 wrap on overflow, try to use range info to prove that wrapping
3401 2018-02-01 Renlin Li <renlin.li@arm.com>
3404 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3406 (aarch64_register_move_cost): Likewise.
3407 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3409 (REG_CLASS_NAMES): Likewise.
3410 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3411 TAILCALL_ADDR_REGS. Remove IP registers.
3412 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3414 2018-02-01 Richard Biener <rguenther@suse.de>
3416 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3417 for specifying RPO order and allow NULL for that.
3418 * domwalk.c (dom_walker::dom_walker): Likewise.
3419 (dom_walker::walk): Handle NULL RPO order.
3420 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3422 (rewrite_update_dom_walker): Likewise.
3423 (mark_def_dom_walker): Likewise.
3425 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3427 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3428 (aarch64_maybe_expand_sve_subreg_move): Declare.
3429 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3430 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3432 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3433 that are semantically a reverse operation.
3434 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3435 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3436 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3438 (aarch64_can_change_mode_class): For big-endian, forbid changes
3439 between two SVE modes if they have different element sizes.
3441 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3443 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3444 the TImode handling for big-endian targets.
3446 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3448 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3449 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3451 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3452 Remove BSWAP handing for big-endian targets and use the form of
3453 LD1RQ appropariate for the mode.
3455 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3457 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3458 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3461 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3464 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3465 check for operands that need to go through aarch64_sve_reload_be.
3467 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3469 PR tree-optimization/81661
3470 PR tree-optimization/84117
3471 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3472 * tree-eh.c: Include gimplify.h.
3473 (find_trapping_overflow, replace_trapping_overflow,
3474 rewrite_to_non_trapping_overflow): New functions.
3475 * tree-vect-loop.c: Include tree-eh.h.
3476 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3477 * tree-data-ref.c: Include tree-eh.h.
3478 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3480 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3482 PR rtl-optimization/84123
3483 * combine.c (change_zero_ext): Check if hard register satisfies
3484 can_change_dest_mode before calling gen_lowpart_SUBREG.
3486 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3489 * ira.c (ira_init_register_move_cost): Remove assert.
3491 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3493 PR rtl-optimization/84071
3494 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3495 * doc/tm.texi: Regenerate.
3497 2018-01-31 Richard Biener <rguenther@suse.de>
3499 PR tree-optimization/84132
3500 * tree-data-ref.c (analyze_miv_subscript): Properly
3501 check whether evolution_function_is_affine_multivariate_p
3502 before calling gcd_of_steps_may_divide_p.
3504 2018-01-31 Julia Koval <julia.koval@intel.com>
3507 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3508 * config/i386/i386.md (rdpid_rex64) New.
3509 (rdpid): Make 32bit only.
3511 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3514 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3515 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3517 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3520 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3522 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3524 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3526 PR rtl-optimization/84071
3527 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3528 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3530 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3532 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3533 (arc_attribute_table): Add 'aux' attribute.
3534 (arc_in_small_data_p): Consider aux like variables.
3535 (arc_is_aux_reg_p): New function.
3536 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3537 (arc_get_aux_arg): New function.
3538 (prepare_move_operands): Handle aux-register access.
3539 (arc_handle_aux_attribute): New function.
3540 * doc/extend.texi (ARC Variable attributes): Add subsection.
3542 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3544 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3545 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3546 (arc_attribute_table): Add 'uncached' attribute.
3547 (arc_print_operand): Print '.di' flag for uncached memory
3549 (arc_in_small_data_p): Do not consider for small data the uncached
3551 (arc_is_uncached_mem_p): New function.
3552 * config/arc/predicates.md (compact_store_memory_operand): Check
3553 for uncached memory accesses.
3554 (nonvol_nonimm_operand): Likewise.
3555 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3557 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3560 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3561 falign-loops=): Add Optimization flag.
3563 2018-01-30 Jeff Law <law@redhat.com>
3566 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3567 INT_REGISTERS_SAVED. Check it prior to calling
3568 get_scratch_register_on_entry.
3569 (ix86_adjust_stack_and_probe): Similarly.
3570 (ix86_emit_probe_stack_range): Similarly.
3571 (ix86_expand_prologue): Corresponding changes.
3573 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3576 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3577 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3579 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3582 * lra-constraints.c (curr_insn_transform): Process AND in the
3585 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3587 PR rtl-optimization/83986
3588 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3589 dependence against last_pending_memory_flush in addition to
3592 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3594 PR tree-optimization/81611
3595 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3598 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3601 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3604 2018-01-30 Richard Biener <rguenther@suse.de>
3605 Jakub Jelinek <jakub@redhat.com>
3607 PR tree-optimization/84111
3608 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3609 inner loops added during recursion, as they don't have up-to-date
3612 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3615 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3616 (can_inline_edge_by_limits_p): ... here.
3617 (can_early_inline_edge_p, check_callers,
3618 update_caller_keys, update_callee_keys, recursive_inlining,
3619 add_new_edges_to_heap, speculation_useful_p,
3620 inline_small_functions,
3621 inline_small_functions, flatten_function,
3622 inline_to_all_callers_1): Update.
3624 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3626 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3629 2018-01-30 Richard Biener <rguenther@suse.de>
3631 PR tree-optimization/83008
3632 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3633 invariant and constant vector uses in stmts when they need
3636 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3639 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3640 * configure: Regenerate.
3642 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3644 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3646 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3647 Use gen_rtx_REG rather than gen_lowpart.
3649 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3651 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3652 rather than 0 when creating partial subregs.
3654 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3656 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3659 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3662 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3663 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3664 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3665 flags. This restores the settings used before the 2017-07-24.
3666 Turning off pre increment/decrement/modify allows IVOPTS to
3667 optimize DF/SF loops where the index is an int.
3669 2018-01-29 Richard Biener <rguenther@suse.de>
3670 Kelvin Nilsen <kelvin@gcc.gnu.org>
3673 * tree-vect-stmts.c (vectorizable_call): Don't call
3674 targetm.vectorize_builtin_md_vectorized_function if callee is
3677 2018-01-22 Carl Love <cel@us.ibm.com>
3679 * doc/extend.tex: Fix typo in second arg in
3680 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3682 2018-01-29 Richard Biener <rguenther@suse.de>
3684 PR tree-optimization/84086
3685 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3686 (flush_ssaname_freelist): When SSA names were released reset
3687 the SCEV hash table.
3689 2018-01-29 Richard Biener <rguenther@suse.de>
3691 PR tree-optimization/84057
3692 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3693 removed paths when removing edges.
3695 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3697 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3698 -mfunction-return=@var{choice}.
3700 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3703 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3705 (layout::print_source_line): Likewise.
3706 (test_get_line_width_without_trailing_whitespace): Add test cases.
3708 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3711 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3714 2018-01-26 Jim Wilson <jimw@sifive.com>
3716 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3718 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3721 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3723 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3724 and CMP + SUB-immediate -> SUBS.
3726 2018-01-26 Martin Sebor <msebor@redhat.com>
3728 PR tree-optimization/83896
3729 * tree-ssa-strlen.c (get_string_len): Rename...
3730 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3731 Avoid assuming length is constant.
3732 (handle_char_store): Use HOST_WIDE_INT for string length.
3734 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3737 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3738 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3740 2018-01-26 Richard Biener <rguenther@suse.de>
3742 PR rtl-optimization/84003
3743 * dse.c (record_store): Only record redundant stores when
3744 the earlier store aliases at least all accesses the later one does.
3746 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3748 PR rtl-optimization/83985
3749 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3750 REG_CFA_RESTORE insns.
3751 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3752 REG_CFA_RESTORE insns here.
3755 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3756 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3758 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3760 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3761 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3762 (arc_init): Likewise.
3763 (arc_override_options): Likewise.
3764 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3766 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3768 * config/arc/arc.h (TARGET_DBNZ): Define.
3769 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3770 properly set the tune attribute.
3771 (dbnz): Use TARGET_DBNZ guard.
3772 * config/arc/arc.opt (mtune): Add core3 option.
3774 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3776 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3777 recognize new pic like addresses.
3778 (arc_delegitimize_address): Clean up.
3780 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3782 * config/arc/arc-arches.def: Option mrf16 valid for all
3784 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3785 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3786 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3787 * config/arc/arc-tables.opt: Regenerate.
3788 * config/arc/arc.c (arc_conditional_register_usage): Handle
3789 reduced register file case.
3790 (arc_file_start): Set must have build attributes.
3791 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3793 * config/arc/arc.opt (mrf16): Add new option.
3794 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3795 * config/arc/genmultilib.awk: Handle new mrf16 option.
3796 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3797 * config/arc/t-multilib: Regenerate.
3798 * doc/invoke.texi (ARC Options): Document mrf16 option.
3800 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3802 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3803 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3804 (arc_attribute_table): Add 'secure_call' attribute.
3805 (arc_print_operand): Print secure call operand.
3806 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3808 (arc_is_secure_call_p): New function. * config/arc/arc.md
3809 (call_i): Add support for sjli instruction.
3810 (call_value_i): Likewise.
3811 * config/arc/constraints.md (Csc): New constraint.
3813 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3814 John Eric Martin <John.Martin@emmicro-us.com>
3816 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3817 * config/arc/arc.c (_arc_jli_section): New struct.
3818 (arc_jli_section): New type.
3819 (rc_jli_sections): New static variable.
3820 (arc_handle_jli_attribute): New function.
3821 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3822 (arc_file_end): New function.
3823 (TARGET_ASM_FILE_END): Define.
3824 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3825 (arc_add_jli_section): New function.
3826 (jli_call_scan): Likewise.
3827 (arc_reorg): Call jli_call_scan.
3828 (arc_output_addsi): Remove 'S' from printing asm operand.
3829 (arc_is_jli_call_p): New function.
3830 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3832 (movhi_insn): Likewise.
3833 (movsi_insn): Likewise.
3834 (movsi_set_cc_insn): Likewise.
3835 (loadqi_update): Likewise.
3836 (load_zeroextendqisi_update): Likewise.
3837 (load_signextendqisi_update): Likewise.
3838 (loadhi_update): Likewise.
3839 (load_zeroextendhisi_update): Likewise.
3840 (load_signextendhisi_update): Likewise.
3841 (loadsi_update): Likewise.
3842 (loadsf_update): Likewise.
3843 (movsicc_insn): Likewise.
3844 (bset_insn): Likewise.
3845 (bxor_insn): Likewise.
3846 (bclr_insn): Likewise.
3847 (bmsk_insn): Likewise.
3848 (bicsi3_insn): Likewise.
3849 (cmpsi_cc_c_insn): Likewise.
3850 (movsi_ne): Likewise.
3851 (movsi_cond_exec): Likewise.
3852 (clrsbsi2): Likewise.
3863 (call_i): Remove 'S' asm letter, add jli instruction.
3864 (call_value_i): Likewise.
3865 * config/arc/arc.op (mjli-always): New option.
3866 * config/arc/constraints.md (Cji): New constraint.
3867 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3869 (subsf3_fpx): Likewise.
3870 (mulsf3_fpx): Likewise.
3871 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3873 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3875 * doc/invoke.texi (ARC): Document mjli-always option.
3877 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3879 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3880 avoid addition with 0 and use incw and decw where possible.
3882 2018-01-26 Richard Biener <rguenther@suse.de>
3884 PR tree-optimization/81082
3885 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3886 association if it requires casting to unsigned.
3887 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3888 from fold_plusminus_mult_expr to catch important cases late when
3889 range info is available.
3891 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3893 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3894 * configure.ac (hidden_linkonce): New test.
3895 * configure: Regenerate.
3896 * config.in: Regenerate.
3898 2018-01-26 Julia Koval <julia.koval@intel.com>
3900 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3901 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3902 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3903 _mm_mask_bitshuffle_epi64_mask): Fix type.
3904 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3905 USI_FTYPE_V4DI_V4DI_USI): Remove.
3906 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3907 __builtin_ia32_vpshufbitqmb256_mask,
3908 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3909 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3910 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3912 2018-01-26 Alan Modra <amodra@gmail.com>
3915 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3916 UNSPEC_VBPERMQ. Sort other unspecs.
3918 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3920 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3922 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3925 * predict.c (drop_profile): Do not push/pop cfun; update also
3927 (handle_missing_profiles): Fix logic looking for zero profiles.
3929 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3932 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3933 on functions with #pragma omp declare simd or functions with simd
3935 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3936 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3937 Remove trailing \n from warning_at calls.
3939 2018-01-25 Tom de Vries <tom@codesourcery.com>
3942 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3943 for neutered workers.
3945 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3948 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3949 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3951 2018-01-24 Jeff Law <law@redhat.com>
3954 * i386.c (get_probe_interval): Move to earlier point.
3955 (ix86_compute_frame_layout): If -fstack-clash-protection and
3956 the frame is larger than the probe interval, then use pushes
3957 to save registers rather than reg->mem moves.
3958 (ix86_expand_prologue): Remove conditional for int_registers_saved
3961 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3964 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3965 min/max for never referenced object.
3967 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3970 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3972 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3973 attributes from DECL_ATTRIBUTES (decl) without affecting
3974 DECL_ATTRIBUTES (current_function_decl).
3975 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3976 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3978 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3980 PR tree-optimization/83979
3981 * fold-const.c (fold_comparison): Use constant_boolean_node
3982 instead of boolean_{true,false}_node.
3984 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3986 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3989 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3991 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3992 Simplify the clause that sets the length attribute.
3993 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3994 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3995 clause that sets the length attribute.
3996 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3998 2018-01-24 Tom de Vries <tom@codesourcery.com>
4001 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4002 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4003 Add strict parameter.
4004 (prevent_branch_around_nothing): Insert dummy insn between branch to
4005 label and label with no ptx insn inbetween.
4006 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4008 2018-01-24 Tom de Vries <tom@codesourcery.com>
4011 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4012 for neutered threads in warp.
4013 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4015 2018-01-24 Richard Biener <rguenther@suse.de>
4017 PR tree-optimization/83176
4018 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4021 2018-01-24 Richard Biener <rguenther@suse.de>
4023 PR tree-optimization/82819
4024 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4025 code generating pluses that are no-ops in the target precision.
4027 2018-01-24 Richard Biener <rguenther@suse.de>
4030 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4032 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4034 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4035 to merge probabilities.
4036 * predict.c (probably_never_executed): Also mark as cold functions
4037 with global 0 profile and guessed local profile.
4038 * profile-count.c (profile_probability::combine_with_count): New
4040 * profile-count.h (profile_probability::operator*,
4041 profile_probability::operator*=, profile_probability::operator/,
4042 profile_probability::operator/=): Reduce precision to adjusted
4043 and set value to guessed on contradictory divisions.
4044 (profile_probability::combine_with_freq): Remove.
4045 (profile_probability::combine_wiht_count): Declare.
4046 (profile_count::force_nonzero):: Set to adjusted.
4047 (profile_count::probability_in):: Set quality to adjusted.
4048 * tree-ssa-tail-merge.c (replace_block_by): Use
4051 2018-01-23 Andrew Waterman <andrew@sifive.com>
4052 Jim Wilson <jimw@sifive.com>
4054 * config/riscv/riscv.c (riscv_stack_boundary): New.
4055 (riscv_option_override): Set riscv_stack_boundary. Handle
4056 riscv_preferred_stack_boundary_arg.
4057 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4058 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4059 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4060 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4061 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4062 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4064 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4067 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4068 of struct ix86_frame.
4069 (ix86_expand_epilogue): Likewise. Add a local variable for
4070 the reg_save_offset field in struct ix86_frame.
4072 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4074 PR tree-optimization/82604
4075 * tree-loop-distribution.c (enum partition_kind): New enum item
4076 PKIND_PARTIAL_MEMSET.
4077 (partition_builtin_p): Support above new enum item.
4078 (generate_code_for_partition): Ditto.
4079 (compute_access_range): Differentiate cases that equality can be
4080 proven at all loops, the innermost loops or no loops.
4081 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4082 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4083 (finalize_partitions, distribute_loop): Don't fuse partition of
4084 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4085 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4088 2018-01-23 Martin Liska <mliska@suse.cz>
4090 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4091 order to ignore the predictor.
4092 (PRED_POLYMORPHIC_CALL): Likewise.
4093 (PRED_RECURSIVE_CALL): Likewise.
4095 2018-01-23 Martin Liska <mliska@suse.cz>
4097 * tree-profile.c (tree_profiling): Print function header to
4098 aware reader which function we are working on.
4099 * value-prof.c (gimple_find_values_to_profile): Do not print
4100 not interesting value histograms.
4102 2018-01-23 Martin Liska <mliska@suse.cz>
4104 * profile-count.h (enum profile_quality): Add
4105 profile_uninitialized as the first value. Do not number values
4106 as they are zero based.
4107 (profile_count::verify): Update sanity check.
4108 (profile_probability::verify): Likewise.
4110 2018-01-23 Nathan Sidwell <nathan@acm.org>
4112 * doc/invoke.texi (ffor-scope): Deprecate.
4114 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4116 PR tree-optimization/83510
4117 * domwalk.c (set_all_edges_as_executable): New function.
4118 (dom_walker::dom_walker): Convert bool param
4119 "skip_unreachable_blocks" to enum reachability. Move setup of
4120 edge flags to set_all_edges_as_executable and only do it when
4121 reachability is REACHABLE_BLOCKS.
4122 * domwalk.h (enum dom_walker::reachability): New enum.
4123 (dom_walker::dom_walker): Convert bool param
4124 "skip_unreachable_blocks" to enum reachability.
4125 (set_all_edges_as_executable): New decl.
4126 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4127 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4129 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4130 but converting true to REACHABLE_BLOCKS.
4131 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4133 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4134 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4135 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4137 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4138 if check_all_array_refs will be called.
4140 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4142 * tree.c (selftest::test_location_wrappers): Add more test
4145 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4147 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4148 (selftest::test_bit_in_range): Likewise.
4150 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4153 * doc/sourcebuild.texi (vect_float): Say that the selector
4154 only describes the situation when -funsafe-math-optimizations is on.
4155 (vect_float_strict): Document.
4157 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4159 PR tree-optimization/83965
4160 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4161 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4162 instead of checking only for a reduction.
4163 (vect_recog_widen_sum_pattern): Likewise.
4165 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4167 * predict.c (probably_never_executed): Only use precise profile info.
4168 (compute_function_frequency): Skip after inlining hack since we now
4169 have quality checking.
4171 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4173 * profile-count.h (profile_probability::very_unlikely,
4174 profile_probability::unlikely, profile_probability::even): Set
4175 precision to guessed.
4177 2018-01-23 Richard Biener <rguenther@suse.de>
4179 PR tree-optimization/83963
4180 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4181 Properly terminate dominator walk when crossing the exit edge not
4182 when visiting its source block.
4184 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4187 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4188 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4190 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4192 PR tree-optimization/83957
4193 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4194 semicolon after for body surrounded by braces.
4196 PR tree-optimization/83081
4197 * profile-count.h (profile_probability::split): New method.
4198 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4199 Use profile_probability::split.
4200 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4201 when splitting a single conditional jump into 2.
4203 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4205 PR tree-optimization/69452
4206 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4209 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4211 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4212 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4213 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4215 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4217 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4218 * config/rl78/rl78.md (movdi): New define_expand.
4219 * config/rl78/rl78.c (rl78_split_movdi): New function.
4221 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4224 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4226 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4227 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4228 128-bit to produce an UNSPEC move to get the double word with the
4229 signbit and then a shift directly to do signbit.
4230 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4231 implementation with a new version that just does either a direct
4232 move or a regular move. Move memory interface to separate insns.
4233 Move insns so they are next to the expander.
4234 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4235 with signbit move. Split big and little endian case.
4236 (signbit<mode>2_dm_mem_le): Likewise.
4237 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4238 (signbit<mode>2_dm2): Likewise.
4240 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4242 * config/rl78/rl78.md (anddi3): New define_expand.
4244 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4246 * config/rl78/rl78.md (umindi3): New define_expand.
4248 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4250 * config/rl78/rl78.md (smindi3): New define_expand.
4252 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4254 * config/rl78/rl78.md (smaxdi3): New define_expand.
4256 2018-01-22 Carl Love <cel@us.ibm.com>
4258 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4259 LVX_V1TI): Add macro expansion.
4260 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4261 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4262 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4263 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4264 Change check to determine if the instruction is a byte reversing
4265 entry. Fix typo in comment.
4266 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4267 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4268 Add def_builtin calls for new builtins.
4269 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4270 Add define_insn expansion.
4272 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4274 * config/rl78/rl78.md (umaxdi3): New define_expand.
4276 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4278 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4279 for non-QImode registers.
4281 2018-01-22 Richard Biener <rguenther@suse.de>
4283 PR tree-optimization/83963
4284 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4285 including the loop exit block.
4286 (scop_detection::merge_sese): Likewise.
4287 (scop_detection::add_scop): Do it here instead.
4289 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4291 * doc/sourcebuild.texi (arm_softfloat): Document.
