1 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/84982
4 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5 by flipping the least significant bit rather than all bits from
6 bitpos to bitpos + bitsize - 1.
8 2018-03-21 Nathan Sidwell <nathan@acm.org>
10 * doc/extend.texi (Deprecated Features): Remove mention of
11 long-deleted deprecations.
13 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
17 * configure: Regenerate.
19 2018-03-21 Tom de Vries <tom@codesourcery.com>
21 PR tree-optimization/83126
22 * tree-parloops.c (num_phis): New function.
23 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
25 2018-03-21 Nathan Sidwell <nathan@acm.org>
27 * doc/extend.texi (Deprecated Features): Update deprecated flags,
28 mention anon-struct/union members and trailing attributes.
30 2018-03-21 Bin Cheng <bin.cheng@arm.com>
32 PR tree-optimization/84969
33 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
34 builtin memset partitions if they set different rhs values.
36 2018-03-21 Jakub Jelinek <jakub@redhat.com>
38 PR rtl-optimization/84989
39 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
40 VEC_DUPLICATE with scalar result mode.
42 2018-03-21 Martin Liska <mliska@suse.cz>
45 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
46 not intended return statement.
48 2018-03-21 Martin Liska <mliska@suse.cz>
51 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
52 (chkp_find_bound_slots_1): Limit number of iterations.
54 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
57 * Minor grammar fixes for x86 options.
59 2018-03-20 Jakub Jelinek <jakub@redhat.com>
62 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
63 holding REG_CFA_RESTORE notes, instead turn them into a USE.
65 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
68 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
69 (altivec_lvx_<mode>_1op): Likewise.
70 (altivec_stvx_<mode>_2op): Likewise.
71 (altivec_stvx_<mode>_1op): Likewise.
72 (altivec_lvx_<VM2:mode>): New define_expand.
73 (altivec_stvx_<VM2:mode>): Likewise.
74 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
75 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
76 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
77 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
78 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
79 (rs6000_gen_lvx): Likewise.
80 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
81 (altivec_expand_stv_builtin): Likewise.
82 (altivec_expand_builtin): Likewise.
83 * config/rs6000/vector.md: Likewise.
85 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
88 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
91 2018-03-20 Richard Biener <rguenther@suse.de>
94 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
95 sign-conversions as zero, fall back to standard scalar_stmt
98 2018-03-20 Martin Liska <mliska@suse.cz>
101 * predict.c (rebuild_frequencies): Handle case when we have
102 PROFILE_ABSENT, but flag_guess_branch_prob is false.
104 2018-03-20 Jakub Jelinek <jakub@redhat.com>
107 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
108 flag_section_anchors.
109 * varasm.c (use_blocks_for_decl_p): Remove hack for
113 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
115 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
116 be created, use lowpart_subreg of operands[0] rather than operands[0]
118 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
119 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
120 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
121 and n constraint instead of aarch64_shift_imm_di and Usd.
122 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
123 (*aarch64_<optab>_reg_minus<mode>3): ... this.
125 2018-03-20 Sudakshina Das <sudi.das@arm.com>
128 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
129 to favor GPR over NEON registers.
130 (<shift>di3_neon): Likewise.
132 2018-03-20 Tom de Vries <tom@codesourcery.com>
135 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
136 (nvptx_process_pars): Emit bar.sync asap and alap.
138 2018-03-20 Tom de Vries <tom@codesourcery.com>
141 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
142 seen_label if seen_label is already set.
144 2018-03-20 Jakub Jelinek <jakub@redhat.com>
147 * config/i386/i386.c (fold_builtin_cpu): For features above 31
148 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
149 Use 1U instead of 1. Formatting fixes.
152 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
153 instead of TREE_TYPE (s1) for the return value.
155 2018-03-19 Jakub Jelinek <jakub@redhat.com>
157 PR tree-optimization/84946
158 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
159 bitsize + bitsize in poly_uint64 rather than poly_int64.
162 * dwarf2asm.c: Include fold-const.c.
163 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
164 of decl rather than decl itself.
166 PR rtl-optimization/84643
167 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
169 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
172 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
173 calling assemble_variable.
175 2018-03-19 Sudakshina Das <sudi.das@arm.com>
178 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
179 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
181 2018-03-19 Jim Wilson <jimw@sifive.com>
184 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
185 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
186 (riscv_first_stack_step): Likewise.
187 (riscv_option_override): Use STACK_BOUNDARY instead of
189 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
191 (BIGGEST_ALIGNMENT): Set to 128.
192 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
193 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
196 2018-03-19 Richard Biener <rguenther@suse.de>
198 PR tree-optimization/84933
199 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
200 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
202 2018-03-19 Richard Biener <rguenther@suse.de>
204 PR tree-optimization/84859
205 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
206 (cond_if_else_store_replacement): Perform sinking operation on
207 single-store BBs regardless of MAX_STORES_TO_SINK setting.
208 Generalize what a BB with a single eligible store is.
210 2018-03-19 Richard Biener <rguenther@suse.de>
212 PR tree-optimization/84929
213 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
214 chrec_is_positive against non-chrec arg.
216 2018-03-19 Tamar Christina <tamar.christina@arm.com>
219 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
221 2018-03-18 Martin Liska <mliska@suse.cz>
223 PR rtl-optimization/84635
224 * regrename.c (build_def_use): Use matches_mode only when
227 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
229 PR tree-optimization/84913
230 * tree-vect-loop.c (vectorizable_reduction): Don't try to
231 vectorize chains of COND_EXPRs.
233 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
235 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
237 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
239 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
241 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
243 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
245 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
246 Kito Cheng <kito.cheng@gmail.com>
248 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
249 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
250 (nds32_adjust_reg_alloc_order): New function.
251 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
253 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
255 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
256 nds32_print_operand, nds32_print_operand_address): Use
257 HOST_WIDE_INT_PRINT_DEC instead.
259 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
261 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
263 2018-03-17 Jakub Jelinek <jakub@redhat.com>
266 * config/i386/i386.c (initial_ix86_tune_features,
267 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
269 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
270 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
271 rather than 1u << ix86_tune. Formatting fix.
272 (ix86_option_override_internal): Change ix86_arch_mask from
273 unsigned int to unsigned HOST_WIDE_INT, initialize to
274 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
275 (ix86_function_specific_restore): Likewise.
277 2018-03-16 Jakub Jelinek <jakub@redhat.com>
280 * postreload.c (reload_combine_recognize_pattern): Perform
281 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
282 truncate_int_for_mode the result for the destination's mode.
285 * hsa-gen.c (mem_type_for_type): Fix comment typo.
286 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
288 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
291 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
294 * lra-assigns.c (lra_split_hard_reg_for): Don't use
295 regno_allocno_class_array and sorted_pseudos.
296 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
297 insns where regno is used.
299 2018-03-16 Martin Liska <mliska@suse.cz>
302 * multiple_target.c (create_dispatcher_calls): Redirect
303 reference in the symbol table.
305 2018-03-16 Martin Liska <mliska@suse.cz>
308 * multiple_target.c (create_dispatcher_calls): Redirect also
311 2018-03-16 Jakub Jelinek <jakub@redhat.com>
315 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
316 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
317 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
319 2018-03-16 Julia Koval <julia.koval@intel.com>
321 * doc/invoke.texi (Skylake Server): Add CLWB.
322 Cannonlake): Remove CLWB.
324 2018-03-16 Jakub Jelinek <jakub@redhat.com>
326 PR tree-optimization/84841
327 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
329 (FLOAT_ONE_CONST_TYPE): Define.
330 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
331 (sort_by_operand_rank): Put entries with higher constant_type last
332 rather than first to match comments.
334 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
336 * config/nios2/nios2.md (movsi_internal): Fix thinko in
339 2018-03-15 Jakub Jelinek <jakub@redhat.com>
342 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
343 check and use address of target always.
345 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
348 * config/i386/i386.c (indirect_thunk_needed): Update comments.
349 (indirect_thunk_bnd_needed): Likewise.
350 (indirect_thunks_used): Likewise.
351 (indirect_thunks_bnd_used): Likewise.
352 (indirect_return_needed): New.
353 (indirect_return_bnd_needed): Likewise.
354 (output_indirect_thunk_function): Add a bool argument for
356 (output_indirect_thunk_function): Don't generate alias for
357 function return thunk.
358 (ix86_code_end): Call output_indirect_thunk_function to generate
359 function return thunks.
360 (ix86_output_function_return): Set indirect_return_bnd_needed
361 and indirect_return_needed instead of indirect_thunk_bnd_needed
362 and indirect_thunk_needed.
364 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
366 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
367 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
368 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
370 2018-03-15 David Malcolm <dmalcolm@redhat.com>
371 Paul Hua <paul.hua.gm@gmail.com>
374 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
376 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
378 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
379 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
382 2018-03-15 Tamar Christina <tamar.christina@arm.com>
385 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
386 instead of GET_MODE_SIZE when comparing Units.
388 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
391 * varasm.c (hash_section): Return an unchangeble hash value
392 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
393 Return !aarch64_can_use_per_function_literal_pools_p ().
395 2018-03-15 Jakub Jelinek <jakub@redhat.com>
398 * optabs.c (emit_conditional_move): Pass address of cmode's copy
399 rather than address of cmode as last argument to prepare_cmp_insn.
401 2018-03-15 Julia Koval <julia.koval@intel.com>
403 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
404 F_AVX512VNNI, F_AVX512BITALG): New.
406 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
409 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
410 insn for floating-point loads and stores.
412 2018-03-14 Carl Love <cel@us.ibm.com>
414 * config/rs6000/rs6000-c.c: Add macro definitions for
415 ALTIVEC_BUILTIN_VEC_PERMXOR.
416 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
417 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
418 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
419 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
421 * config/doc/extend.texi: Add prototypes for vec_permxor.
423 2018-03-14 David Malcolm <dmalcolm@redhat.com>
426 * diagnostic-show-locus.c (class layout_point): Convert m_line
427 from int to linenum_type.
428 (line_span::comparator): Use linenum "compare" function when
429 comparing line numbers.
430 (test_line_span): New function.
431 (layout_range::contains_point): Convert param "row" from int to
433 (layout_range::intersects_line_p): Likewise.
434 (layout::will_show_line_p): Likewise.
435 (layout::print_source_line): Likewise.
436 (layout::should_print_annotation_line_p): Likewise.
437 (layout::print_annotation_line): Likewise.
438 (layout::print_leading_fixits): Likewise.
439 (layout::annotation_line_showed_range_p): Likewise.
440 (struct line_corrections): Likewise for field m_row.
441 (line_corrections::line_corrections): Likewise for param "row".
442 (layout::print_trailing_fixits): Likewise.
443 (layout::get_state_at_point): Likewise.
444 (layout::get_x_bound_for_row): Likewise.
445 (layout::print_line): Likewise.
446 (diagnostic_show_locus): Likewise for locals "last_line" and
448 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
449 * input.c (selftest::test_linenum_comparisons): New function.
450 (selftest::input_c_tests): Call it.
451 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
452 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
453 * selftest.h (ASSERT_GT): New macro.
454 (ASSERT_GT_AT): New macro.
455 (ASSERT_LT): New macro.
456 (ASSERT_LT_AT): New macro.
458 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
460 PR rtl-optimization/84780
461 * combine.c (distribute_links): Don't make a link based on pc_rtx.
463 2018-03-14 Martin Liska <mliska@suse.cz>
465 * tree.c (record_node_allocation_statistics): Use
467 (get_stats_node_kind): New function extracted from
468 record_node_allocation_statistics.
469 (free_node): Use get_stats_node_kind.
471 2018-03-14 Richard Biener <rguenther@suse.de>
473 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
474 that the value-set of ANTIC_IN doesn't grow.
477 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
479 (BB_VISITED_WITH_VISITED_SUCCS): New define.
480 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
482 2018-03-14 Julia Koval <julia.koval@intel.com>
484 * config.gcc (icelake-client, icelake-server): New.
486 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
487 (initial_ix86_arch_features): Ditto.
488 (PTA_SKYLAKE): Add SGX.
489 (PTA_ICELAKE): Remove.
490 (PTA_ICELAKE_CLIENT): New.
491 (PTA_ICELAKE_SERVER): New.
492 (ix86_option_override_internal): Split up icelake on icelake client and
494 (get_builtin_code_for_version): Ditto.
495 (fold_builtin_cpu): Ditto.
496 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
497 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
498 * config/i386/i386.h (processor_type): Ditto.
499 * doc/invoke.texi: Ditto.
501 2018-03-14 Jakub Jelinek <jakub@redhat.com>
504 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
505 INTEGER_CST offset, add it together with bitpos / 8 and
506 sign extend based on POINTER_SIZE.
510 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
513 * config/i386/constraints.md (Yc): New register constraint.
514 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
515 Use Yc constraint for alternative 2 of operand 0. Remove
516 preferred_for_speed attribute.
518 2018-03-14 Richard Biener <rguenther@suse.de>
520 PR tree-optimization/84830
521 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
522 with the old one to avoid oscillations.
524 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
527 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
529 (assign_by_spills): Return a flag of reload assignment failure.
530 Do not process the reload assignment failures. Do not spill other
531 reload pseudos if they has the same reg class. Update n if
533 (lra_assign): Add a return arg. Set up from the result of
534 assign_by_spills call.
535 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
536 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
537 usage_insns if it is not NULL.
538 (spill_hard_reg_in_range): New function.
539 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
540 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
542 (lra_assign): Change prototype.
543 * lra.c (lra): Add code to deal with fails by splitting hard reg
546 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
548 * config/riscv/riscv.opt (mrelax): New option.
549 * config/riscv/riscv.c (riscv_file_start): Emit ".option
550 "norelax" when riscv_mrelax is disabled.
551 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
553 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
556 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
557 reassociation for int modes.
559 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
561 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
562 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
564 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
565 * config/aarch64/aarch64-sve.md
566 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
567 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
568 (*extend<mode><Vwide>2): Rename to...
569 (aarch64_sve_extend<mode><Vwide>2): ...this.
570 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
571 renaming the old pattern to...
572 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
574 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
575 define_expand, renaming the old pattern to...
576 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
577 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
578 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
579 account when deciding which SVE instruction the optab should use.
580 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
582 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
584 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
585 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
586 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
587 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
588 (tlsdesc_small_<mode>): Turn a define_expand and use
589 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
590 (tlsdesc_small_advsimd_<mode>): ...this.
591 (tlsdesc_small_sve_<mode>): New pattern.
593 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
595 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
596 (UNSPEC_UMUL_HIGHPART): New constants.
597 (MUL_HIGHPART): New int iteraor.
598 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
599 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
601 (*<su>mul<mode>3_highpart): New define_insn.
603 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
606 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
609 2018-03-13 Martin Liska <mliska@suse.cz>
612 * (sem_item_optimizer::sem_item_optimizer): Initialize new
614 (sem_item_optimizer::~sem_item_optimizer): Release it.
615 (sem_item_optimizer::merge_classes): Register variable aliases.
616 (sem_item_optimizer::fixup_pt_set): New function.
617 (sem_item_optimizer::fixup_points_to_sets): Likewise.
618 * ipa-icf.h: Declare new variables and functions.
620 2018-03-13 Jakub Jelinek <jakub@redhat.com>
623 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
624 integer_pow2p@2 and test integer_pow2p in condition.
625 (A < 0 ? C : 0): Similarly for @1.
628 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
629 characters starting at p contain '\0' character, don't look beyond
633 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
634 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
637 * reg-stack.c (change_stack): Change update_end var from int to
638 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
639 also call set_block_for_insn on the newly added insns and rescan.
642 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
646 * tree.c (stabilize_reference_1): Return save_expr (e) for
647 STATEMENT_LIST even if it doesn't have side-effects.
649 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
651 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
653 2018-03-12 Renlin Li <renlin.li@arm.com>
655 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
656 aarch64_output_scalar_simd_mov_immediate.
658 2018-03-12 Martin Sebor <msebor@redhat.com>
660 PR tree-optimization/83456
661 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
662 for perfectly overlapping calls to memcpy.
663 (gimple_fold_builtin_memory_chk): Same.
664 (gimple_fold_builtin_strcpy): Handle no-warning.
665 (gimple_fold_builtin_stxcpy_chk): Same.
666 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
668 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
670 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
671 parameter. Use it for SFmode.
672 (rs6000_function_arg_advance_1): Adjust.
673 (rs6000_function_arg): Adjust.
674 (rs6000_gimplify_va_arg): Pass false for that new parameter.
676 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
678 PR rtl-optimization/84169
679 PR rtl-optimization/84780
680 * combine.c (can_combine_p): Check for a 2-insn combination whether
681 the destination register is used between the two insns, too.
683 2018-03-12 Richard Biener <rguenther@suse.de>
685 PR tree-optimization/84803
686 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
687 for refs DR analysis didn't process.
689 2018-03-12 Richard Biener <rguenther@suse.de>
691 PR tree-optimization/84777
692 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
693 force-vectorize loops ignore whether we are optimizing for size.
695 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
697 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
698 (TARGET_MD_ASM_ADJUST): Define.
700 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
701 Kito Cheng <kito.cheng@gmail.com>
702 Chung-Ju Wu <jasonwucj@gmail.com>
704 * config/nds32/nds32.c (nds32_compute_stack_frame,
705 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
706 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
707 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
708 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
709 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
710 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
711 * config/nds32/nds32.md (prologue, epilogue): Use macro
712 NDS32_V3PUSH_AVAILABLE_P to do checking.
714 2018-03-11 Jakub Jelinek <jakub@redhat.com>
717 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
718 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
719 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
720 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
721 addition of most attributes on !orig_type_die or the attribute not
722 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
724 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
725 Chung-Ju Wu <jasonwucj@gmail.com>
727 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
729 * config/nds32/nds32.opt (mvh): New option.
731 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
732 Chung-Ju Wu <jasonwucj@gmail.com>
734 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
736 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
737 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
740 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
741 Chung-Ju Wu <jasonwucj@gmail.com>
743 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
745 * config/nds32/nds32-multiple.md (strlensi): New pattern.
746 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
748 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
749 Kito Cheng <kito.cheng@gmail.com>
750 Chung-Ju Wu <jasonwucj@gmail.com>
752 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
753 UNSPEC_FFMISM and UNSPEC_FLMISM.
754 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
755 for ffb, ffmism and flmism.
756 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
757 (unspec_ffmism): Ditto.
758 (unspec_flmism): Ditto.
759 (nds32_expand_builtin_impl): Check if string extension is available.
760 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
761 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
763 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
766 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
769 * lra-assigns.c (assign_by_spills): Return a flag of reload
770 assignment failure. Do not process the reload assignment
771 failures. Do not spill other reload pseudos if they has the same
773 (lra_assign): Add a return arg. Set up from the result of
774 assign_by_spills call.
775 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
776 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
777 usage_insns if it is not NULL.
778 (spill_hard_reg_in_range): New function.
779 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
780 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
782 (lra_assign): Change prototype.
783 * lra.c (lra): Add code to deal with fails by splitting hard reg
786 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
789 * config/i386/i386.opt: Replace Enforcment with Enforcement.
791 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
794 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
795 (dw_val_node): Add val_symbolic_view.
796 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
797 (symview_upper_bound): New.
798 (new_line_info_table): Initialize symviews_since_reset.
799 (dwarf2out_source_line): Count symviews_since_reset and set
801 (dw_val_equal_p): Handle symview.
802 (add_AT_symview): New.
803 (print_dw_val): Handle symview.
804 (attr_checksum, attr_checksum_ordered): Likewise.
805 (same_dw_val_p, size_of_die): Likewise.
806 (value_format, output_die): Likewise.
807 (add_high_low_attributes): Use add_AT_symview for entry_view.
808 (dwarf2out_finish): Reset symview_upper_bound, clear
811 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
814 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
815 Add strict argument and use it.
816 (rs6000_split_multireg_move): Update for new strict argument.
817 (mem_operand_gpr): Disallow all non-offsettable addresses.
818 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
820 2018-03-09 Jakub Jelinek <jakub@redhat.com>
823 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
824 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
825 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
828 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
829 decl, use remap_type if we want to use the type.
831 2018-03-09 Martin Sebor <msebor@redhat.com>
833 PR tree-optimization/84526
834 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
836 (builtin_access::generic_overlap): Be prepared to handle non-array
839 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
841 PR rtl-optimization/84682
842 * lra-constraints.c (process_address_1): Check is_address flag
843 for address constraints.
844 (process_alt_operands): Likewise.
845 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
846 preprocess_constraints.
847 * recog.h (preprocess_constraints): Add oploc parameter.
849 * recog.c (preprocess_constraints): Test address_operand for
850 CT_ADDRESS constraints.
852 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
855 * lra-assigns.c (assign_by_spills): Return a flag of reload
856 assignment failure. Do not process the reload assignment
857 failures. Do not spill other reload pseudos if they has the same
859 (lra_assign): Add a return arg. Set up from the result of
860 assign_by_spills call.
861 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
862 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
863 usage_insns if it is not NULL.
864 (spill_hard_reg_in_range): New function.
865 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
866 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
868 (lra_assign): Change prototype.
869 * lra.c (lra): Add code to deal with fails by splitting hard reg
872 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
875 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
876 Accept complain bool parameter. Only emit errors if it is true.
877 (arm_parse_cpu_option_name): Likewise.
878 (arm_target_thumb_only): Adjust callers of the above.
879 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
880 prototype to take a default true bool parameter.
881 (arm_parse_arch_option_name): Likewise.
883 2018-03-09 David Malcolm <dmalcolm@redhat.com>
884 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
888 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
889 * configure: Regenerate.
890 * configure.ac ("linker --version-script option"): New.
891 ("linker soname option"): New.
893 2018-03-09 Richard Biener <rguenther@suse.de>
895 PR tree-optimization/84775
896 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
897 immediate uses of predicate stmts and mark them modified.
900 PR tree-optimization/84178
901 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
903 (version_loop_for_if_conversion): Delay update_ssa call.
904 (tree_if_conversion): Delay update_ssa until after predicate
907 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
910 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
911 when the function accesses prior frames.
913 2018-03-08 Jakub Jelinek <jakub@redhat.com>
916 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
917 gen_llsym, otherwise call maybe_gen_llsym.
920 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
921 has ',' character inside of it.
923 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
926 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
927 as clobbering CC_REGNUM.
929 2018-03-08 Richard Biener <rguenther@suse.de>
932 * tree-scalar-evolution.c: Include tree-into-ssa.h.
933 (follow_copies_to_constant): Do not follow SSA names registered
936 2018-03-08 Richard Biener <rguenther@suse.de>
938 PR tree-optimization/84178
939 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
941 (version_loop_for_if_conversion): Delay update_ssa call.
942 (tree_if_conversion): Delay update_ssa until after predicate
945 2018-03-08 David Malcolm <dmalcolm@redhat.com>
947 PR tree-optimization/84178
948 * tree-if-conv.c (release_bb_predicate): Remove the
949 the assertion that the stmts have NULL use_ops.
950 Discard the statements, asserting that they haven't
951 yet been added to a BB.
953 2018-03-08 Richard Biener <rguenther@suse.de>
955 PR tree-optimization/84746
956 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
957 (phi_translate): Pass in destination ANTIC_OUT set.
958 (phi_translate_1): Likewise. For a simplified result lookup
959 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
960 (phi_translate_set): Adjust.
961 (do_pre_regular_insertion): Likewise.
962 (do_pre_partial_partial_insertion): Likewise.
964 2018-03-08 Martin Liska <mliska@suse.cz>
966 PR gcov-profile/84735
967 * doc/gcov.texi: Document usage of profile files.
968 * gcov-io.h: Document changes in the format.
970 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
974 * dwarf2out.c (struct dw_line_info_table): Update comments for
976 (FORCE_RESET_NEXT_VIEW): New.
977 (FORCE_RESETTING_VIEW_P): New.
978 (RESETTING_VIEW_P): Check for -1 too.
979 (ZERO_VIEW_P): Likewise.
980 (new_line_info_table): Force-reset next view.
981 (dwarf2out_begin_function): Likewise.
982 (dwarf2out_source_line): Simplify zero_view_p initialization.
983 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
984 view directly. Omit view when omitting .loc at line 0.
986 2018-03-08 Jakub Jelinek <jakub@redhat.com>
988 PR tree-optimization/84740
989 * tree-switch-conversion.c (process_switch): Call build_constructors
990 only if info.phi_count is non-zero.
992 PR tree-optimization/84739
993 * tree-tailcall.c (find_tail_calls): Check call arguments against
994 DECL_ARGUMENTS (current_function_decl) rather than
995 DECL_ARGUMENTS (func) when checking for tail recursion.
997 2018-03-07 Jakub Jelinek <jakub@redhat.com>
999 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1000 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1001 Volker Reichelt's entry and add entries for people that perform
1002 GCC fuzzy testing and report numerous bugs.
1004 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1007 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1008 readonly data in sdata, if that is disabled.