4293 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4296 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4297 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4298 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4300 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4301 David Edelsohn <dje.gcc@gmail.com>
4304 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4305 Change "crset eq" to "crset 2".
4306 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4307 (*call_indirect_aix<mode>_nospec): Likewise.
4308 (*call_value_indirect_aix<mode>_nospec): Likewise.
4309 (*call_indirect_elfv2<mode>_nospec): Likewise.
4310 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4311 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4312 change assembly output from . to $.
4313 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4314 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4315 (*tablejump<mode>_internal1_nospec): Likewise.
4317 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4320 * config/sh/sh_optimize_sett_clrt.cc:
4321 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4323 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4325 PR tree-optimization/83940
4326 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4327 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4328 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4329 use it to pass back the definition type.
4330 (vect_check_store_rhs): Likewise rhs_dt_out.
4331 (vect_build_gather_load_calls): Add a mask_dt argument and use
4332 it instead of a call to vect_is_simple_use.
4333 (vectorizable_store): Update calls to vect_check_load_store_mask
4334 and vect_check_store_rhs. Use the dt returned by the latter instead
4335 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4336 instead of calls to vect_is_simple_use. Pass the scalar rather
4337 than the vector operand to vect_is_simple_use when handling
4338 second and subsequent copies of an rhs value.
4339 (vectorizable_load): Update calls to vect_check_load_store_mask
4340 and vect_build_gather_load_calls. Use the cached mask_dt and
4341 gs_info.offset_dt instead of calls to vect_is_simple_use.
4343 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4346 * tree-emutls.c: Include gimplify.h.
4347 (lower_emutls_2): New function.
4348 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4349 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4350 it before further processing.
4353 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4354 UINTVAL (trueop1) instead of INTVAL (op1).
4356 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4360 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4361 INCOMING_FRAME_SP_OFFSET if not defined.
4362 (scan_trace): Add ENTRY argument. If true and
4363 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4364 emit a note to adjust the CFA offset.
4365 (create_cfi_notes): Adjust scan_trace callers.
4366 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4367 INCOMING_FRAME_SP_OFFSET in the CIE.
4368 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4369 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4371 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4372 * doc/tm.texi: Regenerated.
4374 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4376 PR rtl-optimization/83147
4377 * lra-constraints.c (remove_inheritance_pseudos): Use
4378 lra_substitute_pseudo_within_insn.
4380 2018-01-19 Tom de Vries <tom@codesourcery.com>
4381 Cesar Philippidis <cesar@codesourcery.com>
4384 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4386 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4389 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4390 spaces for function labels.
4392 2018-01-19 Martin Liska <mliska@suse.cz>
4394 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4395 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4396 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4397 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4398 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4399 (PRED_CONST_RETURN): Change from 69 to 65.
4400 (PRED_NULL_RETURN): Change from 91 to 71.
4401 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4402 (PRED_LOOP_GUARD): Change from 66 to 73.
4404 2018-01-19 Martin Liska <mliska@suse.cz>
4406 * predict.c (predict_insn_def): Add new assert.
4407 (struct branch_predictor): Change type to signed integer.
4408 (test_prediction_value_range): Amend test to cover
4410 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4411 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4412 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4413 (PRED_LOOP_IV_COMPARE): Likewise.
4414 * predict.h (PROB_UNINITIALIZED): Define new constant.
4416 2018-01-19 Martin Liska <mliska@suse.cz>
4418 * predict.c (dump_prediction): Add new format for
4419 analyze_brprob.py script which is enabled with -details
4421 * profile-count.h (precise_p): New function.
4423 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4425 PR tree-optimization/83922
4426 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4427 there are no statements that need masking.
4428 (vect_active_double_reduction_p): New function.
4429 (vect_analyze_loop_operations): Use it when handling phis that
4430 are not in the loop header.
4432 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4434 PR tree-optimization/83914
4435 * tree-vect-loop.c (vectorizable_induction): Don't convert
4436 init_expr or apply the peeling adjustment for inductions
4437 that are nested within the vectorized loop.
4439 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4441 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4444 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4448 * function.h (gimplify_parameters): Add gimple_seq * argument.
4449 * function.c: Include gimple.h and options.h.
4450 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4451 for the added local temporaries if needed.
4452 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4453 if there are any parameter cleanups, wrap whole body into a
4454 try/finally with the cleanups.
4456 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4459 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4460 Use GET_MODE_CLASS for scalar floating point.
4462 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4466 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4467 Fix call of call_cgraph_insertion_hooks.
4469 2018-01-18 Martin Sebor <msebor@redhat.com>
4471 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4473 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4476 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4479 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4482 * common.opt: (-ffile-prefix-map): New option.
4483 * opts.c (common_handle_option): Defer it.
4484 * opts-global.c (handle_common_deferred_options): Handle it.
4485 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4486 * file-prefix-map.h: New file.
4487 (remap_debug_filename, add_debug_prefix_map): ...here.
4488 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4489 * final.c (debug_prefix_map, add_debug_prefix_map
4490 remap_debug_filename): Move to...
4491 * file-prefix-map.c: New file.
4492 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4493 generalize, get rid of alloca(), use strrchr() instead of strchr().
4494 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4495 Implement in terms of add_prefix_map().
4496 (remap_macro_filename, remap_debug_filename): Implement in term of
4498 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4499 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4500 * dbxout.c: Include file-prefix-map.h.
4501 * varasm.c: Likewise.
4502 * vmsdbgout.c: Likewise.
4503 * xcoffout.c: Likewise.
4504 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4505 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4506 * doc/invoke.texi (-ffile-prefix-map): Document.
4507 (-fdebug-prefix-map): Update description.
4509 2018-01-18 Martin Liska <mliska@suse.cz>
4511 * config/i386/i386.c (indirect_thunk_name): Document that also
4513 (output_indirect_thunk): Document why both instructions
4514 (pause and lfence) are generated.
4516 2018-01-18 Richard Biener <rguenther@suse.de>
4518 PR tree-optimization/83887
4519 * graphite-scop-detection.c
4520 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4521 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4522 (scop_detection::merge_sese): Re-implement with a flood-fill
4523 algorithm that properly finds a SESE region if it exists.
4525 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4528 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4529 pointer_diff optimizations use view_convert instead of convert.
4531 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4533 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4534 Generate different code for -mno-speculate-indirect-jumps.
4535 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4536 (*call_indirect_aix<mode>): Disable for
4537 -mno-speculate-indirect-jumps.
4538 (*call_indirect_aix<mode>_nospec): New define_insn.
4539 (*call_value_indirect_aix<mode>): Disable for
4540 -mno-speculate-indirect-jumps.
4541 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4542 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4543 -mno-speculate-indirect-jumps.
4544 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4546 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4548 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4549 long double type, set the flags for noting the default long double
4550 type, even if we don't pass or return a long double type.
4552 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4555 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4558 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4560 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4561 support for merge[hl].
4562 (fold_mergehl_helper): New helper function.
4563 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4564 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4565 (altivec_vmrglw_direct): Add xxmrglw insn.
4567 2018-01-17 Andrew Waterman <andrew@sifive.com>
4569 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4570 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4572 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4575 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4576 call the lto_location_cache before reading the
4577 DECL_SOURCE_LOCATION of the types.
4579 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4580 Richard Sandiford <richard.sandiford@linaro.org>
4582 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4583 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4584 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4585 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4586 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4588 * config/aarch64/constraints.md (aarch64_movti_operand):
4590 * config/aarch64/predicates.md (Uti): Add new constraint.
4592 2018-01-17 Carl Love <cel@us.ibm.com>
4594 * config/rs6000/vsx.md (define_expand xl_len_r,
4595 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4596 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4598 (define_expand, define_insn): Move the shift left from the
4599 define_insn to the define_expand for lxvl and stxvl instructions.
4600 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4601 and XL_LEN_R definitions to PURE.
4603 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4605 * config/i386/i386.c (indirect_thunk_name): Declare regno
4606 as unsigned int. Compare regno with INVALID_REGNUM.
4607 (output_indirect_thunk): Ditto.
4608 (output_indirect_thunk_function): Ditto.
4609 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4610 in the call to output_indirect_thunk_function.
4612 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4615 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4616 rather than the size of inner_type to determine the stack slot size
4617 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4619 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4622 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4625 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4627 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4628 endian Linux systems to optionally enable multilibs for selecting
4629 the long double type if the user configured an explicit type.
4630 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4631 have no long double multilibs if not defined.
4632 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4633 warn if the user used -mabi={ieee,ibm}longdouble and we built
4634 multilibs for long double.
4635 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4636 appropriate multilib option.
4637 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4639 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4640 for building long double multilibs.
4641 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4643 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4645 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4648 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4650 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4653 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4656 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4659 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4661 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4662 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4664 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4666 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4667 different rtl trees depending on TARGET_64BIT.
4668 (rs6000_gen_lvx): Likewise.
4670 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4672 * config/visium/visium.md (nop): Tweak comment.
4673 (hazard_nop): Likewise.
4675 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4677 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4678 -mspeculate-indirect-jumps.
4679 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4680 for -mno-speculate-indirect-jumps.
4681 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4682 (*call_value_indirect_elfv2<mode>): Disable for
4683 -mno-speculate-indirect-jumps.
4684 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4685 (indirect_jump): Emit different RTL for
4686 -mno-speculate-indirect-jumps.
4687 (*indirect_jump<mode>): Disable for
4688 -mno-speculate-indirect-jumps.
4689 (*indirect_jump<mode>_nospec): New define_insn.
4690 (tablejump): Emit different RTL for
4691 -mno-speculate-indirect-jumps.
4692 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4693 (tablejumpsi_nospec): New define_expand.
4694 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4695 (tablejumpdi_nospec): New define_expand.
4696 (*tablejump<mode>_internal1): Disable for
4697 -mno-speculate-indirect-jumps.
4698 (*tablejump<mode>_internal1_nospec): New define_insn.
4699 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4702 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4704 * caller-save.c (insert_save): Drop unnecessary parameter. All
4707 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4708 Richard Biener <rguenth@suse.de>
4711 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4712 return early, inline manually is_gimple_sizepos. Make sure if we
4713 call gimplify_expr we don't end up with a gimple constant.
4714 * tree.c (variably_modified_type_p): Don't return true for
4715 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4716 * gimplify.h (is_gimple_sizepos): Remove.
4718 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4720 PR tree-optimization/83857
4721 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4722 vectorizable_live_operation for pure SLP statements.
4723 (vectorizable_live_operation): Handle PHIs.
4725 2018-01-16 Richard Biener <rguenther@suse.de>
4727 PR tree-optimization/83867
4728 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4729 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4731 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4734 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4735 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4736 If off is not INTEGER_CST, issue a may not be aligned warning
4737 rather than isn't aligned. Use isn%'t rather than isn't.
4738 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4740 <case MULT_EXPR>: Improve the case when bottom and one of the
4741 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4742 operand, in that case check if the other operand is multiple of
4743 bottom divided by the INTEGER_CST operand.
4745 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4748 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4749 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4750 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4751 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4752 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4753 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4754 (pa_function_arg_size): New function.
4756 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4758 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4759 in a separate statement.
4761 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4763 PR tree-optimization/83847
4764 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4765 group gathers and scatters.
4767 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4769 PR rtl-optimization/86620
4770 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4772 PR rtl-optimization/83213
4773 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4774 to last if both are JUMP_INSNs.
4776 PR tree-optimization/83843
4777 * gimple-ssa-store-merging.c
4778 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4779 store_immediate_info for bswap/nop orig_stores.
4781 2018-01-15 Andrew Waterman <andrew@sifive.com>
4783 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4785 <UDIV>: Increase cost if !TARGET_DIV.
4787 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4789 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4790 (define_attr "cr_logical_3op"): New.
4791 (cceq_ior_compare): Adjust.
4792 (cceq_ior_compare_complement): Adjust.
4793 (*cceq_rev_compare): Adjust.
4794 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4795 (is_cracked_insn): Adjust.
4796 (insn_must_be_first_in_group): Adjust.
4797 * config/rs6000/40x.md: Adjust.
4798 * config/rs6000/440.md: Adjust.
4799 * config/rs6000/476.md: Adjust.
4800 * config/rs6000/601.md: Adjust.
4801 * config/rs6000/603.md: Adjust.
4802 * config/rs6000/6xx.md: Adjust.
4803 * config/rs6000/7450.md: Adjust.
4804 * config/rs6000/7xx.md: Adjust.
4805 * config/rs6000/8540.md: Adjust.
4806 * config/rs6000/cell.md: Adjust.
4807 * config/rs6000/e300c2c3.md: Adjust.
4808 * config/rs6000/e500mc.md: Adjust.
4809 * config/rs6000/e500mc64.md: Adjust.
4810 * config/rs6000/e5500.md: Adjust.
4811 * config/rs6000/e6500.md: Adjust.
4812 * config/rs6000/mpc.md: Adjust.
4813 * config/rs6000/power4.md: Adjust.
4814 * config/rs6000/power5.md: Adjust.
4815 * config/rs6000/power6.md: Adjust.
4816 * config/rs6000/power7.md: Adjust.
4817 * config/rs6000/power8.md: Adjust.
4818 * config/rs6000/power9.md: Adjust.
4819 * config/rs6000/rs64.md: Adjust.
4820 * config/rs6000/titan.md: Adjust.
4822 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4824 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4825 ix86_indirect_branch_register logic.
4827 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4829 * config/i386/constraints.md (Bs): Update
4830 ix86_indirect_branch_register check. Don't check
4831 ix86_indirect_branch_register with GOT_memory_operand.
4833 * config/i386/predicates.md (GOT_memory_operand): Don't check
4834 ix86_indirect_branch_register here.
4835 (GOT32_symbol_operand): Likewise.
4837 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4839 * config/i386/predicates.md (constant_call_address_operand):
4840 Rewrite ix86_indirect_branch_register logic.
4841 (sibcall_insn_operand): Likewise.
4843 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4845 * config/i386/constraints.md (Bs): Replace
4846 ix86_indirect_branch_thunk_register with
4847 ix86_indirect_branch_register.
4849 * config/i386/i386.md (indirect_jump): Likewise.
4850 (tablejump): Likewise.
4851 (*sibcall_memory): Likewise.
4852 (*sibcall_value_memory): Likewise.
4853 Peepholes of indirect call and jump via memory: Likewise.
4854 * config/i386/i386.opt: Likewise.
4855 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4856 (GOT_memory_operand): Likewise.
4857 (call_insn_operand): Likewise.
4858 (sibcall_insn_operand): Likewise.
4859 (GOT32_symbol_operand): Likewise.
4861 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4864 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4865 type rather than type addr's type points to.
4866 (expand_omp_atomic_mutex): Likewise.
4867 (expand_omp_atomic): Likewise.
4869 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4872 * config/i386/i386.c (output_indirect_thunk_function): Use
4873 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4874 for __x86_return_thunk.
4876 2018-01-15 Richard Biener <rguenther@suse.de>
4879 * expmed.c (extract_bit_field_1): Fix typo.
4881 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4884 * config/arm/iterators.md (VF): New mode iterator.
4885 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4886 Remove integer-related logic from pattern.
4887 (neon_vabd<mode>_3): Likewise.
4889 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4892 * common.opt (fstrict-overflow): No longer an alias.
4893 (fwrapv-pointer): New option.
4894 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4895 also for pointer types based on flag_wrapv_pointer.
4896 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4897 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4898 opts->x_flag_wrapv got set.
4899 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4900 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4901 POINTER_TYPE_OVERFLOW_UNDEFINED.
4902 * match.pd: Likewise in address comparison pattern.
4903 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4905 2018-01-15 Richard Biener <rguenther@suse.de>
4908 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4909 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4910 Reset type names to their identifier if their TYPE_DECL doesn't
4911 have linkage (and thus is used for ODR and devirt).
4912 (save_debug_info_for_decl): Remove.
4913 (save_debug_info_for_type): Likewise.
4914 (add_tree_to_fld_list): Adjust.
4915 * tree-pretty-print.c (dump_generic_node): Make dumping of
4916 type names more robust.
4918 2018-01-15 Richard Biener <rguenther@suse.de>
4920 * BASE-VER: Bump to 8.0.1.
4922 2018-01-14 Martin Sebor <msebor@redhat.com>
4925 * builtins.c (check_access): Avoid warning when the no-warning bit
4928 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4930 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4931 * ira-color (allocno_hard_regs_compare): Likewise.
4933 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4936 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4937 Use .pushsection/.popsection.
4939 2018-01-14 Martin Sebor <msebor@redhat.com>
4942 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4944 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4946 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4947 entry from extra_headers.
4948 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4949 extra_headers, make the list bitwise identical to the i?86-*-* one.
4951 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4953 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4954 -mcmodel=large with -mindirect-branch=thunk,
4955 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4956 -mfunction-return=thunk-extern.
4957 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4958 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4959 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4961 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4963 * config/i386/i386.c (print_reg): Print the name of the full
4964 integer register without '%'.
4965 (ix86_print_operand): Handle 'V'.
4966 * doc/extend.texi: Document 'V' modifier.
4968 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4970 * config/i386/constraints.md (Bs): Disallow memory operand for
4971 -mindirect-branch-register.
4973 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4974 (GOT_memory_operand): Likewise.
4975 (call_insn_operand): Likewise.
4976 (sibcall_insn_operand): Likewise.
4977 (GOT32_symbol_operand): Likewise.
4978 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4979 for -mindirect-branch-register.
4980 (tablejump): Likewise.
4981 (*sibcall_memory): Likewise.
4982 (*sibcall_value_memory): Likewise.
4983 Disallow peepholes of indirect call and jump via memory for
4984 -mindirect-branch-register.
4985 (*call_pop): Replace m with Bw.
4986 (*call_value_pop): Likewise.
4987 (*sibcall_pop_memory): Replace m with Bs.
4988 * config/i386/i386.opt (mindirect-branch-register): New option.
4989 * doc/invoke.texi: Document -mindirect-branch-register option.
4991 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4993 * config/i386/i386-protos.h (ix86_output_function_return): New.
4994 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4995 set function_return_type.
4996 (indirect_thunk_name): Add ret_p to indicate thunk for function
4998 (output_indirect_thunk_function): Pass false to
4999 indirect_thunk_name.
5000 (ix86_output_indirect_branch_via_reg): Likewise.
5001 (ix86_output_indirect_branch_via_push): Likewise.
5002 (output_indirect_thunk_function): Create alias for function
5003 return thunk if regno < 0.
5004 (ix86_output_function_return): New function.
5005 (ix86_handle_fndecl_attribute): Handle function_return.
5006 (ix86_attribute_table): Add function_return.
5007 * config/i386/i386.h (machine_function): Add
5008 function_return_type.
5009 * config/i386/i386.md (simple_return_internal): Use
5010 ix86_output_function_return.
5011 (simple_return_internal_long): Likewise.
5012 * config/i386/i386.opt (mfunction-return=): New option.
5013 (indirect_branch): Mention -mfunction-return=.
5014 * doc/extend.texi: Document function_return function attribute.
5015 * doc/invoke.texi: Document -mfunction-return= option.
5017 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5019 * config/i386/i386-opts.h (indirect_branch): New.
5020 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5021 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5022 with local indirect jump when converting indirect call and jump.
5023 (ix86_set_indirect_branch_type): New.
5024 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5025 (indirectlabelno): New.
5026 (indirect_thunk_needed): Likewise.
5027 (indirect_thunk_bnd_needed): Likewise.
5028 (indirect_thunks_used): Likewise.
5029 (indirect_thunks_bnd_used): Likewise.
5030 (INDIRECT_LABEL): Likewise.
5031 (indirect_thunk_name): Likewise.
5032 (output_indirect_thunk): Likewise.
5033 (output_indirect_thunk_function): Likewise.
5034 (ix86_output_indirect_branch_via_reg): Likewise.
5035 (ix86_output_indirect_branch_via_push): Likewise.
5036 (ix86_output_indirect_branch): Likewise.
5037 (ix86_output_indirect_jmp): Likewise.
5038 (ix86_code_end): Call output_indirect_thunk_function if needed.
5039 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5041 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5042 (ix86_attribute_table): Add indirect_branch.
5043 * config/i386/i386.h (machine_function): Add indirect_branch_type
5044 and has_local_indirect_jump.
5045 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5047 (tablejump): Likewise.
5048 (*indirect_jump): Use ix86_output_indirect_jmp.
5049 (*tablejump_1): Likewise.
5050 (simple_return_indirect_internal): Likewise.
5051 * config/i386/i386.opt (mindirect-branch=): New option.
5052 (indirect_branch): New.
5055 (thunk-inline): Likewise.
5056 (thunk-extern): Likewise.
5057 * doc/extend.texi: Document indirect_branch function attribute.
5058 * doc/invoke.texi: Document -mindirect-branch= option.
5060 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5063 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5065 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5067 * ipa-inline.c (want_inline_small_function_p): Return false if
5068 inlining has already failed with CIF_FINAL_ERROR.
5069 (update_caller_keys): Call want_inline_small_function_p before
5071 (update_callee_keys): Likewise.
5073 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5075 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5077 (rs6000_quadword_masked_address_p): Likewise.
5078 (quad_aligned_load_p): Likewise.
5079 (quad_aligned_store_p): Likewise.
5080 (const_load_sequence_p): Add comment to describe the outer-most loop.
5081 (mimic_memory_attributes_and_flags): New function.
5082 (rs6000_gen_stvx): Likewise.
5083 (replace_swapped_aligned_store): Likewise.
5084 (rs6000_gen_lvx): Likewise.
5085 (replace_swapped_aligned_load): Likewise.
5086 (replace_swapped_load_constant): Capitalize argument name in
5087 comment describing this function.
5088 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5089 and stores that access quad-word aligned addresses and replace
5090 with stvx or lvx instructions when appropriate.
5091 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5092 New function prototype.
5093 (rs6000_quadword_masked_address_p): Likewise.
5094 (rs6000_gen_lvx): Likewise.
5095 (rs6000_gen_stvx): Likewise.
5096 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5097 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5098 when memory address is aligned.
5099 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5100 this split to select lvx instruction when memory address is aligned.
5101 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5102 instruction when memory address is aligned.
5103 (*vsx_le_perm_load_v16qi): Likewise.
5104 (four unnamed splitters): Modify to select the stvx instruction
5105 when memory is aligned.
5107 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5109 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5110 which appears in the queue multiple times.
5112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5113 Alan Hayward <alan.hayward@arm.com>
5114 David Sherwood <david.sherwood@arm.com>
5116 * tree-vectorizer.h (vec_lower_bound): New structure.
5117 (_loop_vec_info): Add check_nonzero and lower_bounds.
5118 (LOOP_VINFO_CHECK_NONZERO): New macro.
5119 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5120 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5121 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5122 fields. Make seg_len the distance travelled, not including the
5124 (dr_direction_indicator): Declare.
5125 (dr_zero_step_indicator): Likewise.
5126 (dr_known_forward_stride_p): Likewise.
5127 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5129 (runtime_alias_check_p): Allow runtime alias checks with
5131 (operator ==): Compare access_size and align.
5132 (prune_runtime_alias_test_list): Rework for new distinction between
5133 the access_size and seg_len.
5134 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5136 (get_segment_min_max): New function.
5137 (create_intersect_range_checks): Use it.
5138 (dr_step_indicator): New function.
5139 (dr_direction_indicator): Likewise.
5140 (dr_zero_step_indicator): Likewise.
5141 (dr_known_forward_stride_p): Likewise.
5142 * tree-loop-distribution.c (data_ref_segment_size): Return
5143 DR_STEP * (niters - 1).
5144 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5146 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5147 (vect_preserves_scalar_order_p): New function, split out from...
5148 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5149 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5150 (vect_vfa_access_size): New function.
5151 (vect_vfa_align): Likewise.
5152 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5153 (dump_lower_bound): New function.
5154 (vect_check_lower_bound): Likewise.
5155 (vect_small_gap_p): Likewise.
5156 (vectorizable_with_step_bound_p): Likewise.
5157 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5158 depencies if the vectorization factor is 1. Convert the checks
5159 for nonzero steps into checks on the bounds of DR_STEP. Try using
5160 a bunds check for variable steps if the minimum required step is
5161 relatively small. Update calls to the dr_with_seg_len
5162 constructor and to vect_compile_time_alias.
5163 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5165 (vect_loop_versioning): Call it.
5166 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5168 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5170 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5171 Alan Hayward <alan.hayward@arm.com>
5172 David Sherwood <david.sherwood@arm.com>
5174 * doc/sourcebuild.texi (vect_scatter_store): Document.
5175 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5177 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5179 * genopinit.c (main): Add supports_vec_scatter_store and
5180 supports_vec_scatter_store_cached to target_optabs.
5181 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5182 IFN_MASK_SCATTER_STORE.
5183 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5185 * internal-fn.h (internal_store_fn_p): Declare.
5186 (internal_fn_stored_value_index): Likewise.
5187 * internal-fn.c (scatter_store_direct): New macro.
5188 (expand_scatter_store_optab_fn): New function.
5189 (direct_scatter_store_optab_supported_p): New macro.
5190 (internal_store_fn_p): New function.
5191 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5192 IFN_MASK_SCATTER_STORE.
5193 (internal_fn_mask_index): Likewise.
5194 (internal_fn_stored_value_index): New function.
5195 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5197 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5198 * optabs-query.c (supports_vec_scatter_store_p): New function.
5199 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5200 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5201 true for scatter stores.
5202 (vect_gather_scatter_fn_p): Handle scatter stores too.
5203 (vect_check_gather_scatter): Consider using scatter stores if
5204 supports_vec_scatter_store_p.
5205 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5207 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5208 internal_fn_stored_value_index.
5209 (check_load_store_masking): Handle scatter stores too.
5210 (vect_get_store_rhs): Make public.
5211 (vectorizable_call): Use internal_store_fn_p.
5212 (vectorizable_store): Handle scatter store internal functions.
5213 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5214 when deciding whether the end of the group has been reached.
5215 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5216 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5217 (mask_scatter_store<mode>): New insns.
5219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5220 Alan Hayward <alan.hayward@arm.com>
5221 David Sherwood <david.sherwood@arm.com>
5223 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5224 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5225 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5227 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5228 Use vect_truncate_gather_scatter_offset if we can't treat the
5229 operation as a normal gather load or scatter store.
5230 (get_group_load_store_type): Take the gather_scatter_info
5231 as argument. Try using a gather load or scatter store for
5232 single-element groups.
5233 (get_load_store_type): Update calls to get_group_load_store_type
5234 and vect_use_strided_gather_scatters_p.
5236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5237 Alan Hayward <alan.hayward@arm.com>
5238 David Sherwood <david.sherwood@arm.com>
5240 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5241 optional tree argument.
5242 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5244 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5245 but continue to use the current value as a fallback.
5246 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5247 to compare the updates.
5248 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5249 (get_load_store_type): Use it when handling a strided access.
5250 (vect_get_strided_load_store_ops): New function.
5251 (vect_get_data_ptr_increment): Likewise.
5252 (vectorizable_load): Handle strided gather loads. Always pass
5253 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5256 Alan Hayward <alan.hayward@arm.com>
5257 David Sherwood <david.sherwood@arm.com>
5259 * doc/md.texi (gather_load@var{m}): Document.
5260 (mask_gather_load@var{m}): Likewise.
5261 * genopinit.c (main): Add supports_vec_gather_load and
5262 supports_vec_gather_load_cached to target_optabs.
5263 * optabs-tree.c (init_tree_optimization_optabs): Use
5264 ggc_cleared_alloc to allocate target_optabs.
5265 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5266 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5268 * internal-fn.h (internal_load_fn_p): Declare.
5269 (internal_gather_scatter_fn_p): Likewise.
5270 (internal_fn_mask_index): Likewise.
5271 (internal_gather_scatter_fn_supported_p): Likewise.
5272 * internal-fn.c (gather_load_direct): New macro.
5273 (expand_gather_load_optab_fn): New function.
5274 (direct_gather_load_optab_supported_p): New macro.
5275 (direct_internal_fn_optab): New function.
5276 (internal_load_fn_p): Likewise.
5277 (internal_gather_scatter_fn_p): Likewise.
5278 (internal_fn_mask_index): Likewise.
5279 (internal_gather_scatter_fn_supported_p): Likewise.
5280 * optabs-query.c (supports_at_least_one_mode_p): New function.
5281 (supports_vec_gather_load_p): Likewise.
5282 * optabs-query.h (supports_vec_gather_load_p): Declare.
5283 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5284 and memory_type field.
5285 (NUM_PATTERNS): Bump to 15.
5286 * tree-vect-data-refs.c: Include internal-fn.h.
5287 (vect_gather_scatter_fn_p): New function.
5288 (vect_describe_gather_scatter_call): Likewise.
5289 (vect_check_gather_scatter): Try using internal functions for
5290 gather loads. Recognize existing calls to a gather load function.
5291 (vect_analyze_data_refs): Consider using gather loads if
5292 supports_vec_gather_load_p.
5293 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5294 (vect_get_gather_scatter_offset_type): Likewise.
5295 (vect_convert_mask_for_vectype): Likewise.
5296 (vect_add_conversion_to_patterm): Likewise.
5297 (vect_try_gather_scatter_pattern): Likewise.
5298 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5299 (vect_vect_recog_func_ptrs): Add it.
5300 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5301 internal_fn_mask_index and internal_gather_scatter_fn_p.
5302 (check_load_store_masking): Take the gather_scatter_info as an
5303 argument and handle gather loads.
5304 (vect_get_gather_scatter_ops): New function.
5305 (vectorizable_call): Check internal_load_fn_p.
5306 (vectorizable_load): Likewise. Handle gather load internal
5308 (vectorizable_store): Update call to check_load_store_masking.
5309 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5310 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5311 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5312 (aarch64_gather_scale_operand_d): New predicates.
5313 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5314 (mask_gather_load<mode>): New insns.
5316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5317 Alan Hayward <alan.hayward@arm.com>
5318 David Sherwood <david.sherwood@arm.com>
5320 * optabs.def (fold_left_plus_optab): New optab.
5321 * doc/md.texi (fold_left_plus_@var{m}): Document.
5322 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5323 * internal-fn.c (fold_left_direct): Define.
5324 (expand_fold_left_optab_fn): Likewise.
5325 (direct_fold_left_optab_supported_p): Likewise.
5326 * fold-const-call.c (fold_const_fold_left): New function.
5327 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5328 * tree-parloops.c (valid_reduction_p): New function.
5329 (gather_scalar_reductions): Use it.
5330 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5331 (vect_finish_replace_stmt): Declare.
5332 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5333 (needs_fold_left_reduction_p): New function, split out from...
5334 (vect_is_simple_reduction): ...here. Accept reductions that
5335 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5336 (vect_force_simple_reduction): Also store the reduction type in
5337 the assignment's STMT_VINFO_REDUC_TYPE.
5338 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5339 (merge_with_identity): New function.
5340 (vect_expand_fold_left): Likewise.
5341 (vectorize_fold_left_reduction): Likewise.
5342 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5343 scalar phi in place for it. Check for target support and reject
5344 cases that would reassociate the operation. Defer the transform
5345 phase to vectorize_fold_left_reduction.
5346 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5347 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5348 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5352 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5353 call to ifc_temp_var.
5355 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5356 Alan Hayward <alan.hayward@arm.com>
5357 David Sherwood <david.sherwood@arm.com>
5359 * target.def (legitimize_address_displacement): Take the original
5360 offset as a poly_int.
5361 * targhooks.h (default_legitimize_address_displacement): Update
5363 * targhooks.c (default_legitimize_address_displacement): Likewise.
5364 * doc/tm.texi: Regenerate.
5365 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5366 as an argument, moving assert of ad->disp == ad->disp_term to...
5367 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5368 Try calling targetm.legitimize_address_displacement before expanding
5369 the address rather than afterwards, and adjust for the new interface.
5370 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5371 Match the new hook interface. Handle SVE addresses.
5372 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5377 * Makefile.in (OBJS): Add early-remat.o.
5378 * target.def (select_early_remat_modes): New hook.
5379 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5380 * doc/tm.texi: Regenerate.
5381 * targhooks.h (default_select_early_remat_modes): Declare.
5382 * targhooks.c (default_select_early_remat_modes): New function.
5383 * timevar.def (TV_EARLY_REMAT): New timevar.
5384 * passes.def (pass_early_remat): New pass.
5385 * tree-pass.h (make_pass_early_remat): Declare.
5386 * early-remat.c: New file.
5387 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5389 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5392 Alan Hayward <alan.hayward@arm.com>
5393 David Sherwood <david.sherwood@arm.com>
5395 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5396 vfm1 with a bound_epilog parameter.
5397 (vect_do_peeling): Update calls accordingly, and move the prologue
5398 call earlier in the function. Treat the base bound_epilog as 0 for
5399 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5400 this base when peeling for gaps.
5401 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5402 with fully-masked loops.
5403 (vect_estimate_min_profitable_iters): Handle the single peeled
5404 iteration in that case.
5406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5407 Alan Hayward <alan.hayward@arm.com>
5408 David Sherwood <david.sherwood@arm.com>
5410 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5411 single-element interleaving even if the size is not a power of 2.
5412 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5413 accesses for single-element interleaving if the group size is
5416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5417 Alan Hayward <alan.hayward@arm.com>
5418 David Sherwood <david.sherwood@arm.com>
5420 * doc/md.texi (fold_extract_last_@var{m}): Document.
5421 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5422 * optabs.def (fold_extract_last_optab): New optab.
5423 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5424 * internal-fn.c (fold_extract_direct): New macro.
5425 (expand_fold_extract_optab_fn): Likewise.
5426 (direct_fold_extract_optab_supported_p): Likewise.
5427 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5428 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5429 EXTRACT_LAST_REDUCTION.
5430 (get_initial_def_for_reduction): Do not create an initial vector
5431 for EXTRACT_LAST_REDUCTION reductions.
5432 (vectorizable_reduction): Leave the scalar phi in place for
5433 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5434 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5435 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5436 transform phase to vectorizable_condition.
5437 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5439 (vect_finish_stmt_generation): ...here.
5440 (vect_finish_replace_stmt): New function.
5441 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5442 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5444 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5447 Alan Hayward <alan.hayward@arm.com>
5448 David Sherwood <david.sherwood@arm.com>
5450 * doc/md.texi (extract_last_@var{m}): Document.
5451 * optabs.def (extract_last_optab): New optab.
5452 * internal-fn.def (EXTRACT_LAST): New internal function.
5453 * internal-fn.c (cond_unary_direct): New macro.
5454 (expand_cond_unary_optab_fn): Likewise.
5455 (direct_cond_unary_optab_supported_p): Likewise.
5456 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5457 loops using EXTRACT_LAST.
5458 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5459 (extract_last_<mode>): ...this optab.
5460 (vec_extract<mode><Vel>): Update accordingly.
5462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5463 Alan Hayward <alan.hayward@arm.com>
5464 David Sherwood <david.sherwood@arm.com>
5466 * target.def (empty_mask_is_expensive): New hook.
5467 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5468 * doc/tm.texi: Regenerate.
5469 * targhooks.h (default_empty_mask_is_expensive): Declare.
5470 * targhooks.c (default_empty_mask_is_expensive): New function.
5471 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5472 if the target says that empty masks are expensive.
5473 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5475 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5478 Alan Hayward <alan.hayward@arm.com>
5479 David Sherwood <david.sherwood@arm.com>
5481 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5482 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5483 (vect_use_loop_mask_for_alignment_p): New function.
5484 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5485 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5486 niters_skip argument. Make sure that the first niters_skip elements
5487 of the first iteration are inactive.
5488 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5489 Update call to vect_set_loop_masks_directly.
5490 (get_misalign_in_elems): New function, split out from...
5491 (vect_gen_prolog_loop_niters): ...here.
5492 (vect_update_init_of_dr): Take a code argument that specifies whether
5493 the adjustment should be added or subtracted.
5494 (vect_update_init_of_drs): Likewise.
5495 (vect_prepare_for_masked_peels): New function.
5496 (vect_do_peeling): Skip prologue peeling if we're using a mask
5497 instead. Update call to vect_update_inits_of_drs.
5498 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5500 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5501 alignment. Do not include the number of peeled iterations in
5502 the minimum threshold in that case.
5503 (vectorizable_induction): Adjust the start value down by
5504 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5505 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5506 Take the number of skipped iterations into account when calculating
5508 * tree-vect-stmts.c (vect_gen_while_not): New function.
5510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5511 Alan Hayward <alan.hayward@arm.com>
5512 David Sherwood <david.sherwood@arm.com>
5514 * doc/sourcebuild.texi (vect_fully_masked): Document.
5515 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5517 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5519 (vect_analyze_loop_2): ...here. Don't check the vectorization
5520 factor against the number of loop iterations if the loop is
5523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5524 Alan Hayward <alan.hayward@arm.com>
5525 David Sherwood <david.sherwood@arm.com>
5527 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5528 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5529 (dump_groups): Update accordingly.
5530 (iv_use::mem_type): New member variable.
5531 (address_p): New function.
5532 (record_use): Add a mem_type argument and initialize the new
5534 (record_group_use): Add a mem_type argument. Use address_p.
5535 Remove obsolete null checks of base_object. Update call to record_use.
5536 (find_interesting_uses_op): Update call to record_group_use.
5537 (find_interesting_uses_cond): Likewise.
5538 (find_interesting_uses_address): Likewise.
5539 (get_mem_type_for_internal_fn): New function.
5540 (find_address_like_use): Likewise.
5541 (find_interesting_uses_stmt): Try find_address_like_use before
5542 calling find_interesting_uses_op.
5543 (addr_offset_valid_p): Use the iv mem_type field as the type
5544 of the addressed memory.
5545 (add_autoinc_candidates): Likewise.
5546 (get_address_cost): Likewise.
5547 (split_small_address_groups_p): Use address_p.
5548 (split_address_groups): Likewise.
5549 (add_iv_candidate_for_use): Likewise.
5550 (autoinc_possible_for_pair): Likewise.
5551 (rewrite_groups): Likewise.
5552 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5553 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5554 (get_alias_ptr_type_for_ptr_address): New function.
5555 (rewrite_use_address): Rewrite address uses in calls that were
5556 identified by find_address_like_use.
5558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5559 Alan Hayward <alan.hayward@arm.com>
5560 David Sherwood <david.sherwood@arm.com>
5562 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5564 * gimple-expr.h (is_gimple_addressable: Likewise.
5565 * gimple-expr.c (is_gimple_address): Likewise.
5566 * internal-fn.c (expand_call_mem_ref): New function.
5567 (expand_mask_load_optab_fn): Use it.
5568 (expand_mask_store_optab_fn): Likewise.
5570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5571 Alan Hayward <alan.hayward@arm.com>
5572 David Sherwood <david.sherwood@arm.com>
5574 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5575 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5576 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5577 (cond_umax@var{mode}): Document.
5578 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5579 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5580 (cond_umin_optab, cond_umax_optab): New optabs.
5581 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5582 (COND_IOR, COND_XOR): New internal functions.
5583 * internal-fn.h (get_conditional_internal_fn): Declare.
5584 * internal-fn.c (cond_binary_direct): New macro.
5585 (expand_cond_binary_optab_fn): Likewise.
5586 (direct_cond_binary_optab_supported_p): Likewise.
5587 (get_conditional_internal_fn): New function.
5588 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5589 Cope with reduction statements that are vectorized as calls rather
5591 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5592 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5593 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5594 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5595 (UNSPEC_COND_EOR): New unspecs.
5596 (optab): Add mappings for them.
5597 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5598 (sve_int_op, sve_fp_op): New int attributes.
5600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5601 Alan Hayward <alan.hayward@arm.com>
5602 David Sherwood <david.sherwood@arm.com>
5604 * optabs.def (while_ult_optab): New optab.
5605 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5606 * internal-fn.def (WHILE_ULT): New internal function.
5607 * internal-fn.h (direct_internal_fn_supported_p): New override
5608 that takes two types as argument.
5609 * internal-fn.c (while_direct): New macro.
5610 (expand_while_optab_fn): New function.
5611 (convert_optab_supported_p): Likewise.
5612 (direct_while_optab_supported_p): New macro.
5613 * wide-int.h (wi::udiv_ceil): New function.
5614 * tree-vectorizer.h (rgroup_masks): New structure.
5615 (vec_loop_masks): New typedef.
5616 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5618 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5619 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5620 (vect_max_vf): New function.
5621 (slpeel_make_loop_iterate_ntimes): Delete.
5622 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5623 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5624 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5625 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5626 internal-fn.h, stor-layout.h and optabs-query.h.
5627 (vect_set_loop_mask): New function.
5628 (add_preheader_seq): Likewise.
5629 (add_header_seq): Likewise.
5630 (interleave_supported_p): Likewise.
5631 (vect_maybe_permute_loop_masks): Likewise.
5632 (vect_set_loop_masks_directly): Likewise.
5633 (vect_set_loop_condition_masked): Likewise.
5634 (vect_set_loop_condition_unmasked): New function, split out from
5635 slpeel_make_loop_iterate_ntimes.
5636 (slpeel_make_loop_iterate_ntimes): Rename to..
5637 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5638 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5639 (vect_do_peeling): Update call accordingly.
5640 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5642 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5643 mask_compare_type, can_fully_mask_p and fully_masked_p.
5644 (release_vec_loop_masks): New function.
5645 (_loop_vec_info): Use it to free the loop masks.
5646 (can_produce_all_loop_masks_p): New function.
5647 (vect_get_max_nscalars_per_iter): Likewise.
5648 (vect_verify_full_masking): Likewise.
5649 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5650 retries, and free the mask rgroups before retrying. Check loop-wide
5651 reasons for disallowing fully-masked loops. Make the final decision
5652 about whether use a fully-masked loop or not.