1009 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1010 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1011 -mreadonly-in-sdata option.
1013 2018-03-07 Martin Sebor <msebor@redhat.com>
1015 PR tree-optimization/84468
1016 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1017 basic block when looking for nul assignment.
1019 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1022 * except.h (output_function_exception_table): Adjust prototype.
1023 * except.c (output_function_exception_table): Remove FNNAME parameter
1024 and add SECTION parameter. Ouput one part of the table at a time.
1025 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1026 the first part of the exception table and emit unwind directives.
1027 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1028 (i386_pe_seh_cold_init): Likewise.
1029 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1030 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1031 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1032 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1033 * config/i386/winnt.c: Include except.h.
1034 (struct seh_frame_state): Add reg_offset, after_prologue and
1035 in_cold_section fields.
1036 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1037 (i386_pe_seh_cold_init): New function.
1038 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1039 to seh->in_cold_section.
1040 (seh_emit_push): Record the offset of the push.
1041 (seh_emit_save): Record the offet of the save.
1042 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1043 Test seh->after_prologue to disregard the epilogue.
1044 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1045 (i386_pe_end_cold_function): New function.
1047 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1050 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1051 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1054 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1056 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1057 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1060 * multiple_target.c: Include tree-inline.h and intl.h.
1061 (expand_target_clones): Diagnose and fail if node->definition and
1062 !tree_versionable_function_p (node->decl).
1064 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1066 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1068 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1069 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1070 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1072 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1075 * combine.c (try_combine): Use reg_or_subregno instead of handling
1076 just paradoxical SUBREGs and REGs.
1078 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1080 * config/arc/arc.c (arc_finalize_pic): Remove function.
1081 (arc_must_save_register): We use single base PIC register, remove
1082 checks to save/restore the PIC register.
1083 (arc_expand_prologue): Likewise.
1084 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1086 (arc_verify_short): Likewise.
1087 (arc_attr_type): Likewise.
1088 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1089 (walk_stores): Likewise.
1090 (arc_address_cost): Make it static.
1091 (arc_verify_short): Likewise.
1092 (branch_dest): Likewise.
1093 (arc_attr_type): Likewise.
1094 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1095 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1096 (arc_final_prescan_insn): Remove inserting the nops due to
1097 hardware hazards. It is done in reorg step.
1098 (insn_length_variant_t): Remove.
1099 (insn_length_parameters_t): Likewise.
1100 (arc_insn_length_parameters): Likewise.
1101 (arc_get_insn_variants): Likewise.
1102 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1104 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1107 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1110 PR tree-optimization/84687
1111 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1113 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1115 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1117 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1118 Rename to ppc_speculation_barrier.
1119 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1120 __builtin_ppc_speculation_barrier.
1122 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1125 * combine.c (combine_simplify_rtx): Don't try to simplify if
1126 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1129 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1131 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1132 to 32 bytes when compiling for POWER9.
1134 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1137 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1138 regparm >= 3 with no arg reg available also for calls with
1139 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1142 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1144 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1146 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1149 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1151 2018-03-05 Richard Biener <rguenther@suse.de>
1153 PR tree-optimization/84486
1154 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1155 When inserting a __builtin_assume_aligned call set the LHS
1156 SSA name alignment info accordingly.
1158 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1160 PR tree-optimization/84114
1161 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1162 Avoid reassociation of FLOAT_MODE addition.
1164 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1166 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1167 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1168 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1169 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1170 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1171 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1172 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1174 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1175 __builtin_ia32_wbinvd): New builtins.
1176 (SPECIAL_ARGS2): New.
1177 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1178 (SPECIAL_ARGS2): New.
1179 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1180 (ix86_valid_target_attribute_inner_p): Ditto.
1181 (ix86_init_mmx_sse_builtins): Add special_args2.
1182 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1183 TARGET_WBNOINVD_P): New.
1184 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1185 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1186 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1187 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1188 * config/i386/pconfigintrin.h: New file.
1189 * config/i386/wbnoinvdintrin.h: Ditto.
1190 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1191 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1193 2018-03-05 Richard Biener <rguenther@suse.de>
1195 PR tree-optimization/84670
1196 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1198 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1199 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1200 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1201 doesn't grow if all successors (recursively) were visited at least
1204 2018-03-05 Richard Biener <rguenther@suse.de>
1206 PR tree-optimization/84650
1207 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1208 if executed in the loop pipeline.
1210 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1212 * doc/configfiles.texi (Configuration Files): Move info about
1213 conditionalizing $target-protos.h to...
1214 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1215 differs from $target-protos.h.
1217 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1218 Chung-Ju Wu <jasonwucj@gmail.com>
1220 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1221 * config/nds32/nds32-multiple.md (setmemsi): Define.
1222 * config/nds32/nds32-memory-manipulation.c
1223 (nds32_gen_dup_4_byte_to_word_value): New.
1224 (emit_setmem_word_loop): New.
1225 (emit_setmem_byte_loop): New.
1226 (nds32_expand_setmem_loop): New.
1227 (nds32_expand_setmem_loop_v3m): New.
1228 (nds32_expand_setmem_unroll): New.
1229 (nds32_expand_setmem): New.
1231 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1232 Chung-Ju Wu <jasonwucj@gmail.com>
1234 * config/nds32/nds32-memory-manipulation.c
1235 (nds32_emit_load_store): New.
1236 (nds32_emit_post_inc_load_store): New.
1237 (nds32_emit_mem_move): New.
1238 (nds32_emit_mem_move_block): New.
1239 (nds32_expand_movmemsi_loop_unknown_size): New.
1240 (nds32_expand_movmemsi_loop_known_size): New.
1241 (nds32_expand_movmemsi_loop): New.
1242 (nds32_expand_movmemsi_unroll): New.
1243 (nds32_expand_movmemqi): Rename ...
1244 (nds32_expand_movmemsi): ... to this.
1245 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1246 (movmemsi): ... to this.
1247 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1248 (nds32_expand_movmemsi): ... to this.
1250 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1251 Monk Chiang <sh.chiang04@gmail.com>
1252 Chung-Ju Wu <jasonwucj@gmail.com>
1254 * config/nds32/nds32-protos.h
1255 (nds32_expand_load_multiple): New arguments.
1256 (nds32_expand_store_multiple): Ditto.
1257 (nds32_valid_multiple_load_store): Rename ...
1258 (nds32_valid_multiple_load_store_p): ... to this.
1259 * config/nds32/nds32-memory-manipulation.c
1260 (nds32_expand_load_multiple): Refine implementation.
1261 (nds32_expand_store_multiple): Ditto.
1262 * config/nds32/nds32-multiple.md
1263 (load_multiple): Update nds32_expand_load_multiple interface.
1264 (store_multiple): Update nds32_expand_store_multiple interface.
1265 * config/nds32/nds32-predicates.c
1266 (nds32_valid_multiple_load_store): Rename ...
1267 (nds32_valid_multiple_load_store_p): ... to this and refine
1269 * config/nds32/predicates.md
1270 (nds32_load_multiple_and_update_address_operation): New predicate.
1271 (nds32_store_multiple_and_update_address_operation): New predicate.
1273 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1274 Chung-Ju Wu <jasonwucj@gmail.com>
1276 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1277 (combo): New attribute.
1278 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1280 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1282 * config/nds32/nds32.opt: Change -mcmodel= default value.
1284 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1285 Monk Chiang <sh.chiang04@gmail.com>
1286 Chung-Ju Wu <jasonwucj@gmail.com>
1288 * config/nds32/constants.md (unspec_element): New enum.
1289 * config/nds32/constraints.md (Umw): New constraint.
1290 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1291 * config/nds32/nds32-intrinsic.md: Likewise.
1292 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1293 (nds32_valid_smw_lwm_base_p): New.
1294 (nds32_output_smw_single_word): New.
1295 (nds32_output_lmw_single_word): New.
1296 (nds32_expand_unaligned_load): New.
1297 (nds32_expand_unaligned_store): New.
1298 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1299 (nds32_output_smw_single_word): Declare.
1300 (nds32_output_lmw_single_word): Declare.
1301 (nds32_expand_unaligned_load): Declare.
1302 (nds32_expand_unaligned_store): Declare.
1303 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1304 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1305 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1306 NDS32_BUILTIN_UASTORE_DW.
1307 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1310 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1311 Kito Cheng <kito.cheng@gmail.com>
1312 Chung-Ju Wu <jasonwucj@gmail.com>
1314 * config/nds32/nds32-intrinsic.c
1315 (nds32_expand_builtin_null_ftype_reg): Delete.
1316 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1317 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1318 (nds32_read_argument): New.
1319 (nds32_legitimize_target): Ditto.
1320 (nds32_legitimize_argument): Ditto.
1321 (nds32_check_constant_argument): Ditto.
1322 (nds32_expand_unop_builtin): Ditto.
1323 (nds32_expand_unopimm_builtin): Ditto.
1324 (nds32_expand_binop_builtin): Ditto.
1325 (nds32_builtin_decl_impl): Ditto.
1326 (builtin_description): Ditto.
1327 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1328 (nds32_init_builtins_impl): Ditto.
1329 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1330 (nds32_builtin_decl): New.
1331 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1332 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1334 2018-03-02 Jeff Law <law@redhat.com>
1336 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1337 (redundant_insn, fill_simple_delay_slots): Likewise.
1338 (fill_slots_from_thread): Likewise.
1339 * resource.c (mark_referenced_resources): Likewise.
1340 (mark_set_resources, find_dead_or_set_registers): Likewise.
1342 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1344 * substring-locations.h (format_warning_va): Formatting fix for
1346 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1348 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1349 * substring-locations.c: Include intl.h.
1350 (format_warning_va): Turned into small wrapper around
1351 format_warning_n_va, renamed to ...
1352 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1353 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1355 (format_warning_at_substring_n): New function.
1356 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1357 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1358 format_warning_at_substring with just a shorter name instead of
1359 const function pointer.
1360 (fmtwarn_n): New function.
1361 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1362 appropriate, get rid of all the fmtstr temporaries, move conditionals
1363 with G_() wrapped string literals directly into fmtwarn arguments,
1364 cast dir.len to (int), formatting fixes.
1366 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1368 * doc/invoke.texi: Remove "Cilk Plus" references.
1370 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1371 Richard Biener <rguenther@suse.de>
1374 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1375 for error or warning attributes if CALL_FROM_THUNK_P is set.
1378 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1381 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1382 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1385 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1386 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1388 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1389 instead of -1U in last predictors element's probability member.
1391 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1394 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1395 arguments if they are comparable.
1397 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1399 PR tree-optimization/84634
1400 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1401 masks and masked_loop_p with a single loop_masks, making sure it's
1402 null for bb vectorization.
1404 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1406 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1407 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1408 loop->force_vectorize to check whether there is no alias.
1410 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1413 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1415 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1417 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1418 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1419 instead of a loop around prev_real_insn.
1420 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1424 * config/i386/i386.c (ix86_print_operand): Use conditional
1425 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1428 2018-03-02 Richard Biener <rguenther@suse.de>
1430 PR tree-optimization/84427
1431 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1432 (bitmap_set_subtract_values): Rewrite to handle multiple
1435 (prune_clobbered_mems): Likewise.
1436 (phi_translate): Take edge instead of pred/phiblock.
1437 (phi_translate_1): Likewise.
1438 (phi_translate_set): Likewise. Insert all translated
1439 exprs for a value into the set, keeping possibly multiple
1440 expressions per value.
1441 (compute_antic_aux): Adjust for phi_translate changes.
1442 When intersecting union the expressions and prune those
1443 not in the final value set, keeping possibly multiple
1444 expressions per value. Do not use value-insertion
1445 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1446 all expressions. Add verification that the value-sets
1447 only shrink during iteration.
1448 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1449 (do_pre_regular_insertion): Likewise.
1450 (do_pre_partial_partial_insertion): Likewise.
1452 2018-03-02 Richard Biener <rguenther@suse.de>
1455 * config/darwin.c (saved_debug_info_level): New static global.
1456 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1457 (darwin_asm_lto_end): Restore debug info generation settings.
1459 2018-03-01 Martin Liska <mliska@suse.cz>
1462 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1465 2018-03-01 Richard Biener <rguenther@suse.de>
1468 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1469 type annotation with LTO when debug was disabled at compile-time.
1471 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1473 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1475 (mips_final_postscan_insn): Likewise.
1477 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1479 PR rtl-optimization/84528
1480 * alias.c (init_alias_target): Add commentary.
1481 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1482 a unique base value if the frame pointer is not eliminated
1483 to the stack pointer.
1485 2018-03-01 Tom de Vries <tom@codesourcery.com>
1487 PR rtl-optimization/83327
1488 * lra-int.h (hard_regs_spilled_into): Declare.
1489 * lra.c (hard_regs_spilled_into): Define.
1490 (init_reg_info): Init hard_regs_spilled_into.
1491 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1492 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1493 (process_bb_lives): Handle hard_regs_spilled_into.
1494 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1495 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1497 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1499 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1500 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1501 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1502 * config/rs6000/aix72.h: New file.
1504 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1506 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1507 instead of warning_at with conditional singular and plural messages
1511 * stor-layout.c (update_alignment_for_field): For
1512 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1513 && !DECL_PACKED (field), do the alignment update, just use
1514 only desired_align instead of MAX (type_align, desired_align)
1516 (place_field): Don't do known_align < desired_align handling
1517 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1518 is non-NULL, instead do it after rli->prev_field handling and
1519 only if not within a bitfield word. For DECL_PACKED (field)
1520 use type_align of BITS_PER_UNIT.
1522 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1524 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1525 superfluous parentheses and trailing spaces.
1527 2018-02-28 Richard Biener <rguenther@suse.de>
1529 PR tree-optimization/84584
1530 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1531 SCoPs with fake exit edge.
1533 2018-02-28 Martin Liska <mliska@suse.cz>
1536 * timevar.c (timer::print): Fix format to properly print 100%
1539 2018-02-28 Richard Biener <rguenther@suse.de>
1542 * genmatch.c (capture_info::walk_match): Do not mark
1543 captured expressions without operands as expr_p given
1544 they act more like predicates and should be subject to
1545 "lost tail" side-effect preserving.
1547 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1549 PR rtl-optimization/81611
1550 * auto-inc-dec.c (attempt_change): Move dead note from
1551 mem_insn if it's the next use of regno
1552 (find_address): Take address use of reg holding
1553 non-incremented value. Add parm to limit search to the named
1555 (merge_in_block): Attempt to use a mem insn that is the next
1556 use of the original regno.
1558 2018-02-27 Martin Sebor <msebor@redhat.com>
1561 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1562 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1564 2018-02-27 Martin Sebor <msebor@redhat.com>
1566 PR translation/84207
1567 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1568 n argument to unsigned HOST_WIDE_INT.
1569 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1570 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1571 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1572 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1574 2018-02-27 Richard Biener <rguenther@suse.de>
1576 PR tree-optimization/84512
1577 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1578 Do not use the estimate returned from record_stmt_cost for
1579 the scalar iteration cost but sum properly using add_stmt_cost.
1581 2018-02-27 Richard Biener <rguenther@suse.de>
1583 PR tree-optimization/84466
1584 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1585 Adjust last change to less strictly validate use operands.
1587 2018-02-27 Martin Liska <mliska@suse.cz>
1589 PR gcov-profile/84548
1590 * gcov.c (process_file): Allow partial overlap and consider it
1591 also as group functions.
1592 (output_lines): Properly calculate range of lines for a group.
1594 2018-02-27 Martin Liska <mliska@suse.cz>
1596 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1597 'ggc' suffixes. Change first column width.
1598 (timer::print): Fix formatting of the column.
1600 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1602 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1603 preserve inline entry blocks for the sake of debug inline
1604 entry point markers alone.
1605 (remove_unused_locals): Suggest in comments a better place to
1606 force the preservation of inline entry blocks that are
1607 otherwise unused, but do not preserve them.
1609 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1611 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1613 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1616 * config/i386/constraints.md (Bs): Replace
1617 ix86_indirect_branch_register with
1618 TARGET_INDIRECT_BRANCH_REGISTER.
1620 * config/i386/i386.md (indirect_jump): Likewise.
1621 (tablejump): Likewise.
1622 (*sibcall_memory): Likewise.
1623 (*sibcall_value_memory): Likewise.
1624 Peepholes of indirect call and jump via memory: Likewise.
1625 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1626 (*sibcall_value_GOT_32): Likewise.
1627 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1628 (GOT_memory_operand): Likewise.
1629 (call_insn_operand): Likewise.
1630 (sibcall_insn_operand): Likewise.
1631 (GOT32_symbol_operand): Likewise.
1632 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1634 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1636 PR rtl-optimization/83496
1637 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1638 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1639 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1640 redundant insn, if any.
1641 (relax_delay_slots): Likewise.
1642 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1644 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1646 PR tree-optimization/83965
1647 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1648 that grouped statements are part of a reduction chain. Return
1649 true if the statement is not marked as a reduction itself but
1651 (vect_recog_dot_prod_pattern): Don't check whether the statement
1652 is part of a group here.
1653 (vect_recog_sad_pattern): Likewise.
1654 (vect_recog_widen_sum_pattern): Likewise.
1656 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1659 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1661 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1664 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1666 (ix86_output_indirect_function_return): New prototype.
1667 (ix86_split_simple_return_pop_internal): Likewise.
1668 * config/i386/i386.c (indirect_return_via_cx): New.
1669 (indirect_return_via_cx_bnd): Likewise.
1670 (indirect_thunk_name): Handle return va CX_REG.
1671 (output_indirect_thunk_function): Create alias for
1672 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1673 (ix86_output_indirect_jmp): Remove the bool argument.
1674 (ix86_output_indirect_function_return): New function.
1675 (ix86_split_simple_return_pop_internal): Likewise.
1676 * config/i386/i386.md (*indirect_jump): Don't pass false
1677 to ix86_output_indirect_jmp.
1678 (*tablejump_1): Likewise.
1679 (simple_return_pop_internal): Change it to define_insn_and_split.
1680 Call ix86_split_simple_return_pop_internal to split it for
1682 (simple_return_indirect_internal): Call
1683 ix86_output_indirect_function_return instead of
1684 ix86_output_indirect_jmp.
1686 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1689 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1690 memset and value initialization afterwards.
1692 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1694 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1696 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1699 * common/config/aarch64/aarch64-common.c
1700 (aarch_option_optimization_table[]): Switch
1701 off fomit-frame-pointer
1703 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1704 Chung-Ju Wu <jasonwucj@gmail.com>
1706 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1708 (store_multiple): Ditto.
1710 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1712 * config.gcc: Add --with-cpu support for nds32 target.
1713 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1714 * config/nds32/nds32.opt: Add -mcpu= option.
1716 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1718 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1719 isel=yes): Warn for these deprecated options.
1721 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1723 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1724 ISA_2_5_MASKS_EMBEDDED.
1726 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1728 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1729 p->max as pointers rather than using iterative_hash_expr.
1731 2018-02-23 Carl Love <cel@us.ibm.com>
1733 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1734 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1736 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1737 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1738 P8V_BUILTIN_VEC_VUNSIGNED2.
1740 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1743 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1744 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1746 * lra-constraints.c (curr_insn_transform): Set up
1747 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1749 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1750 * lra-eliminations.c (spill_pseudos): Ditto.
1751 (process_insn_for_elimination): Ditto.
1752 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1753 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1754 LRA_NON_CLOBBERED_ALT.
1756 2018-02-22 Martin Sebor <msebor@redhat.com>
1758 PR tree-optimization/84480
1759 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1760 to maybe_diag_stxncpy_trunc. Call it.
1761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1762 from gimple_fold_builtin_strcpy. Print inlining stack.
1763 (handle_builtin_stxncpy): Print inlining stack.
1764 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1766 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1769 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1770 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1771 and -fcheck-pointer-bounds are used together.
1772 (indirect_thunk_prefix): New enum.
1773 (indirect_thunk_need_prefix): New function.
1774 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1775 "_nt" instead of "_bnd" for NOTRACK prefix.
1776 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1777 (output_indirect_thunk_function): Likewise.
1779 (ix86_code_end): Update output_indirect_thunk_function calls.
1780 (ix86_output_indirect_branch_via_reg): Replace
1781 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1782 (ix86_output_indirect_branch_via_push): Likewise.
1783 (ix86_output_function_return): Likewise.
1784 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1785 incompatible with -fcf-protection=branch and
1786 -fcheck-pointer-bounds.
1788 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1791 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1792 Change gcc_assert call to output_operand_lossage.
1794 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1796 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1798 2018-02-22 DJ Delorie <dj@redhat.com>
1799 Sebastian Perta <sebastian.perta@renesas.com>
1800 Oleg Endo <olegendo@gcc.gnu.org>
1802 * config/rx/rx.c (rx_rtx_costs): New function.
1803 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1805 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1807 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1809 2018-02-22 Martin Liska <mliska@suse.cz>
1812 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1813 Add "native" as a possible value.
1815 2018-02-22 Martin Liska <mliska@suse.cz>
1818 * config/i386/i386.c (ix86_option_override_internal):
1819 Add "native" as a possible value for -march and -mtune.
1821 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1824 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1825 to all type variants.
1827 PR tree-optimization/84503
1828 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1829 width as info->bitpos + info->bitsize - start.
1830 (merged_store_group::merge_overlapping): Simplify width computation.
1831 (check_no_overlap): New function.
1832 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1833 start + width and last_order of the group, fail if check_no_overlap
1835 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1836 to group if check_no_overlap fails.
1838 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1840 * config/rs6000/altivec.md: Delete contraint arguments to
1841 define_expand, define_split, and define_peephole2, and in
1842 define_insn_and_split if always unused.
1843 * config/rs6000/darwin.md: Ditto.
1844 * config/rs6000/dfp.md: Ditto.
1845 * config/rs6000/rs6000.md: Ditto.
1846 * config/rs6000/sync.md: Ditto.
1847 * config/rs6000/vector.md: Ditto.
1848 * config/rs6000/vsx.md: Ditto.
1850 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1852 * config/rs6000/altivec.md: Write output control strings as braced
1853 blocks instead of double-quoted strings.
1854 * config/rs6000/darwin.md: Ditto.
1855 * config/rs6000/rs6000.md: Ditto.
1856 * config/rs6000/vector.md: Ditto.
1857 * config/rs6000/vsx.md: Ditto.
1859 2018-02-21 Jason Merrill <jason@redhat.com>
1861 PR c++/84314 - ICE with templates and fastcall attribute.
1862 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1864 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1866 * ipa-cp.c (determine_versionability): Fix comment typos.
1868 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1871 * ipa-cp.c (determine_versionability): Do not version functions caling
1874 2018-02-21 Martin Liska <mliska@suse.cz>
1877 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1878 Add "native" as a possible value.
1879 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1880 the macro when native cpu detection is available.
1882 2018-02-21 Martin Liska <mliska@suse.cz>
1885 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1886 Add "native" as a possible value.
1887 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1888 when native cpu detection is available.
1890 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1891 Martin Sebor <msebor@redhat.com>
1893 PR tree-optimization/84478
1894 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1896 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1897 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1898 support which is conservatively correct, for 2 only stay conservative
1899 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1900 argument to the 2 argument get_range_strlen, adjust 6 arg
1901 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1903 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1904 (gimple_fold_builtin_strlen): Pass true as last argument to
1907 2018-02-20 Martin Sebor <msebor@redhat.com>
1910 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1911 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1912 (builtin_memref::builtin_memref): Factor out parts into
1913 set_base_and_offset and call it.
1915 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1918 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1919 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1920 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1921 search at the associated MODE_INT.
1923 2018-02-20 Jeff Law <law@redhat.com>
1926 PR tree-optimization/81592
1928 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1929 for range data rather than using global data.
1930 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1931 range data rather than using global data.
1932 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1933 pass it to children as needed.
1934 (struct directive::fmtresult): Similarly.
1935 (struct directive::set_width): Similarly.
1936 (struct directive::set_precision): Similarly.
1937 (format_integer, format_directive, parse_directive): Similarly.
1938 (format_none): Accept unnamed vr_values parameter.
1939 (format_percent, format_floating, format_character): Similarly.
1940 (format_string, format_plain): Similarly.
1941 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1942 the EVRP range analyzer for range data rather than using global data.
1943 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1944 gimple-ssa-evrp-analyze.h
1945 (class sprintf_dom_walker): Add after_dom_children member function.
1946 Add evrp_range_analyzer member.
1947 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1948 range analyzer as needed.
1949 (sprintf_dom_walker::after_dom_children): New member function.
1950 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1952 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1953 (evrp_range_analyzer::pop_to_marker): Likewise.