5653 (vect_estimate_min_profitable_iters): Do not assume that peeling
5654 for the number of iterations will be needed for fully-masked loops.
5655 (vectorizable_reduction): Disable fully-masked loops.
5656 (vectorizable_live_operation): Likewise.
5657 (vect_halve_mask_nunits): New function.
5658 (vect_double_mask_nunits): Likewise.
5659 (vect_record_loop_mask): Likewise.
5660 (vect_get_loop_mask): Likewise.
5661 (vect_transform_loop): Handle the case in which the final loop
5662 iteration might handle a partial vector. Call vect_set_loop_condition
5663 instead of slpeel_make_loop_iterate_ntimes.
5664 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5665 (check_load_store_masking): New function.
5666 (prepare_load_store_mask): Likewise.
5667 (vectorizable_store): Handle fully-masked loops.
5668 (vectorizable_load): Likewise.
5669 (supportable_widening_operation): Use vect_halve_mask_nunits for
5671 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5672 (vect_gen_while): New function.
5673 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5674 (aarch64_uqdec<mode>): New insn.
5676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5677 Alan Hayward <alan.hayward@arm.com>
5678 David Sherwood <david.sherwood@arm.com>
5680 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5681 (reduc_xor_scal_optab): New optabs.
5682 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5683 (reduc_xor_scal_@var{m}): Document.
5684 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5685 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5687 * fold-const-call.c (fold_const_call): Handle them.
5688 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5689 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5690 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5691 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5692 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5693 (UNSPEC_XORV): New unspecs.
5694 (optab): Add entries for them.
5695 (BITWISEV): New int iterator.
5696 (bit_reduc_op): New int attributes.
5698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5699 Alan Hayward <alan.hayward@arm.com>
5700 David Sherwood <david.sherwood@arm.com>
5702 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5703 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5704 * optabs.def (vec_shl_insert_optab): New optab.
5705 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5706 (duplicate_and_interleave): Likewise.
5707 * tree-vect-loop.c: Include internal-fn.h.
5708 (neutral_op_for_slp_reduction): New function, split out from
5709 get_initial_defs_for_reduction.
5710 (get_initial_def_for_reduction): Handle option 2 for variable-length
5711 vectors by loading the neutral value into a vector and then shifting
5712 the initial value into element 0.
5713 (get_initial_defs_for_reduction): Replace the code argument with
5714 the neutral value calculated by neutral_op_for_slp_reduction.
5715 Use gimple_build_vector for constant-length vectors.
5716 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5717 but the first group_size elements have a neutral value.
5718 Use duplicate_and_interleave otherwise.
5719 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5720 Update call to get_initial_defs_for_reduction. Handle SLP
5721 reductions for variable-length vectors by creating one vector
5722 result for each scalar result, with the elements associated
5723 with other scalar results stubbed out with the neutral value.
5724 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5725 Require IFN_VEC_SHL_INSERT for double reductions on
5726 variable-length vectors, or SLP reductions that have
5727 a neutral value. Require can_duplicate_and_interleave_p
5728 support for variable-length unchained SLP reductions if there
5729 is no neutral value, such as for MIN/MAX reductions. Also require
5730 the number of vector elements to be a multiple of the number of
5731 SLP statements when doing variable-length unchained SLP reductions.
5732 Update call to vect_create_epilog_for_reduction.
5733 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5734 and remove initial values.
5735 (duplicate_and_interleave): Make public.
5736 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5737 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5740 Alan Hayward <alan.hayward@arm.com>
5741 David Sherwood <david.sherwood@arm.com>
5743 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5744 (can_duplicate_and_interleave_p): New function.
5745 (vect_get_and_check_slp_defs): Take the vector of statements
5746 rather than just the current one. Remove excess parentheses.
5747 Restriction rejectinon of vect_constant_def and vect_external_def
5748 for variable-length vectors to boolean types, or types for which
5749 can_duplicate_and_interleave_p is false.
5750 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5751 (duplicate_and_interleave): New function.
5752 (vect_get_constant_vectors): Use gimple_build_vector for
5753 constant-length vectors and suitable variable-length constant
5754 vectors. Use duplicate_and_interleave for other variable-length
5755 vectors. Don't defer the update when inserting new statements.
5757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5758 Alan Hayward <alan.hayward@arm.com>
5759 David Sherwood <david.sherwood@arm.com>
5761 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5762 min_profitable_iters doesn't go negative.
5764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5765 Alan Hayward <alan.hayward@arm.com>
5766 David Sherwood <david.sherwood@arm.com>
5768 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5769 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5770 * optabs.def (vec_mask_load_lanes_optab): New optab.
5771 (vec_mask_store_lanes_optab): Likewise.
5772 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5773 (MASK_STORE_LANES): Likewise.
5774 * internal-fn.c (mask_load_lanes_direct): New macro.
5775 (mask_store_lanes_direct): Likewise.
5776 (expand_mask_load_optab_fn): Handle masked operations.
5777 (expand_mask_load_lanes_optab_fn): New macro.
5778 (expand_mask_store_optab_fn): Handle masked operations.
5779 (expand_mask_store_lanes_optab_fn): New macro.
5780 (direct_mask_load_lanes_optab_supported_p): Likewise.
5781 (direct_mask_store_lanes_optab_supported_p): Likewise.
5782 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5784 (vect_load_lanes_supported): Likewise.
5785 * tree-vect-data-refs.c (strip_conversion): New function.
5786 (can_group_stmts_p): Likewise.
5787 (vect_analyze_data_ref_accesses): Use it instead of checking
5788 for a pair of assignments.
5789 (vect_store_lanes_supported): Take a masked_p parameter.
5790 (vect_load_lanes_supported): Likewise.
5791 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5792 vect_store_lanes_supported and vect_load_lanes_supported.
5793 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5794 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5795 parameter. Don't allow gaps for masked accesses.
5796 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5797 and vect_load_lanes_supported.
5798 (get_load_store_type): Take a masked_p parameter and update
5799 call to get_group_load_store_type.
5800 (vectorizable_store): Update call to get_load_store_type.
5801 Handle IFN_MASK_STORE_LANES.
5802 (vectorizable_load): Update call to get_load_store_type.
5803 Handle IFN_MASK_LOAD_LANES.
5805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5806 Alan Hayward <alan.hayward@arm.com>
5807 David Sherwood <david.sherwood@arm.com>
5809 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5811 * config/aarch64/aarch64-protos.h
5812 (aarch64_sve_struct_memory_operand_p): Declare.
5813 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5814 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5815 (VPRED, vpred): Handle SVE structure modes.
5816 * config/aarch64/constraints.md (Utx): New constraint.
5817 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5818 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5819 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5820 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5821 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5822 structure modes. Split into pieces after RA.
5823 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5824 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5826 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5827 SVE structure modes.
5828 (aarch64_classify_address): Likewise.
5829 (sizetochar): Move earlier in file.
5830 (aarch64_print_operand): Handle SVE register lists.
5831 (aarch64_array_mode): New function.
5832 (aarch64_sve_struct_memory_operand_p): Likewise.
5833 (TARGET_ARRAY_MODE): Redefine.
5835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5836 Alan Hayward <alan.hayward@arm.com>
5837 David Sherwood <david.sherwood@arm.com>
5839 * target.def (array_mode): New target hook.
5840 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5841 * doc/tm.texi: Regenerate.
5842 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5843 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5844 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5846 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5850 Alan Hayward <alan.hayward@arm.com>
5851 David Sherwood <david.sherwood@arm.com>
5853 * fold-const.c (fold_binary_loc): Check the argument types
5854 rather than the result type when testing for a vector operation.
5856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5858 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5859 * doc/tm.texi: Regenerate.
5861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5862 Alan Hayward <alan.hayward@arm.com>
5863 David Sherwood <david.sherwood@arm.com>
5865 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5866 (sve): Document new AArch64 extension.
5867 * doc/md.texi (w): Extend the description of the AArch64
5868 constraint to include SVE vectors.
5869 (Upl, Upa): Document new AArch64 predicate constraints.
5870 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5872 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5873 (msve-vector-bits=): New option.
5874 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5875 SVE when these are disabled.
5876 (sve): New extension.
5877 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5878 modes. Adjust their number of units based on aarch64_sve_vg.
5879 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5880 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5881 aarch64_addr_query_type.
5882 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5883 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5884 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5885 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5886 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5887 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5888 (aarch64_simd_imm_zero_p): Delete.
5889 (aarch64_check_zero_based_sve_index_immediate): Declare.
5890 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5891 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5892 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5893 (aarch64_sve_float_mul_immediate_p): Likewise.
5894 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5896 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5897 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5898 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5899 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5900 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5901 (aarch64_regmode_natural_size): Likewise.
5902 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5903 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5905 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5906 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5907 for VG and the SVE predicate registers.
5908 (V_ALIASES): Add a "z"-prefixed alias.
5909 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5910 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5911 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5912 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5913 (REG_CLASS_NAMES): Add entries for them.
5914 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5915 and the predicate registers.
5916 (aarch64_sve_vg): Declare.
5917 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5918 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5919 (REGMODE_NATURAL_SIZE): Define.
5920 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5922 * config/aarch64/aarch64.c: Include cfgrtl.h.
5923 (simd_immediate_info): Add a constructor for series vectors,
5924 and an associated step field.
5925 (aarch64_sve_vg): New variable.
5926 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5927 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5928 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5929 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5930 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5931 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5932 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5933 (aarch64_get_mask_mode): New functions.
5934 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5935 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5936 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5937 predicate modes and predicate registers. Explicitly restrict
5938 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5939 to store a vector mode if it is recognized by
5940 aarch64_classify_vector_mode.
5941 (aarch64_regmode_natural_size): New function.
5942 (aarch64_hard_regno_caller_save_mode): Return the original mode
5944 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5945 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5946 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5947 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5949 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5950 does not overlap dest if the function is frame-related. Handle
5952 (aarch64_split_add_offset): New function.
5953 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5954 them aarch64_add_offset.
5955 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5956 and update call to aarch64_sub_sp.
5957 (aarch64_add_cfa_expression): New function.
5958 (aarch64_expand_prologue): Pass extra temporary registers to the
5959 functions above. Handle the case in which we need to emit new
5960 DW_CFA_expressions for registers that were originally saved
5961 relative to the stack pointer, but now have to be expressed
5962 relative to the frame pointer.
5963 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5965 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5966 IP0 and IP1 values for SVE frames.
5967 (aarch64_expand_vec_series): New function.
5968 (aarch64_expand_sve_widened_duplicate): Likewise.
5969 (aarch64_expand_sve_const_vector): Likewise.
5970 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5971 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5972 into the register, rather than emitting a SET directly.
5973 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5974 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5975 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5976 (offset_9bit_signed_scaled_p): New functions.
5977 (aarch64_replicate_bitmask_imm): New function.
5978 (aarch64_bitmask_imm): Use it.
5979 (aarch64_cannot_force_const_mem): Reject expressions involving
5980 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5981 (aarch64_classify_index): Handle SVE indices, by requiring
5982 a plain register index with a scale that matches the element size.
5983 (aarch64_classify_address): Handle SVE addresses. Assert that
5984 the mode of the address is VOIDmode or an integer mode.
5985 Update call to aarch64_classify_symbol.
5986 (aarch64_classify_symbolic_expression): Update call to
5987 aarch64_classify_symbol.
5988 (aarch64_const_vec_all_in_range_p): New function.
5989 (aarch64_print_vector_float_operand): Likewise.
5990 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5991 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5992 and the FP immediates 1.0 and 0.5.
5993 (aarch64_print_address_internal): Handle SVE addresses.
5994 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5995 (aarch64_regno_regclass): Handle predicate registers.
5996 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5998 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5999 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6000 (aarch64_convert_sve_vector_bits): New function.
6001 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6002 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6004 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6005 Handle SVE vector and predicate modes. Accept VL-based constants
6006 that need only one temporary register, and VL offsets that require
6007 no temporary registers.
6008 (aarch64_conditional_register_usage): Mark the predicate registers
6009 as fixed if SVE isn't available.
6010 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6011 Return true for SVE vector and predicate modes.
6012 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6013 rather than an unsigned int. Handle SVE modes.
6014 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6016 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6018 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6019 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6020 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6021 (aarch64_sve_float_mul_immediate_p): New functions.
6022 (aarch64_sve_valid_immediate): New function.
6023 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6024 Explicitly reject structure modes. Check for INDEX constants.
6025 Handle PTRUE and PFALSE constants.
6026 (aarch64_check_zero_based_sve_index_immediate): New function.
6027 (aarch64_simd_imm_zero_p): Delete.
6028 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6029 vector modes. Accept constants in the range of CNT[BHWD].
6030 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6031 ask for an Advanced SIMD mode.
6032 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6033 (aarch64_simd_vector_alignment): Handle SVE predicates.
6034 (aarch64_vectorize_preferred_vector_alignment): New function.
6035 (aarch64_simd_vector_alignment_reachable): Use it instead of
6037 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6038 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6040 (MAX_VECT_LEN): Delete.
6041 (expand_vec_perm_d): Add a vec_flags field.
6042 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6043 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6044 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6046 (aarch64_evpc_rev): Rename to...
6047 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6048 (aarch64_evpc_rev_global): New function.
6049 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6050 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6052 (aarch64_evpc_sve_tbl): New function.
6053 (aarch64_expand_vec_perm_const_1): Update after rename of
6054 aarch64_evpc_rev. Handle SVE permutes too, trying
6055 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6056 than aarch64_evpc_tbl.
6057 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6058 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6059 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6060 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6061 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6062 (aarch64_expand_sve_vcond): New functions.
6063 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6064 of aarch64_vector_mode_p.
6065 (aarch64_dwarf_poly_indeterminate_value): New function.
6066 (aarch64_compute_pressure_classes): Likewise.
6067 (aarch64_can_change_mode_class): Likewise.
6068 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6069 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6070 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6071 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6072 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6073 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6074 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6075 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6077 (Dn, Dl, Dr): Accept const as well as const_vector.
6078 (Dz): Likewise. Compare against CONST0_RTX.
6079 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6080 of "vector" where appropriate.
6081 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6082 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6083 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6084 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6085 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6086 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6087 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6088 (v_int_equiv): Extend to SVE modes.
6089 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6091 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6092 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6093 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6094 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6095 (SVE_COND_FP_CMP): New int iterators.
6096 (perm_hilo): Handle the new unpack unspecs.
6097 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6099 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6100 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6101 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6102 (aarch64_equality_operator, aarch64_constant_vector_operand)
6103 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6104 (aarch64_sve_nonimmediate_operand): Likewise.
6105 (aarch64_sve_general_operand): Likewise.
6106 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6107 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6108 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6109 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6110 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6111 (aarch64_sve_float_arith_immediate): Likewise.
6112 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6113 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6114 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6115 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6116 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6117 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6118 (aarch64_sve_float_arith_operand): Likewise.
6119 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6120 (aarch64_sve_float_mul_operand): Likewise.
6121 (aarch64_sve_vec_perm_operand): Likewise.
6122 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6123 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6124 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6125 as well as const_vector.
6126 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6127 in file. Use CONST0_RTX and CONSTM1_RTX.
6128 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6129 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6130 Use aarch64_simd_imm_zero.
6131 * config/aarch64/aarch64-sve.md: New file.
6132 * config/aarch64/aarch64.md: Include it.
6133 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6134 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6135 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6136 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6137 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6138 (sve): New attribute.
6139 (enabled): Disable instructions with the sve attribute unless
6141 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6142 aarch64_expand_mov_immediate.
6143 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6144 CNT[BHSD] immediates.
6145 (movti): Split CONST_POLY_INT moves into two halves.
6146 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6147 Split additions that need a temporary here if the destination
6148 is the stack pointer.
6149 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6150 (*add<mode>3_poly_1): New instruction.
6151 (set_clobber_cc): New expander.
6153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6155 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6156 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6157 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6158 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6159 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6160 Change innermode from fixed_mode_size to machine_mode.
6161 (simplify_subreg): Update call accordingly. Handle a constant-sized
6162 subreg of a variable-length CONST_VECTOR.
6164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6165 Alan Hayward <alan.hayward@arm.com>
6166 David Sherwood <david.sherwood@arm.com>
6168 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6169 (add_offset_to_base): New function, split out from...
6170 (create_mem_ref): ...here. When handling a scale other than 1,
6171 check first whether the address is valid without the offset.
6172 Add it into the base if so, leaving the index and scale as-is.
6174 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6177 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6178 fold_for_warn before checking if arg2 is INTEGER_CST.
6180 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6182 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6183 (store_multiple_operation): Delete.
6184 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6185 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6186 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6187 guarded by TARGET_STRING.
6188 (rs6000_output_load_multiple): Delete.
6189 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6190 OPTION_MASK_STRING / TARGET_STRING handling.
6191 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6192 (const rs6000_opt_masks) <"string">: Change mask to 0.
6193 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6194 (MASK_STRING): Delete.
6195 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6197 (load_multiple): Delete.
6204 (store_multiple): Delete.
6211 (movmemsi_8reg): Delete.
6212 (corresponding unnamed define_insn): Delete.
6213 (movmemsi_6reg): Delete.
6214 (corresponding unnamed define_insn): Delete.
6215 (movmemsi_4reg): Delete.
6216 (corresponding unnamed define_insn): Delete.
6217 (movmemsi_2reg): Delete.
6218 (corresponding unnamed define_insn): Delete.
6219 (movmemsi_1reg): Delete.
6220 (corresponding unnamed define_insn): Delete.
6221 * config/rs6000/rs6000.opt (mno-string): New.
6222 (mstring): Replace by deprecation warning stub.
6223 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6225 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6227 * regrename.c (regrename_do_replace): If replacing the same
6228 reg multiple times, try to reuse last created gen_raw_REG.
6231 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6232 main to workaround a bug in GDB.
6234 2018-01-12 Tom de Vries <tom@codesourcery.com>
6237 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6239 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6241 PR rtl-optimization/80481
6242 * ira-color.c (get_cap_member): New function.
6243 (allocnos_conflict_by_live_ranges_p): Use it.
6244 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6245 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6247 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6250 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6251 (*saddl_se_1): Ditto.
6253 (*ssubl_se_1): Ditto.
6255 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6257 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6258 rather than wi::to_widest for DR_INITs.
6259 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6260 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6261 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6263 (vect_analyze_group_access_1): Note that here.
6265 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6267 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6268 polynomial type sizes.
6270 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6272 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6273 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6274 (gimple_add_tmp_var): Likewise.
6276 2018-01-12 Martin Liska <mliska@suse.cz>
6278 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6279 (gimple_alloc_sizes): Likewise.
6280 (dump_gimple_statistics): Use PRIu64 in printf format.
6281 * gimple.h: Change uint64_t to int.
6283 2018-01-12 Martin Liska <mliska@suse.cz>
6285 * tree-core.h: Use uint64_t instead of int.
6286 * tree.c (tree_node_counts): Likewise.
6287 (tree_node_sizes): Likewise.
6288 (dump_tree_statistics): Use PRIu64 in printf format.
6290 2018-01-12 Martin Liska <mliska@suse.cz>
6292 * Makefile.in: As qsort_chk is implemented in vec.c, add
6293 vec.o to linkage of gencfn-macros.
6294 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6295 passing the info to record_node_allocation_statistics.
6296 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6298 * ggc-common.c (struct ggc_usage): Add operator== and use
6299 it in operator< and compare function.
6300 * mem-stats.h (struct mem_usage): Likewise.
6301 * vec.c (struct vec_usage): Remove operator< and compare
6302 function. Can be simply inherited.
6304 2018-01-12 Martin Jambor <mjambor@suse.cz>
6307 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6308 * tree-ssa-math-opts.c: Include domwalk.h.
6309 (convert_mult_to_fma_1): New function.
6310 (fma_transformation_info): New type.
6311 (fma_deferring_state): Likewise.
6312 (cancel_fma_deferring): New function.
6313 (result_of_phi): Likewise.
6314 (last_fma_candidate_feeds_initial_phi): Likewise.
6315 (convert_mult_to_fma): Added deferring logic, split actual
6316 transformation to convert_mult_to_fma_1.
6317 (math_opts_dom_walker): New type.
6318 (math_opts_dom_walker::after_dom_children): New method, body moved
6319 here from pass_optimize_widening_mul::execute, added deferring logic
6321 (pass_optimize_widening_mul::execute): Moved most of code to
6322 math_opts_dom_walker::after_dom_children.
6323 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6324 * config/i386/i386.c (ix86_option_override_internal): Added
6325 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6327 2018-01-12 Richard Biener <rguenther@suse.de>
6330 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6331 inline instance vars.
6333 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6336 * config/rx/rx.c (rx_is_restricted_memory_address):
6339 2018-01-12 Richard Biener <rguenther@suse.de>
6341 PR tree-optimization/80846
6342 * target.def (split_reduction): New target hook.
6343 * targhooks.c (default_split_reduction): New function.
6344 * targhooks.h (default_split_reduction): Declare.
6345 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6346 target requests first reduce vectors by combining low and high
6348 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6349 (get_vectype_for_scalar_type_and_size): Export.
6350 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6351 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6352 * doc/tm.texi: Regenerate.
6353 * config/i386/i386.c (ix86_split_reduction): Implement
6354 TARGET_VECTORIZE_SPLIT_REDUCTION.
6356 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6359 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6360 in PIC mode except for TARGET_VXWORKS_RTP.
6361 * config/sparc/sparc.c: Include cfgrtl.h.
6362 (TARGET_INIT_PIC_REG): Define.
6363 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6364 (sparc_pic_register_p): New predicate.
6365 (sparc_legitimate_address_p): Use it.
6366 (sparc_legitimize_pic_address): Likewise.
6367 (sparc_delegitimize_address): Likewise.
6368 (sparc_mode_dependent_address_p): Likewise.
6369 (gen_load_pcrel_sym): Remove 4th parameter.
6370 (load_got_register): Adjust call to above. Remove obsolete stuff.
6371 (sparc_expand_prologue): Do not call load_got_register here.
6372 (sparc_flat_expand_prologue): Likewise.
6373 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6374 (sparc_use_pseudo_pic_reg): New function.
6375 (sparc_init_pic_reg): Likewise.
6376 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6377 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6379 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6381 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6382 Add item for branch_cost.
6384 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6386 PR rtl-optimization/83565
6387 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6388 not extend the result to a larger mode for rotate operations.
6389 (num_sign_bit_copies1): Likewise.
6391 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6394 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6396 Use values-Xc.o for -pedantic.
6397 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6399 2018-01-12 Martin Liska <mliska@suse.cz>
6402 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6404 (possible_polymorphic_call_targets): Use it.
6405 (ipa_devirt): Likewise.
6407 2018-01-12 Martin Liska <mliska@suse.cz>
6409 * profile-count.h (enum profile_quality): Use 0 as invalid
6410 enum value of profile_quality.
6412 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6414 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6415 -mext-string options.
6417 2018-01-12 Richard Biener <rguenther@suse.de>
6419 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6420 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6421 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6423 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6425 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6427 * configure.ac (--with-long-double-format): Add support for the
6428 configuration option to change the default long double format on
6430 * config.gcc (powerpc*-linux*-*): Likewise.
6431 * configure: Regenerate.
6432 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6433 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6434 used without modification.
6436 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6438 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6439 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6440 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6441 MISC_BUILTIN_SPEC_BARRIER.
6442 (rs6000_init_builtins): Likewise.
6443 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6445 (speculation_barrier): New define_insn.
6446 * doc/extend.texi: Document __builtin_speculation_barrier.
6448 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6451 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6452 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6453 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6455 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6456 integral modes instead of "ss" and "sd".
6457 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6458 vectors with 32-bit and 64-bit elements.
6459 (vecdupssescalarmodesuffix): New mode attribute.
6460 (vec_dup<mode>): Use it.
6462 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6465 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6466 frame if argument is passed on stack.
6468 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6471 * ree.c (combine_reaching_defs): Optimize also
6472 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6473 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6475 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6478 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6480 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6483 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6484 after they are computed.
6486 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6488 PR tree-optimization/83695
6489 * gimple-loop-linterchange.cc
6490 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6491 reset cached scev information after interchange.
6492 (pass_linterchange::execute): Remove call to scev_reset_htab.
6494 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6496 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6497 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6498 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6499 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6500 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6501 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6502 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6503 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6504 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6505 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6506 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6507 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6508 (V_lane_reg): Likewise.
6509 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6511 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6512 (vfmal_lane_low<mode>_intrinsic,
6513 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6514 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6515 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6516 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6517 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6518 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6520 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6522 * config/arm/arm-cpus.in (fp16fml): New feature.
6523 (ALL_SIMD): Add fp16fml.
6524 (armv8.2-a): Add fp16fml as an option.
6525 (armv8.3-a): Likewise.
6526 (armv8.4-a): Add fp16fml as part of fp16.
6527 * config/arm/arm.h (TARGET_FP16FML): Define.
6528 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6530 * config/arm/arm-modes.def (V2HF): Define.
6531 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6532 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6533 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6534 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6535 vfmsl_low, vfmsl_high): New set of builtins.
6536 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6537 (vfml_op): New code attribute.
6538 (VFMLHALVES): New int iterator.
6539 (VFML, VFMLSEL): New mode attributes.
6540 (V_reg): Define mapping for V2HF.
6541 (V_hi, V_lo): New mode attributes.
6542 (VF_constraint): Likewise.
6543 (vfml_half, vfml_half_selector): New int attributes.
6544 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6546 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6547 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6549 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6550 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6551 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6552 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6554 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6555 Document new effective target and option set.
6557 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6559 * config/arm/arm-cpus.in (armv8_4): New feature.
6560 (ARMv8_4a): New fgroup.
6561 (armv8.4-a): New arch.
6562 * config/arm/arm-tables.opt: Regenerate.
6563 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6564 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6565 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6566 Add matching rules for -march=armv8.4-a and extensions.
6567 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6569 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6572 * config/rx/rx.md (BW): New mode attribute.
6573 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6575 2018-01-11 Richard Biener <rguenther@suse.de>
6577 PR tree-optimization/83435
6578 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6579 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6580 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6582 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6583 Alan Hayward <alan.hayward@arm.com>
6584 David Sherwood <david.sherwood@arm.com>
6586 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6588 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6589 (aarch64_print_address_internal): Use it to check for a zero offset.
6591 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6592 Alan Hayward <alan.hayward@arm.com>
6593 David Sherwood <david.sherwood@arm.com>
6595 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6596 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6597 Return a poly_int64 rather than a HOST_WIDE_INT.
6598 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6599 rather than a HOST_WIDE_INT.
6600 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6601 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6602 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6603 final_offset from HOST_WIDE_INT to poly_int64.
6604 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6605 to_constant when getting the number of units in an Advanced SIMD
6607 (aarch64_builtin_vectorized_function): Check for a constant number
6609 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6611 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6612 attribute instead of GET_MODE_NUNITS.
6613 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6614 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6615 GET_MODE_SIZE for fixed-size registers.
6616 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6617 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6618 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6619 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6620 (aarch64_print_operand, aarch64_print_address_internal)
6621 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6622 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6623 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6624 Handle polynomial GET_MODE_SIZE.
6625 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6626 wider than SImode without modification.
6627 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6628 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6629 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6630 passing and returning SVE modes.
6631 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6632 rather than GEN_INT.
6633 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6634 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6635 (aarch64_allocate_and_probe_stack_space): Likewise.
6636 (aarch64_layout_frame): Cope with polynomial offsets.
6637 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6638 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6640 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6641 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6642 poly_int64 rather than a HOST_WIDE_INT.
6643 (aarch64_get_separate_components, aarch64_process_components)
6644 (aarch64_expand_prologue, aarch64_expand_epilogue)
6645 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6646 (aarch64_anchor_offset): New function, split out from...
6647 (aarch64_legitimize_address): ...here.
6648 (aarch64_builtin_vectorization_cost): Handle polynomial
6649 TYPE_VECTOR_SUBPARTS.
6650 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6652 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6653 number of elements from the PARALLEL rather than the mode.
6654 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6655 rather than GET_MODE_BITSIZE.
6656 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6657 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6658 (aarch64_expand_vec_perm_const_1): Handle polynomial
6659 d->perm.length () and d->perm elements.
6660 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6661 Apply to_constant to d->perm elements.
6662 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6663 polynomial CONST_VECTOR_NUNITS.
6664 (aarch64_move_pointer): Take amount as a poly_int64 rather
6666 (aarch64_progress_pointer): Avoid temporary variable.
6667 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6668 the mode attribute instead of GET_MODE.
6670 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6671 Alan Hayward <alan.hayward@arm.com>
6672 David Sherwood <david.sherwood@arm.com>
6674 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6675 x exists before using it.
6676 (aarch64_add_constant_internal): Rename to...
6677 (aarch64_add_offset_1): ...this. Replace regnum with separate
6678 src and dest rtxes. Handle the case in which they're different,
6679 including when the offset is zero. Replace scratchreg with an rtx.
6680 Use 2 additions if there is no spare register into which we can
6681 move a 16-bit constant.
6682 (aarch64_add_constant): Delete.
6683 (aarch64_add_offset): Replace reg with separate src and dest
6684 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6685 Use aarch64_add_offset_1.
6686 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6687 an rtx rather than an int. Take the delta as a poly_int64
6688 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6689 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6690 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6691 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6692 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6694 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6695 aarch64_add_constant.
6697 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6699 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6700 Use scalar_float_mode.
6702 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6704 * config/aarch64/aarch64-simd.md
6705 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6706 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6707 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6708 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6709 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6710 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6711 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6712 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6713 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6714 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6716 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6719 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6720 targ_options->x_arm_arch_string is non NULL.
6722 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6724 * config/aarch64/aarch64.h
6725 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6727 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6730 * expmed.c (emit_store_flag_force): Swap if const op0
6731 and change VOIDmode to mode of op0.
6733 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6735 PR rtl-optimization/83761
6736 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6737 than bytes to mode_for_size.
6739 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6742 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6743 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6746 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6749 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6750 when in layout mode.
6751 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6752 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6755 2018-01-10 Michael Collison <michael.collison@arm.com>
6757 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6758 * config/aarch64/aarch64-option-extension.def: Add
6759 AARCH64_OPT_EXTENSION of 'fp16fml'.
6760 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6761 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6762 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6763 * config/aarch64/constraints.md (Ui7): New constraint.
6764 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6765 (VFMLA_SEL_W): Ditto.
6768 (VFMLA16_LOW): New int iterator.
6769 (VFMLA16_HIGH): Ditto.
6770 (UNSPEC_FMLAL): New unspec.
6771 (UNSPEC_FMLSL): Ditto.
6772 (UNSPEC_FMLAL2): Ditto.
6773 (UNSPEC_FMLSL2): Ditto.
6774 (f16mac): New code attribute.
6775 * config/aarch64/aarch64-simd-builtins.def
6776 (aarch64_fmlal_lowv2sf): Ditto.
6777 (aarch64_fmlsl_lowv2sf): Ditto.
6778 (aarch64_fmlalq_lowv4sf): Ditto.
6779 (aarch64_fmlslq_lowv4sf): Ditto.
6780 (aarch64_fmlal_highv2sf): Ditto.
6781 (aarch64_fmlsl_highv2sf): Ditto.
6782 (aarch64_fmlalq_highv4sf): Ditto.
6783 (aarch64_fmlslq_highv4sf): Ditto.
6784 (aarch64_fmlal_lane_lowv2sf): Ditto.
6785 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6786 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6787 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6788 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6789 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6790 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6791 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6792 (aarch64_fmlal_lane_highv2sf): Ditto.
6793 (aarch64_fmlsl_lane_highv2sf): Ditto.
6794 (aarch64_fmlal_laneq_highv2sf): Ditto.
6795 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6796 (aarch64_fmlalq_lane_highv4sf): Ditto.
6797 (aarch64_fmlsl_lane_highv4sf): Ditto.
6798 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6799 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6800 * config/aarch64/aarch64-simd.md:
6801 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6802 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6803 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6804 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6805 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6806 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6807 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6808 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6809 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6810 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6811 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6812 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6813 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6814 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6815 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6816 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6817 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6818 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6819 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6820 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6821 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6822 (vfmlsl_low_u32): Ditto.
6823 (vfmlalq_low_u32): Ditto.
6824 (vfmlslq_low_u32): Ditto.
6825 (vfmlal_high_u32): Ditto.
6826 (vfmlsl_high_u32): Ditto.
6827 (vfmlalq_high_u32): Ditto.
6828 (vfmlslq_high_u32): Ditto.
6829 (vfmlal_lane_low_u32): Ditto.
6830 (vfmlsl_lane_low_u32): Ditto.
6831 (vfmlal_laneq_low_u32): Ditto.
6832 (vfmlsl_laneq_low_u32): Ditto.
6833 (vfmlalq_lane_low_u32): Ditto.
6834 (vfmlslq_lane_low_u32): Ditto.
6835 (vfmlalq_laneq_low_u32): Ditto.
6836 (vfmlslq_laneq_low_u32): Ditto.
6837 (vfmlal_lane_high_u32): Ditto.
6838 (vfmlsl_lane_high_u32): Ditto.
6839 (vfmlal_laneq_high_u32): Ditto.
6840 (vfmlsl_laneq_high_u32): Ditto.
6841 (vfmlalq_lane_high_u32): Ditto.
6842 (vfmlslq_lane_high_u32): Ditto.
6843 (vfmlalq_laneq_high_u32): Ditto.
6844 (vfmlslq_laneq_high_u32): Ditto.
6845 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6846 (AARCH64_FL_FOR_ARCH8_4): New.
6847 (AARCH64_ISA_F16FML): New ISA flag.
6848 (TARGET_F16FML): New feature flag for fp16fml.
6849 (doc/invoke.texi): Document new fp16fml option.
6851 2018-01-10 Michael Collison <michael.collison@arm.com>
6853 * config/aarch64/aarch64-builtins.c:
6854 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6855 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6856 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6857 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6858 (AARCH64_ISA_SHA3): New ISA flag.
6859 (TARGET_SHA3): New feature flag for sha3.
6860 * config/aarch64/iterators.md (sha512_op): New int attribute.
6861 (CRYPTO_SHA512): New int iterator.
6862 (UNSPEC_SHA512H): New unspec.
6863 (UNSPEC_SHA512H2): Ditto.
6864 (UNSPEC_SHA512SU0): Ditto.
6865 (UNSPEC_SHA512SU1): Ditto.
6866 * config/aarch64/aarch64-simd-builtins.def
6867 (aarch64_crypto_sha512hqv2di): New builtin.
6868 (aarch64_crypto_sha512h2qv2di): Ditto.
6869 (aarch64_crypto_sha512su0qv2di): Ditto.
6870 (aarch64_crypto_sha512su1qv2di): Ditto.
6871 (aarch64_eor3qv8hi): Ditto.
6872 (aarch64_rax1qv2di): Ditto.
6873 (aarch64_xarqv2di): Ditto.
6874 (aarch64_bcaxqv8hi): Ditto.
6875 * config/aarch64/aarch64-simd.md:
6876 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6877 (aarch64_crypto_sha512su0qv2di): Ditto.
6878 (aarch64_crypto_sha512su1qv2di): Ditto.
6879 (aarch64_eor3qv8hi): Ditto.
6880 (aarch64_rax1qv2di): Ditto.
6881 (aarch64_xarqv2di): Ditto.
6882 (aarch64_bcaxqv8hi): Ditto.
6883 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6884 (vsha512h2q_u64): Ditto.
6885 (vsha512su0q_u64): Ditto.
6886 (vsha512su1q_u64): Ditto.
6887 (veor3q_u16): Ditto.
6888 (vrax1q_u64): Ditto.
6890 (vbcaxq_u16): Ditto.
6891 * config/arm/types.md (crypto_sha512): New type attribute.
6892 (crypto_sha3): Ditto.
6893 (doc/invoke.texi): Document new sha3 option.
6895 2018-01-10 Michael Collison <michael.collison@arm.com>
6897 * config/aarch64/aarch64-builtins.c:
6898 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6899 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6900 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6901 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6902 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6903 (AARCH64_ISA_SM4): New ISA flag.
6904 (TARGET_SM4): New feature flag for sm4.
6905 * config/aarch64/aarch64-simd-builtins.def
6906 (aarch64_sm3ss1qv4si): Ditto.
6907 (aarch64_sm3tt1aq4si): Ditto.
6908 (aarch64_sm3tt1bq4si): Ditto.
6909 (aarch64_sm3tt2aq4si): Ditto.
6910 (aarch64_sm3tt2bq4si): Ditto.
6911 (aarch64_sm3partw1qv4si): Ditto.
6912 (aarch64_sm3partw2qv4si): Ditto.
6913 (aarch64_sm4eqv4si): Ditto.
6914 (aarch64_sm4ekeyqv4si): Ditto.
6915 * config/aarch64/aarch64-simd.md:
6916 (aarch64_sm3ss1qv4si): Ditto.
6917 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6918 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6919 (aarch64_sm4eqv4si): Ditto.
6920 (aarch64_sm4ekeyqv4si): Ditto.
6921 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6922 (sm3part_op): Ditto.
6923 (CRYPTO_SM3TT): Ditto.
6924 (CRYPTO_SM3PART): Ditto.
6925 (UNSPEC_SM3SS1): New unspec.
6926 (UNSPEC_SM3TT1A): Ditto.
6927 (UNSPEC_SM3TT1B): Ditto.
6928 (UNSPEC_SM3TT2A): Ditto.
6929 (UNSPEC_SM3TT2B): Ditto.
6930 (UNSPEC_SM3PARTW1): Ditto.
6931 (UNSPEC_SM3PARTW2): Ditto.
6932 (UNSPEC_SM4E): Ditto.
6933 (UNSPEC_SM4EKEY): Ditto.
6934 * config/aarch64/constraints.md (Ui2): New constraint.
6935 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6936 * config/arm/types.md (crypto_sm3): New type attribute.
6937 (crypto_sm4): Ditto.
6938 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6939 (vsm3tt1aq_u32): Ditto.
6940 (vsm3tt1bq_u32): Ditto.
6941 (vsm3tt2aq_u32): Ditto.
6942 (vsm3tt2bq_u32): Ditto.
6943 (vsm3partw1q_u32): Ditto.
6944 (vsm3partw2q_u32): Ditto.
6945 (vsm4eq_u32): Ditto.
6946 (vsm4ekeyq_u32): Ditto.
6947 (doc/invoke.texi): Document new sm4 option.
6949 2018-01-10 Michael Collison <michael.collison@arm.com>
6951 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6952 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6953 (AARCH64_FL_FOR_ARCH8_4): New.
6954 (AARCH64_FL_V8_4): New flag.
6955 (doc/invoke.texi): Document new armv8.4-a option.
6957 2018-01-10 Michael Collison <michael.collison@arm.com>
6959 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6960 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6961 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6962 * config/aarch64/aarch64-option-extension.def: Add
6963 AARCH64_OPT_EXTENSION of 'sha2'.
6964 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6965 (crypto): Disable sha2 and aes if crypto disabled.
6966 (crypto): Enable aes and sha2 if enabled.
6967 (simd): Disable sha2 and aes if simd disabled.
6968 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6970 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6971 (TARGET_SHA2): New feature flag for sha2.
6972 (TARGET_AES): New feature flag for aes.
6973 * config/aarch64/aarch64-simd.md:
6974 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6975 conditional on TARGET_AES.
6976 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6977 (aarch64_crypto_sha1hsi): Make pattern conditional
6979 (aarch64_crypto_sha1hv4si): Ditto.
6980 (aarch64_be_crypto_sha1hv4si): Ditto.
6981 (aarch64_crypto_sha1su1v4si): Ditto.
6982 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6983 (aarch64_crypto_sha1su0v4si): Ditto.
6984 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6985 (aarch64_crypto_sha256su0v4si): Ditto.
6986 (aarch64_crypto_sha256su1v4si): Ditto.
6987 (doc/invoke.texi): Document new aes and sha2 options.
6989 2018-01-10 Martin Sebor <msebor@redhat.com>
6991 PR tree-optimization/83781
6992 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6995 2018-01-11 Martin Sebor <msebor@gmail.com>
6996 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6998 PR tree-optimization/83501
6999 PR tree-optimization/81703
7001 * tree-ssa-strlen.c (get_string_cst): Rename...
7002 (get_string_len): ...to this. Handle global constants.
7003 (handle_char_store): Adjust.
7005 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7006 Jim Wilson <jimw@sifive.com>
7008 * config/riscv/riscv-protos.h (riscv_output_return): New.
7009 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7010 (riscv_attribute_table, riscv_output_return),
7011 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7012 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7013 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7015 (riscv_expand_prologue): Add early return for naked function.
7016 (riscv_expand_epilogue): Likewise.
7017 (riscv_function_ok_for_sibcall): Return false for naked function.
7018 (riscv_set_current_function): New.
7019 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7020 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7021 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7022 * doc/extend.texi (RISC-V Function Attributes): New.
7024 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7026 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7027 check for 128-bit long double before checking TCmode.
7028 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7029 128-bit long doubles before checking TFmode or TCmode.
7030 (FLOAT128_IBM_P): Likewise.
7032 2018-01-10 Martin Sebor <msebor@redhat.com>
7034 PR tree-optimization/83671
7035 * builtins.c (c_strlen): Unconditionally return zero for the empty
7037 Use -Warray-bounds for warnings.
7038 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7039 for non-constant array indices with COMPONENT_REF, arrays of
7040 arrays, and pointers to arrays.
7041 (gimple_fold_builtin_strlen): Determine and set length range for
7042 non-constant character arrays.
7044 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7047 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7050 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7052 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7054 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7057 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7058 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7059 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7060 indexed_or_indirect_operand predicate.