1955 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1957 PR tree-optimization/84419
1958 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1959 with the required type if its current type is compatible but
1962 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1965 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1966 after vectorization.
1968 2018-02-20 Martin Liska <mliska@suse.cz>
1971 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1972 possible values if we don't have a hint.
1974 2018-02-20 Martin Liska <mliska@suse.cz>
1978 * final.c (shorten_branches): Build align_tab array with one
1980 * opts.c (finish_options): Add alignment option limit check.
1981 (MAX_CODE_ALIGN): Likewise.
1982 (MAX_CODE_ALIGN_VALUE): Likewise.
1983 * doc/invoke.texi: Document maximum allowed option value for
1984 all -falign-* options.
1986 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1989 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1990 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1991 * var-tracking.c (emit_note_insn_var_location): Remove all references
1992 to NOTE_INSN_CALL_ARG_LOCATION.
1993 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1994 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1995 Use copy_rtx_if_shared.
1996 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1997 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1998 (dwarf2out_var_location): Remove handling of
1999 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2001 * final.c (final_scan_insn): Remove all references to
2002 NOTE_INSN_CALL_ARG_LOCATION.
2003 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2004 before dumping final insns.
2005 * except.c (emit_note_eh_region_end): Remove all references to
2006 NOTE_INSN_CALL_ARG_LOCATION.
2007 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2008 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2009 * config/arc/arc.c (hwloop_optimize): Likewise.
2010 * config/arm/arm.c (create_fix_barrier): Likewise.
2011 * config/s390/s390.c (s390_chunkify_start): Likewise.
2012 * config/sh/sh.c (find_barrier): Likewise.
2013 * config/i386/i386.c (rest_of_insert_endbranch,
2014 ix86_seh_fixup_eh_fallthru): Likewise.
2015 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2016 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2017 * config/frv/frv.c (frv_function_prologue): Likewise.
2018 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2020 (note_outside_basic_block_p): Remove all references to
2021 NOTE_INSN_CALL_ARG_LOCATION.
2022 * gengtype.c (adjust_field_rtx_def): Likewise.
2023 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2025 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2026 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2029 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2032 PR tree-optimization/84452
2033 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2034 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2037 2018-02-19 Martin Liska <mliska@suse.cz>
2040 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2042 2018-02-19 Martin Liska <mliska@suse.cz>
2043 Richard Sandiford <richard.sandiford@linaro.org>
2045 PR tree-optimization/82491
2046 * gimple-fold.c (get_base_constructor): Make earlier bail out
2049 2018-02-19 Carl Love <cel@us.ibm.com>
2051 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2052 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2054 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2055 P8V_BUILTIN_VEC_NEG.
2057 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2059 * config/rl78/rl78.md (movdf): New define expand.
2061 2018-02-19 Martin Liska <mliska@suse.cz>
2064 * doc/invoke.texi: Fix typo.
2065 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2067 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2069 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2070 handle rs6000_single_float and rs6000_double_float specially for
2073 2018-02-16 Jeff Law <law@redhat.com>
2075 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2078 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2081 * ipa-inline.c (inline_small_functions): Fix a typo.
2083 2018-02-16 Nathan Sidwell <nathan@acm.org>
2085 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2087 2018-02-16 Carl Love <cel@us.ibm.com>
2089 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2090 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2091 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2092 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2093 expansion to P8V_BUILTIN_VEC_FLOAT2.
2095 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2097 PR rtl-optimization/70023
2098 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2099 src_regno into account.
2101 2018-02-16 Carl Love <cel@us.ibm.com>
2103 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2104 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2105 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2106 * config/rs6000/rs6000.c: Remove case statements for
2107 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2108 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2109 and P9V_BUILTIN_VEC_VINSERT4B.
2110 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2111 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2112 * config/rs6000/vsx.md:
2113 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2116 2018-02-16 Carl Love <cel@us.ibm.com>
2118 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2120 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2122 * config/rs6000/rs6000-c.c: Add the definitions for
2123 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2124 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2125 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2126 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2127 definition for insert4b and define insn *insert3b_internal.
2128 * doc/extend.texi: Add documentation for vec_extract4b.
2130 2018-02-16 Nathan Sidwell <nathan@acm.org>
2132 * doc/extend.texi (Backwards Compatibility): Mention friend
2133 injection. Note for-scope is deprecated.
2134 * doc/invoke.texi (-ffriend-injection): Deprecate.
2136 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2138 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2139 that moved to I2, also allow destinations that are a paradoxical
2140 subreg (instead of a normal reg).
2142 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2145 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2148 2018-02-16 Richard Biener <rguenther@suse.de>
2150 PR tree-optimization/84037
2151 PR tree-optimization/84016
2153 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2154 Adjust vec_construct for the fact we need additional higher latency
2155 128bit inserts for AVX256 and AVX512 vector builds.
2156 (ix86_add_stmt_cost): Scale vector construction cost for
2159 2018-02-16 Richard Biener <rguenther@suse.de>
2161 PR tree-optimization/84417
2162 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2163 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2164 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2166 2018-02-16 Martin Liska <mliska@suse.cz>
2169 * internal-fn.def (ASAN_CHECK): Set proper flags.
2170 (ASAN_MARK): Likewise.
2172 2018-02-16 Julia Koval <julia.koval@intel.com>
2174 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2175 from PTA_CANNONLAKE.
2177 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2180 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2181 Use ++iter rather than iter++ for std::list iterators.
2182 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2183 defer deleting them until all nodes in the forest are processed. Do
2184 free even leaf nodes. Change to_process into auto_vec.
2187 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2188 * vec.h (vec_default_construct): Use memset instead of placement new
2189 if BROKEN_VALUE_INITIALIZATION is defined.
2190 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2191 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2194 PR rtl-optimization/83723
2195 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2196 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2197 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2199 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2201 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2203 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2205 PR rtl-optimization/81443
2206 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2207 from inner REGs to paradoxical SUBREGs.
2209 2018-02-16 Richard Biener <rguenther@suse.de>
2211 PR tree-optimization/84399
2212 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2213 For operands we can analyze at their definition make sure we can
2214 analyze them at each use as well.
2216 2018-02-16 Richard Biener <rguenther@suse.de>
2218 PR tree-optimization/84190
2219 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2220 volatile accesses if the decl isn't volatile.
2222 2018-02-15 Jason Merrill <jason@redhat.com>
2224 PR c++/84314 - ICE with templates and fastcall attribute.
2225 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2226 TYPE_CANONICAL on an existing type.
2228 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2230 PR tree-optimization/84383
2231 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2232 dstoff nor call operand_equal_p if dstbase is NULL.
2234 PR tree-optimization/84334
2235 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2236 also a CONSTANT_CLASS_P, punt.
2238 2018-02-14 Jim Wilson <jimw@sifive.com>
2240 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2241 first SMALL_OPERAND check. New local min_second_step. Move assert
2242 to where locals are set. Add TARGET_RVC support.
2243 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2245 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2247 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2249 2018-02-14 Martin Sebor <msebor@redhat.com>
2251 PR tree-optimization/83698
2252 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2253 arrays constrain the offset range to their bounds.
2254 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2255 (builtin_access::overlap): Avoid setting the size of overlap if it's
2257 (maybe_diag_overlap): Also consider arrays when deciding what values
2258 of offsets to include in diagnostics.
2260 2018-02-14 Martin Sebor <msebor@redhat.com>
2263 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2264 that correspond to the kind of a declaration.
2266 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2269 * config/pa/pa.md: Load address of PIC label using the linkage table
2270 if the label is nonlocal.
2272 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2274 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2275 warning message if user requests -maltivec=be.
2276 * doc/invoke.texi: Document deprecation of -maltivec=be.
2278 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2281 * config/rs6000/rs6000-c.c: Update definitions for
2282 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2283 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2285 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2288 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2289 add _get_ssp intrinsics. Remove argument from
2290 __builtin_ia32_rdssp[d|q].
2291 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2292 * config/i386/i386-builtin.def: Remove argument from
2293 __builtin_ia32_rdssp[d|q].
2294 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2295 ix86_expand_special_args_builtin for _rdssp[d|q].
2296 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2297 Clear register before usage.
2298 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2299 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2301 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2303 PR tree-optimization/84357
2304 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2305 operand 1 of an ARRAY_REF too.
2307 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2310 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2311 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2313 (set_of_reg): New struct.
2314 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2315 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2316 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2318 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2319 Split into bitclr, bitset, bitinvert patterns if appropriate.
2320 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2321 use rx_fuse_in_memory_bitop.
2322 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2323 to named insn, correct maximum insn length.
2325 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2328 * machmode.def: Define a complex mode for PARTIAL_INT.
2329 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2331 * doc/rtl.texi: Document CSPImode.
2332 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2334 (msp430_hard_regno_nregs_with_padding): Likewise.
2336 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2339 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2341 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2343 PR rtl-optimization/84169
2344 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2345 we generated a parallel as new i3 and we split that to new i2 and i3
2346 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2347 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2348 those to i2, not i1. Partially rewrite this scan code.
2350 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2353 * stor-layout.c (place_field): For variable length fields, adjust
2354 offset_align afterwards not just based on the field's alignment,
2355 but also on the size.
2358 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2359 of exps and logs in the use_exp2 case.
2361 2018-02-13 Jeff Law <law@redhat.com>
2363 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2366 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2369 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2373 * common.opt (gas-loc-support, gas-locview-support): New.
2374 (ginline-points, ginternal-reset-location-views): New.
2375 * doc/invoke.texi: Document them. Use @itemx where intended.
2376 (gvariable-location-views): Adjust.
2377 * target.def (reset_location_view): New.
2378 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2379 (TARGET_RESET_LOCATION_VIEW): New.
2380 * doc/tm.texi: Rebuilt.
2381 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2382 (dwarf2out_default_as_locview_support): New.
2383 (output_asm_line_debug_info): Use option variables.
2384 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2385 (output_loc_list): Likewise.
2386 (add_high_low_attributes): Check option variables.
2387 Don't output entry view attribute in strict mode.
2388 (gen_inlined_subroutine_die): Check option variables.
2389 (dwarf2out_inline_entry): Likewise.
2390 (init_sections_and_labels): Likewise.
2391 (dwarf2out_early_finish): Likewise.
2392 (maybe_reset_location_view): New, from...
2393 (dwarf2out_var_location): ... here. Call it.
2394 * debug.h (dwarf2out_default_as_loc_support): Declare.
2395 (dwarf2out_default_as_locview_support): Declare.
2396 * hooks.c (hook_int_rtx_insn_0): New.
2397 * hooks.h (hook_int_rtx_insn_0): Declare.
2398 * toplev.c (process_options): Take -gas-loc-support and
2399 -gas-locview-support from dwarf2out. Enable
2400 -gvariable-location-views by default only with locview
2401 assembler support. Enable -ginternal-reset-location-views by
2402 default only if the target defines the corresponding hook.
2403 Enable -ginline-points by default if location views are
2404 enabled; force it disabled if statement frontiers are
2406 * tree-inline.c (expand_call_inline): Check option variables.
2407 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2409 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2411 PR tree-optimization/84321
2412 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2413 handling. Also check whether the anti-range contains any values
2414 that satisfy the mask; switch to a VR_RANGE if not.
2416 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2419 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2421 2018-02-13 Martin Jambor <mjambor@suse.cz>
2424 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2425 of call statements, also set location of a load to a temporary.
2427 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2429 * config/rl78/rl78.c (add_vector_labels): New function.
2430 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2431 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2432 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2433 which checks that no arguments are passed.
2434 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2435 * doc/extend.texi: Documentation for the new attribute.
2437 2018-02-13 Andreas Schwab <schwab@suse.de>
2439 * config/riscv/linux.h (CPP_SPEC): Define.
2441 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2444 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2445 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2446 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2447 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2448 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2449 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2450 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2451 temporarily for AES and PCLMUL builtins.
2453 PR tree-optimization/84339
2454 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2455 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2459 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2460 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2461 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2463 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2465 * omp-simd-clone.h: New file.
2466 * omp-simd-clone.c: Include omp-simd-clone.h.
2467 (expand_simd_clones): No longer static.
2468 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2469 cgraph.h and omp-simd-clone.h.
2470 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2471 (vect_recog_widen_shift_pattern): Formatting fix.
2472 (vect_pattern_recog_1): Don't check optab for calls.
2475 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2476 operands[2] into a REG before using gen_lowpart on it.
2478 2018-02-12 Jeff Law <law@redhat.com>
2481 * config/sh/sh.c (find_barrier): Consider a sibling call
2484 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2485 successfully back substituting a reg.
2487 2018-02-12 Richard Biener <rguenther@suse.de>
2489 PR tree-optimization/84037
2490 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2491 parameter, move visited init to caller.
2492 (vect_slp_analyze_operations): Separate cost from validity
2493 check, initialize visited once for all instances.
2494 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2496 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2498 (vect_model_promotion_demotion_cost): Likewise.
2499 (vectorizable_bswap): Guard cost modeling with !slp_node
2500 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2502 (vectorizable_call): Likewise.
2503 (vectorizable_conversion): Likewise.
2504 (vectorizable_assignment): Likewise.
2505 (vectorizable_shift): Likewise.
2506 (vectorizable_operation): Likewise.
2507 (vectorizable_store): Likewise.
2508 (vectorizable_load): Likewise.
2509 (vectorizable_condition): Likewise.
2510 (vectorizable_comparison): Likewise.
2512 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2515 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2516 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2517 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2519 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2522 * params.def (inline-min-speedup): Increase from 8 to 15.
2523 (max-inline-insns-auto): Decrease from 40 to 30.
2524 * ipa-split.c (consider_split): Add some buffer for function to
2525 be considered inlining candidate.
2526 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2529 2018-02-12 Richard Biener <rguenther@suse.de>
2531 PR tree-optimization/84037
2532 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2533 matched stmts if we cannot swap the non-matched ones.
2535 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2537 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2538 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2539 _mm_maskz_scalef_round_ss): New intrinsics.
2540 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2541 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2542 __builtin_ia32_scalefss_round): Remove.
2543 (__builtin_ia32_scalefsd_mask_round,
2544 __builtin_ia32_scalefss_mask_round): New intrinsics.
2545 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2546 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2547 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2548 "<round_constraint>")): Changed to ...
2549 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2550 "<round_scalar_constraint>")): ... this.
2551 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2552 %0, %1, %2<round_op3>}"): Changed to ...
2553 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2554 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2555 %2<round_scalar_mask_op3>}"): ... this.
2556 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2558 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2560 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2561 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2562 (_mm_maskz_sqrt_round_ss): New intrinsics.
2563 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2564 (__builtin_ia32_sqrtsd_mask_round)
2565 (__builtin_ia32_sqrtss_mask_round): New builtins.
2566 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2567 (__builtin_ia32_sqrtss_round): Remove.
2568 (__builtin_ia32_sqrtsd_mask_round)
2569 (__builtin_ia32_sqrtss_mask_round): New builtins.
2570 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2571 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2572 ((match_operand:VF_128 1 "vector_operand"
2573 "xBm,<round_constraint>")): Changed to ...
2574 ((match_operand:VF_128 1 "vector_operand"
2575 "xBm,<round_scalar_constraint>")): ... this.
2576 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2577 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2578 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2579 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2580 %<iptr>1<round_scalar_mask_op3>}): ... this.
2581 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2582 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2584 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2587 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2588 Cast vec_cmpeq result to correct type.
2589 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2590 Cast vec_cmpgt result to correct type.
2592 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2594 * final.c (final_scan_insn_1): Renamed from...
2595 (final_scan_insn): ... this. New wrapper, to recover
2596 seen from the outermost call in recursive ones.
2597 * config/sparc/sparc.c (output_return): Drop seen from call.
2598 (output_sibcall): Likewise.
2599 * config/visium/visium.c (output_branch): Likewise.
2601 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2603 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2606 2018-02-10 Alan Modra <amodra@gmail.com>
2609 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2610 Specify LR as an input.
2612 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2615 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2616 remove_member_access_dummy_vars): New functions.
2617 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2618 lower_omp_1, execute_lower_omp): Use them.
2620 PR rtl-optimization/84308
2621 * shrink-wrap.c (spread_components): Release todo vector.
2623 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2625 PR rtl-optimization/57193
2626 * ira-color.c (struct allocno_color_data): Add member
2627 conflict_allocno_hard_prefs.
2628 (update_conflict_allocno_hard_prefs): New.
2629 (bucket_allocno_compare_func): Add a preference based on
2630 conflict_allocno_hard_prefs.
2631 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2632 (color_allocnos): Remove a dead code. Initiate
2633 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2635 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2638 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2639 constraint from =wa to wa. Avoid a subreg on the output operand,
2640 instead use a pseudo and subreg it in a move.
2641 (p9_xxbrd_<mode>): Changed to ...
2642 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2643 (p9_xxbrd_v2df): New expander.
2644 (p9_xxbrw_<mode>): Changed to ...
2645 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2646 (p9_xxbrw_v4sf): New expander.
2648 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2650 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2652 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2655 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2656 multiply in 32-bit mode.
2657 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2658 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2661 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2663 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2664 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2665 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2666 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2668 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2671 * dwarf2out.c (is_trivial_indirect_ref): New function.
2672 (dwarf2out_late_global_decl): Do not generate a location
2673 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2674 and that are not defined in the current unit.
2676 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2678 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2679 instead of a libcall for UNORDERED.
2681 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2684 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2685 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2687 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2690 * config/s390/s390.c (s390_set_current_function): Invoke
2691 s390_indirect_branch_settings also if fndecl didn't change.
2693 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2695 * config/rs6000/rs6000.md (blockage): Set length to zero.
2697 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2699 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2701 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2704 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2705 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2709 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2710 PARALLEL incoming that failed vt_get_decl_and_offset check.
2713 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2714 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2716 (get_variable_section): For decls in named .bss* sections pass true as
2717 second argument to bss_initializer_p.
2719 2018-02-09 Marek Polacek <polacek@redhat.com>
2720 Jakub Jelinek <jakub@redhat.com>
2723 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2724 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2725 Sync some changes from cxx_fold_indirect_ref.
2727 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2729 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2731 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2732 (BLOCK_INLINE_ENTRY_LABEL): New.
2733 (dwarf2out_var_location): Disregard inline entry markers.
2734 (inline_entry_data): New struct.
2735 (inline_entry_data_hasher): New hashtable type.
2736 (inline_entry_data_hasher::hash): New.
2737 (inline_entry_data_hasher::equal): New.
2738 (inline_entry_data_table): New variable.
2739 (add_high_low_attributes): Add DW_AT_entry_pc and
2740 DW_AT_GNU_entry_view attributes if a pending entry is found
2741 in inline_entry_data_table. Add old entry_pc attribute only
2742 if debug nonbinding markers are disabled.
2743 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2744 markers are enabled.
2745 (block_within_block_p, dwarf2out_inline_entry): New.
2746 (dwarf2out_finish): Check that no entries remained in
2747 inline_entry_data_table.
2748 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2749 (final_scan_insn, notice_source_line): Likewise.
2750 (rest_of_clean_state): Skip inline entry markers.
2751 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2753 * gimple.c (gimple_build_debug_inline_entry): New.
2754 * gimple.h (enum gimple_debug_subcode): Add
2755 GIMPLE_DEBUG_INLINE_ENTRY.
2756 (gimple_build_debug_inline_entry): Declare.
2757 (gimple_debug_inline_entry_p): New.
2758 (gimple_debug_nonbind_marker_p): Adjust.
2759 * insn-notes.def (INLINE_ENTRY): New.
2760 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2761 inline entry marker notes.
2762 (print_insn): Likewise.
2763 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2764 (INSN_DEBUG_MARKER_KIND): Likewise.
2765 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2766 * tree-inline.c (expand_call_inline): Build and insert
2767 debug_inline_entry stmt.
2768 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2769 inline entry blocks early, if nonbind markers are enabled.
2770 (dump_scope_block): Dump fragment info.
2771 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2772 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2773 (gimple_build_debug_inline_entry): New.
2774 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2775 Enable/disable inline entry points too.
2776 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2777 (DEBUG_INSN): Describe inline entry markers.
2779 * common.opt (gvariable-location-views): New.
2780 (gvariable-location-views=incompat5): New.
2781 * config.in: Rebuilt.
2782 * configure: Rebuilt.
2783 * configure.ac: Test assembler for view support.
2784 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2785 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2786 * dwarf2out.c (var_loc_view): New typedef.
2787 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2788 (dwarf2out_locviews_in_attribute): New.
2789 (dwarf2out_locviews_in_loclist): New.
2790 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2791 (enum dw_line_info_opcode): Add LI_adv_address.
2792 (struct dw_line_info_table): Add view.
2793 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2794 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2795 (zero_view_p): New variable.
2796 (ZERO_VIEW_P): New macro.
2797 (output_asm_line_debug_info): New.
2798 (struct var_loc_node): Add view.
2799 (add_AT_view_list, AT_loc_list): New.
2800 (add_var_loc_to_decl): Add view param. Test it against last.
2801 (new_loc_list): Add view params. Record them.
2802 (AT_loc_list_ptr): Handle loc and view lists.
2803 (view_list_to_loc_list_val_node): New.
2804 (print_dw_val): Handle dw_val_class_view_list.
2805 (size_of_die): Likewise.
2806 (value_format): Likewise.
2807 (loc_list_has_views): New.
2808 (gen_llsym): Set vl_symbol too.
2809 (maybe_gen_llsym, skip_loc_list_entry): New.
2810 (dwarf2out_maybe_output_loclist_view_pair): New.
2811 (output_loc_list): Output view list or entries too.
2812 (output_view_list_offset): New.
2813 (output_die): Handle dw_val_class_view_list.
2814 (output_dwarf_version): New.
2815 (output_compilation_unit_header): Use it.
2816 (output_skeleton_debug_sections): Likewise.
2817 (output_rnglists, output_line_info): Likewise.
2818 (output_pubnames, output_aranges): Update version comments.
2819 (output_one_line_info_table): Output view numbers in asm comments.
2820 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2821 Call maybe_gen_llsym.
2822 (loc_list_from_tree_1): Adjust.
2823 (add_AT_location_description): Create view list attribute if
2824 needed, check it's absent otherwise.
2825 (convert_cfa_to_fb_loc_list): Adjust.
2826 (maybe_emit_file): Call output_asm_line_debug_info for test.
2827 (dwarf2out_var_location): Reset views as needed. Precompute
2828 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2829 attribute. Set view.
2830 (new_line_info_table): Reset next view.
2831 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2832 (dwarf2out_source_line): Likewise. Output view resets and labels to
2833 the assembler, or select appropriate line info opcodes.
2834 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2835 (optimize_string_length): Catch it. Adjust.
2836 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2837 dw_val_class_view_list, and remove it if no longer needed.
2838 (hash_loc_list): Hash view numbers.
2839 (loc_list_hasher::equal): Compare them.
2840 (optimize_location_lists): Check whether a view list symbol is
2841 needed, and whether the locview attribute is present, and
2842 whether they match. Remove the locview attribute if no longer
2844 (index_location_lists): Call skip_loc_list_entry for test.
2845 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2846 Use output_dwarf_version.
2847 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2848 (struct dw_val_node): Add val_view_list.
2849 * final.c (SEEN_NEXT_VIEW): New.
2850 (set_next_view_needed): New.
2851 (clear_next_view_needed): New.
2852 (maybe_output_next_view): New.
2853 (final_start_function): Rename to...
2854 (final_start_function_1): ... this. Take pointer to FIRST,
2855 add SEEN parameter. Emit param bindings in the initial view.
2856 (final_start_function): Reintroduce SEEN-less interface.
2857 (final): Rename to...
2858 (final_1): ... this. Take SEEN parameter. Output final pending
2859 next view at the end.
2860 (final): Reintroduce seen-less interface.
2861 (final_scan_insn): Output pending next view before switching
2862 sections or ending a block. Mark the next view as needed when
2863 outputting variable locations. Notify debug backend of section
2864 changes, and of location view changes.
2865 (rest_of_handle_final): Adjust.
2866 * toplev.c (process_options): Autodetect value for debug variable
2867 location views option. Warn on incompat5 without -gdwarf-5.
2868 * doc/invoke.texi (gvariable-location-views): New.
2869 (gvariable-location-views=incompat5): New.
2870 (gno-variable-location-views): New.
2872 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2874 PR tree-optimization/84136
2875 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2876 that the result of find_edge is non-NULL.
2878 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2881 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2882 storing integer register in SImode. Fix cost of 256 and 512
2883 byte aligned SSE register store.
2885 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2887 * config/i386/i386.c (ix86_multiplication_cost): Fix
2888 multiplication cost for TARGET_AVX512DQ.
2890 2018-02-08 Marek Polacek <polacek@redhat.com>
2892 PR tree-optimization/84238
2893 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2896 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2898 PR tree-optimization/84265
2899 * tree-vect-stmts.c (vectorizable_store): Don't treat
2900 VMAT_CONTIGUOUS accesses as grouped.
2901 (vectorizable_load): Likewise.
2903 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2905 PR tree-optimization/81635
2906 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2907 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2908 (test_round_for_mask): New functions.
2909 (wide_int_cc_tests): Call test_round_for_mask.
2910 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2911 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2912 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2913 range returned by get_range_info.
2915 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2918 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2919 * symtab.c: Include builtins.h
2920 (symtab_node::output_to_lto_symbol_table_p): Move here
2921 from lto-streamer-out.c:output_symbol_p.
2922 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2923 (output_symbol_p): Move all logic to symtab.c
2924 (produce_symtab): Update.
2926 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2928 * config/s390/s390-opts.h (enum indirect_branch): Define.
2929 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2930 (s390_indirect_branch_via_thunk)
2931 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2932 (enum s390_indirect_branch_type): Define.
2933 * config/s390/s390.c (struct s390_frame_layout, struct
2934 machine_function): Remove.
2935 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2936 (indirect_branch_table_label_no, indirect_branch_table_name):
2938 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2939 (enum s390_indirect_branch_option): Define.
2940 (s390_return_addr_from_memory): New function.
2941 (s390_handle_string_attribute): New function.
2942 (s390_attribute_table): Add new attribute handler.
2943 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2944 (s390_indirect_branch_via_thunk): New function.
2945 (s390_indirect_branch_via_inline_thunk): New function.
2946 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2947 sibling call optimization for non z10 compiles.
2948 (s390_emit_call): Force indirect branch target to be a single
2949 register. Add r1 clobber for non-z10 compiles.
2950 (s390_emit_epilogue): Emit return jump via return_use expander.
2951 (s390_reorg): Handle JUMP_INSNs as execute targets.
2952 (s390_option_override_internal): Perform validity checks for the
2953 new command line options.
2954 (s390_indirect_branch_attrvalue): New function.
2955 (s390_indirect_branch_settings): New function.
2956 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2957 (s390_output_indirect_thunk_function): New function.
2958 (s390_code_end): Implement target hook.
2959 (s390_case_values_threshold): Implement target hook.
2960 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2962 * config/s390/s390.h (struct s390_frame_layout)
2963 (struct machine_function): Move here from s390.c.
2964 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2965 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2966 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2967 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2968 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2969 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2970 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2971 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2972 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2973 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2974 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2975 (mnemonic attribute): Add values which aren't recognized
2977 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2978 pattern for branch conversion. Fix mnemonic attribute.
2979 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2980 indirect branch via thunk if requested.
2981 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2982 ("*indirect_jump"): Disable for branch conversion using out of
2984 ("indirect_jump_via_thunk<mode>_z10")
2985 ("indirect_jump_via_thunk<mode>")
2986 ("indirect_jump_via_inlinethunk<mode>_z10")
2987 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2988 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2989 ("casesi_jump_via_inlinethunk<mode>_z10")
2990 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2991 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2992 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2993 ("*indirect2_jump"): Disable for branch conversion.
2994 ("casesi_jump"): Turn into expander and expand patterns for branch
2996 ("return_use"): New expander.
2997 ("*return"): Emit return via thunk and rename it to ...
2998 ("*return<mode>"): ... this one.
2999 * config/s390/s390.opt: Add new options and and enum for the
3002 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3004 * lra-constraints.c (match_reload): Unconditionally use
3005 gen_lowpart_SUBREG, rather than selecting between that
3006 and equivalent gen_rtx_SUBREG code.
3008 2018-02-08 Richard Biener <rguenther@suse.de>
3010 PR tree-optimization/84233
3011 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3012 changed flag instead of boguously re-using phi_inserted.
3014 2018-02-08 Martin Jambor <mjambor@suse.cz>
3016 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3017 static local variables.
3019 2018-02-08 Richard Biener <rguenther@suse.de>
3021 PR tree-optimization/84278
3022 * tree-vect-stmts.c (vectorizable_store): When looking for
3023 smaller vector types to perform grouped strided loads/stores
3024 make sure the mode is supported by the target.
3025 (vectorizable_load): Likewise.
3027 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3029 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3030 Increase LDP/STP opportunities by adding adjacent callee-saves.
3032 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3034 PR rtl-optimization/84068
3035 PR rtl-optimization/83459
3036 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3038 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3040 PR tree-optimization/84224
3041 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3042 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3045 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3048 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3049 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3050 count, remove one USE.
3052 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3054 * doc/install.texi (Configuration): Document the
3055 --with-long-double-format={ibm,ieee} PowerPC configuration
3059 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3060 Convert from define_expand to be define_insn_and_split. Rework
3061 float/double/_Float128 conversions to QI/HI/SImode to work with
3062 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3063 conversions to QI/HImode types did a store and then a load to
3064 truncate the value. For conversions to VSX registers, don't split
3065 the insn, instead emit the code directly. Use the code iterator
3066 any_fix to combine signed and unsigned conversions.
3067 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3068 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3069 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3070 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3071 (fix_<mode>di2_hw): Likewise.
3072 (fixuns_<mode>di2_hw): Likewise.
3073 (fix_<mode>si2_hw): Likewise.
3074 (fixuns_<mode>si2_hw): Likewise.
3075 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3076 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3077 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3078 fix<uns>_trunc<SFDF:mode>si2_p8.
3079 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3081 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3082 (fix<uns>_<mode>_mem): Likewise.
3083 (fctiw<u>z_<mode>_mem): Likewise.
3084 (fix<uns>_<mode>_mem): Likewise.
3085 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3086 the register allocator from doing a direct move to the GPRs to do
3087 a store, and instead use the ISA 3.0 store byte/half-word from
3088 vector register instruction. For IEEE 128-bit floating point,
3089 also optimize stores of 32-bit ints.
3090 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3092 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3094 * genextract.c (push_pathstr_operand): New function to support
3096 (walk_rtx): Call push_pathstr_operand.
3097 (print_path): Support [a-zA-Z].
3099 2018-02-07 Richard Biener <rguenther@suse.de>
3101 PR tree-optimization/84037
3102 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3103 (cse_and_gimplify_to_preheader): Declare.
3104 (vect_get_place_in_interleaving_chain): Likewise.
3105 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3107 (_loop_vec_info::~_loop_vec_info): Delete it.
3108 (cse_and_gimplify_to_preheader): New function.
3109 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3110 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3111 (vectorizable_load): Likewise. For grouped stores always base
3112 the IV on the first element.
3113 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3114 condition before gimplifying.
3116 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3118 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3119 *DIV_EXPR and *MOD_EXPR.
3121 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3124 * config/i386/i386.c (ix86_option_override_internal): Mask out
3125 the CF_SET bit when checking -fcf-protection.
3127 2018-02-07 Tom de Vries <tom@codesourcery.com>
3130 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3133 2018-02-07 Richard Biener <rguenther@suse.de>
3135 PR tree-optimization/84204
3136 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3139 PR tree-optimization/84205
3140 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3141 special-case isl_ast_op_zdiv_r.
3143 PR tree-optimization/84223
3144 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3145 Only add conditions from within the region.
3146 (gather_bbs::after_dom_children): Adjust.
3148 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3151 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3152 * config/avr/avr.md: Only post-reload split REG-REG moves if
3153 either register is GENERAL_REG_P.
3155 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3157 PR tree-optimization/84235
3158 * tree-ssa-scopedtables.c
3159 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3160 if the subtraction is performed in floating point type where NaNs are
3161 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3162 build 1. Formatting fix.
3164 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3167 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3168 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3169 and skip it regardless of bb boundaries. Use CALL_P macro,
3170 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3171 unnecessarily, formatting fix.
3173 2018-02-06 Michael Collison <michael.collison@arm.com>
3175 * config/arm/thumb2.md:
3176 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3177 (*thumb_mov_notscc): Ditto.
3179 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3182 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3183 unsigned_fix, not "s".
3185 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3187 * configure.ac (gcc_fn_eh_frame_ro): New function.
3188 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3189 correct .eh_frame permissions.
3190 * configure: Regenerate.
3192 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3194 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3197 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3199 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3200 Display warning message for -mno-speculate-indirect-jumps.
3202 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3204 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3206 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3208 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3210 PR tree-optimization/84225
3211 * tree-eh.c (find_trapping_overflow): Only call
3212 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3214 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3217 * config/i386/i386.c: Reimplement the check of possible options
3218 -mibt/-mshstk conbination. Change error messages.
3219 * doc/invoke.texi: Fix a typo: remove extra '='.
3221 2018-02-06 Marek Polacek <polacek@redhat.com>
3223 PR tree-optimization/84228
3224 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3226 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3229 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3230 emitted arch directives.
3231 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3232 __ARM_FEATURE_COPROC before changing architectures.
3234 2018-02-06 Richard Biener <rguenther@suse.de>
3236 * config/i386/i386.c (print_reg): Fix typo.
3237 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3239 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3241 * configure: Regenerate.
3243 2018-02-05 Martin Sebor <msebor@redhat.com>
3245 PR tree-optimization/83369
3246 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3249 2018-02-05 Martin Liska <mliska@suse.cz>
3251 * doc/invoke.texi: Cherry-pick upstream r323995.
3253 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3255 * ira.c (ira_init_register_move_cost): Adjust comment.
3257 2018-02-05 Martin Liska <mliska@suse.cz>
3259 PR gcov-profile/84137
3260 * doc/gcov.texi: Fix typo in documentation.
3262 2018-02-05 Martin Liska <mliska@suse.cz>
3264 PR gcov-profile/83879
3265 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3266 using dlopen functionality.
3268 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3270 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3271 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3272 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3273 _mm_maskz_range_round_ss): New intrinsics.
3274 (__builtin_ia32_rangesd128_round)
3275 (__builtin_ia32_rangess128_round): Remove.
3276 (__builtin_ia32_rangesd128_mask_round,
3277 __builtin_ia32_rangess128_mask_round): New builtins.
3278 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3279 __builtin_ia32_rangess128_round): Remove.
3280 (__builtin_ia32_rangesd128_mask_round,
3281 __builtin_ia32_rangess128_mask_round): New builtins.
3282 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3283 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3284 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3285 "<round_saeonly_constraint>")): Changed to ...
3286 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3287 "<round_saeonly_scalar_constraint>")): ... this.
3288 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3289 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3290 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3291 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3292 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3294 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3296 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3298 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3299 Remove all values except native, 8540 and 8548.
3301 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3303 * config/i386/i386.c (ix86_output_function_return): Pass
3304 INVALID_REGNUM, instead of -1, as invalid register number to
3305 indirect_thunk_name and output_indirect_thunk.
3307 2018-02-02 Julia Koval <julia.koval@intel.com>
3309 * config.gcc: Add -march=icelake.
3310 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3311 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3312 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3313 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3314 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3315 (processor_target_table): Add icelake.
3316 (ix86_option_override_internal): Handle new PTAs.
3317 (get_builtin_code_for_version): Handle icelake.
3318 (M_INTEL_COREI7_ICELAKE): New.
3319 (fold_builtin_cpu): Handle icelake.
3320 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3321 * doc/invoke.texi: Add -march=icelake.
3323 2018-02-02 Julia Koval <julia.koval@intel.com>
3325 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3326 to wide_int_bitmask.
3327 * wide-int-bitmask.h: New.
3329 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3332 * config/i386/i386.md: Replace Pmode with word_mode in
3333 builtin_setjmp_setup and builtin_longjmp to support x32.
3335 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3339 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3341 (rs6000_supported_cpu_names): New static variable.
3342 (linux_cpu_translation_table): Likewise.
3343 (elf_platform) <cpu>: Define new static variable and use it.
3344 Translate kernel AT_PLATFORM name to canonical name if needed.
3345 Error if platform name is unknown.
3347 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3350 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3352 2018-02-01 Jeff Law <law@redhat.com>
3355 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3356 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3357 the scratch if RELEASE_VIA_POP is false.
3358 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3359 If we have to save a temporary register, decrement SIZE appropriately.
3360 Pass new arguments to release_scratch_register_on_entry.
3361 (ix86_adjust_stack_and_probe): Likewise.
3362 (ix86_emit_probe_stack_range): Pass new arguments to
3363 release_scratch_register_on_entry.
3365 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3367 PR rtl-optimization/84157
3368 * combine.c (change_zero_ext): Use REG_P predicate in
3369 front of HARD_REGISTER_P predicate.
3371 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3373 * config/avr/avr.c (avr_option_override): Move disabling of
3374 -fdelete-null-pointer-checks to...
3375 * common/config/avr/avr-common.c (avr_option_optimization_table):
3378 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3380 PR tree-optimization/81635
3381 * tree-data-ref.c (split_constant_offset_1): For types that
3382 wrap on overflow, try to use range info to prove that wrapping
3385 2018-02-01 Renlin Li <renlin.li@arm.com>
3388 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3390 (aarch64_register_move_cost): Likewise.
3391 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3393 (REG_CLASS_NAMES): Likewise.
3394 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3395 TAILCALL_ADDR_REGS. Remove IP registers.
3396 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3398 2018-02-01 Richard Biener <rguenther@suse.de>
3400 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3401 for specifying RPO order and allow NULL for that.
3402 * domwalk.c (dom_walker::dom_walker): Likewise.
3403 (dom_walker::walk): Handle NULL RPO order.
3404 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3406 (rewrite_update_dom_walker): Likewise.
3407 (mark_def_dom_walker): Likewise.
3409 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3411 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3412 (aarch64_maybe_expand_sve_subreg_move): Declare.
3413 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3414 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3416 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3417 that are semantically a reverse operation.
3418 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3419 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3420 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3422 (aarch64_can_change_mode_class): For big-endian, forbid changes
3423 between two SVE modes if they have different element sizes.
3425 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3427 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3428 the TImode handling for big-endian targets.
3430 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3432 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3433 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3435 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3436 Remove BSWAP handing for big-endian targets and use the form of
3437 LD1RQ appropariate for the mode.
3439 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3441 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3442 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3445 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3448 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3449 check for operands that need to go through aarch64_sve_reload_be.
3451 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3453 PR tree-optimization/81661
3454 PR tree-optimization/84117
3455 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3456 * tree-eh.c: Include gimplify.h.
3457 (find_trapping_overflow, replace_trapping_overflow,
3458 rewrite_to_non_trapping_overflow): New functions.
3459 * tree-vect-loop.c: Include tree-eh.h.
3460 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3461 * tree-data-ref.c: Include tree-eh.h.
3462 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3464 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3466 PR rtl-optimization/84123
3467 * combine.c (change_zero_ext): Check if hard register satisfies
3468 can_change_dest_mode before calling gen_lowpart_SUBREG.
3470 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3473 * ira.c (ira_init_register_move_cost): Remove assert.
3475 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3477 PR rtl-optimization/84071
3478 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3479 * doc/tm.texi: Regenerate.
3481 2018-01-31 Richard Biener <rguenther@suse.de>
3483 PR tree-optimization/84132
3484 * tree-data-ref.c (analyze_miv_subscript): Properly
3485 check whether evolution_function_is_affine_multivariate_p
3486 before calling gcd_of_steps_may_divide_p.
3488 2018-01-31 Julia Koval <julia.koval@intel.com>
3491 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3492 * config/i386/i386.md (rdpid_rex64) New.
3493 (rdpid): Make 32bit only.
3495 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3498 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3499 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3501 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3504 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3506 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3508 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3510 PR rtl-optimization/84071
3511 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3512 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3514 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3516 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3517 (arc_attribute_table): Add 'aux' attribute.
3518 (arc_in_small_data_p): Consider aux like variables.
3519 (arc_is_aux_reg_p): New function.
3520 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3521 (arc_get_aux_arg): New function.
3522 (prepare_move_operands): Handle aux-register access.
3523 (arc_handle_aux_attribute): New function.
3524 * doc/extend.texi (ARC Variable attributes): Add subsection.
3526 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3528 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3529 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3530 (arc_attribute_table): Add 'uncached' attribute.
3531 (arc_print_operand): Print '.di' flag for uncached memory
3533 (arc_in_small_data_p): Do not consider for small data the uncached
3535 (arc_is_uncached_mem_p): New function.
3536 * config/arc/predicates.md (compact_store_memory_operand): Check
3537 for uncached memory accesses.
3538 (nonvol_nonimm_operand): Likewise.
3539 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3541 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3544 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3545 falign-loops=): Add Optimization flag.
3547 2018-01-30 Jeff Law <law@redhat.com>
3550 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3551 INT_REGISTERS_SAVED. Check it prior to calling
3552 get_scratch_register_on_entry.
3553 (ix86_adjust_stack_and_probe): Similarly.
3554 (ix86_emit_probe_stack_range): Similarly.
3555 (ix86_expand_prologue): Corresponding changes.
3557 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3560 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3561 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3563 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3566 * lra-constraints.c (curr_insn_transform): Process AND in the
3569 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3571 PR rtl-optimization/83986
3572 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3573 dependence against last_pending_memory_flush in addition to
3576 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3578 PR tree-optimization/81611
3579 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3582 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3585 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3588 2018-01-30 Richard Biener <rguenther@suse.de>
3589 Jakub Jelinek <jakub@redhat.com>
3591 PR tree-optimization/84111
3592 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3593 inner loops added during recursion, as they don't have up-to-date
3596 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3599 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3600 (can_inline_edge_by_limits_p): ... here.
3601 (can_early_inline_edge_p, check_callers,
3602 update_caller_keys, update_callee_keys, recursive_inlining,
3603 add_new_edges_to_heap, speculation_useful_p,
3604 inline_small_functions,
3605 inline_small_functions, flatten_function,
3606 inline_to_all_callers_1): Update.
3608 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3610 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3613 2018-01-30 Richard Biener <rguenther@suse.de>
3615 PR tree-optimization/83008
3616 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3617 invariant and constant vector uses in stmts when they need
3620 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3623 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3624 * configure: Regenerate.
3626 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3628 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3630 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3631 Use gen_rtx_REG rather than gen_lowpart.
3633 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3635 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3636 rather than 0 when creating partial subregs.
3638 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3640 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3643 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3646 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3647 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3648 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3649 flags. This restores the settings used before the 2017-07-24.
3650 Turning off pre increment/decrement/modify allows IVOPTS to
3651 optimize DF/SF loops where the index is an int.
3653 2018-01-29 Richard Biener <rguenther@suse.de>
3654 Kelvin Nilsen <kelvin@gcc.gnu.org>
3657 * tree-vect-stmts.c (vectorizable_call): Don't call
3658 targetm.vectorize_builtin_md_vectorized_function if callee is
3661 2018-01-22 Carl Love <cel@us.ibm.com>
3663 * doc/extend.tex: Fix typo in second arg in
3664 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3666 2018-01-29 Richard Biener <rguenther@suse.de>
3668 PR tree-optimization/84086
3669 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3670 (flush_ssaname_freelist): When SSA names were released reset
3671 the SCEV hash table.
3673 2018-01-29 Richard Biener <rguenther@suse.de>
3675 PR tree-optimization/84057
3676 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3677 removed paths when removing edges.
3679 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3681 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3682 -mfunction-return=@var{choice}.
3684 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3687 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3689 (layout::print_source_line): Likewise.
3690 (test_get_line_width_without_trailing_whitespace): Add test cases.
3692 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3695 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3698 2018-01-26 Jim Wilson <jimw@sifive.com>
3700 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3702 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3705 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3707 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3708 and CMP + SUB-immediate -> SUBS.
3710 2018-01-26 Martin Sebor <msebor@redhat.com>
3712 PR tree-optimization/83896
3713 * tree-ssa-strlen.c (get_string_len): Rename...
3714 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3715 Avoid assuming length is constant.
3716 (handle_char_store): Use HOST_WIDE_INT for string length.
3718 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3721 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3722 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3724 2018-01-26 Richard Biener <rguenther@suse.de>
3726 PR rtl-optimization/84003
3727 * dse.c (record_store): Only record redundant stores when
3728 the earlier store aliases at least all accesses the later one does.
3730 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3732 PR rtl-optimization/83985
3733 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3734 REG_CFA_RESTORE insns.
3735 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3736 REG_CFA_RESTORE insns here.
3739 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3740 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3742 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3744 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3745 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3746 (arc_init): Likewise.
3747 (arc_override_options): Likewise.
3748 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3750 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3752 * config/arc/arc.h (TARGET_DBNZ): Define.
3753 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3754 properly set the tune attribute.
3755 (dbnz): Use TARGET_DBNZ guard.
3756 * config/arc/arc.opt (mtune): Add core3 option.
3758 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3760 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3761 recognize new pic like addresses.
3762 (arc_delegitimize_address): Clean up.
3764 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3766 * config/arc/arc-arches.def: Option mrf16 valid for all
3768 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3769 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3770 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3771 * config/arc/arc-tables.opt: Regenerate.
3772 * config/arc/arc.c (arc_conditional_register_usage): Handle
3773 reduced register file case.
3774 (arc_file_start): Set must have build attributes.
3775 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3777 * config/arc/arc.opt (mrf16): Add new option.
3778 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3779 * config/arc/genmultilib.awk: Handle new mrf16 option.
3780 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3781 * config/arc/t-multilib: Regenerate.
3782 * doc/invoke.texi (ARC Options): Document mrf16 option.
3784 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3786 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3787 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3788 (arc_attribute_table): Add 'secure_call' attribute.
3789 (arc_print_operand): Print secure call operand.
3790 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3792 (arc_is_secure_call_p): New function. * config/arc/arc.md
3793 (call_i): Add support for sjli instruction.
3794 (call_value_i): Likewise.
3795 * config/arc/constraints.md (Csc): New constraint.
3797 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3798 John Eric Martin <John.Martin@emmicro-us.com>
3800 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3801 * config/arc/arc.c (_arc_jli_section): New struct.
3802 (arc_jli_section): New type.
3803 (rc_jli_sections): New static variable.
3804 (arc_handle_jli_attribute): New function.
3805 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3806 (arc_file_end): New function.
3807 (TARGET_ASM_FILE_END): Define.
3808 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3809 (arc_add_jli_section): New function.
3810 (jli_call_scan): Likewise.
3811 (arc_reorg): Call jli_call_scan.
3812 (arc_output_addsi): Remove 'S' from printing asm operand.
3813 (arc_is_jli_call_p): New function.
3814 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3816 (movhi_insn): Likewise.
3817 (movsi_insn): Likewise.
3818 (movsi_set_cc_insn): Likewise.
3819 (loadqi_update): Likewise.
3820 (load_zeroextendqisi_update): Likewise.
3821 (load_signextendqisi_update): Likewise.
3822 (loadhi_update): Likewise.
3823 (load_zeroextendhisi_update): Likewise.
3824 (load_signextendhisi_update): Likewise.
3825 (loadsi_update): Likewise.
3826 (loadsf_update): Likewise.
3827 (movsicc_insn): Likewise.
3828 (bset_insn): Likewise.
3829 (bxor_insn): Likewise.
3830 (bclr_insn): Likewise.
3831 (bmsk_insn): Likewise.
3832 (bicsi3_insn): Likewise.
3833 (cmpsi_cc_c_insn): Likewise.
3834 (movsi_ne): Likewise.
3835 (movsi_cond_exec): Likewise.
3836 (clrsbsi2): Likewise.
3847 (call_i): Remove 'S' asm letter, add jli instruction.
3848 (call_value_i): Likewise.
3849 * config/arc/arc.op (mjli-always): New option.
3850 * config/arc/constraints.md (Cji): New constraint.
3851 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3853 (subsf3_fpx): Likewise.
3854 (mulsf3_fpx): Likewise.
3855 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3857 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3859 * doc/invoke.texi (ARC): Document mjli-always option.
3861 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3863 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3864 avoid addition with 0 and use incw and decw where possible.
3866 2018-01-26 Richard Biener <rguenther@suse.de>
3868 PR tree-optimization/81082
3869 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3870 association if it requires casting to unsigned.
3871 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3872 from fold_plusminus_mult_expr to catch important cases late when
3873 range info is available.
3875 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3877 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3878 * configure.ac (hidden_linkonce): New test.
3879 * configure: Regenerate.
3880 * config.in: Regenerate.
3882 2018-01-26 Julia Koval <julia.koval@intel.com>
3884 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3885 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3886 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3887 _mm_mask_bitshuffle_epi64_mask): Fix type.
3888 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3889 USI_FTYPE_V4DI_V4DI_USI): Remove.
3890 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3891 __builtin_ia32_vpshufbitqmb256_mask,
3892 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3893 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3894 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3896 2018-01-26 Alan Modra <amodra@gmail.com>
3899 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3900 UNSPEC_VBPERMQ. Sort other unspecs.
3902 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3904 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3906 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3909 * predict.c (drop_profile): Do not push/pop cfun; update also
3911 (handle_missing_profiles): Fix logic looking for zero profiles.