7061 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7062 (*vsx_le_perm_load_v8hi): Likewise.
7063 (*vsx_le_perm_load_v16qi): Likewise.
7064 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7065 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7066 (*vsx_le_perm_store_v8hi): Likewise.
7067 (*vsx_le_perm_store_v16qi): Likewise.
7068 (eight unnamed splitters): Likewise.
7070 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7072 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7073 * config/rs6000/emmintrin.h: Likewise.
7074 * config/rs6000/mmintrin.h: Likewise.
7075 * config/rs6000/xmmintrin.h: Likewise.
7077 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7080 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7082 * tree.c (tree_nop_conversion): Return true for location wrapper
7084 (maybe_wrap_with_location): New function.
7085 (selftest::check_strip_nops): New function.
7086 (selftest::test_location_wrappers): New function.
7087 (selftest::tree_c_tests): Call it.
7088 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7089 (maybe_wrap_with_location): New decl.
7090 (EXPR_LOCATION_WRAPPER_P): New macro.
7091 (location_wrapper_p): New inline function.
7092 (tree_strip_any_location_wrapper): New inline function.
7094 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7097 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7098 stack_realign_offset for the largest alignment of stack slot
7100 (ix86_find_max_used_stack_alignment): New function.
7101 (ix86_finalize_stack_frame_flags): Use it. Set
7102 max_used_stack_alignment if we don't realign stack.
7103 * config/i386/i386.h (machine_function): Add
7104 max_used_stack_alignment.
7106 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7108 * config/arm/arm.opt (-mbranch-cost): New option.
7109 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7112 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7115 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7116 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7118 2018-01-10 Richard Biener <rguenther@suse.de>
7121 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7122 early out so it also covers the case where we have a non-NULL
7125 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7127 PR tree-optimization/83753
7128 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7129 for non-strided grouped accesses if the number of elements is 1.
7131 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7134 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7135 * i386.h (TARGET_USE_GATHER): Define.
7136 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7138 2018-01-10 Martin Liska <mliska@suse.cz>
7141 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7142 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7144 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7145 CLEANUP_NO_PARTITIONING is not set.
7147 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7149 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7150 for vectors, as a partial revert of r254296.
7151 * rtl.h (const_vec_p): Delete.
7152 (const_vec_duplicate_p): Don't test for vector CONSTs.
7153 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7154 * expmed.c (make_tree): Likewise.
7157 * common.md (E, F): Use CONSTANT_P instead of checking for
7159 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7160 checking for CONST_VECTOR.
7162 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7165 * predict.c (force_edge_cold): Handle in more sane way edges
7168 2018-01-09 Carl Love <cel@us.ibm.com>
7170 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7172 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7173 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7174 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7175 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7176 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7177 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7178 * config/rs6000/rs6000-protos.h: Add extern defition for
7179 rs6000_generate_float2_double_code.
7180 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7182 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7183 (float2_v2df): Add define_expand.
7185 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7188 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7189 op_mode in the force_to_mode call.
7191 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7193 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7194 instead of checking each element individually.
7195 (aarch64_evpc_uzp): Likewise.
7196 (aarch64_evpc_zip): Likewise.
7197 (aarch64_evpc_ext): Likewise.
7198 (aarch64_evpc_rev): Likewise.
7199 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7200 instead of checking each element individually. Return true without
7202 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7203 whether all selected elements come from the same input, instead of
7204 checking each element individually. Remove calls to gen_rtx_REG,
7205 start_sequence and end_sequence and instead assert that no rtl is
7208 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7210 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7211 order of HIGH and CONST checks.
7213 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7215 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7216 if the destination isn't an SSA_NAME.
7218 2018-01-09 Richard Biener <rguenther@suse.de>
7220 PR tree-optimization/83668
7221 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7223 (canonicalize_loop_form): ... here, renamed from ...
7224 (canonicalize_loop_closed_ssa_form): ... this and amended to
7225 swap successor edges for loop exit blocks to make us use
7226 the RPO order we need for initial schedule generation.
7228 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7230 PR tree-optimization/64811
7231 * match.pd: When optimizing comparisons with Inf, avoid
7232 introducing or losing exceptions from comparisons with NaN.
7234 2018-01-09 Martin Liska <mliska@suse.cz>
7237 * asan.c (shadow_mem_size): Add gcc_assert.
7239 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7241 Don't save registers in main().
7244 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7245 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7246 * config/avr/avr.c (avr_set_current_function): Don't error if
7247 naked, OS_task or OS_main are specified at the same time.
7248 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7250 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7252 * common/config/avr/avr-common.c (avr_option_optimization_table):
7253 Switch on -mmain-is-OS_task for optimizing compilations.
7255 2018-01-09 Richard Biener <rguenther@suse.de>
7257 PR tree-optimization/83572
7258 * graphite.c: Include cfganal.h.
7259 (graphite_transform_loops): Connect infinite loops to exit
7260 and remove fake edges at the end.
7262 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7264 * ipa-inline.c (edge_badness): Revert accidental checkin.
7266 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7269 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7270 symbols; not inline clones.
7272 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7275 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7276 hard registers. Formatting fixes.
7278 PR preprocessor/83722
7279 * gcc.c (try_generate_repro): Pass
7280 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7281 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7284 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7285 Kito Cheng <kito.cheng@gmail.com>
7287 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7288 (riscv_leaf_function_p): Delete.
7289 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7291 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7293 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7295 (do_ifelse): New function.
7296 (do_isel): New function.
7297 (do_sub3): New function.
7298 (do_add3): New function.
7299 (do_load_mask_compare): New function.
7300 (do_overlap_load_compare): New function.
7301 (expand_compare_loop): New function.
7302 (expand_block_compare): Call expand_compare_loop() when appropriate.
7303 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7305 (-mblock-compare-inline-loop-limit): New option.
7307 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7310 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7311 Reverse order of second and third operands in first alternative.
7312 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7313 of first and second elements in UNSPEC_VPERMR vector.
7314 (altivec_expand_vec_perm_le): Likewise.
7316 2018-01-08 Jeff Law <law@redhat.com>
7318 PR rtl-optimizatin/81308
7319 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7320 (process_switch): If group_case_labels makes a change, then set
7322 (pass_convert_switch::execute): If a switch is converted, then
7323 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7325 PR rtl-optimization/81308
7326 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7329 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7331 PR target/83663 - Revert r255946
7332 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7333 generation for cases where splatting a value is not useful.
7334 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7335 across a vec_duplicate and a paradoxical subreg forming a vector
7336 mode to a vec_concat.
7338 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7340 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7341 -march=armv8.3-a variants.
7342 * config/arm/t-multilib: Likewise.
7343 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7345 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7347 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7349 (cceq_ior_compare_complement): Give it a name so I can use it, and
7350 change boolean_or_operator predicate to boolean_operator so it can
7351 be used to generate a crand.
7352 (eqne): New code iterator.
7353 (bd/bd_neg): New code_attrs.
7354 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7355 a single define_insn.
7356 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7357 decrement (bdnzt/bdnzf/bdzt/bdzf).
7358 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7359 with the new names of the branch decrement patterns, and added the
7360 names of the branch decrement conditional patterns.
7362 2018-01-08 Richard Biener <rguenther@suse.de>
7364 PR tree-optimization/83563
7365 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7368 2018-01-08 Richard Biener <rguenther@suse.de>
7371 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7373 2018-01-08 Richard Biener <rguenther@suse.de>
7375 PR tree-optimization/83685
7376 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7377 references to abnormals.
7379 2018-01-08 Richard Biener <rguenther@suse.de>
7382 * dwarf2out.c (output_indirect_strings): Handle empty
7383 skeleton_debug_str_hash.
7384 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7386 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7388 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7389 (emit_store_direct): Likewise.
7390 (arc_trampoline_adjust_address): Likewise.
7391 (arc_asm_trampoline_template): New function.
7392 (arc_initialize_trampoline): Use asm_trampoline_template.
7393 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7394 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7395 * config/arc/arc.md (flush_icache): Delete pattern.
7397 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7399 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7400 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7403 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7406 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7407 by not USED_FOR_TARGET.
7408 (make_pass_resolve_sw_modes): Likewise.
7410 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7412 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7415 2018-01-08 Richard Biener <rguenther@suse.de>
7418 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7420 2018-01-08 Richard Biener <rguenther@suse.de>
7423 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7425 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7428 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7429 basic blocks with a small number of successors.
7430 (convert_control_dep_chain_into_preds): Improve handling of
7432 (dump_predicates): Split apart into...
7433 (dump_pred_chain): ...here...
7434 (dump_pred_info): ...and here.
7435 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7436 (can_chain_union_be_invalidated_p): Improve check for invalidation
7438 (uninit_uses_cannot_happen): Avoid unnecessary if
7439 convert_control_dep_chain_into_preds yielded nothing.
7441 2018-01-06 Martin Sebor <msebor@redhat.com>
7443 PR tree-optimization/83640
7444 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7445 subtracting negative offset from size.
7446 (builtin_access::overlap): Adjust offset bounds of the access to fall
7447 within the size of the object if possible.
7449 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7451 PR rtl-optimization/83699
7452 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7453 extract_bit_field_as_subreg to cases in which the extracted
7454 value is also a vector.
7456 * lra-constraints.c (process_alt_operands): Test for the equivalence
7457 substitutions when detecting a possible reload cycle.
7459 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7462 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7463 by default if flag_selective_schedling{,2}. Formatting fixes.
7465 PR rtl-optimization/83682
7466 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7467 if it has non-VECTOR_MODE element mode.
7468 (vec_duplicate_p): Likewise.
7471 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7472 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7474 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7477 * config/i386/i386-builtin.def
7478 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7479 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7480 Require also OPTION_MASK_ISA_AVX512F in addition to
7481 OPTION_MASK_ISA_GFNI.
7482 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7483 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7484 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7485 to OPTION_MASK_ISA_GFNI.
7486 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7487 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7488 OPTION_MASK_ISA_AVX512BW.
7489 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7490 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7491 addition to OPTION_MASK_ISA_GFNI.
7492 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7493 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7494 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7495 to OPTION_MASK_ISA_GFNI.
7496 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7497 a requirement for all ISAs rather than any of them with a few
7499 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7501 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7502 bitmasks to be enabled with 3 exceptions, instead of requiring any
7503 enabled ISA with lots of exceptions.
7504 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7505 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7506 Change avx512bw in isa attribute to avx512f.
7507 * config/i386/sgxintrin.h: Add license boilerplate.
7508 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7509 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7510 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7511 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7513 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7514 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7515 temporarily sse2 rather than sse if not enabled already.
7518 * config/i386/sse.md (VI248_VLBW): Rename to ...
7519 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7520 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7521 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7522 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7523 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7524 mode iterator instead of VI248_VLBW.
7526 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7528 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7529 (record_modified): Skip clobbers; add debug output.
7530 (param_change_prob): Use sreal frequencies.
7532 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7534 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7535 punt for user-aligned variables.
7537 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7539 * tree-chrec.c (chrec_contains_symbols): Return true for
7542 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7545 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7546 of (x|y) == x for BICS pattern.
7548 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7550 PR tree-optimization/83605
7551 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7552 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7555 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7557 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7558 * config/epiphany/rtems.h: New file.
7560 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7561 Uros Bizjak <ubizjak@gmail.com>
7564 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7565 QIreg_operand instead of register_operand predicate.
7566 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7567 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7568 comments instead of -fmitigate[-_]rop.
7570 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7573 * cgraphunit.c (symbol_table::compile): Switch to text_section
7574 before calling assembly_start debug hook.
7575 * run-rtl-passes.c (run_rtl_passes): Likewise.
7578 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7580 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7581 range_int_cst_p rather than !symbolic_range_p before calling
7582 extract_range_from_multiplicative_op_1.
7584 2018-01-04 Jeff Law <law@redhat.com>
7586 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7587 redundant test in assertion.
7589 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7591 * doc/rtl.texi: Document machine_mode wrapper classes.
7593 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7595 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7598 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7600 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7601 the VEC_PERM_EXPR fold to fail.
7603 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7606 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7607 to switched_sections.
7609 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7612 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7615 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7618 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7619 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7621 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7624 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7625 is BLKmode and bitpos not zero or mode change is needed.
7627 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7630 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7633 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7636 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7637 instead of MULT rtx. Update all corresponding splitters.
7639 (*ssub<modesuffix>): Ditto.
7641 (*cmp_sadd_di): Update split patterns.
7642 (*cmp_sadd_si): Ditto.
7643 (*cmp_sadd_sidi): Ditto.
7644 (*cmp_ssub_di): Ditto.
7645 (*cmp_ssub_si): Ditto.
7646 (*cmp_ssub_sidi): Ditto.
7647 * config/alpha/predicates.md (const23_operand): New predicate.
7648 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7649 Look for ASHIFT, not MULT inner operand.
7650 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7652 2018-01-04 Martin Liska <mliska@suse.cz>
7654 PR gcov-profile/83669
7655 * gcov.c (output_intermediate_file): Add version to intermediate
7657 * doc/gcov.texi: Document new field 'version' in intermediate
7658 file format. Fix location of '-k' option of gcov command.
7660 2018-01-04 Martin Liska <mliska@suse.cz>
7663 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7665 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7667 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7669 2018-01-03 Martin Sebor <msebor@redhat.com>
7671 PR tree-optimization/83655
7672 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7673 checking calls with invalid arguments.
7675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7677 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7678 (vectorizable_mask_load_store): Delete.
7679 (vectorizable_call): Return false for masked loads and stores.
7680 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7681 instead of gimple_assign_rhs1.
7682 (vectorizable_load): Handle IFN_MASK_LOAD.
7683 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7687 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7689 (vectorizable_mask_load_store): ...here.
7690 (vectorizable_load): ...and here.
7692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7694 * tree-vect-stmts.c (vect_build_all_ones_mask)
7695 (vect_build_zero_merge_argument): New functions, split out from...
7696 (vectorizable_load): ...here.
7698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7700 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7702 (vectorizable_mask_load_store): ...here.
7703 (vectorizable_store): ...and here.
7705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7707 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7709 (vectorizable_mask_load_store): ...here.
7711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7713 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7714 (vect_model_store_cost): Take a vec_load_store_type instead of a
7716 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7717 (vect_model_store_cost): Take a vec_load_store_type instead of a
7719 (vectorizable_mask_load_store): Update accordingly.
7720 (vectorizable_store): Likewise.
7721 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7725 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7726 IFN_MASK_LOAD calls here rather than...
7727 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7730 Alan Hayward <alan.hayward@arm.com>
7731 David Sherwood <david.sherwood@arm.com>
7733 * expmed.c (extract_bit_field_1): For vector extracts,
7734 fall back to extract_bit_field_as_subreg if vec_extract
7737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7738 Alan Hayward <alan.hayward@arm.com>
7739 David Sherwood <david.sherwood@arm.com>
7741 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7742 they are variable or constant sized.
7743 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7744 slots for constant-sized data.
7746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7747 Alan Hayward <alan.hayward@arm.com>
7748 David Sherwood <david.sherwood@arm.com>
7750 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7751 handling COND_EXPRs with boolean comparisons, try to find a better
7752 basis for the mask type than the boolean itself.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7756 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7757 is calculated and how it can be overridden.
7758 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7759 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7761 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7765 Alan Hayward <alan.hayward@arm.com>
7766 David Sherwood <david.sherwood@arm.com>
7768 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7769 Remove the mode argument.
7770 (aarch64_simd_valid_immediate): Remove the mode and inverse
7772 * config/aarch64/iterators.md (bitsize): New iterator.
7773 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7774 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7775 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7776 aarch64_simd_valid_immediate.
7777 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7778 (aarch64_reg_or_bic_imm): Likewise.
7779 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7780 with an insn_type enum and msl with a modifier_type enum.
7781 Replace element_width with a scalar_mode. Change the shift
7782 to unsigned int. Add constructors for scalar_float_mode and
7783 scalar_int_mode elements.
7784 (aarch64_vect_float_const_representable_p): Delete.
7785 (aarch64_can_const_movi_rtx_p)
7786 (aarch64_simd_scalar_immediate_valid_for_move)
7787 (aarch64_simd_make_constant): Update call to
7788 aarch64_simd_valid_immediate.
7789 (aarch64_advsimd_valid_immediate_hs): New function.
7790 (aarch64_advsimd_valid_immediate): Likewise.
7791 (aarch64_simd_valid_immediate): Remove mode and inverse
7792 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7793 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7794 and aarch64_float_const_representable_p on the result.
7795 (aarch64_output_simd_mov_immediate): Remove mode argument.
7796 Update call to aarch64_simd_valid_immediate and use of
7797 simd_immediate_info.
7798 (aarch64_output_scalar_simd_mov_immediate): Update call
7801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7802 Alan Hayward <alan.hayward@arm.com>
7803 David Sherwood <david.sherwood@arm.com>
7805 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7806 (mode_nunits): Likewise CONST_MODE_NUNITS.
7807 * machmode.def (ADJUST_NUNITS): Document.
7808 * genmodes.c (mode_data::need_nunits_adj): New field.
7809 (blank_mode): Update accordingly.
7810 (adj_nunits): New variable.
7811 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7813 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7814 listed in adj_nunits.
7815 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7816 listed in adj_nunits. Don't emit case statements for such modes.
7817 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7818 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7819 nothing if adj_nunits is nonnull.
7820 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7821 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7822 (emit_mode_fbit): Update use of print_maybe_const_decl.
7823 (emit_move_size): Likewise. Treat the array as non-const
7825 (emit_mode_adjustments): Handle adj_nunits.
7827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7829 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7830 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7831 (VECTOR_MODES): Use it.
7832 (make_vector_modes): Take the prefix as an argument.
7834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7835 Alan Hayward <alan.hayward@arm.com>
7836 David Sherwood <david.sherwood@arm.com>
7838 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7839 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7840 for MODE_VECTOR_BOOL.
7841 * machmode.def (VECTOR_BOOL_MODE): Document.
7842 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7843 (make_vector_bool_mode): New function.
7844 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7846 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7847 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7849 * stor-layout.c (int_mode_for_mode): Likewise.
7850 * tree.c (build_vector_type_for_mode): Likewise.
7851 * varasm.c (output_constant_pool_2): Likewise.
7852 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7853 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7854 for MODE_VECTOR_BOOL.
7855 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7856 of mode class checks.
7857 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7858 instead of a list of mode class checks.
7859 (expand_vector_scalar_condition): Likewise.
7860 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7863 Alan Hayward <alan.hayward@arm.com>
7864 David Sherwood <david.sherwood@arm.com>
7866 * machmode.h (mode_size): Change from unsigned short to
7868 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7869 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7870 or if measurement_type is not polynomial.
7871 (fixed_size_mode::includes_p): Check for constant-sized modes.
7872 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7873 return a poly_uint16 rather than an unsigned short.
7874 (emit_mode_size): Change the type of mode_size from unsigned short
7875 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7876 (emit_mode_adjustments): Cope with polynomial vector sizes.
7877 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7879 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7881 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7882 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7883 * caller-save.c (setup_save_areas): Likewise.
7884 (replace_reg_with_saved_mem): Likewise.
7885 * calls.c (emit_library_call_value_1): Likewise.
7886 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7887 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7888 (gen_lowpart_for_combine): Likewise.
7889 * convert.c (convert_to_integer_1): Likewise.
7890 * cse.c (equiv_constant, cse_insn): Likewise.
7891 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7892 (cselib_subst_to_values): Likewise.
7893 * dce.c (word_dce_process_block): Likewise.
7894 * df-problems.c (df_word_lr_mark_ref): Likewise.
7895 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7896 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7897 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7898 (rtl_for_decl_location): Likewise.
7899 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7900 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7901 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7902 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7903 (expand_expr_real_1): Likewise.
7904 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7905 (pad_below): Likewise.
7906 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7907 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7908 * ira.c (get_subreg_tracking_sizes): Likewise.
7909 * ira-build.c (ira_create_allocno_objects): Likewise.
7910 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7911 (ira_sort_regnos_for_alter_reg): Likewise.
7912 * ira-costs.c (record_operand_costs): Likewise.
7913 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7914 (resolve_simple_move): Likewise.
7915 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7916 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7917 (lra_constraints): Likewise.
7918 (CONST_POOL_OK_P): Reject variable-sized modes.
7919 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7920 (add_pseudo_to_slot, lra_spill): Likewise.
7921 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7922 * optabs-query.c (get_best_extraction_insn): Likewise.
7923 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7924 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7925 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7926 * recog.c (offsettable_address_addr_space_p): Likewise.
7927 * regcprop.c (maybe_mode_change): Likewise.
7928 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7929 * regrename.c (build_def_use): Likewise.
7930 * regstat.c (dump_reg_info): Likewise.
7931 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7932 (find_reloads, find_reloads_subreg_address): Likewise.
7933 * reload1.c (eliminate_regs_1): Likewise.
7934 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7935 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7936 (simplify_binary_operation_1, simplify_subreg): Likewise.