3913 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3916 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3917 on functions with #pragma omp declare simd or functions with simd
3919 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3920 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3921 Remove trailing \n from warning_at calls.
3923 2018-01-25 Tom de Vries <tom@codesourcery.com>
3926 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3927 for neutered workers.
3929 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3932 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3933 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3935 2018-01-24 Jeff Law <law@redhat.com>
3938 * i386.c (get_probe_interval): Move to earlier point.
3939 (ix86_compute_frame_layout): If -fstack-clash-protection and
3940 the frame is larger than the probe interval, then use pushes
3941 to save registers rather than reg->mem moves.
3942 (ix86_expand_prologue): Remove conditional for int_registers_saved
3945 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3948 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3949 min/max for never referenced object.
3951 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3954 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3956 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3957 attributes from DECL_ATTRIBUTES (decl) without affecting
3958 DECL_ATTRIBUTES (current_function_decl).
3959 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3960 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3962 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3964 PR tree-optimization/83979
3965 * fold-const.c (fold_comparison): Use constant_boolean_node
3966 instead of boolean_{true,false}_node.
3968 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3970 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3973 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3975 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3976 Simplify the clause that sets the length attribute.
3977 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3978 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3979 clause that sets the length attribute.
3980 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3982 2018-01-24 Tom de Vries <tom@codesourcery.com>
3985 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3986 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3987 Add strict parameter.
3988 (prevent_branch_around_nothing): Insert dummy insn between branch to
3989 label and label with no ptx insn inbetween.
3990 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3992 2018-01-24 Tom de Vries <tom@codesourcery.com>
3995 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3996 for neutered threads in warp.
3997 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3999 2018-01-24 Richard Biener <rguenther@suse.de>
4001 PR tree-optimization/83176
4002 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4005 2018-01-24 Richard Biener <rguenther@suse.de>
4007 PR tree-optimization/82819
4008 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4009 code generating pluses that are no-ops in the target precision.
4011 2018-01-24 Richard Biener <rguenther@suse.de>
4014 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4016 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4018 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4019 to merge probabilities.
4020 * predict.c (probably_never_executed): Also mark as cold functions
4021 with global 0 profile and guessed local profile.
4022 * profile-count.c (profile_probability::combine_with_count): New
4024 * profile-count.h (profile_probability::operator*,
4025 profile_probability::operator*=, profile_probability::operator/,
4026 profile_probability::operator/=): Reduce precision to adjusted
4027 and set value to guessed on contradictory divisions.
4028 (profile_probability::combine_with_freq): Remove.
4029 (profile_probability::combine_wiht_count): Declare.
4030 (profile_count::force_nonzero):: Set to adjusted.
4031 (profile_count::probability_in):: Set quality to adjusted.
4032 * tree-ssa-tail-merge.c (replace_block_by): Use
4035 2018-01-23 Andrew Waterman <andrew@sifive.com>
4036 Jim Wilson <jimw@sifive.com>
4038 * config/riscv/riscv.c (riscv_stack_boundary): New.
4039 (riscv_option_override): Set riscv_stack_boundary. Handle
4040 riscv_preferred_stack_boundary_arg.
4041 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4042 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4043 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4044 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4045 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4046 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4048 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4051 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4052 of struct ix86_frame.
4053 (ix86_expand_epilogue): Likewise. Add a local variable for
4054 the reg_save_offset field in struct ix86_frame.
4056 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4058 PR tree-optimization/82604
4059 * tree-loop-distribution.c (enum partition_kind): New enum item
4060 PKIND_PARTIAL_MEMSET.
4061 (partition_builtin_p): Support above new enum item.
4062 (generate_code_for_partition): Ditto.
4063 (compute_access_range): Differentiate cases that equality can be
4064 proven at all loops, the innermost loops or no loops.
4065 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4066 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4067 (finalize_partitions, distribute_loop): Don't fuse partition of
4068 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4069 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4072 2018-01-23 Martin Liska <mliska@suse.cz>
4074 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4075 order to ignore the predictor.
4076 (PRED_POLYMORPHIC_CALL): Likewise.
4077 (PRED_RECURSIVE_CALL): Likewise.
4079 2018-01-23 Martin Liska <mliska@suse.cz>
4081 * tree-profile.c (tree_profiling): Print function header to
4082 aware reader which function we are working on.
4083 * value-prof.c (gimple_find_values_to_profile): Do not print
4084 not interesting value histograms.
4086 2018-01-23 Martin Liska <mliska@suse.cz>
4088 * profile-count.h (enum profile_quality): Add
4089 profile_uninitialized as the first value. Do not number values
4090 as they are zero based.
4091 (profile_count::verify): Update sanity check.
4092 (profile_probability::verify): Likewise.
4094 2018-01-23 Nathan Sidwell <nathan@acm.org>
4096 * doc/invoke.texi (ffor-scope): Deprecate.
4098 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4100 PR tree-optimization/83510
4101 * domwalk.c (set_all_edges_as_executable): New function.
4102 (dom_walker::dom_walker): Convert bool param
4103 "skip_unreachable_blocks" to enum reachability. Move setup of
4104 edge flags to set_all_edges_as_executable and only do it when
4105 reachability is REACHABLE_BLOCKS.
4106 * domwalk.h (enum dom_walker::reachability): New enum.
4107 (dom_walker::dom_walker): Convert bool param
4108 "skip_unreachable_blocks" to enum reachability.
4109 (set_all_edges_as_executable): New decl.
4110 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4111 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4113 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4114 but converting true to REACHABLE_BLOCKS.
4115 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4117 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4118 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4119 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4121 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4122 if check_all_array_refs will be called.
4124 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4126 * tree.c (selftest::test_location_wrappers): Add more test
4129 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4131 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4132 (selftest::test_bit_in_range): Likewise.
4134 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4137 * doc/sourcebuild.texi (vect_float): Say that the selector
4138 only describes the situation when -funsafe-math-optimizations is on.
4139 (vect_float_strict): Document.
4141 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4143 PR tree-optimization/83965
4144 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4145 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4146 instead of checking only for a reduction.
4147 (vect_recog_widen_sum_pattern): Likewise.
4149 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4151 * predict.c (probably_never_executed): Only use precise profile info.
4152 (compute_function_frequency): Skip after inlining hack since we now
4153 have quality checking.
4155 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4157 * profile-count.h (profile_probability::very_unlikely,
4158 profile_probability::unlikely, profile_probability::even): Set
4159 precision to guessed.
4161 2018-01-23 Richard Biener <rguenther@suse.de>
4163 PR tree-optimization/83963
4164 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4165 Properly terminate dominator walk when crossing the exit edge not
4166 when visiting its source block.
4168 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4171 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4172 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4174 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4176 PR tree-optimization/83957
4177 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4178 semicolon after for body surrounded by braces.
4180 PR tree-optimization/83081
4181 * profile-count.h (profile_probability::split): New method.
4182 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4183 Use profile_probability::split.
4184 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4185 when splitting a single conditional jump into 2.
4187 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4189 PR tree-optimization/69452
4190 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4193 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4195 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4196 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4197 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4199 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4201 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4202 * config/rl78/rl78.md (movdi): New define_expand.
4203 * config/rl78/rl78.c (rl78_split_movdi): New function.
4205 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4208 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4210 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4211 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4212 128-bit to produce an UNSPEC move to get the double word with the
4213 signbit and then a shift directly to do signbit.
4214 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4215 implementation with a new version that just does either a direct
4216 move or a regular move. Move memory interface to separate insns.
4217 Move insns so they are next to the expander.
4218 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4219 with signbit move. Split big and little endian case.
4220 (signbit<mode>2_dm_mem_le): Likewise.
4221 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4222 (signbit<mode>2_dm2): Likewise.
4224 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4226 * config/rl78/rl78.md (anddi3): New define_expand.
4228 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4230 * config/rl78/rl78.md (umindi3): New define_expand.
4232 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4234 * config/rl78/rl78.md (smindi3): New define_expand.
4236 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4238 * config/rl78/rl78.md (smaxdi3): New define_expand.
4240 2018-01-22 Carl Love <cel@us.ibm.com>
4242 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4243 LVX_V1TI): Add macro expansion.
4244 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4245 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4246 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4247 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4248 Change check to determine if the instruction is a byte reversing
4249 entry. Fix typo in comment.
4250 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4251 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4252 Add def_builtin calls for new builtins.
4253 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4254 Add define_insn expansion.
4256 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4258 * config/rl78/rl78.md (umaxdi3): New define_expand.
4260 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4262 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4263 for non-QImode registers.
4265 2018-01-22 Richard Biener <rguenther@suse.de>
4267 PR tree-optimization/83963
4268 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4269 including the loop exit block.
4270 (scop_detection::merge_sese): Likewise.
4271 (scop_detection::add_scop): Do it here instead.
4273 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4275 * doc/sourcebuild.texi (arm_softfloat): Document.
4277 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4280 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4281 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4282 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4284 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4285 David Edelsohn <dje.gcc@gmail.com>
4288 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4289 Change "crset eq" to "crset 2".
4290 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4291 (*call_indirect_aix<mode>_nospec): Likewise.
4292 (*call_value_indirect_aix<mode>_nospec): Likewise.
4293 (*call_indirect_elfv2<mode>_nospec): Likewise.
4294 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4295 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4296 change assembly output from . to $.
4297 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4298 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4299 (*tablejump<mode>_internal1_nospec): Likewise.
4301 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4304 * config/sh/sh_optimize_sett_clrt.cc:
4305 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4307 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4309 PR tree-optimization/83940
4310 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4311 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4312 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4313 use it to pass back the definition type.
4314 (vect_check_store_rhs): Likewise rhs_dt_out.
4315 (vect_build_gather_load_calls): Add a mask_dt argument and use
4316 it instead of a call to vect_is_simple_use.
4317 (vectorizable_store): Update calls to vect_check_load_store_mask
4318 and vect_check_store_rhs. Use the dt returned by the latter instead
4319 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4320 instead of calls to vect_is_simple_use. Pass the scalar rather
4321 than the vector operand to vect_is_simple_use when handling
4322 second and subsequent copies of an rhs value.
4323 (vectorizable_load): Update calls to vect_check_load_store_mask
4324 and vect_build_gather_load_calls. Use the cached mask_dt and
4325 gs_info.offset_dt instead of calls to vect_is_simple_use.
4327 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4330 * tree-emutls.c: Include gimplify.h.
4331 (lower_emutls_2): New function.
4332 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4333 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4334 it before further processing.
4337 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4338 UINTVAL (trueop1) instead of INTVAL (op1).
4340 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4344 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4345 INCOMING_FRAME_SP_OFFSET if not defined.
4346 (scan_trace): Add ENTRY argument. If true and
4347 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4348 emit a note to adjust the CFA offset.
4349 (create_cfi_notes): Adjust scan_trace callers.
4350 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4351 INCOMING_FRAME_SP_OFFSET in the CIE.
4352 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4353 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4355 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4356 * doc/tm.texi: Regenerated.
4358 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4360 PR rtl-optimization/83147
4361 * lra-constraints.c (remove_inheritance_pseudos): Use
4362 lra_substitute_pseudo_within_insn.
4364 2018-01-19 Tom de Vries <tom@codesourcery.com>
4365 Cesar Philippidis <cesar@codesourcery.com>
4368 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4370 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4373 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4374 spaces for function labels.
4376 2018-01-19 Martin Liska <mliska@suse.cz>
4378 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4379 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4380 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4381 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4382 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4383 (PRED_CONST_RETURN): Change from 69 to 65.
4384 (PRED_NULL_RETURN): Change from 91 to 71.
4385 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4386 (PRED_LOOP_GUARD): Change from 66 to 73.
4388 2018-01-19 Martin Liska <mliska@suse.cz>
4390 * predict.c (predict_insn_def): Add new assert.
4391 (struct branch_predictor): Change type to signed integer.
4392 (test_prediction_value_range): Amend test to cover
4394 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4395 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4396 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4397 (PRED_LOOP_IV_COMPARE): Likewise.
4398 * predict.h (PROB_UNINITIALIZED): Define new constant.
4400 2018-01-19 Martin Liska <mliska@suse.cz>
4402 * predict.c (dump_prediction): Add new format for
4403 analyze_brprob.py script which is enabled with -details
4405 * profile-count.h (precise_p): New function.
4407 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4409 PR tree-optimization/83922
4410 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4411 there are no statements that need masking.
4412 (vect_active_double_reduction_p): New function.
4413 (vect_analyze_loop_operations): Use it when handling phis that
4414 are not in the loop header.
4416 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4418 PR tree-optimization/83914
4419 * tree-vect-loop.c (vectorizable_induction): Don't convert
4420 init_expr or apply the peeling adjustment for inductions
4421 that are nested within the vectorized loop.
4423 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4425 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4428 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4432 * function.h (gimplify_parameters): Add gimple_seq * argument.
4433 * function.c: Include gimple.h and options.h.
4434 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4435 for the added local temporaries if needed.
4436 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4437 if there are any parameter cleanups, wrap whole body into a
4438 try/finally with the cleanups.
4440 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4443 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4444 Use GET_MODE_CLASS for scalar floating point.
4446 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4450 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4451 Fix call of call_cgraph_insertion_hooks.
4453 2018-01-18 Martin Sebor <msebor@redhat.com>
4455 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4457 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4460 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4463 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4466 * common.opt: (-ffile-prefix-map): New option.
4467 * opts.c (common_handle_option): Defer it.
4468 * opts-global.c (handle_common_deferred_options): Handle it.
4469 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4470 * file-prefix-map.h: New file.
4471 (remap_debug_filename, add_debug_prefix_map): ...here.
4472 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4473 * final.c (debug_prefix_map, add_debug_prefix_map
4474 remap_debug_filename): Move to...
4475 * file-prefix-map.c: New file.
4476 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4477 generalize, get rid of alloca(), use strrchr() instead of strchr().
4478 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4479 Implement in terms of add_prefix_map().
4480 (remap_macro_filename, remap_debug_filename): Implement in term of
4482 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4483 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4484 * dbxout.c: Include file-prefix-map.h.
4485 * varasm.c: Likewise.
4486 * vmsdbgout.c: Likewise.
4487 * xcoffout.c: Likewise.
4488 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4489 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4490 * doc/invoke.texi (-ffile-prefix-map): Document.
4491 (-fdebug-prefix-map): Update description.
4493 2018-01-18 Martin Liska <mliska@suse.cz>
4495 * config/i386/i386.c (indirect_thunk_name): Document that also
4497 (output_indirect_thunk): Document why both instructions
4498 (pause and lfence) are generated.
4500 2018-01-18 Richard Biener <rguenther@suse.de>
4502 PR tree-optimization/83887
4503 * graphite-scop-detection.c
4504 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4505 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4506 (scop_detection::merge_sese): Re-implement with a flood-fill
4507 algorithm that properly finds a SESE region if it exists.
4509 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4512 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4513 pointer_diff optimizations use view_convert instead of convert.
4515 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4517 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4518 Generate different code for -mno-speculate-indirect-jumps.
4519 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4520 (*call_indirect_aix<mode>): Disable for
4521 -mno-speculate-indirect-jumps.
4522 (*call_indirect_aix<mode>_nospec): New define_insn.
4523 (*call_value_indirect_aix<mode>): Disable for
4524 -mno-speculate-indirect-jumps.
4525 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4526 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4527 -mno-speculate-indirect-jumps.
4528 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4530 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4532 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4533 long double type, set the flags for noting the default long double
4534 type, even if we don't pass or return a long double type.
4536 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4539 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4542 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4544 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4545 support for merge[hl].
4546 (fold_mergehl_helper): New helper function.
4547 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4548 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4549 (altivec_vmrglw_direct): Add xxmrglw insn.
4551 2018-01-17 Andrew Waterman <andrew@sifive.com>
4553 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4554 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4556 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4559 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4560 call the lto_location_cache before reading the
4561 DECL_SOURCE_LOCATION of the types.
4563 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4564 Richard Sandiford <richard.sandiford@linaro.org>
4566 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4567 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4568 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4569 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4570 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4572 * config/aarch64/constraints.md (aarch64_movti_operand):
4574 * config/aarch64/predicates.md (Uti): Add new constraint.
4576 2018-01-17 Carl Love <cel@us.ibm.com>
4578 * config/rs6000/vsx.md (define_expand xl_len_r,
4579 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4580 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4582 (define_expand, define_insn): Move the shift left from the
4583 define_insn to the define_expand for lxvl and stxvl instructions.
4584 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4585 and XL_LEN_R definitions to PURE.
4587 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4589 * config/i386/i386.c (indirect_thunk_name): Declare regno
4590 as unsigned int. Compare regno with INVALID_REGNUM.
4591 (output_indirect_thunk): Ditto.
4592 (output_indirect_thunk_function): Ditto.
4593 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4594 in the call to output_indirect_thunk_function.
4596 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4599 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4600 rather than the size of inner_type to determine the stack slot size
4601 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4603 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4606 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4609 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4611 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4612 endian Linux systems to optionally enable multilibs for selecting
4613 the long double type if the user configured an explicit type.
4614 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4615 have no long double multilibs if not defined.
4616 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4617 warn if the user used -mabi={ieee,ibm}longdouble and we built
4618 multilibs for long double.
4619 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4620 appropriate multilib option.
4621 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4623 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4624 for building long double multilibs.
4625 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4627 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4629 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4632 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4634 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4637 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4640 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4643 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4645 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4646 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4648 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4650 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4651 different rtl trees depending on TARGET_64BIT.
4652 (rs6000_gen_lvx): Likewise.
4654 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4656 * config/visium/visium.md (nop): Tweak comment.
4657 (hazard_nop): Likewise.
4659 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4661 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4662 -mspeculate-indirect-jumps.
4663 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4664 for -mno-speculate-indirect-jumps.
4665 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4666 (*call_value_indirect_elfv2<mode>): Disable for
4667 -mno-speculate-indirect-jumps.
4668 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4669 (indirect_jump): Emit different RTL for
4670 -mno-speculate-indirect-jumps.
4671 (*indirect_jump<mode>): Disable for
4672 -mno-speculate-indirect-jumps.
4673 (*indirect_jump<mode>_nospec): New define_insn.
4674 (tablejump): Emit different RTL for
4675 -mno-speculate-indirect-jumps.
4676 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4677 (tablejumpsi_nospec): New define_expand.
4678 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4679 (tablejumpdi_nospec): New define_expand.
4680 (*tablejump<mode>_internal1): Disable for
4681 -mno-speculate-indirect-jumps.
4682 (*tablejump<mode>_internal1_nospec): New define_insn.
4683 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4686 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4688 * caller-save.c (insert_save): Drop unnecessary parameter. All
4691 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4692 Richard Biener <rguenth@suse.de>
4695 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4696 return early, inline manually is_gimple_sizepos. Make sure if we
4697 call gimplify_expr we don't end up with a gimple constant.
4698 * tree.c (variably_modified_type_p): Don't return true for
4699 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4700 * gimplify.h (is_gimple_sizepos): Remove.
4702 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4704 PR tree-optimization/83857
4705 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4706 vectorizable_live_operation for pure SLP statements.
4707 (vectorizable_live_operation): Handle PHIs.
4709 2018-01-16 Richard Biener <rguenther@suse.de>
4711 PR tree-optimization/83867
4712 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4713 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4715 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4718 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4719 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4720 If off is not INTEGER_CST, issue a may not be aligned warning
4721 rather than isn't aligned. Use isn%'t rather than isn't.
4722 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4724 <case MULT_EXPR>: Improve the case when bottom and one of the
4725 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4726 operand, in that case check if the other operand is multiple of
4727 bottom divided by the INTEGER_CST operand.
4729 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4732 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4733 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4734 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4735 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4736 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4737 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4738 (pa_function_arg_size): New function.
4740 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4742 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4743 in a separate statement.
4745 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4747 PR tree-optimization/83847
4748 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4749 group gathers and scatters.
4751 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4753 PR rtl-optimization/86620
4754 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4756 PR rtl-optimization/83213
4757 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4758 to last if both are JUMP_INSNs.
4760 PR tree-optimization/83843
4761 * gimple-ssa-store-merging.c
4762 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4763 store_immediate_info for bswap/nop orig_stores.
4765 2018-01-15 Andrew Waterman <andrew@sifive.com>
4767 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4769 <UDIV>: Increase cost if !TARGET_DIV.
4771 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4773 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4774 (define_attr "cr_logical_3op"): New.
4775 (cceq_ior_compare): Adjust.
4776 (cceq_ior_compare_complement): Adjust.
4777 (*cceq_rev_compare): Adjust.
4778 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4779 (is_cracked_insn): Adjust.
4780 (insn_must_be_first_in_group): Adjust.
4781 * config/rs6000/40x.md: Adjust.
4782 * config/rs6000/440.md: Adjust.
4783 * config/rs6000/476.md: Adjust.
4784 * config/rs6000/601.md: Adjust.
4785 * config/rs6000/603.md: Adjust.
4786 * config/rs6000/6xx.md: Adjust.
4787 * config/rs6000/7450.md: Adjust.
4788 * config/rs6000/7xx.md: Adjust.
4789 * config/rs6000/8540.md: Adjust.
4790 * config/rs6000/cell.md: Adjust.
4791 * config/rs6000/e300c2c3.md: Adjust.
4792 * config/rs6000/e500mc.md: Adjust.
4793 * config/rs6000/e500mc64.md: Adjust.
4794 * config/rs6000/e5500.md: Adjust.
4795 * config/rs6000/e6500.md: Adjust.
4796 * config/rs6000/mpc.md: Adjust.
4797 * config/rs6000/power4.md: Adjust.
4798 * config/rs6000/power5.md: Adjust.
4799 * config/rs6000/power6.md: Adjust.
4800 * config/rs6000/power7.md: Adjust.
4801 * config/rs6000/power8.md: Adjust.
4802 * config/rs6000/power9.md: Adjust.
4803 * config/rs6000/rs64.md: Adjust.
4804 * config/rs6000/titan.md: Adjust.
4806 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4808 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4809 ix86_indirect_branch_register logic.
4811 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4813 * config/i386/constraints.md (Bs): Update
4814 ix86_indirect_branch_register check. Don't check
4815 ix86_indirect_branch_register with GOT_memory_operand.
4817 * config/i386/predicates.md (GOT_memory_operand): Don't check
4818 ix86_indirect_branch_register here.
4819 (GOT32_symbol_operand): Likewise.
4821 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4823 * config/i386/predicates.md (constant_call_address_operand):
4824 Rewrite ix86_indirect_branch_register logic.
4825 (sibcall_insn_operand): Likewise.
4827 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4829 * config/i386/constraints.md (Bs): Replace
4830 ix86_indirect_branch_thunk_register with
4831 ix86_indirect_branch_register.
4833 * config/i386/i386.md (indirect_jump): Likewise.
4834 (tablejump): Likewise.
4835 (*sibcall_memory): Likewise.
4836 (*sibcall_value_memory): Likewise.
4837 Peepholes of indirect call and jump via memory: Likewise.
4838 * config/i386/i386.opt: Likewise.
4839 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4840 (GOT_memory_operand): Likewise.
4841 (call_insn_operand): Likewise.
4842 (sibcall_insn_operand): Likewise.
4843 (GOT32_symbol_operand): Likewise.
4845 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4848 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4849 type rather than type addr's type points to.
4850 (expand_omp_atomic_mutex): Likewise.
4851 (expand_omp_atomic): Likewise.
4853 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4856 * config/i386/i386.c (output_indirect_thunk_function): Use
4857 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4858 for __x86_return_thunk.
4860 2018-01-15 Richard Biener <rguenther@suse.de>
4863 * expmed.c (extract_bit_field_1): Fix typo.
4865 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4868 * config/arm/iterators.md (VF): New mode iterator.
4869 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4870 Remove integer-related logic from pattern.
4871 (neon_vabd<mode>_3): Likewise.
4873 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4876 * common.opt (fstrict-overflow): No longer an alias.
4877 (fwrapv-pointer): New option.
4878 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4879 also for pointer types based on flag_wrapv_pointer.
4880 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4881 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4882 opts->x_flag_wrapv got set.
4883 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4884 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4885 POINTER_TYPE_OVERFLOW_UNDEFINED.
4886 * match.pd: Likewise in address comparison pattern.
4887 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4889 2018-01-15 Richard Biener <rguenther@suse.de>
4892 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4893 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4894 Reset type names to their identifier if their TYPE_DECL doesn't
4895 have linkage (and thus is used for ODR and devirt).
4896 (save_debug_info_for_decl): Remove.
4897 (save_debug_info_for_type): Likewise.
4898 (add_tree_to_fld_list): Adjust.
4899 * tree-pretty-print.c (dump_generic_node): Make dumping of
4900 type names more robust.
4902 2018-01-15 Richard Biener <rguenther@suse.de>
4904 * BASE-VER: Bump to 8.0.1.
4906 2018-01-14 Martin Sebor <msebor@redhat.com>
4909 * builtins.c (check_access): Avoid warning when the no-warning bit
4912 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4914 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4915 * ira-color (allocno_hard_regs_compare): Likewise.
4917 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4920 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4921 Use .pushsection/.popsection.
4923 2018-01-14 Martin Sebor <msebor@redhat.com>
4926 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4928 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4930 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4931 entry from extra_headers.
4932 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4933 extra_headers, make the list bitwise identical to the i?86-*-* one.
4935 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4937 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4938 -mcmodel=large with -mindirect-branch=thunk,
4939 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4940 -mfunction-return=thunk-extern.
4941 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4942 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4943 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4945 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4947 * config/i386/i386.c (print_reg): Print the name of the full
4948 integer register without '%'.
4949 (ix86_print_operand): Handle 'V'.
4950 * doc/extend.texi: Document 'V' modifier.
4952 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4954 * config/i386/constraints.md (Bs): Disallow memory operand for
4955 -mindirect-branch-register.
4957 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4958 (GOT_memory_operand): Likewise.
4959 (call_insn_operand): Likewise.
4960 (sibcall_insn_operand): Likewise.
4961 (GOT32_symbol_operand): Likewise.
4962 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4963 for -mindirect-branch-register.
4964 (tablejump): Likewise.
4965 (*sibcall_memory): Likewise.
4966 (*sibcall_value_memory): Likewise.
4967 Disallow peepholes of indirect call and jump via memory for
4968 -mindirect-branch-register.
4969 (*call_pop): Replace m with Bw.
4970 (*call_value_pop): Likewise.
4971 (*sibcall_pop_memory): Replace m with Bs.
4972 * config/i386/i386.opt (mindirect-branch-register): New option.
4973 * doc/invoke.texi: Document -mindirect-branch-register option.
4975 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4977 * config/i386/i386-protos.h (ix86_output_function_return): New.
4978 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4979 set function_return_type.
4980 (indirect_thunk_name): Add ret_p to indicate thunk for function
4982 (output_indirect_thunk_function): Pass false to
4983 indirect_thunk_name.
4984 (ix86_output_indirect_branch_via_reg): Likewise.
4985 (ix86_output_indirect_branch_via_push): Likewise.
4986 (output_indirect_thunk_function): Create alias for function
4987 return thunk if regno < 0.
4988 (ix86_output_function_return): New function.
4989 (ix86_handle_fndecl_attribute): Handle function_return.
4990 (ix86_attribute_table): Add function_return.
4991 * config/i386/i386.h (machine_function): Add
4992 function_return_type.
4993 * config/i386/i386.md (simple_return_internal): Use
4994 ix86_output_function_return.
4995 (simple_return_internal_long): Likewise.
4996 * config/i386/i386.opt (mfunction-return=): New option.
4997 (indirect_branch): Mention -mfunction-return=.
4998 * doc/extend.texi: Document function_return function attribute.
4999 * doc/invoke.texi: Document -mfunction-return= option.
5001 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5003 * config/i386/i386-opts.h (indirect_branch): New.
5004 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5005 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5006 with local indirect jump when converting indirect call and jump.
5007 (ix86_set_indirect_branch_type): New.
5008 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5009 (indirectlabelno): New.
5010 (indirect_thunk_needed): Likewise.
5011 (indirect_thunk_bnd_needed): Likewise.
5012 (indirect_thunks_used): Likewise.
5013 (indirect_thunks_bnd_used): Likewise.
5014 (INDIRECT_LABEL): Likewise.
5015 (indirect_thunk_name): Likewise.
5016 (output_indirect_thunk): Likewise.
5017 (output_indirect_thunk_function): Likewise.
5018 (ix86_output_indirect_branch_via_reg): Likewise.
5019 (ix86_output_indirect_branch_via_push): Likewise.
5020 (ix86_output_indirect_branch): Likewise.
5021 (ix86_output_indirect_jmp): Likewise.
5022 (ix86_code_end): Call output_indirect_thunk_function if needed.
5023 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5025 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5026 (ix86_attribute_table): Add indirect_branch.
5027 * config/i386/i386.h (machine_function): Add indirect_branch_type
5028 and has_local_indirect_jump.
5029 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5031 (tablejump): Likewise.
5032 (*indirect_jump): Use ix86_output_indirect_jmp.
5033 (*tablejump_1): Likewise.
5034 (simple_return_indirect_internal): Likewise.
5035 * config/i386/i386.opt (mindirect-branch=): New option.
5036 (indirect_branch): New.
5039 (thunk-inline): Likewise.
5040 (thunk-extern): Likewise.
5041 * doc/extend.texi: Document indirect_branch function attribute.
5042 * doc/invoke.texi: Document -mindirect-branch= option.
5044 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5047 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5049 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5051 * ipa-inline.c (want_inline_small_function_p): Return false if
5052 inlining has already failed with CIF_FINAL_ERROR.
5053 (update_caller_keys): Call want_inline_small_function_p before
5055 (update_callee_keys): Likewise.
5057 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5059 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5061 (rs6000_quadword_masked_address_p): Likewise.
5062 (quad_aligned_load_p): Likewise.
5063 (quad_aligned_store_p): Likewise.
5064 (const_load_sequence_p): Add comment to describe the outer-most loop.
5065 (mimic_memory_attributes_and_flags): New function.
5066 (rs6000_gen_stvx): Likewise.
5067 (replace_swapped_aligned_store): Likewise.
5068 (rs6000_gen_lvx): Likewise.
5069 (replace_swapped_aligned_load): Likewise.
5070 (replace_swapped_load_constant): Capitalize argument name in
5071 comment describing this function.
5072 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5073 and stores that access quad-word aligned addresses and replace
5074 with stvx or lvx instructions when appropriate.
5075 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5076 New function prototype.
5077 (rs6000_quadword_masked_address_p): Likewise.
5078 (rs6000_gen_lvx): Likewise.
5079 (rs6000_gen_stvx): Likewise.
5080 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5081 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5082 when memory address is aligned.
5083 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5084 this split to select lvx instruction when memory address is aligned.
5085 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5086 instruction when memory address is aligned.
5087 (*vsx_le_perm_load_v16qi): Likewise.
5088 (four unnamed splitters): Modify to select the stvx instruction
5089 when memory is aligned.
5091 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5093 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5094 which appears in the queue multiple times.
5096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5097 Alan Hayward <alan.hayward@arm.com>
5098 David Sherwood <david.sherwood@arm.com>
5100 * tree-vectorizer.h (vec_lower_bound): New structure.
5101 (_loop_vec_info): Add check_nonzero and lower_bounds.
5102 (LOOP_VINFO_CHECK_NONZERO): New macro.
5103 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5104 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5105 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5106 fields. Make seg_len the distance travelled, not including the
5108 (dr_direction_indicator): Declare.
5109 (dr_zero_step_indicator): Likewise.
5110 (dr_known_forward_stride_p): Likewise.
5111 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5113 (runtime_alias_check_p): Allow runtime alias checks with
5115 (operator ==): Compare access_size and align.
5116 (prune_runtime_alias_test_list): Rework for new distinction between
5117 the access_size and seg_len.
5118 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5120 (get_segment_min_max): New function.
5121 (create_intersect_range_checks): Use it.
5122 (dr_step_indicator): New function.
5123 (dr_direction_indicator): Likewise.
5124 (dr_zero_step_indicator): Likewise.
5125 (dr_known_forward_stride_p): Likewise.
5126 * tree-loop-distribution.c (data_ref_segment_size): Return
5127 DR_STEP * (niters - 1).
5128 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5130 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5131 (vect_preserves_scalar_order_p): New function, split out from...
5132 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5133 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5134 (vect_vfa_access_size): New function.
5135 (vect_vfa_align): Likewise.
5136 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5137 (dump_lower_bound): New function.
5138 (vect_check_lower_bound): Likewise.
5139 (vect_small_gap_p): Likewise.
5140 (vectorizable_with_step_bound_p): Likewise.
5141 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5142 depencies if the vectorization factor is 1. Convert the checks
5143 for nonzero steps into checks on the bounds of DR_STEP. Try using
5144 a bunds check for variable steps if the minimum required step is
5145 relatively small. Update calls to the dr_with_seg_len
5146 constructor and to vect_compile_time_alias.
5147 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5149 (vect_loop_versioning): Call it.
5150 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5152 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5155 Alan Hayward <alan.hayward@arm.com>
5156 David Sherwood <david.sherwood@arm.com>
5158 * doc/sourcebuild.texi (vect_scatter_store): Document.
5159 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5161 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5163 * genopinit.c (main): Add supports_vec_scatter_store and
5164 supports_vec_scatter_store_cached to target_optabs.
5165 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5166 IFN_MASK_SCATTER_STORE.
5167 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5169 * internal-fn.h (internal_store_fn_p): Declare.
5170 (internal_fn_stored_value_index): Likewise.
5171 * internal-fn.c (scatter_store_direct): New macro.
5172 (expand_scatter_store_optab_fn): New function.
5173 (direct_scatter_store_optab_supported_p): New macro.
5174 (internal_store_fn_p): New function.
5175 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5176 IFN_MASK_SCATTER_STORE.
5177 (internal_fn_mask_index): Likewise.
5178 (internal_fn_stored_value_index): New function.
5179 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5181 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5182 * optabs-query.c (supports_vec_scatter_store_p): New function.
5183 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5184 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5185 true for scatter stores.
5186 (vect_gather_scatter_fn_p): Handle scatter stores too.
5187 (vect_check_gather_scatter): Consider using scatter stores if
5188 supports_vec_scatter_store_p.
5189 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5191 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5192 internal_fn_stored_value_index.
5193 (check_load_store_masking): Handle scatter stores too.
5194 (vect_get_store_rhs): Make public.
5195 (vectorizable_call): Use internal_store_fn_p.
5196 (vectorizable_store): Handle scatter store internal functions.
5197 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5198 when deciding whether the end of the group has been reached.
5199 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5200 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5201 (mask_scatter_store<mode>): New insns.
5203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5204 Alan Hayward <alan.hayward@arm.com>
5205 David Sherwood <david.sherwood@arm.com>
5207 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5208 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5209 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5211 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5212 Use vect_truncate_gather_scatter_offset if we can't treat the
5213 operation as a normal gather load or scatter store.
5214 (get_group_load_store_type): Take the gather_scatter_info
5215 as argument. Try using a gather load or scatter store for
5216 single-element groups.
5217 (get_load_store_type): Update calls to get_group_load_store_type
5218 and vect_use_strided_gather_scatters_p.
5220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5221 Alan Hayward <alan.hayward@arm.com>
5222 David Sherwood <david.sherwood@arm.com>
5224 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5225 optional tree argument.
5226 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5228 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5229 but continue to use the current value as a fallback.
5230 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5231 to compare the updates.
5232 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5233 (get_load_store_type): Use it when handling a strided access.
5234 (vect_get_strided_load_store_ops): New function.
5235 (vect_get_data_ptr_increment): Likewise.
5236 (vectorizable_load): Handle strided gather loads. Always pass
5237 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5240 Alan Hayward <alan.hayward@arm.com>
5241 David Sherwood <david.sherwood@arm.com>
5243 * doc/md.texi (gather_load@var{m}): Document.
5244 (mask_gather_load@var{m}): Likewise.
5245 * genopinit.c (main): Add supports_vec_gather_load and
5246 supports_vec_gather_load_cached to target_optabs.
5247 * optabs-tree.c (init_tree_optimization_optabs): Use
5248 ggc_cleared_alloc to allocate target_optabs.
5249 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5250 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5252 * internal-fn.h (internal_load_fn_p): Declare.
5253 (internal_gather_scatter_fn_p): Likewise.
5254 (internal_fn_mask_index): Likewise.
5255 (internal_gather_scatter_fn_supported_p): Likewise.
5256 * internal-fn.c (gather_load_direct): New macro.
5257 (expand_gather_load_optab_fn): New function.
5258 (direct_gather_load_optab_supported_p): New macro.
5259 (direct_internal_fn_optab): New function.
5260 (internal_load_fn_p): Likewise.
5261 (internal_gather_scatter_fn_p): Likewise.
5262 (internal_fn_mask_index): Likewise.
5263 (internal_gather_scatter_fn_supported_p): Likewise.
5264 * optabs-query.c (supports_at_least_one_mode_p): New function.
5265 (supports_vec_gather_load_p): Likewise.
5266 * optabs-query.h (supports_vec_gather_load_p): Declare.
5267 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5268 and memory_type field.
5269 (NUM_PATTERNS): Bump to 15.
5270 * tree-vect-data-refs.c: Include internal-fn.h.
5271 (vect_gather_scatter_fn_p): New function.
5272 (vect_describe_gather_scatter_call): Likewise.
5273 (vect_check_gather_scatter): Try using internal functions for
5274 gather loads. Recognize existing calls to a gather load function.
5275 (vect_analyze_data_refs): Consider using gather loads if
5276 supports_vec_gather_load_p.
5277 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5278 (vect_get_gather_scatter_offset_type): Likewise.
5279 (vect_convert_mask_for_vectype): Likewise.
5280 (vect_add_conversion_to_patterm): Likewise.
5281 (vect_try_gather_scatter_pattern): Likewise.
5282 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5283 (vect_vect_recog_func_ptrs): Add it.
5284 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5285 internal_fn_mask_index and internal_gather_scatter_fn_p.
5286 (check_load_store_masking): Take the gather_scatter_info as an
5287 argument and handle gather loads.
5288 (vect_get_gather_scatter_ops): New function.
5289 (vectorizable_call): Check internal_load_fn_p.
5290 (vectorizable_load): Likewise. Handle gather load internal
5292 (vectorizable_store): Update call to check_load_store_masking.
5293 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5294 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5295 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5296 (aarch64_gather_scale_operand_d): New predicates.
5297 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5298 (mask_gather_load<mode>): New insns.
5300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5301 Alan Hayward <alan.hayward@arm.com>
5302 David Sherwood <david.sherwood@arm.com>
5304 * optabs.def (fold_left_plus_optab): New optab.
5305 * doc/md.texi (fold_left_plus_@var{m}): Document.
5306 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5307 * internal-fn.c (fold_left_direct): Define.
5308 (expand_fold_left_optab_fn): Likewise.
5309 (direct_fold_left_optab_supported_p): Likewise.
5310 * fold-const-call.c (fold_const_fold_left): New function.
5311 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5312 * tree-parloops.c (valid_reduction_p): New function.
5313 (gather_scalar_reductions): Use it.
5314 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5315 (vect_finish_replace_stmt): Declare.
5316 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5317 (needs_fold_left_reduction_p): New function, split out from...
5318 (vect_is_simple_reduction): ...here. Accept reductions that
5319 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5320 (vect_force_simple_reduction): Also store the reduction type in
5321 the assignment's STMT_VINFO_REDUC_TYPE.
5322 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5323 (merge_with_identity): New function.
5324 (vect_expand_fold_left): Likewise.
5325 (vectorize_fold_left_reduction): Likewise.
5326 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5327 scalar phi in place for it. Check for target support and reject
5328 cases that would reassociate the operation. Defer the transform
5329 phase to vectorize_fold_left_reduction.
5330 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5331 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5332 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5336 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5337 call to ifc_temp_var.
5339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5340 Alan Hayward <alan.hayward@arm.com>
5341 David Sherwood <david.sherwood@arm.com>
5343 * target.def (legitimize_address_displacement): Take the original
5344 offset as a poly_int.
5345 * targhooks.h (default_legitimize_address_displacement): Update
5347 * targhooks.c (default_legitimize_address_displacement): Likewise.
5348 * doc/tm.texi: Regenerate.
5349 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5350 as an argument, moving assert of ad->disp == ad->disp_term to...
5351 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5352 Try calling targetm.legitimize_address_displacement before expanding
5353 the address rather than afterwards, and adjust for the new interface.
5354 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5355 Match the new hook interface. Handle SVE addresses.
5356 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5361 * Makefile.in (OBJS): Add early-remat.o.
5362 * target.def (select_early_remat_modes): New hook.
5363 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5364 * doc/tm.texi: Regenerate.
5365 * targhooks.h (default_select_early_remat_modes): Declare.
5366 * targhooks.c (default_select_early_remat_modes): New function.
5367 * timevar.def (TV_EARLY_REMAT): New timevar.
5368 * passes.def (pass_early_remat): New pass.
5369 * tree-pass.h (make_pass_early_remat): Declare.
5370 * early-remat.c: New file.
5371 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5373 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5376 Alan Hayward <alan.hayward@arm.com>
5377 David Sherwood <david.sherwood@arm.com>
5379 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5380 vfm1 with a bound_epilog parameter.
5381 (vect_do_peeling): Update calls accordingly, and move the prologue
5382 call earlier in the function. Treat the base bound_epilog as 0 for
5383 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5384 this base when peeling for gaps.
5385 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5386 with fully-masked loops.
5387 (vect_estimate_min_profitable_iters): Handle the single peeled
5388 iteration in that case.
5390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5391 Alan Hayward <alan.hayward@arm.com>
5392 David Sherwood <david.sherwood@arm.com>
5394 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5395 single-element interleaving even if the size is not a power of 2.
5396 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5397 accesses for single-element interleaving if the group size is
5400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5401 Alan Hayward <alan.hayward@arm.com>
5402 David Sherwood <david.sherwood@arm.com>
5404 * doc/md.texi (fold_extract_last_@var{m}): Document.
5405 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5406 * optabs.def (fold_extract_last_optab): New optab.
5407 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5408 * internal-fn.c (fold_extract_direct): New macro.
5409 (expand_fold_extract_optab_fn): Likewise.
5410 (direct_fold_extract_optab_supported_p): Likewise.
5411 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5412 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5413 EXTRACT_LAST_REDUCTION.
5414 (get_initial_def_for_reduction): Do not create an initial vector
5415 for EXTRACT_LAST_REDUCTION reductions.
5416 (vectorizable_reduction): Leave the scalar phi in place for
5417 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5418 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5419 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5420 transform phase to vectorizable_condition.
5421 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5423 (vect_finish_stmt_generation): ...here.
5424 (vect_finish_replace_stmt): New function.
5425 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5426 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5428 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5431 Alan Hayward <alan.hayward@arm.com>
5432 David Sherwood <david.sherwood@arm.com>
5434 * doc/md.texi (extract_last_@var{m}): Document.
5435 * optabs.def (extract_last_optab): New optab.
5436 * internal-fn.def (EXTRACT_LAST): New internal function.
5437 * internal-fn.c (cond_unary_direct): New macro.
5438 (expand_cond_unary_optab_fn): Likewise.
5439 (direct_cond_unary_optab_supported_p): Likewise.
5440 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5441 loops using EXTRACT_LAST.
5442 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5443 (extract_last_<mode>): ...this optab.
5444 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
5451 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5452 * doc/tm.texi: Regenerate.
5453 * targhooks.h (default_empty_mask_is_expensive): Declare.
5454 * targhooks.c (default_empty_mask_is_expensive): New function.
5455 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5456 if the target says that empty masks are expensive.
5457 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5459 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5462 Alan Hayward <alan.hayward@arm.com>
5463 David Sherwood <david.sherwood@arm.com>
5465 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5466 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5467 (vect_use_loop_mask_for_alignment_p): New function.
5468 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5469 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5470 niters_skip argument. Make sure that the first niters_skip elements
5471 of the first iteration are inactive.
5472 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5473 Update call to vect_set_loop_masks_directly.
5474 (get_misalign_in_elems): New function, split out from...
5475 (vect_gen_prolog_loop_niters): ...here.
5476 (vect_update_init_of_dr): Take a code argument that specifies whether
5477 the adjustment should be added or subtracted.
5478 (vect_update_init_of_drs): Likewise.
5479 (vect_prepare_for_masked_peels): New function.
5480 (vect_do_peeling): Skip prologue peeling if we're using a mask
5481 instead. Update call to vect_update_inits_of_drs.
5482 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5484 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5485 alignment. Do not include the number of peeled iterations in
5486 the minimum threshold in that case.
5487 (vectorizable_induction): Adjust the start value down by
5488 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5489 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5490 Take the number of skipped iterations into account when calculating
5492 * tree-vect-stmts.c (vect_gen_while_not): New function.
5494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5495 Alan Hayward <alan.hayward@arm.com>
5496 David Sherwood <david.sherwood@arm.com>
5498 * doc/sourcebuild.texi (vect_fully_masked): Document.
5499 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5501 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5503 (vect_analyze_loop_2): ...here. Don't check the vectorization
5504 factor against the number of loop iterations if the loop is
5507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5508 Alan Hayward <alan.hayward@arm.com>
5509 David Sherwood <david.sherwood@arm.com>
5511 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5512 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5513 (dump_groups): Update accordingly.
5514 (iv_use::mem_type): New member variable.
5515 (address_p): New function.
5516 (record_use): Add a mem_type argument and initialize the new
5518 (record_group_use): Add a mem_type argument. Use address_p.
5519 Remove obsolete null checks of base_object. Update call to record_use.
5520 (find_interesting_uses_op): Update call to record_group_use.
5521 (find_interesting_uses_cond): Likewise.
5522 (find_interesting_uses_address): Likewise.
5523 (get_mem_type_for_internal_fn): New function.
5524 (find_address_like_use): Likewise.
5525 (find_interesting_uses_stmt): Try find_address_like_use before
5526 calling find_interesting_uses_op.
5527 (addr_offset_valid_p): Use the iv mem_type field as the type
5528 of the addressed memory.
5529 (add_autoinc_candidates): Likewise.
5530 (get_address_cost): Likewise.
5531 (split_small_address_groups_p): Use address_p.
5532 (split_address_groups): Likewise.
5533 (add_iv_candidate_for_use): Likewise.
5534 (autoinc_possible_for_pair): Likewise.
5535 (rewrite_groups): Likewise.
5536 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5537 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5538 (get_alias_ptr_type_for_ptr_address): New function.
5539 (rewrite_use_address): Rewrite address uses in calls that were
5540 identified by find_address_like_use.
5542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5543 Alan Hayward <alan.hayward@arm.com>
5544 David Sherwood <david.sherwood@arm.com>
5546 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5548 * gimple-expr.h (is_gimple_addressable: Likewise.
5549 * gimple-expr.c (is_gimple_address): Likewise.
5550 * internal-fn.c (expand_call_mem_ref): New function.
5551 (expand_mask_load_optab_fn): Use it.
5552 (expand_mask_store_optab_fn): Likewise.
5554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5555 Alan Hayward <alan.hayward@arm.com>
5556 David Sherwood <david.sherwood@arm.com>
5558 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5559 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5560 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5561 (cond_umax@var{mode}): Document.
5562 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5563 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5564 (cond_umin_optab, cond_umax_optab): New optabs.
5565 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5566 (COND_IOR, COND_XOR): New internal functions.
5567 * internal-fn.h (get_conditional_internal_fn): Declare.
5568 * internal-fn.c (cond_binary_direct): New macro.
5569 (expand_cond_binary_optab_fn): Likewise.
5570 (direct_cond_binary_optab_supported_p): Likewise.
5571 (get_conditional_internal_fn): New function.
5572 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5573 Cope with reduction statements that are vectorized as calls rather
5575 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5576 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5577 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5578 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5579 (UNSPEC_COND_EOR): New unspecs.
5580 (optab): Add mappings for them.
5581 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5582 (sve_int_op, sve_fp_op): New int attributes.
5584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5585 Alan Hayward <alan.hayward@arm.com>
5586 David Sherwood <david.sherwood@arm.com>
5588 * optabs.def (while_ult_optab): New optab.
5589 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5590 * internal-fn.def (WHILE_ULT): New internal function.
5591 * internal-fn.h (direct_internal_fn_supported_p): New override
5592 that takes two types as argument.
5593 * internal-fn.c (while_direct): New macro.
5594 (expand_while_optab_fn): New function.
5595 (convert_optab_supported_p): Likewise.
5596 (direct_while_optab_supported_p): New macro.
5597 * wide-int.h (wi::udiv_ceil): New function.
5598 * tree-vectorizer.h (rgroup_masks): New structure.
5599 (vec_loop_masks): New typedef.
5600 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5602 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5603 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5604 (vect_max_vf): New function.
5605 (slpeel_make_loop_iterate_ntimes): Delete.
5606 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5607 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5608 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5609 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5610 internal-fn.h, stor-layout.h and optabs-query.h.
5611 (vect_set_loop_mask): New function.
5612 (add_preheader_seq): Likewise.
5613 (add_header_seq): Likewise.
5614 (interleave_supported_p): Likewise.
5615 (vect_maybe_permute_loop_masks): Likewise.
5616 (vect_set_loop_masks_directly): Likewise.
5617 (vect_set_loop_condition_masked): Likewise.