7937 * targhooks.c (default_function_arg_padding): Likewise.
7938 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7939 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7940 (verify_gimple_assign_ternary): Likewise.
7941 * tree-inline.c (estimate_move_cost): Likewise.
7942 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7943 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7944 (get_address_cost_ainc): Likewise.
7945 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7946 (vect_supportable_dr_alignment): Likewise.
7947 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7948 (vectorizable_reduction): Likewise.
7949 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7950 (vectorizable_operation, vectorizable_load): Likewise.
7951 * tree.c (build_same_sized_truth_vector_type): Likewise.
7952 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7953 * var-tracking.c (emit_note_insn_var_location): Likewise.
7954 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7955 (ADDR_VEC_ALIGN): Likewise.
7957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7958 Alan Hayward <alan.hayward@arm.com>
7959 David Sherwood <david.sherwood@arm.com>
7961 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7963 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7964 or if measurement_type is polynomial.
7965 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7966 * combine.c (make_extraction): Likewise.
7967 * dse.c (find_shift_sequence): Likewise.
7968 * dwarf2out.c (mem_loc_descriptor): Likewise.
7969 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7970 (extract_bit_field, extract_low_bits): Likewise.
7971 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7972 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7973 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7974 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7975 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7976 * reload.c (find_reloads): Likewise.
7977 * reload1.c (alter_reg): Likewise.
7978 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7979 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7980 * tree-if-conv.c (predicate_mem_writes): Likewise.
7981 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7982 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7983 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7984 * valtrack.c (dead_debug_insert_temp): Likewise.
7985 * varasm.c (mergeable_constant_section): Likewise.
7986 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7989 Alan Hayward <alan.hayward@arm.com>
7990 David Sherwood <david.sherwood@arm.com>
7992 * expr.c (expand_assignment): Cope with polynomial mode sizes
7993 when assigning to a CONCAT.
7995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7996 Alan Hayward <alan.hayward@arm.com>
7997 David Sherwood <david.sherwood@arm.com>
7999 * machmode.h (mode_precision): Change from unsigned short to
8001 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8003 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8004 or if measurement_type is not polynomial.
8005 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8006 in which the mode is already known to be a scalar_int_mode.
8007 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8008 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8010 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8011 for GET_MODE_PRECISION.
8012 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8013 for GET_MODE_PRECISION.
8014 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8016 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8017 (expand_field_assignment, make_extraction): Likewise.
8018 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8019 (get_last_value): Likewise.
8020 * convert.c (convert_to_integer_1): Likewise.
8021 * cse.c (cse_insn): Likewise.
8022 * expr.c (expand_expr_real_1): Likewise.
8023 * lra-constraints.c (simplify_operand_subreg): Likewise.
8024 * optabs-query.c (can_atomic_load_p): Likewise.
8025 * optabs.c (expand_atomic_load): Likewise.
8026 (expand_atomic_store): Likewise.
8027 * ree.c (combine_reaching_defs): Likewise.
8028 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8029 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8030 * tree.h (type_has_mode_precision_p): Likewise.
8031 * ubsan.c (instrument_si_overflow): Likewise.
8033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8034 Alan Hayward <alan.hayward@arm.com>
8035 David Sherwood <david.sherwood@arm.com>
8037 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8038 polynomial numbers of units.
8039 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8040 (valid_vector_subparts_p): New function.
8041 (build_vector_type): Remove temporary shim and take the number
8042 of units as a poly_uint64 rather than an int.
8043 (build_opaque_vector_type): Take the number of units as a
8044 poly_uint64 rather than an int.
8045 * tree.c (build_vector_from_ctor): Handle polynomial
8046 TYPE_VECTOR_SUBPARTS.
8047 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8048 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8049 (build_vector_from_val): If the number of units is variable,
8050 use build_vec_duplicate_cst for constant operands and
8051 VEC_DUPLICATE_EXPR otherwise.
8052 (make_vector_type): Remove temporary is_constant ().
8053 (build_vector_type, build_opaque_vector_type): Take the number of
8054 units as a poly_uint64 rather than an int.
8055 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8057 * cfgexpand.c (expand_debug_expr): Likewise.
8058 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8059 (store_constructor, expand_expr_real_1): Likewise.
8060 (const_scalar_mask_from_tree): Likewise.
8061 * fold-const-call.c (fold_const_reduction): Likewise.
8062 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8063 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8064 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8065 (fold_relational_const): Likewise.
8066 (native_interpret_vector): Likewise. Change the size from an
8067 int to an unsigned int.
8068 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8069 TYPE_VECTOR_SUBPARTS.
8070 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8071 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8072 duplicating a non-constant operand into a variable-length vector.
8073 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8074 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8075 * ipa-icf.c (sem_variable::equals): Likewise.
8076 * match.pd: Likewise.
8077 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8078 * print-tree.c (print_node): Likewise.
8079 * stor-layout.c (layout_type): Likewise.
8080 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8081 * tree-cfg.c (verify_gimple_comparison): Likewise.
8082 (verify_gimple_assign_binary): Likewise.
8083 (verify_gimple_assign_ternary): Likewise.
8084 (verify_gimple_assign_single): Likewise.
8085 * tree-pretty-print.c (dump_generic_node): Likewise.
8086 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8087 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8088 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8089 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8090 (vect_shift_permute_load_chain): Likewise.
8091 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8092 (expand_vector_condition, optimize_vector_constructor): Likewise.
8093 (lower_vec_perm, get_compute_type): Likewise.
8094 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8095 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8096 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8097 (vect_recog_mask_conversion_pattern): Likewise.
8098 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8099 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8100 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8101 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8102 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8103 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8104 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8105 (supportable_widening_operation): Likewise.
8106 (supportable_narrowing_operation): Likewise.
8107 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8109 * varasm.c (output_constant): Likewise.
8111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8112 Alan Hayward <alan.hayward@arm.com>
8113 David Sherwood <david.sherwood@arm.com>
8115 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8116 so that both the length == 3 and length != 3 cases set up their
8117 own permute vectors. Add comments explaining why we know the
8118 number of elements is constant.
8119 (vect_permute_load_chain): Likewise.
8121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8122 Alan Hayward <alan.hayward@arm.com>
8123 David Sherwood <david.sherwood@arm.com>
8125 * machmode.h (mode_nunits): Change from unsigned char to
8127 (ONLY_FIXED_SIZE_MODES): New macro.
8128 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8129 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8130 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8132 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8133 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8134 or if measurement_type is not polynomial.
8135 * genmodes.c (ZERO_COEFFS): New macro.
8136 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8138 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8139 Use ZERO_COEFFS when emitting initializers.
8140 * data-streamer.h (bp_pack_poly_value): New function.
8141 (bp_unpack_poly_value): Likewise.
8142 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8143 for GET_MODE_NUNITS.
8144 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8145 for GET_MODE_NUNITS.
8146 * tree.c (make_vector_type): Remove temporary shim and make
8147 the real function take the number of units as a poly_uint64
8149 (build_vector_type_for_mode): Handle polynomial nunits.
8150 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8151 * emit-rtl.c (const_vec_series_p_1): Likewise.
8152 (gen_rtx_CONST_VECTOR): Likewise.
8153 * fold-const.c (test_vec_duplicate_folding): Likewise.
8154 * genrecog.c (validate_pattern): Likewise.
8155 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8156 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8157 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8158 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8159 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8160 * rtlanal.c (subreg_get_info): Likewise.
8161 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8162 (vect_grouped_load_supported): Likewise.
8163 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8164 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8165 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8166 (simplify_const_unary_operation, simplify_binary_operation_1)
8167 (simplify_const_binary_operation, simplify_ternary_operation)
8168 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8169 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8170 instead of CONST_VECTOR_NUNITS.
8171 * varasm.c (output_constant_pool_2): Likewise.
8172 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8173 explicit-encoded elements in the XVEC for variable-length vectors.
8175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8177 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8180 Alan Hayward <alan.hayward@arm.com>
8181 David Sherwood <david.sherwood@arm.com>
8183 * coretypes.h (fixed_size_mode): Declare.
8184 (fixed_size_mode_pod): New typedef.
8185 * builtins.h (target_builtins::x_apply_args_mode)
8186 (target_builtins::x_apply_result_mode): Change type to
8187 fixed_size_mode_pod.
8188 * builtins.c (apply_args_size, apply_result_size, result_vector)
8189 (expand_builtin_apply_args_1, expand_builtin_apply)
8190 (expand_builtin_return): Update accordingly.
8192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8194 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8195 * cselib.c (cselib_hash_rtx): Likewise.
8196 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8197 CONST_VECTOR encoding.
8199 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8200 Jeff Law <law@redhat.com>
8203 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8204 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8205 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8206 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8209 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8210 explicitly probe *sp in a noreturn function if there were any callee
8211 register saves or frame pointer is needed.
8213 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8216 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8217 BLKmode for ternary, binary or unary expressions.
8220 * var-tracking.c (delete_vta_debug_insn): New inline function.
8221 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8222 insns from get_insns () to NULL instead of each bb separately.
8223 Use delete_vta_debug_insn. No longer static.
8224 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8225 delete_vta_debug_insns callers.
8226 * rtl.h (delete_vta_debug_insns): Declare.
8227 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8228 instead of variable_tracking_main.
8230 2018-01-03 Martin Sebor <msebor@redhat.com>
8232 PR tree-optimization/83603
8233 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8234 arguments past the endof the argument list in functions declared
8235 without a prototype.
8236 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8237 Avoid checking when arguments are null.
8239 2018-01-03 Martin Sebor <msebor@redhat.com>
8242 * doc/extend.texi (attribute const): Fix a typo.
8243 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8244 issuing -Wsuggest-attribute for void functions.
8246 2018-01-03 Martin Sebor <msebor@redhat.com>
8248 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8249 offset_int::from instead of wide_int::to_shwi.
8250 (maybe_diag_overlap): Remove assertion.
8251 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8252 * gimple-ssa-sprintf.c (format_directive): Same.
8253 (parse_directive): Same.
8254 (sprintf_dom_walker::compute_format_length): Same.
8255 (try_substitute_return_value): Same.
8257 2018-01-03 Jeff Law <law@redhat.com>
8260 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8261 non-constant residual for zero at runtime and avoid probing in
8262 that case. Reorganize code for trailing problem to mirror handling
8265 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8267 PR tree-optimization/83501
8268 * tree-ssa-strlen.c (get_string_cst): New.
8269 (handle_char_store): Call get_string_cst.
8271 2018-01-03 Martin Liska <mliska@suse.cz>
8273 PR tree-optimization/83593
8274 * tree-ssa-strlen.c: Include tree-cfg.h.
8275 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8276 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8277 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8279 (strlen_dom_walker::before_dom_children): Call
8280 gimple_purge_dead_eh_edges. Dump tranformation with details
8282 (strlen_dom_walker::before_dom_children): Update call by adding
8283 new argument cleanup_eh.
8284 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8286 2018-01-03 Martin Liska <mliska@suse.cz>
8289 * cif-code.def (VARIADIC_THUNK): New enum value.
8290 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8293 2018-01-03 Jan Beulich <jbeulich@suse.com>
8295 * sse.md (mov<mode>_internal): Tighten condition for when to use
8296 vmovdqu<ssescalarsize> for TI and OI modes.
8298 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8300 Update copyright years.
8302 2018-01-03 Martin Liska <mliska@suse.cz>
8305 * ipa-visibility.c (function_and_variable_visibility): Skip
8306 functions with noipa attribure.
8308 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8310 * gcc.c (process_command): Update copyright notice dates.
8311 * gcov-dump.c (print_version): Ditto.
8312 * gcov.c (print_version): Ditto.
8313 * gcov-tool.c (print_version): Ditto.
8314 * gengtype.c (create_file): Ditto.
8315 * doc/cpp.texi: Bump @copying's copyright year.
8316 * doc/cppinternals.texi: Ditto.
8317 * doc/gcc.texi: Ditto.
8318 * doc/gccint.texi: Ditto.
8319 * doc/gcov.texi: Ditto.
8320 * doc/install.texi: Ditto.
8321 * doc/invoke.texi: Ditto.
8323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8325 * vector-builder.h (vector_builder::m_full_nelts): Change from
8326 unsigned int to poly_uint64.
8327 (vector_builder::full_nelts): Update prototype accordingly.
8328 (vector_builder::new_vector): Likewise.
8329 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8330 (vector_builder::operator ==): Likewise.
8331 (vector_builder::finalize): Likewise.
8332 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8333 Take the number of elements as a poly_uint64 rather than an
8335 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8336 from unsigned int to poly_uint64.
8337 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8338 (vec_perm_indices::new_vector): Likewise.
8339 (vec_perm_indices::length): Likewise.
8340 (vec_perm_indices::nelts_per_input): Likewise.
8341 (vec_perm_indices::input_nelts): Likewise.
8342 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8343 number of elements per input as a poly_uint64 rather than an
8344 unsigned int. Use the original encoding for variable-length
8345 vectors, rather than clamping each individual element.
8346 For the second and subsequent elements in each pattern,
8347 clamp the step and base before clamping their sum.
8348 (vec_perm_indices::series_p): Handle polynomial element counts.
8349 (vec_perm_indices::all_in_range_p): Likewise.
8350 (vec_perm_indices_to_tree): Likewise.
8351 (vec_perm_indices_to_rtx): Likewise.
8352 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8353 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8354 (tree_vector_builder::new_binary_operation): Handle polynomial
8355 element counts. Return false if we need to know the number
8356 of elements at compile time.
8357 * fold-const.c (fold_vec_perm): Punt if the number of elements
8358 isn't known at compile time.
8360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8362 * vec-perm-indices.h (vec_perm_builder): Change element type
8363 from HOST_WIDE_INT to poly_int64.
8364 (vec_perm_indices::element_type): Update accordingly.
8365 (vec_perm_indices::clamp): Handle polynomial element_types.
8366 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8367 (vec_perm_indices::all_in_range_p): Likewise.
8368 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8370 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8371 polynomial vec_perm_indices element types.
8372 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8373 * fold-const.c (fold_vec_perm): Likewise.
8374 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8375 * tree-vect-generic.c (lower_vec_perm): Likewise.
8376 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8377 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8378 element type to HOST_WIDE_INT.
8380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8381 Alan Hayward <alan.hayward@arm.com>
8382 David Sherwood <david.sherwood@arm.com>
8384 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8385 rather than an int. Use plus_constant.
8386 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8387 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8390 Alan Hayward <alan.hayward@arm.com>
8391 David Sherwood <david.sherwood@arm.com>
8393 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8394 a HOST_WIDE_INT to a poly_int64.
8396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8397 Alan Hayward <alan.hayward@arm.com>
8398 David Sherwood <david.sherwood@arm.com>
8400 * calls.c (load_register_parameters): Cope with polynomial
8401 mode sizes. Require a constant size for BLKmode parameters
8402 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8403 forces a parameter to be padded at the lsb end in order to
8404 fill a complete number of words, require the parameter size
8405 to be ordered wrt UNITS_PER_WORD.
8407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8408 Alan Hayward <alan.hayward@arm.com>
8409 David Sherwood <david.sherwood@arm.com>
8411 * reload1.c (spill_stack_slot_width): Change element type
8412 from unsigned int to poly_uint64_pod.
8413 (alter_reg): Treat mode sizes as polynomial.
8415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8416 Alan Hayward <alan.hayward@arm.com>
8417 David Sherwood <david.sherwood@arm.com>
8419 * reload.c (complex_word_subreg_p): New function.
8420 (reload_inner_reg_of_subreg, push_reload): Use it.
8422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8423 Alan Hayward <alan.hayward@arm.com>
8424 David Sherwood <david.sherwood@arm.com>
8426 * lra-constraints.c (process_alt_operands): Reject matched
8427 operands whose sizes aren't ordered.
8428 (match_reload): Refer to this check here.
8430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8431 Alan Hayward <alan.hayward@arm.com>
8432 David Sherwood <david.sherwood@arm.com>
8434 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8435 that the mode size is in the set {1, 2, 4, 8, 16}.
8437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8438 Alan Hayward <alan.hayward@arm.com>
8439 David Sherwood <david.sherwood@arm.com>
8441 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8442 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8449 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8450 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8451 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8452 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8453 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8454 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8455 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8456 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8457 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8459 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8460 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8461 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8462 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8463 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8464 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8465 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8466 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8467 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8468 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8470 * expr.c (emit_move_resolve_push): Treat the input and result
8471 of PUSH_ROUNDING as a poly_int64.
8472 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8473 (emit_push_insn): Likewise.
8474 * lra-eliminations.c (mark_not_eliminable): Likewise.
8475 * recog.c (push_operand): Likewise.
8476 * reload1.c (elimination_effects): Likewise.
8477 * rtlanal.c (nonzero_bits1): Likewise.
8478 * calls.c (store_one_arg): Likewise. Require the padding to be
8479 known at compile time.
8481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8482 Alan Hayward <alan.hayward@arm.com>
8483 David Sherwood <david.sherwood@arm.com>
8485 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8486 Use plus_constant instead of gen_rtx_PLUS.
8488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8489 Alan Hayward <alan.hayward@arm.com>
8490 David Sherwood <david.sherwood@arm.com>
8492 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8496 Alan Hayward <alan.hayward@arm.com>
8497 David Sherwood <david.sherwood@arm.com>
8499 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8500 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8501 via stack temporaries. Treat the mode size as polynomial too.
8503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8504 Alan Hayward <alan.hayward@arm.com>
8505 David Sherwood <david.sherwood@arm.com>
8507 * expr.c (expand_expr_real_2): When handling conversions involving
8508 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8509 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8510 as a poly_uint64 too.
8512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8513 Alan Hayward <alan.hayward@arm.com>
8514 David Sherwood <david.sherwood@arm.com>
8516 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8519 Alan Hayward <alan.hayward@arm.com>
8520 David Sherwood <david.sherwood@arm.com>
8522 * combine.c (can_change_dest_mode): Handle polynomial
8523 REGMODE_NATURAL_SIZE.
8524 * expmed.c (store_bit_field_1): Likewise.
8525 * expr.c (store_constructor): Likewise.
8526 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8527 and polynomial REGMODE_NATURAL_SIZE.
8528 (gen_lowpart_common): Likewise.
8529 * reginfo.c (record_subregs_of_mode): Likewise.
8530 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8537 numbers of elements.
8539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8540 Alan Hayward <alan.hayward@arm.com>
8541 David Sherwood <david.sherwood@arm.com>
8543 * match.pd: Cope with polynomial numbers of vector elements.
8545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8546 Alan Hayward <alan.hayward@arm.com>
8547 David Sherwood <david.sherwood@arm.com>
8549 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8550 in a POINTER_PLUS_EXPR.
8552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8553 Alan Hayward <alan.hayward@arm.com>
8554 David Sherwood <david.sherwood@arm.com>
8556 * omp-simd-clone.c (simd_clone_subparts): New function.
8557 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8558 (ipa_simd_modify_function_body): Likewise.
8560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8561 Alan Hayward <alan.hayward@arm.com>
8562 David Sherwood <david.sherwood@arm.com>
8564 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8565 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8566 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8567 (expand_vector_condition, vector_element): Likewise.
8568 (subparts_gt): New function.
8569 (get_compute_type): Use subparts_gt.
8570 (count_type_subparts): Delete.
8571 (expand_vector_operations_1): Use subparts_gt instead of
8572 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8579 (vect_compile_time_alias): ...this new function. Do the calculation
8580 on poly_ints rather than trees.
8581 (vect_prune_runtime_alias_test_list): Update call accordingly.
8583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8584 Alan Hayward <alan.hayward@arm.com>
8585 David Sherwood <david.sherwood@arm.com>
8587 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8589 (vect_schedule_slp_instance): Likewise.
8591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8592 Alan Hayward <alan.hayward@arm.com>
8593 David Sherwood <david.sherwood@arm.com>
8595 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8596 constant and extern definitions for variable-length vectors.
8597 (vect_get_constant_vectors): Note that the number of units
8598 is known to be constant.
8600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8601 Alan Hayward <alan.hayward@arm.com>
8602 David Sherwood <david.sherwood@arm.com>
8604 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8605 of units as polynomial. Choose between WIDE and NARROW based
8608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8609 Alan Hayward <alan.hayward@arm.com>
8610 David Sherwood <david.sherwood@arm.com>
8612 * tree-vect-stmts.c (simd_clone_subparts): New function.
8613 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8616 Alan Hayward <alan.hayward@arm.com>
8617 David Sherwood <david.sherwood@arm.com>
8619 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8620 vectors as polynomial. Use build_index_vector for
8623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8624 Alan Hayward <alan.hayward@arm.com>
8625 David Sherwood <david.sherwood@arm.com>
8627 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8628 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8629 for variable-length vectors.
8630 (vectorizable_mask_load_store): Treat the number of units as
8631 polynomial, asserting that it is constant if the condition has
8632 already been enforced.
8633 (vectorizable_store, vectorizable_load): Likewise.