5618 (vect_set_loop_condition_unmasked): New function, split out from
5619 slpeel_make_loop_iterate_ntimes.
5620 (slpeel_make_loop_iterate_ntimes): Rename to..
5621 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5622 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5623 (vect_do_peeling): Update call accordingly.
5624 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5626 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5627 mask_compare_type, can_fully_mask_p and fully_masked_p.
5628 (release_vec_loop_masks): New function.
5629 (_loop_vec_info): Use it to free the loop masks.
5630 (can_produce_all_loop_masks_p): New function.
5631 (vect_get_max_nscalars_per_iter): Likewise.
5632 (vect_verify_full_masking): Likewise.
5633 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5634 retries, and free the mask rgroups before retrying. Check loop-wide
5635 reasons for disallowing fully-masked loops. Make the final decision
5636 about whether use a fully-masked loop or not.
5637 (vect_estimate_min_profitable_iters): Do not assume that peeling
5638 for the number of iterations will be needed for fully-masked loops.
5639 (vectorizable_reduction): Disable fully-masked loops.
5640 (vectorizable_live_operation): Likewise.
5641 (vect_halve_mask_nunits): New function.
5642 (vect_double_mask_nunits): Likewise.
5643 (vect_record_loop_mask): Likewise.
5644 (vect_get_loop_mask): Likewise.
5645 (vect_transform_loop): Handle the case in which the final loop
5646 iteration might handle a partial vector. Call vect_set_loop_condition
5647 instead of slpeel_make_loop_iterate_ntimes.
5648 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5649 (check_load_store_masking): New function.
5650 (prepare_load_store_mask): Likewise.
5651 (vectorizable_store): Handle fully-masked loops.
5652 (vectorizable_load): Likewise.
5653 (supportable_widening_operation): Use vect_halve_mask_nunits for
5655 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5656 (vect_gen_while): New function.
5657 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5658 (aarch64_uqdec<mode>): New insn.
5660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5661 Alan Hayward <alan.hayward@arm.com>
5662 David Sherwood <david.sherwood@arm.com>
5664 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5665 (reduc_xor_scal_optab): New optabs.
5666 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5667 (reduc_xor_scal_@var{m}): Document.
5668 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5669 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5671 * fold-const-call.c (fold_const_call): Handle them.
5672 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5673 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5674 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5675 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5676 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5677 (UNSPEC_XORV): New unspecs.
5678 (optab): Add entries for them.
5679 (BITWISEV): New int iterator.
5680 (bit_reduc_op): New int attributes.
5682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5683 Alan Hayward <alan.hayward@arm.com>
5684 David Sherwood <david.sherwood@arm.com>
5686 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5687 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5688 * optabs.def (vec_shl_insert_optab): New optab.
5689 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5690 (duplicate_and_interleave): Likewise.
5691 * tree-vect-loop.c: Include internal-fn.h.
5692 (neutral_op_for_slp_reduction): New function, split out from
5693 get_initial_defs_for_reduction.
5694 (get_initial_def_for_reduction): Handle option 2 for variable-length
5695 vectors by loading the neutral value into a vector and then shifting
5696 the initial value into element 0.
5697 (get_initial_defs_for_reduction): Replace the code argument with
5698 the neutral value calculated by neutral_op_for_slp_reduction.
5699 Use gimple_build_vector for constant-length vectors.
5700 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5701 but the first group_size elements have a neutral value.
5702 Use duplicate_and_interleave otherwise.
5703 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5704 Update call to get_initial_defs_for_reduction. Handle SLP
5705 reductions for variable-length vectors by creating one vector
5706 result for each scalar result, with the elements associated
5707 with other scalar results stubbed out with the neutral value.
5708 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5709 Require IFN_VEC_SHL_INSERT for double reductions on
5710 variable-length vectors, or SLP reductions that have
5711 a neutral value. Require can_duplicate_and_interleave_p
5712 support for variable-length unchained SLP reductions if there
5713 is no neutral value, such as for MIN/MAX reductions. Also require
5714 the number of vector elements to be a multiple of the number of
5715 SLP statements when doing variable-length unchained SLP reductions.
5716 Update call to vect_create_epilog_for_reduction.
5717 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5718 and remove initial values.
5719 (duplicate_and_interleave): Make public.
5720 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5721 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5724 Alan Hayward <alan.hayward@arm.com>
5725 David Sherwood <david.sherwood@arm.com>
5727 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5728 (can_duplicate_and_interleave_p): New function.
5729 (vect_get_and_check_slp_defs): Take the vector of statements
5730 rather than just the current one. Remove excess parentheses.
5731 Restriction rejectinon of vect_constant_def and vect_external_def
5732 for variable-length vectors to boolean types, or types for which
5733 can_duplicate_and_interleave_p is false.
5734 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5735 (duplicate_and_interleave): New function.
5736 (vect_get_constant_vectors): Use gimple_build_vector for
5737 constant-length vectors and suitable variable-length constant
5738 vectors. Use duplicate_and_interleave for other variable-length
5739 vectors. Don't defer the update when inserting new statements.
5741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5742 Alan Hayward <alan.hayward@arm.com>
5743 David Sherwood <david.sherwood@arm.com>
5745 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5746 min_profitable_iters doesn't go negative.
5748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5749 Alan Hayward <alan.hayward@arm.com>
5750 David Sherwood <david.sherwood@arm.com>
5752 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5753 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5754 * optabs.def (vec_mask_load_lanes_optab): New optab.
5755 (vec_mask_store_lanes_optab): Likewise.
5756 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5757 (MASK_STORE_LANES): Likewise.
5758 * internal-fn.c (mask_load_lanes_direct): New macro.
5759 (mask_store_lanes_direct): Likewise.
5760 (expand_mask_load_optab_fn): Handle masked operations.
5761 (expand_mask_load_lanes_optab_fn): New macro.
5762 (expand_mask_store_optab_fn): Handle masked operations.
5763 (expand_mask_store_lanes_optab_fn): New macro.
5764 (direct_mask_load_lanes_optab_supported_p): Likewise.
5765 (direct_mask_store_lanes_optab_supported_p): Likewise.
5766 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5768 (vect_load_lanes_supported): Likewise.
5769 * tree-vect-data-refs.c (strip_conversion): New function.
5770 (can_group_stmts_p): Likewise.
5771 (vect_analyze_data_ref_accesses): Use it instead of checking
5772 for a pair of assignments.
5773 (vect_store_lanes_supported): Take a masked_p parameter.
5774 (vect_load_lanes_supported): Likewise.
5775 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5776 vect_store_lanes_supported and vect_load_lanes_supported.
5777 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5778 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5779 parameter. Don't allow gaps for masked accesses.
5780 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5781 and vect_load_lanes_supported.
5782 (get_load_store_type): Take a masked_p parameter and update
5783 call to get_group_load_store_type.
5784 (vectorizable_store): Update call to get_load_store_type.
5785 Handle IFN_MASK_STORE_LANES.
5786 (vectorizable_load): Update call to get_load_store_type.
5787 Handle IFN_MASK_LOAD_LANES.
5789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5790 Alan Hayward <alan.hayward@arm.com>
5791 David Sherwood <david.sherwood@arm.com>
5793 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5795 * config/aarch64/aarch64-protos.h
5796 (aarch64_sve_struct_memory_operand_p): Declare.
5797 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5798 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5799 (VPRED, vpred): Handle SVE structure modes.
5800 * config/aarch64/constraints.md (Utx): New constraint.
5801 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5802 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5803 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5804 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5805 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5806 structure modes. Split into pieces after RA.
5807 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5808 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5810 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5811 SVE structure modes.
5812 (aarch64_classify_address): Likewise.
5813 (sizetochar): Move earlier in file.
5814 (aarch64_print_operand): Handle SVE register lists.
5815 (aarch64_array_mode): New function.
5816 (aarch64_sve_struct_memory_operand_p): Likewise.
5817 (TARGET_ARRAY_MODE): Redefine.
5819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5820 Alan Hayward <alan.hayward@arm.com>
5821 David Sherwood <david.sherwood@arm.com>
5823 * target.def (array_mode): New target hook.
5824 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5825 * doc/tm.texi: Regenerate.
5826 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5827 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5828 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5830 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5834 Alan Hayward <alan.hayward@arm.com>
5835 David Sherwood <david.sherwood@arm.com>
5837 * fold-const.c (fold_binary_loc): Check the argument types
5838 rather than the result type when testing for a vector operation.
5840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5842 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5843 * doc/tm.texi: Regenerate.
5845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5846 Alan Hayward <alan.hayward@arm.com>
5847 David Sherwood <david.sherwood@arm.com>
5849 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5850 (sve): Document new AArch64 extension.
5851 * doc/md.texi (w): Extend the description of the AArch64
5852 constraint to include SVE vectors.
5853 (Upl, Upa): Document new AArch64 predicate constraints.
5854 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5856 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5857 (msve-vector-bits=): New option.
5858 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5859 SVE when these are disabled.
5860 (sve): New extension.
5861 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5862 modes. Adjust their number of units based on aarch64_sve_vg.
5863 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5864 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5865 aarch64_addr_query_type.
5866 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5867 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5868 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5869 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5870 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5871 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5872 (aarch64_simd_imm_zero_p): Delete.
5873 (aarch64_check_zero_based_sve_index_immediate): Declare.
5874 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5875 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5876 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5877 (aarch64_sve_float_mul_immediate_p): Likewise.
5878 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5880 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5881 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5882 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5883 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5884 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5885 (aarch64_regmode_natural_size): Likewise.
5886 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5887 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5889 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5890 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5891 for VG and the SVE predicate registers.
5892 (V_ALIASES): Add a "z"-prefixed alias.
5893 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5894 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5895 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5896 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5897 (REG_CLASS_NAMES): Add entries for them.
5898 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5899 and the predicate registers.
5900 (aarch64_sve_vg): Declare.
5901 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5902 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5903 (REGMODE_NATURAL_SIZE): Define.
5904 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5906 * config/aarch64/aarch64.c: Include cfgrtl.h.
5907 (simd_immediate_info): Add a constructor for series vectors,
5908 and an associated step field.
5909 (aarch64_sve_vg): New variable.
5910 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5911 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5912 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5913 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5914 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5915 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5916 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5917 (aarch64_get_mask_mode): New functions.
5918 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5919 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5920 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5921 predicate modes and predicate registers. Explicitly restrict
5922 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5923 to store a vector mode if it is recognized by
5924 aarch64_classify_vector_mode.
5925 (aarch64_regmode_natural_size): New function.
5926 (aarch64_hard_regno_caller_save_mode): Return the original mode
5928 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5929 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5930 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5931 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5933 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5934 does not overlap dest if the function is frame-related. Handle
5936 (aarch64_split_add_offset): New function.
5937 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5938 them aarch64_add_offset.
5939 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5940 and update call to aarch64_sub_sp.
5941 (aarch64_add_cfa_expression): New function.
5942 (aarch64_expand_prologue): Pass extra temporary registers to the
5943 functions above. Handle the case in which we need to emit new
5944 DW_CFA_expressions for registers that were originally saved
5945 relative to the stack pointer, but now have to be expressed
5946 relative to the frame pointer.
5947 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5949 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5950 IP0 and IP1 values for SVE frames.
5951 (aarch64_expand_vec_series): New function.
5952 (aarch64_expand_sve_widened_duplicate): Likewise.
5953 (aarch64_expand_sve_const_vector): Likewise.
5954 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5955 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5956 into the register, rather than emitting a SET directly.
5957 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5958 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5959 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5960 (offset_9bit_signed_scaled_p): New functions.
5961 (aarch64_replicate_bitmask_imm): New function.
5962 (aarch64_bitmask_imm): Use it.
5963 (aarch64_cannot_force_const_mem): Reject expressions involving
5964 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5965 (aarch64_classify_index): Handle SVE indices, by requiring
5966 a plain register index with a scale that matches the element size.
5967 (aarch64_classify_address): Handle SVE addresses. Assert that
5968 the mode of the address is VOIDmode or an integer mode.
5969 Update call to aarch64_classify_symbol.
5970 (aarch64_classify_symbolic_expression): Update call to
5971 aarch64_classify_symbol.
5972 (aarch64_const_vec_all_in_range_p): New function.
5973 (aarch64_print_vector_float_operand): Likewise.
5974 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5975 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5976 and the FP immediates 1.0 and 0.5.
5977 (aarch64_print_address_internal): Handle SVE addresses.
5978 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5979 (aarch64_regno_regclass): Handle predicate registers.
5980 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5982 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5983 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5984 (aarch64_convert_sve_vector_bits): New function.
5985 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5986 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5988 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5989 Handle SVE vector and predicate modes. Accept VL-based constants
5990 that need only one temporary register, and VL offsets that require
5991 no temporary registers.
5992 (aarch64_conditional_register_usage): Mark the predicate registers
5993 as fixed if SVE isn't available.
5994 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5995 Return true for SVE vector and predicate modes.
5996 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5997 rather than an unsigned int. Handle SVE modes.
5998 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6000 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6002 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6003 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6004 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6005 (aarch64_sve_float_mul_immediate_p): New functions.
6006 (aarch64_sve_valid_immediate): New function.
6007 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6008 Explicitly reject structure modes. Check for INDEX constants.
6009 Handle PTRUE and PFALSE constants.
6010 (aarch64_check_zero_based_sve_index_immediate): New function.
6011 (aarch64_simd_imm_zero_p): Delete.
6012 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6013 vector modes. Accept constants in the range of CNT[BHWD].
6014 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6015 ask for an Advanced SIMD mode.
6016 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6017 (aarch64_simd_vector_alignment): Handle SVE predicates.
6018 (aarch64_vectorize_preferred_vector_alignment): New function.
6019 (aarch64_simd_vector_alignment_reachable): Use it instead of
6021 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6022 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6024 (MAX_VECT_LEN): Delete.
6025 (expand_vec_perm_d): Add a vec_flags field.
6026 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6027 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6028 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6030 (aarch64_evpc_rev): Rename to...
6031 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6032 (aarch64_evpc_rev_global): New function.
6033 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6034 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6036 (aarch64_evpc_sve_tbl): New function.
6037 (aarch64_expand_vec_perm_const_1): Update after rename of
6038 aarch64_evpc_rev. Handle SVE permutes too, trying
6039 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6040 than aarch64_evpc_tbl.
6041 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6042 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6043 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6044 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6045 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6046 (aarch64_expand_sve_vcond): New functions.
6047 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6048 of aarch64_vector_mode_p.
6049 (aarch64_dwarf_poly_indeterminate_value): New function.
6050 (aarch64_compute_pressure_classes): Likewise.
6051 (aarch64_can_change_mode_class): Likewise.
6052 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6053 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6054 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6055 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6056 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6057 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6058 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6059 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6061 (Dn, Dl, Dr): Accept const as well as const_vector.
6062 (Dz): Likewise. Compare against CONST0_RTX.
6063 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6064 of "vector" where appropriate.
6065 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6066 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6067 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6068 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6069 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6070 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6071 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6072 (v_int_equiv): Extend to SVE modes.
6073 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6075 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6076 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6077 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6078 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6079 (SVE_COND_FP_CMP): New int iterators.
6080 (perm_hilo): Handle the new unpack unspecs.
6081 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6083 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6084 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6085 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6086 (aarch64_equality_operator, aarch64_constant_vector_operand)
6087 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6088 (aarch64_sve_nonimmediate_operand): Likewise.
6089 (aarch64_sve_general_operand): Likewise.
6090 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6091 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6092 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6093 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6094 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6095 (aarch64_sve_float_arith_immediate): Likewise.
6096 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6097 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6098 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6099 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6100 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6101 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6102 (aarch64_sve_float_arith_operand): Likewise.
6103 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6104 (aarch64_sve_float_mul_operand): Likewise.
6105 (aarch64_sve_vec_perm_operand): Likewise.
6106 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6107 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6108 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6109 as well as const_vector.
6110 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6111 in file. Use CONST0_RTX and CONSTM1_RTX.
6112 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6113 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6114 Use aarch64_simd_imm_zero.
6115 * config/aarch64/aarch64-sve.md: New file.
6116 * config/aarch64/aarch64.md: Include it.
6117 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6118 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6119 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6120 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6121 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6122 (sve): New attribute.
6123 (enabled): Disable instructions with the sve attribute unless
6125 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6126 aarch64_expand_mov_immediate.
6127 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6128 CNT[BHSD] immediates.
6129 (movti): Split CONST_POLY_INT moves into two halves.
6130 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6131 Split additions that need a temporary here if the destination
6132 is the stack pointer.
6133 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6134 (*add<mode>3_poly_1): New instruction.
6135 (set_clobber_cc): New expander.
6137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6139 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6140 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6141 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6142 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6143 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6144 Change innermode from fixed_mode_size to machine_mode.
6145 (simplify_subreg): Update call accordingly. Handle a constant-sized
6146 subreg of a variable-length CONST_VECTOR.
6148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6149 Alan Hayward <alan.hayward@arm.com>
6150 David Sherwood <david.sherwood@arm.com>
6152 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6153 (add_offset_to_base): New function, split out from...
6154 (create_mem_ref): ...here. When handling a scale other than 1,
6155 check first whether the address is valid without the offset.
6156 Add it into the base if so, leaving the index and scale as-is.
6158 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6161 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6162 fold_for_warn before checking if arg2 is INTEGER_CST.
6164 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6166 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6167 (store_multiple_operation): Delete.
6168 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6169 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6170 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6171 guarded by TARGET_STRING.
6172 (rs6000_output_load_multiple): Delete.
6173 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6174 OPTION_MASK_STRING / TARGET_STRING handling.
6175 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6176 (const rs6000_opt_masks) <"string">: Change mask to 0.
6177 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6178 (MASK_STRING): Delete.
6179 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6181 (load_multiple): Delete.
6188 (store_multiple): Delete.
6195 (movmemsi_8reg): Delete.
6196 (corresponding unnamed define_insn): Delete.
6197 (movmemsi_6reg): Delete.
6198 (corresponding unnamed define_insn): Delete.
6199 (movmemsi_4reg): Delete.
6200 (corresponding unnamed define_insn): Delete.
6201 (movmemsi_2reg): Delete.
6202 (corresponding unnamed define_insn): Delete.
6203 (movmemsi_1reg): Delete.
6204 (corresponding unnamed define_insn): Delete.
6205 * config/rs6000/rs6000.opt (mno-string): New.
6206 (mstring): Replace by deprecation warning stub.
6207 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6209 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6211 * regrename.c (regrename_do_replace): If replacing the same
6212 reg multiple times, try to reuse last created gen_raw_REG.
6215 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6216 main to workaround a bug in GDB.
6218 2018-01-12 Tom de Vries <tom@codesourcery.com>
6221 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6223 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6225 PR rtl-optimization/80481
6226 * ira-color.c (get_cap_member): New function.
6227 (allocnos_conflict_by_live_ranges_p): Use it.
6228 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6229 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6231 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6234 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6235 (*saddl_se_1): Ditto.
6237 (*ssubl_se_1): Ditto.
6239 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6241 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6242 rather than wi::to_widest for DR_INITs.
6243 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6244 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6245 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6247 (vect_analyze_group_access_1): Note that here.
6249 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6251 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6252 polynomial type sizes.
6254 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6256 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6257 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6258 (gimple_add_tmp_var): Likewise.
6260 2018-01-12 Martin Liska <mliska@suse.cz>
6262 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6263 (gimple_alloc_sizes): Likewise.
6264 (dump_gimple_statistics): Use PRIu64 in printf format.
6265 * gimple.h: Change uint64_t to int.
6267 2018-01-12 Martin Liska <mliska@suse.cz>
6269 * tree-core.h: Use uint64_t instead of int.
6270 * tree.c (tree_node_counts): Likewise.
6271 (tree_node_sizes): Likewise.
6272 (dump_tree_statistics): Use PRIu64 in printf format.
6274 2018-01-12 Martin Liska <mliska@suse.cz>
6276 * Makefile.in: As qsort_chk is implemented in vec.c, add
6277 vec.o to linkage of gencfn-macros.
6278 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6279 passing the info to record_node_allocation_statistics.
6280 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6282 * ggc-common.c (struct ggc_usage): Add operator== and use
6283 it in operator< and compare function.
6284 * mem-stats.h (struct mem_usage): Likewise.
6285 * vec.c (struct vec_usage): Remove operator< and compare
6286 function. Can be simply inherited.
6288 2018-01-12 Martin Jambor <mjambor@suse.cz>
6291 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6292 * tree-ssa-math-opts.c: Include domwalk.h.
6293 (convert_mult_to_fma_1): New function.
6294 (fma_transformation_info): New type.
6295 (fma_deferring_state): Likewise.
6296 (cancel_fma_deferring): New function.
6297 (result_of_phi): Likewise.
6298 (last_fma_candidate_feeds_initial_phi): Likewise.
6299 (convert_mult_to_fma): Added deferring logic, split actual
6300 transformation to convert_mult_to_fma_1.
6301 (math_opts_dom_walker): New type.
6302 (math_opts_dom_walker::after_dom_children): New method, body moved
6303 here from pass_optimize_widening_mul::execute, added deferring logic
6305 (pass_optimize_widening_mul::execute): Moved most of code to
6306 math_opts_dom_walker::after_dom_children.
6307 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6308 * config/i386/i386.c (ix86_option_override_internal): Added
6309 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6311 2018-01-12 Richard Biener <rguenther@suse.de>
6314 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6315 inline instance vars.
6317 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6320 * config/rx/rx.c (rx_is_restricted_memory_address):
6323 2018-01-12 Richard Biener <rguenther@suse.de>
6325 PR tree-optimization/80846
6326 * target.def (split_reduction): New target hook.
6327 * targhooks.c (default_split_reduction): New function.
6328 * targhooks.h (default_split_reduction): Declare.
6329 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6330 target requests first reduce vectors by combining low and high
6332 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6333 (get_vectype_for_scalar_type_and_size): Export.
6334 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6335 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6336 * doc/tm.texi: Regenerate.
6337 * config/i386/i386.c (ix86_split_reduction): Implement
6338 TARGET_VECTORIZE_SPLIT_REDUCTION.
6340 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6343 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6344 in PIC mode except for TARGET_VXWORKS_RTP.
6345 * config/sparc/sparc.c: Include cfgrtl.h.
6346 (TARGET_INIT_PIC_REG): Define.
6347 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6348 (sparc_pic_register_p): New predicate.
6349 (sparc_legitimate_address_p): Use it.
6350 (sparc_legitimize_pic_address): Likewise.
6351 (sparc_delegitimize_address): Likewise.
6352 (sparc_mode_dependent_address_p): Likewise.
6353 (gen_load_pcrel_sym): Remove 4th parameter.
6354 (load_got_register): Adjust call to above. Remove obsolete stuff.
6355 (sparc_expand_prologue): Do not call load_got_register here.
6356 (sparc_flat_expand_prologue): Likewise.
6357 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6358 (sparc_use_pseudo_pic_reg): New function.
6359 (sparc_init_pic_reg): Likewise.
6360 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6361 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6363 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6365 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6366 Add item for branch_cost.
6368 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6370 PR rtl-optimization/83565
6371 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6372 not extend the result to a larger mode for rotate operations.
6373 (num_sign_bit_copies1): Likewise.
6375 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6378 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6380 Use values-Xc.o for -pedantic.
6381 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6383 2018-01-12 Martin Liska <mliska@suse.cz>
6386 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6388 (possible_polymorphic_call_targets): Use it.
6389 (ipa_devirt): Likewise.
6391 2018-01-12 Martin Liska <mliska@suse.cz>
6393 * profile-count.h (enum profile_quality): Use 0 as invalid
6394 enum value of profile_quality.
6396 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6398 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6399 -mext-string options.
6401 2018-01-12 Richard Biener <rguenther@suse.de>
6403 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6404 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6405 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6407 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6409 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6411 * configure.ac (--with-long-double-format): Add support for the
6412 configuration option to change the default long double format on
6414 * config.gcc (powerpc*-linux*-*): Likewise.
6415 * configure: Regenerate.
6416 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6417 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6418 used without modification.
6420 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6422 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6423 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6424 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6425 MISC_BUILTIN_SPEC_BARRIER.
6426 (rs6000_init_builtins): Likewise.
6427 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6429 (speculation_barrier): New define_insn.
6430 * doc/extend.texi: Document __builtin_speculation_barrier.
6432 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6435 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6436 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6437 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6439 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6440 integral modes instead of "ss" and "sd".
6441 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6442 vectors with 32-bit and 64-bit elements.
6443 (vecdupssescalarmodesuffix): New mode attribute.
6444 (vec_dup<mode>): Use it.
6446 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6449 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6450 frame if argument is passed on stack.
6452 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6455 * ree.c (combine_reaching_defs): Optimize also
6456 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6457 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6459 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6462 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6464 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6467 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6468 after they are computed.