8635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8636 Alan Hayward <alan.hayward@arm.com>
8637 David Sherwood <david.sherwood@arm.com>
8639 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8640 of units as polynomial. Punt if we can't tell at compile time
8641 which vector contains the final result.
8643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8644 Alan Hayward <alan.hayward@arm.com>
8645 David Sherwood <david.sherwood@arm.com>
8647 * tree-vect-loop.c (vectorizable_induction): Treat the number
8648 of units as polynomial. Punt on SLP inductions. Use an integer
8649 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8650 cast of such a series for variable-length floating-point
8653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8654 Alan Hayward <alan.hayward@arm.com>
8655 David Sherwood <david.sherwood@arm.com>
8657 * tree.h (build_index_vector): Declare.
8658 * tree.c (build_index_vector): New function.
8659 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8660 of units as polynomial, forcibly converting it to a constant if
8661 vectorizable_reduction has already enforced the condition.
8662 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8663 to create a {1,2,3,...} vector.
8664 (vectorizable_reduction): Treat the number of units as polynomial.
8665 Choose vectype_in based on the largest scalar element size rather
8666 than the smallest number of units. Enforce the restrictions
8669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8670 Alan Hayward <alan.hayward@arm.com>
8671 David Sherwood <david.sherwood@arm.com>
8673 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8674 number of units as polynomial.
8676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8677 Alan Hayward <alan.hayward@arm.com>
8678 David Sherwood <david.sherwood@arm.com>
8680 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8681 * target.def (autovectorize_vector_sizes): Return the vector sizes
8682 by pointer, using vector_sizes rather than a bitmask.
8683 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8684 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8685 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8687 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8688 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8689 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8690 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8691 * omp-general.c (omp_max_vf): Likewise.
8692 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8693 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8694 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8695 * tree-vect-slp.c (vect_slp_bb): Likewise.
8696 * doc/tm.texi: Regenerate.
8697 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8699 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8700 the vector size as a poly_uint64 rather than an unsigned int.
8701 (current_vector_size): Change from an unsigned int to a poly_uint64.
8702 (get_vectype_for_scalar_type): Update accordingly.
8703 * tree.h (build_truth_vector_type): Take the size and number of
8704 units as a poly_uint64 rather than an unsigned int.
8705 (build_vector_type): Add a temporary overload that takes
8706 the number of units as a poly_uint64 rather than an unsigned int.
8707 * tree.c (make_vector_type): Likewise.
8708 (build_truth_vector_type): Take the number of units as a poly_uint64
8709 rather than an unsigned int.
8711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8712 Alan Hayward <alan.hayward@arm.com>
8713 David Sherwood <david.sherwood@arm.com>
8715 * target.def (get_mask_mode): Take the number of units and length
8716 as poly_uint64s rather than unsigned ints.
8717 * targhooks.h (default_get_mask_mode): Update accordingly.
8718 * targhooks.c (default_get_mask_mode): Likewise.
8719 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8720 * doc/tm.texi: Regenerate.
8722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8723 Alan Hayward <alan.hayward@arm.com>
8724 David Sherwood <david.sherwood@arm.com>
8726 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8727 * omp-general.c (omp_max_vf): Likewise.
8728 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8729 (expand_omp_simd): Handle polynomial safelen.
8730 * omp-low.c (omplow_simd_context): Add a default constructor.
8731 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8732 (lower_rec_simd_input_clauses): Update accordingly.
8733 (lower_rec_input_clauses): Likewise.
8735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8736 Alan Hayward <alan.hayward@arm.com>
8737 David Sherwood <david.sherwood@arm.com>
8739 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8740 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8741 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8742 (vect_analyze_slp_cost): Likewise.
8743 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8744 (vect_model_load_cost): Likewise.
8746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8747 Alan Hayward <alan.hayward@arm.com>
8748 David Sherwood <david.sherwood@arm.com>
8750 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8751 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8752 from an unsigned int * to a poly_uint64_pod *.
8753 (calculate_unrolling_factor): New function.
8754 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8757 Alan Hayward <alan.hayward@arm.com>
8758 David Sherwood <david.sherwood@arm.com>
8760 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8761 from an unsigned int to a poly_uint64.
8762 (_loop_vec_info::slp_unrolling_factor): Likewise.
8763 (_loop_vec_info::vectorization_factor): Change from an int
8765 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8766 (vect_get_num_vectors): New function.
8767 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8768 (vect_get_num_copies): Use vect_get_num_vectors.
8769 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8770 to an unsigned int *.
8771 (vect_analyze_data_refs): Change min_vf from an int * to a
8773 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8774 than an unsigned HOST_WIDE_INT.
8775 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8776 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8777 to an unsigned int *.
8778 (vect_analyze_data_ref_dependences): Likewise.
8779 (vect_compute_data_ref_alignment): Handle polynomial vf.
8780 (vect_enhance_data_refs_alignment): Likewise.
8781 (vect_prune_runtime_alias_test_list): Likewise.
8782 (vect_shift_permute_load_chain): Likewise.
8783 (vect_supportable_dr_alignment): Likewise.
8784 (dependence_distance_ge_vf): Take the vectorization factor as a
8785 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8786 (vect_analyze_data_refs): Change min_vf from an int * to a
8788 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8789 vfm1 as a poly_uint64 rather than an int. Make the same change
8790 for the returned bound_scalar.
8791 (vect_gen_vector_loop_niters): Handle polynomial vf.
8792 (vect_do_peeling): Likewise. Update call to
8793 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8794 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8796 * tree-vect-loop.c (vect_determine_vectorization_factor)
8797 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8798 (vect_get_known_peeling_cost): Likewise.
8799 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8800 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8801 (vect_transform_loop): Likewise. Use the lowest possible VF when
8802 updating the upper bounds of the loop.
8803 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8805 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8806 polynomial unroll factors.
8807 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8808 (vect_make_slp_decision): Likewise.
8809 (vect_supported_load_permutation_p): Likewise, and polynomial
8811 (vect_analyze_slp_cost): Handle polynomial vf.
8812 (vect_slp_analyze_node_operations): Likewise.
8813 (vect_slp_analyze_bb_1): Likewise.
8814 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8815 than an unsigned HOST_WIDE_INT.
8816 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8817 (vectorizable_load): Handle polynomial vf.
8818 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8820 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8823 Alan Hayward <alan.hayward@arm.com>
8824 David Sherwood <david.sherwood@arm.com>
8826 * match.pd: Handle bit operations involving three constants
8827 and try to fold one pair.
8829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8831 * tree-vect-loop-manip.c: Include gimple-fold.h.
8832 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8833 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8834 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8835 Add a path that uses a step of VF instead of 1, but disable it
8837 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8838 and niters_no_overflow parameters. Update calls to
8839 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8840 Create a new SSA name if the latter choses to use a ste other
8841 than zero, and return it via niters_vector_mult_vf_var.
8842 * tree-vect-loop.c (vect_transform_loop): Update calls to
8843 vect_do_peeling, vect_gen_vector_loop_niters and
8844 slpeel_make_loop_iterate_ntimes.
8845 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8846 (vect_gen_vector_loop_niters): Update declarations after above changes.
8848 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8850 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8851 128-bit round to integer instructions.
8852 (ceil<mode>2): Likewise.
8853 (btrunc<mode>2): Likewise.
8854 (round<mode>2): Likewise.
8856 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8858 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8859 unaligned VSX load/store on P8/P9.
8860 (expand_block_clear): Allow the use of unaligned VSX
8861 load/store on P8/P9.
8863 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8865 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8867 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8868 swap associated with both a load and a store.
8870 2018-01-02 Andrew Waterman <andrew@sifive.com>
8872 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8873 * config/riscv/riscv.md (clear_cache): Use it.
8875 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8877 * web.c: Remove out-of-date comment.
8879 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8881 * expr.c (fixup_args_size_notes): Check that any existing
8882 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8883 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8884 (emit_single_push_insn): ...here.
8886 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8888 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8889 (const_vector_encoded_nelts): New function.
8890 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8891 (const_vector_int_elt, const_vector_elt): Declare.
8892 * emit-rtl.c (const_vector_int_elt_1): New function.
8893 (const_vector_elt): Likewise.
8894 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8895 of CONST_VECTOR_ELT.
8897 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8899 * expr.c: Include rtx-vector-builder.h.
8900 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8901 directly on the tree encoding.
8902 (const_vector_from_tree): Likewise.
8903 * optabs.c: Include rtx-vector-builder.h.
8904 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8905 sequence of "u" values.
8906 * vec-perm-indices.c: Include rtx-vector-builder.h.
8907 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8908 directly on the vec_perm_indices encoding.
8910 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8912 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8913 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8914 * rtx-vector-builder.h: New file.
8915 * rtx-vector-builder.c: Likewise.
8916 * rtl.h (rtx_def::u2): Add a const_vector field.
8917 (CONST_VECTOR_NPATTERNS): New macro.
8918 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8919 (CONST_VECTOR_DUPLICATE_P): Likewise.
8920 (CONST_VECTOR_STEPPED_P): Likewise.
8921 (CONST_VECTOR_ENCODED_ELT): Likewise.
8922 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8923 (unwrap_const_vec_duplicate): Likewise.
8924 (const_vec_series_p): Check for a non-duplicated vector encoding.
8925 Say that the function only returns true for integer vectors.
8926 * emit-rtl.c: Include rtx-vector-builder.h.
8927 (gen_const_vec_duplicate_1): Delete.
8928 (gen_const_vector): Call gen_const_vec_duplicate instead of
8929 gen_const_vec_duplicate_1.
8930 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8931 (gen_const_vec_duplicate): Use rtx_vector_builder.
8932 (gen_const_vec_series): Likewise.
8933 (gen_rtx_CONST_VECTOR): Likewise.
8934 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8935 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8936 Build a new vector rather than modifying a CONST_VECTOR in-place.
8937 (handle_special_swappables): Update call accordingly.
8938 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8939 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8940 Build a new vector rather than modifying a CONST_VECTOR in-place.
8941 (handle_special_swappables): Update call accordingly.
8943 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8945 * simplify-rtx.c (simplify_const_binary_operation): Use
8946 CONST_VECTOR_ELT instead of XVECEXP.
8948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8950 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8951 the selector elements to be different from the data elements
8952 if the selector is a VECTOR_CST.
8953 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8954 ssizetype for the selector.
8956 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8958 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8959 before testing each element individually.
8960 * tree-vect-generic.c (lower_vec_perm): Likewise.
8962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8964 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8965 * selftest-run-tests.c (selftest::run_tests): Call it.
8966 * vector-builder.h (vector_builder::operator ==): New function.
8967 (vector_builder::operator !=): Likewise.
8968 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8969 (vec_perm_indices::all_from_input_p): New function.
8970 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8971 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8972 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8973 instead of reading the VECTOR_CST directly. Detect whether both
8974 vector inputs are the same before constructing the vec_perm_indices,
8975 and update the number of inputs argument accordingly. Use the
8976 utility functions added above. Only construct sel2 if we need to.
8978 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8980 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8981 the broadcast of the low byte.
8982 (expand_mult_highpart): Use an explicit encoding for the permutes.
8983 * optabs-query.c (can_mult_highpart_p): Likewise.
8984 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8985 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8986 (vectorizable_bswap): Likewise.
8987 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8988 explicit encoding for the power-of-2 permutes.
8989 (vect_permute_store_chain): Likewise.
8990 (vect_grouped_load_supported): Likewise.
8991 (vect_permute_load_chain): Likewise.
8993 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8995 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8996 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8997 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8998 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8999 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9000 (vect_gen_perm_mask_any): Likewise.
9002 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9004 * int-vector-builder.h: New file.
9005 * vec-perm-indices.h: Include int-vector-builder.h.
9006 (vec_perm_indices): Redefine as an int_vector_builder.
9007 (auto_vec_perm_indices): Delete.
9008 (vec_perm_builder): Redefine as a stand-alone class.
9009 (vec_perm_indices::vec_perm_indices): New function.
9010 (vec_perm_indices::clamp): Likewise.
9011 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9012 (vec_perm_indices::new_vector): New function.
9013 (vec_perm_indices::new_expanded_vector): Update for new
9014 vec_perm_indices class.
9015 (vec_perm_indices::rotate_inputs): New function.
9016 (vec_perm_indices::all_in_range_p): Operate directly on the
9017 encoded form, without computing elided elements.
9018 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9019 encoding. Update for new vec_perm_indices class.
9020 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9021 the given vec_perm_builder.
9022 (expand_vec_perm_var): Update vec_perm_builder constructor.
9023 (expand_mult_highpart): Use vec_perm_builder instead of
9024 auto_vec_perm_indices.
9025 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9026 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9027 or double series encoding as appropriate.
9028 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9029 vec_perm_indices instead of auto_vec_perm_indices.
9030 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9031 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9032 (vect_permute_store_chain): Likewise.
9033 (vect_grouped_load_supported): Likewise.
9034 (vect_permute_load_chain): Likewise.
9035 (vect_shift_permute_load_chain): Likewise.
9036 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9037 (vect_transform_slp_perm_load): Likewise.
9038 (vect_schedule_slp_instance): Likewise.
9039 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9040 (vectorizable_mask_load_store): Likewise.
9041 (vectorizable_bswap): Likewise.
9042 (vectorizable_store): Likewise.
9043 (vectorizable_load): Likewise.
9044 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9045 vec_perm_indices instead of auto_vec_perm_indices. Use
9046 tree_to_vec_perm_builder to read the vector from a tree.
9047 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9048 vec_perm_builder instead of a vec_perm_indices.
9049 (have_whole_vector_shift): Use vec_perm_builder and
9050 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9051 truncation to calc_vec_perm_mask_for_shift.
9052 (vect_create_epilog_for_reduction): Likewise.
9053 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9054 from auto_vec_perm_indices to vec_perm_indices.
9055 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9056 instead of changing individual elements.
9057 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9058 the vector in d.perm.
9059 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9060 from auto_vec_perm_indices to vec_perm_indices.
9061 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9062 instead of changing individual elements.
9063 (arm_vectorize_vec_perm_const): Use new_vector to install
9064 the vector in d.perm.
9065 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9066 Update vec_perm_builder constructor.
9067 (rs6000_expand_interleave): Likewise.
9068 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9069 (rs6000_expand_interleave): Likewise.
9071 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9073 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9074 to qimode could truncate the indices.
9075 * optabs.c (expand_vec_perm_var): Likewise.
9077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9079 * Makefile.in (OBJS): Add vec-perm-indices.o.
9080 * vec-perm-indices.h: New file.
9081 * vec-perm-indices.c: Likewise.
9082 * target.h (vec_perm_indices): Replace with a forward class
9084 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9085 * optabs.h: Include vec-perm-indices.h.
9086 (expand_vec_perm): Delete.
9087 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9088 (expand_vec_perm_const): Declare.
9089 * target.def (vec_perm_const_ok): Replace with...
9090 (vec_perm_const): ...this new hook.
9091 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9092 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9093 * doc/tm.texi: Regenerate.
9094 * optabs.def (vec_perm_const): Delete.
9095 * doc/md.texi (vec_perm_const): Likewise.
9096 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9097 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9098 expand_vec_perm for constant permutation vectors. Assert that
9099 the mode of variable permutation vectors is the integer equivalent
9100 of the mode that is being permuted.
9101 * optabs-query.h (selector_fits_mode_p): Declare.
9102 * optabs-query.c: Include vec-perm-indices.h.
9103 (selector_fits_mode_p): New function.
9104 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9105 is defined, instead of checking whether the vec_perm_const_optab
9106 exists. Use targetm.vectorize.vec_perm_const instead of
9107 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9108 fit in the vector mode before using a variable permute.
9109 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9110 vec_perm_indices instead of an rtx.
9111 (expand_vec_perm): Replace with...
9112 (expand_vec_perm_const): ...this new function. Take the selector
9113 as a vec_perm_indices rather than an rtx. Also take the mode of
9114 the selector. Update call to shift_amt_for_vec_perm_mask.
9115 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9116 Use vec_perm_indices::new_expanded_vector to expand the original
9117 selector into bytes. Check whether the indices fit in the vector
9118 mode before using a variable permute.
9119 (expand_vec_perm_var): Make global.
9120 (expand_mult_highpart): Use expand_vec_perm_const.
9121 * fold-const.c: Includes vec-perm-indices.h.
9122 * tree-ssa-forwprop.c: Likewise.
9123 * tree-vect-data-refs.c: Likewise.
9124 * tree-vect-generic.c: Likewise.
9125 * tree-vect-loop.c: Likewise.
9126 * tree-vect-slp.c: Likewise.
9127 * tree-vect-stmts.c: Likewise.
9128 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9130 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9131 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9132 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9133 (aarch64_vectorize_vec_perm_const): ...this new function.
9134 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9135 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9136 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9137 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9138 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9139 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9140 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9142 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9143 check for NEON modes.
9144 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9145 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9146 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9147 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9149 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9150 the old VEC_PERM_CONST conditions.
9151 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9152 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9153 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9154 (ia64_vectorize_vec_perm_const_ok): Merge into...
9155 (ia64_vectorize_vec_perm_const): ...this new function.
9156 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9157 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9158 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9159 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9160 * config/mips/mips.c (mips_expand_vec_perm_const)
9161 (mips_vectorize_vec_perm_const_ok): Merge into...
9162 (mips_vectorize_vec_perm_const): ...this new function.
9163 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9164 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9165 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9166 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9167 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9168 (rs6000_expand_vec_perm_const): Delete.
9169 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9171 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9172 (altivec_expand_vec_perm_const_le): Take each operand individually.
9173 Operate on constant selectors rather than rtxes.
9174 (altivec_expand_vec_perm_const): Likewise. Update call to
9175 altivec_expand_vec_perm_const_le.
9176 (rs6000_expand_vec_perm_const): Delete.
9177 (rs6000_vectorize_vec_perm_const_ok): Delete.
9178 (rs6000_vectorize_vec_perm_const): New function.
9179 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9180 an element count and rtx array.
9181 (rs6000_expand_extract_even): Update call accordingly.
9182 (rs6000_expand_interleave): Likewise.
9183 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9184 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9185 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9186 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9187 (rs6000_expand_vec_perm_const): Delete.
9188 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9189 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9190 (altivec_expand_vec_perm_const_le): Take each operand individually.
9191 Operate on constant selectors rather than rtxes.
9192 (altivec_expand_vec_perm_const): Likewise. Update call to
9193 altivec_expand_vec_perm_const_le.
9194 (rs6000_expand_vec_perm_const): Delete.
9195 (rs6000_vectorize_vec_perm_const_ok): Delete.
9196 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9197 reference to the SPE evmerge intructions.
9198 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9199 an element count and rtx array.
9200 (rs6000_expand_extract_even): Update call accordingly.
9201 (rs6000_expand_interleave): Likewise.
9202 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9203 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9205 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9209 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9210 vector mode and that that mode matches the mode of the data
9212 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9213 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9214 directly using expand_vec_perm_1 when forcing selectors into
9216 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9218 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9220 * optabs-query.h (can_vec_perm_p): Delete.
9221 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9222 * optabs-query.c (can_vec_perm_p): Split into...
9223 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9224 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9225 particular selector is valid.
9226 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9227 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9228 (vect_grouped_load_supported): Likewise.
9229 (vect_shift_permute_load_chain): Likewise.
9230 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9231 (vect_transform_slp_perm_load): Likewise.
9232 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9233 (vectorizable_bswap): Likewise.
9234 (vect_gen_perm_mask_checked): Likewise.
9235 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9236 implementations of variable permutation vectors into account
9237 when deciding which selector to use.
9238 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9239 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9240 with a false third argument.
9241 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9242 to test whether the constant selector is valid and can_vec_perm_var_p
9243 to test whether a variable selector is valid.
9245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9247 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9248 * optabs-query.c (can_vec_perm_p): Likewise.
9249 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9250 instead of vec_perm_indices.
9251 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9252 (vect_gen_perm_mask_checked): Likewise,
9253 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9254 (vect_gen_perm_mask_checked): Likewise,
9256 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9258 * optabs-query.h (qimode_for_vec_perm): Declare.
9259 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9260 (qimode_for_vec_perm): ...this new function.
9261 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9263 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9265 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9266 does not have a conditional at the top.
9268 2018-01-02 Richard Biener <rguenther@suse.de>
9270 * ipa-inline.c (big_speedup_p): Fix expression.
9272 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9275 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9278 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9282 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9283 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9284 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9285 cond_taken_branch_cost 3->4.
9287 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9289 PR tree-optimization/83581
9290 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9291 TODO_cleanup_cfg if any changes have been made.
9294 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9295 convert_modes if target mode has the right side, but different mode
9299 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9300 last argument when extracting from CONCAT. If either from_real or
9301 from_imag is NULL, use expansion through memory. If result is not
9302 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9303 the parts directly to inner mode, if even that fails, use expansion
9307 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9308 check for bswap in mode rather than HImode and use that in expand_unop
9311 Copyright (C) 2018 Free Software Foundation, Inc.
9313 Copying and distribu