6470 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6472 PR tree-optimization/83695
6473 * gimple-loop-linterchange.cc
6474 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6475 reset cached scev information after interchange.
6476 (pass_linterchange::execute): Remove call to scev_reset_htab.
6478 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6480 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6481 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6482 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6483 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6484 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6485 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6486 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6487 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6488 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6489 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6490 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6491 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6492 (V_lane_reg): Likewise.
6493 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6495 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6496 (vfmal_lane_low<mode>_intrinsic,
6497 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6498 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6499 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6500 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6501 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6502 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6504 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6506 * config/arm/arm-cpus.in (fp16fml): New feature.
6507 (ALL_SIMD): Add fp16fml.
6508 (armv8.2-a): Add fp16fml as an option.
6509 (armv8.3-a): Likewise.
6510 (armv8.4-a): Add fp16fml as part of fp16.
6511 * config/arm/arm.h (TARGET_FP16FML): Define.
6512 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6514 * config/arm/arm-modes.def (V2HF): Define.
6515 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6516 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6517 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6518 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6519 vfmsl_low, vfmsl_high): New set of builtins.
6520 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6521 (vfml_op): New code attribute.
6522 (VFMLHALVES): New int iterator.
6523 (VFML, VFMLSEL): New mode attributes.
6524 (V_reg): Define mapping for V2HF.
6525 (V_hi, V_lo): New mode attributes.
6526 (VF_constraint): Likewise.
6527 (vfml_half, vfml_half_selector): New int attributes.
6528 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6530 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6531 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6533 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6534 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6535 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6536 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6538 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6539 Document new effective target and option set.
6541 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6543 * config/arm/arm-cpus.in (armv8_4): New feature.
6544 (ARMv8_4a): New fgroup.
6545 (armv8.4-a): New arch.
6546 * config/arm/arm-tables.opt: Regenerate.
6547 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6548 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6549 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6550 Add matching rules for -march=armv8.4-a and extensions.
6551 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6553 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6556 * config/rx/rx.md (BW): New mode attribute.
6557 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6559 2018-01-11 Richard Biener <rguenther@suse.de>
6561 PR tree-optimization/83435
6562 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6563 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6564 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6566 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6567 Alan Hayward <alan.hayward@arm.com>
6568 David Sherwood <david.sherwood@arm.com>
6570 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6572 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6573 (aarch64_print_address_internal): Use it to check for a zero offset.
6575 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6576 Alan Hayward <alan.hayward@arm.com>
6577 David Sherwood <david.sherwood@arm.com>
6579 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6580 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6581 Return a poly_int64 rather than a HOST_WIDE_INT.
6582 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6583 rather than a HOST_WIDE_INT.
6584 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6585 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6586 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6587 final_offset from HOST_WIDE_INT to poly_int64.
6588 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6589 to_constant when getting the number of units in an Advanced SIMD
6591 (aarch64_builtin_vectorized_function): Check for a constant number
6593 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6595 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6596 attribute instead of GET_MODE_NUNITS.
6597 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6598 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6599 GET_MODE_SIZE for fixed-size registers.
6600 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6601 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6602 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6603 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6604 (aarch64_print_operand, aarch64_print_address_internal)
6605 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6606 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6607 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6608 Handle polynomial GET_MODE_SIZE.
6609 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6610 wider than SImode without modification.
6611 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6612 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6613 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6614 passing and returning SVE modes.
6615 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6616 rather than GEN_INT.
6617 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6618 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6619 (aarch64_allocate_and_probe_stack_space): Likewise.
6620 (aarch64_layout_frame): Cope with polynomial offsets.
6621 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6622 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6624 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6625 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6626 poly_int64 rather than a HOST_WIDE_INT.
6627 (aarch64_get_separate_components, aarch64_process_components)
6628 (aarch64_expand_prologue, aarch64_expand_epilogue)
6629 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6630 (aarch64_anchor_offset): New function, split out from...
6631 (aarch64_legitimize_address): ...here.
6632 (aarch64_builtin_vectorization_cost): Handle polynomial
6633 TYPE_VECTOR_SUBPARTS.
6634 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6636 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6637 number of elements from the PARALLEL rather than the mode.
6638 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6639 rather than GET_MODE_BITSIZE.
6640 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6641 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6642 (aarch64_expand_vec_perm_const_1): Handle polynomial
6643 d->perm.length () and d->perm elements.
6644 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6645 Apply to_constant to d->perm elements.
6646 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6647 polynomial CONST_VECTOR_NUNITS.
6648 (aarch64_move_pointer): Take amount as a poly_int64 rather
6650 (aarch64_progress_pointer): Avoid temporary variable.
6651 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6652 the mode attribute instead of GET_MODE.
6654 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6655 Alan Hayward <alan.hayward@arm.com>
6656 David Sherwood <david.sherwood@arm.com>
6658 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6659 x exists before using it.
6660 (aarch64_add_constant_internal): Rename to...
6661 (aarch64_add_offset_1): ...this. Replace regnum with separate
6662 src and dest rtxes. Handle the case in which they're different,
6663 including when the offset is zero. Replace scratchreg with an rtx.
6664 Use 2 additions if there is no spare register into which we can
6665 move a 16-bit constant.
6666 (aarch64_add_constant): Delete.
6667 (aarch64_add_offset): Replace reg with separate src and dest
6668 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6669 Use aarch64_add_offset_1.
6670 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6671 an rtx rather than an int. Take the delta as a poly_int64
6672 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6673 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6674 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6675 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6676 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6678 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6679 aarch64_add_constant.
6681 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6683 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6684 Use scalar_float_mode.
6686 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6688 * config/aarch64/aarch64-simd.md
6689 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6690 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6691 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6692 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6693 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6694 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6695 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6696 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6697 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6698 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6700 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6703 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6704 targ_options->x_arm_arch_string is non NULL.
6706 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6708 * config/aarch64/aarch64.h
6709 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6711 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6714 * expmed.c (emit_store_flag_force): Swap if const op0
6715 and change VOIDmode to mode of op0.
6717 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6719 PR rtl-optimization/83761
6720 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6721 than bytes to mode_for_size.
6723 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6726 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6727 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6730 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6733 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6734 when in layout mode.
6735 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6736 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6739 2018-01-10 Michael Collison <michael.collison@arm.com>
6741 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6742 * config/aarch64/aarch64-option-extension.def: Add
6743 AARCH64_OPT_EXTENSION of 'fp16fml'.
6744 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6745 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6746 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6747 * config/aarch64/constraints.md (Ui7): New constraint.
6748 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6749 (VFMLA_SEL_W): Ditto.
6752 (VFMLA16_LOW): New int iterator.
6753 (VFMLA16_HIGH): Ditto.
6754 (UNSPEC_FMLAL): New unspec.
6755 (UNSPEC_FMLSL): Ditto.
6756 (UNSPEC_FMLAL2): Ditto.
6757 (UNSPEC_FMLSL2): Ditto.
6758 (f16mac): New code attribute.
6759 * config/aarch64/aarch64-simd-builtins.def
6760 (aarch64_fmlal_lowv2sf): Ditto.
6761 (aarch64_fmlsl_lowv2sf): Ditto.
6762 (aarch64_fmlalq_lowv4sf): Ditto.
6763 (aarch64_fmlslq_lowv4sf): Ditto.
6764 (aarch64_fmlal_highv2sf): Ditto.
6765 (aarch64_fmlsl_highv2sf): Ditto.
6766 (aarch64_fmlalq_highv4sf): Ditto.
6767 (aarch64_fmlslq_highv4sf): Ditto.
6768 (aarch64_fmlal_lane_lowv2sf): Ditto.
6769 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6770 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6771 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6772 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6773 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6774 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6775 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6776 (aarch64_fmlal_lane_highv2sf): Ditto.
6777 (aarch64_fmlsl_lane_highv2sf): Ditto.
6778 (aarch64_fmlal_laneq_highv2sf): Ditto.
6779 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6780 (aarch64_fmlalq_lane_highv4sf): Ditto.
6781 (aarch64_fmlsl_lane_highv4sf): Ditto.
6782 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6783 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6784 * config/aarch64/aarch64-simd.md:
6785 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6786 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6787 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6788 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6789 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6790 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6791 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6792 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6793 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6794 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6795 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6796 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6797 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6798 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6799 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6800 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6801 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6802 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6803 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6804 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6805 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6806 (vfmlsl_low_u32): Ditto.
6807 (vfmlalq_low_u32): Ditto.
6808 (vfmlslq_low_u32): Ditto.
6809 (vfmlal_high_u32): Ditto.
6810 (vfmlsl_high_u32): Ditto.
6811 (vfmlalq_high_u32): Ditto.
6812 (vfmlslq_high_u32): Ditto.
6813 (vfmlal_lane_low_u32): Ditto.
6814 (vfmlsl_lane_low_u32): Ditto.
6815 (vfmlal_laneq_low_u32): Ditto.
6816 (vfmlsl_laneq_low_u32): Ditto.
6817 (vfmlalq_lane_low_u32): Ditto.
6818 (vfmlslq_lane_low_u32): Ditto.
6819 (vfmlalq_laneq_low_u32): Ditto.
6820 (vfmlslq_laneq_low_u32): Ditto.
6821 (vfmlal_lane_high_u32): Ditto.
6822 (vfmlsl_lane_high_u32): Ditto.
6823 (vfmlal_laneq_high_u32): Ditto.
6824 (vfmlsl_laneq_high_u32): Ditto.
6825 (vfmlalq_lane_high_u32): Ditto.
6826 (vfmlslq_lane_high_u32): Ditto.
6827 (vfmlalq_laneq_high_u32): Ditto.
6828 (vfmlslq_laneq_high_u32): Ditto.
6829 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6830 (AARCH64_FL_FOR_ARCH8_4): New.
6831 (AARCH64_ISA_F16FML): New ISA flag.
6832 (TARGET_F16FML): New feature flag for fp16fml.
6833 (doc/invoke.texi): Document new fp16fml option.
6835 2018-01-10 Michael Collison <michael.collison@arm.com>
6837 * config/aarch64/aarch64-builtins.c:
6838 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6839 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6840 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6841 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6842 (AARCH64_ISA_SHA3): New ISA flag.
6843 (TARGET_SHA3): New feature flag for sha3.
6844 * config/aarch64/iterators.md (sha512_op): New int attribute.
6845 (CRYPTO_SHA512): New int iterator.
6846 (UNSPEC_SHA512H): New unspec.
6847 (UNSPEC_SHA512H2): Ditto.
6848 (UNSPEC_SHA512SU0): Ditto.
6849 (UNSPEC_SHA512SU1): Ditto.
6850 * config/aarch64/aarch64-simd-builtins.def
6851 (aarch64_crypto_sha512hqv2di): New builtin.
6852 (aarch64_crypto_sha512h2qv2di): Ditto.
6853 (aarch64_crypto_sha512su0qv2di): Ditto.
6854 (aarch64_crypto_sha512su1qv2di): Ditto.
6855 (aarch64_eor3qv8hi): Ditto.
6856 (aarch64_rax1qv2di): Ditto.
6857 (aarch64_xarqv2di): Ditto.
6858 (aarch64_bcaxqv8hi): Ditto.
6859 * config/aarch64/aarch64-simd.md:
6860 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6861 (aarch64_crypto_sha512su0qv2di): Ditto.
6862 (aarch64_crypto_sha512su1qv2di): Ditto.
6863 (aarch64_eor3qv8hi): Ditto.
6864 (aarch64_rax1qv2di): Ditto.
6865 (aarch64_xarqv2di): Ditto.
6866 (aarch64_bcaxqv8hi): Ditto.
6867 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6868 (vsha512h2q_u64): Ditto.
6869 (vsha512su0q_u64): Ditto.
6870 (vsha512su1q_u64): Ditto.
6871 (veor3q_u16): Ditto.
6872 (vrax1q_u64): Ditto.
6874 (vbcaxq_u16): Ditto.
6875 * config/arm/types.md (crypto_sha512): New type attribute.
6876 (crypto_sha3): Ditto.
6877 (doc/invoke.texi): Document new sha3 option.
6879 2018-01-10 Michael Collison <michael.collison@arm.com>
6881 * config/aarch64/aarch64-builtins.c:
6882 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6883 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6884 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6885 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6886 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6887 (AARCH64_ISA_SM4): New ISA flag.
6888 (TARGET_SM4): New feature flag for sm4.
6889 * config/aarch64/aarch64-simd-builtins.def
6890 (aarch64_sm3ss1qv4si): Ditto.
6891 (aarch64_sm3tt1aq4si): Ditto.
6892 (aarch64_sm3tt1bq4si): Ditto.
6893 (aarch64_sm3tt2aq4si): Ditto.
6894 (aarch64_sm3tt2bq4si): Ditto.
6895 (aarch64_sm3partw1qv4si): Ditto.
6896 (aarch64_sm3partw2qv4si): Ditto.
6897 (aarch64_sm4eqv4si): Ditto.
6898 (aarch64_sm4ekeyqv4si): Ditto.
6899 * config/aarch64/aarch64-simd.md:
6900 (aarch64_sm3ss1qv4si): Ditto.
6901 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6902 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6903 (aarch64_sm4eqv4si): Ditto.
6904 (aarch64_sm4ekeyqv4si): Ditto.
6905 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6906 (sm3part_op): Ditto.
6907 (CRYPTO_SM3TT): Ditto.
6908 (CRYPTO_SM3PART): Ditto.
6909 (UNSPEC_SM3SS1): New unspec.
6910 (UNSPEC_SM3TT1A): Ditto.
6911 (UNSPEC_SM3TT1B): Ditto.
6912 (UNSPEC_SM3TT2A): Ditto.
6913 (UNSPEC_SM3TT2B): Ditto.
6914 (UNSPEC_SM3PARTW1): Ditto.
6915 (UNSPEC_SM3PARTW2): Ditto.
6916 (UNSPEC_SM4E): Ditto.
6917 (UNSPEC_SM4EKEY): Ditto.
6918 * config/aarch64/constraints.md (Ui2): New constraint.
6919 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6920 * config/arm/types.md (crypto_sm3): New type attribute.
6921 (crypto_sm4): Ditto.
6922 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6923 (vsm3tt1aq_u32): Ditto.
6924 (vsm3tt1bq_u32): Ditto.
6925 (vsm3tt2aq_u32): Ditto.
6926 (vsm3tt2bq_u32): Ditto.
6927 (vsm3partw1q_u32): Ditto.
6928 (vsm3partw2q_u32): Ditto.
6929 (vsm4eq_u32): Ditto.
6930 (vsm4ekeyq_u32): Ditto.
6931 (doc/invoke.texi): Document new sm4 option.
6933 2018-01-10 Michael Collison <michael.collison@arm.com>
6935 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6936 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6937 (AARCH64_FL_FOR_ARCH8_4): New.
6938 (AARCH64_FL_V8_4): New flag.
6939 (doc/invoke.texi): Document new armv8.4-a option.
6941 2018-01-10 Michael Collison <michael.collison@arm.com>
6943 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6944 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6945 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6946 * config/aarch64/aarch64-option-extension.def: Add
6947 AARCH64_OPT_EXTENSION of 'sha2'.
6948 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6949 (crypto): Disable sha2 and aes if crypto disabled.
6950 (crypto): Enable aes and sha2 if enabled.
6951 (simd): Disable sha2 and aes if simd disabled.
6952 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6954 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6955 (TARGET_SHA2): New feature flag for sha2.
6956 (TARGET_AES): New feature flag for aes.
6957 * config/aarch64/aarch64-simd.md:
6958 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6959 conditional on TARGET_AES.
6960 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6961 (aarch64_crypto_sha1hsi): Make pattern conditional
6963 (aarch64_crypto_sha1hv4si): Ditto.
6964 (aarch64_be_crypto_sha1hv4si): Ditto.
6965 (aarch64_crypto_sha1su1v4si): Ditto.
6966 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6967 (aarch64_crypto_sha1su0v4si): Ditto.
6968 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6969 (aarch64_crypto_sha256su0v4si): Ditto.
6970 (aarch64_crypto_sha256su1v4si): Ditto.
6971 (doc/invoke.texi): Document new aes and sha2 options.
6973 2018-01-10 Martin Sebor <msebor@redhat.com>
6975 PR tree-optimization/83781
6976 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6979 2018-01-11 Martin Sebor <msebor@gmail.com>
6980 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6982 PR tree-optimization/83501
6983 PR tree-optimization/81703
6985 * tree-ssa-strlen.c (get_string_cst): Rename...
6986 (get_string_len): ...to this. Handle global constants.
6987 (handle_char_store): Adjust.
6989 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6990 Jim Wilson <jimw@sifive.com>
6992 * config/riscv/riscv-protos.h (riscv_output_return): New.
6993 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6994 (riscv_attribute_table, riscv_output_return),
6995 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6996 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6997 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6999 (riscv_expand_prologue): Add early return for naked function.
7000 (riscv_expand_epilogue): Likewise.
7001 (riscv_function_ok_for_sibcall): Return false for naked function.
7002 (riscv_set_current_function): New.
7003 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7004 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7005 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7006 * doc/extend.texi (RISC-V Function Attributes): New.
7008 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7010 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7011 check for 128-bit long double before checking TCmode.
7012 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7013 128-bit long doubles before checking TFmode or TCmode.
7014 (FLOAT128_IBM_P): Likewise.
7016 2018-01-10 Martin Sebor <msebor@redhat.com>
7018 PR tree-optimization/83671
7019 * builtins.c (c_strlen): Unconditionally return zero for the empty
7021 Use -Warray-bounds for warnings.
7022 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7023 for non-constant array indices with COMPONENT_REF, arrays of
7024 arrays, and pointers to arrays.
7025 (gimple_fold_builtin_strlen): Determine and set length range for
7026 non-constant character arrays.
7028 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7031 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7034 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7036 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7038 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7041 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7042 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7043 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7044 indexed_or_indirect_operand predicate.
7045 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7046 (*vsx_le_perm_load_v8hi): Likewise.
7047 (*vsx_le_perm_load_v16qi): Likewise.
7048 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7049 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7050 (*vsx_le_perm_store_v8hi): Likewise.
7051 (*vsx_le_perm_store_v16qi): Likewise.
7052 (eight unnamed splitters): Likewise.
7054 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7056 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7057 * config/rs6000/emmintrin.h: Likewise.
7058 * config/rs6000/mmintrin.h: Likewise.
7059 * config/rs6000/xmmintrin.h: Likewise.
7061 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7064 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7066 * tree.c (tree_nop_conversion): Return true for location wrapper
7068 (maybe_wrap_with_location): New function.
7069 (selftest::check_strip_nops): New function.
7070 (selftest::test_location_wrappers): New function.
7071 (selftest::tree_c_tests): Call it.
7072 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7073 (maybe_wrap_with_location): New decl.
7074 (EXPR_LOCATION_WRAPPER_P): New macro.
7075 (location_wrapper_p): New inline function.
7076 (tree_strip_any_location_wrapper): New inline function.
7078 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7081 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7082 stack_realign_offset for the largest alignment of stack slot
7084 (ix86_find_max_used_stack_alignment): New function.
7085 (ix86_finalize_stack_frame_flags): Use it. Set
7086 max_used_stack_alignment if we don't realign stack.
7087 * config/i386/i386.h (machine_function): Add
7088 max_used_stack_alignment.
7090 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7092 * config/arm/arm.opt (-mbranch-cost): New option.
7093 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7096 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7099 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7100 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7102 2018-01-10 Richard Biener <rguenther@suse.de>
7105 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7106 early out so it also covers the case where we have a non-NULL
7109 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7111 PR tree-optimization/83753
7112 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7113 for non-strided grouped accesses if the number of elements is 1.
7115 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7118 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7119 * i386.h (TARGET_USE_GATHER): Define.
7120 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7122 2018-01-10 Martin Liska <mliska@suse.cz>
7125 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7126 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7128 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7129 CLEANUP_NO_PARTITIONING is not set.
7131 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7133 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7134 for vectors, as a partial revert of r254296.
7135 * rtl.h (const_vec_p): Delete.
7136 (const_vec_duplicate_p): Don't test for vector CONSTs.
7137 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7138 * expmed.c (make_tree): Likewise.
7141 * common.md (E, F): Use CONSTANT_P instead of checking for
7143 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7144 checking for CONST_VECTOR.
7146 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7149 * predict.c (force_edge_cold): Handle in more sane way edges
7152 2018-01-09 Carl Love <cel@us.ibm.com>
7154 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7156 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7157 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7158 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7159 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7160 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7161 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7162 * config/rs6000/rs6000-protos.h: Add extern defition for
7163 rs6000_generate_float2_double_code.
7164 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7166 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7167 (float2_v2df): Add define_expand.
7169 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7172 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7173 op_mode in the force_to_mode call.
7175 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7177 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7178 instead of checking each element individually.
7179 (aarch64_evpc_uzp): Likewise.
7180 (aarch64_evpc_zip): Likewise.
7181 (aarch64_evpc_ext): Likewise.
7182 (aarch64_evpc_rev): Likewise.
7183 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7184 instead of checking each element individually. Return true without
7186 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7187 whether all selected elements come from the same input, instead of
7188 checking each element individually. Remove calls to gen_rtx_REG,
7189 start_sequence and end_sequence and instead assert that no rtl is
7192 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7194 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7195 order of HIGH and CONST checks.
7197 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7199 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7200 if the destination isn't an SSA_NAME.
7202 2018-01-09 Richard Biener <rguenther@suse.de>
7204 PR tree-optimization/83668
7205 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7207 (canonicalize_loop_form): ... here, renamed from ...
7208 (canonicalize_loop_closed_ssa_form): ... this and amended to
7209 swap successor edges for loop exit blocks to make us use
7210 the RPO order we need for initial schedule generation.
7212 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7214 PR tree-optimization/64811
7215 * match.pd: When optimizing comparisons with Inf, avoid
7216 introducing or losing exceptions from comparisons with NaN.
7218 2018-01-09 Martin Liska <mliska@suse.cz>
7221 * asan.c (shadow_mem_size): Add gcc_assert.
7223 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7225 Don't save registers in main().
7228 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7229 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7230 * config/avr/avr.c (avr_set_current_function): Don't error if
7231 naked, OS_task or OS_main are specified at the same time.
7232 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7234 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7236 * common/config/avr/avr-common.c (avr_option_optimization_table):
7237 Switch on -mmain-is-OS_task for optimizing compilations.
7239 2018-01-09 Richard Biener <rguenther@suse.de>
7241 PR tree-optimization/83572
7242 * graphite.c: Include cfganal.h.
7243 (graphite_transform_loops): Connect infinite loops to exit
7244 and remove fake edges at the end.
7246 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7248 * ipa-inline.c (edge_badness): Revert accidental checkin.
7250 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7253 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7254 symbols; not inline clones.
7256 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7259 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7260 hard registers. Formatting fixes.
7262 PR preprocessor/83722
7263 * gcc.c (try_generate_repro): Pass
7264 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7265 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7268 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7269 Kito Cheng <kito.cheng@gmail.com>
7271 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7272 (riscv_leaf_function_p): Delete.
7273 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7275 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7277 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7279 (do_ifelse): New function.
7280 (do_isel): New function.
7281 (do_sub3): New function.
7282 (do_add3): New function.
7283 (do_load_mask_compare): New function.
7284 (do_overlap_load_compare): New function.
7285 (expand_compare_loop): New function.
7286 (expand_block_compare): Call expand_compare_loop() when appropriate.
7287 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7289 (-mblock-compare-inline-loop-limit): New option.
7291 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7294 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7295 Reverse order of second and third operands in first alternative.
7296 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7297 of first and second elements in UNSPEC_VPERMR vector.
7298 (altivec_expand_vec_perm_le): Likewise.
7300 2018-01-08 Jeff Law <law@redhat.com>
7302 PR rtl-optimizatin/81308
7303 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7304 (process_switch): If group_case_labels makes a change, then set
7306 (pass_convert_switch::execute): If a switch is converted, then
7307 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7309 PR rtl-optimization/81308
7310 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7313 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7315 PR target/83663 - Revert r255946
7316 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7317 generation for cases where splatting a value is not useful.
7318 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7319 across a vec_duplicate and a paradoxical subreg forming a vector
7320 mode to a vec_concat.
7322 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7324 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7325 -march=armv8.3-a variants.
7326 * config/arm/t-multilib: Likewise.
7327 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7329 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7331 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7333 (cceq_ior_compare_complement): Give it a name so I can use it, and
7334 change boolean_or_operator predicate to boolean_operator so it can
7335 be used to generate a crand.
7336 (eqne): New code iterator.
7337 (bd/bd_neg): New code_attrs.
7338 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7339 a single define_insn.
7340 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7341 decrement (bdnzt/bdnzf/bdzt/bdzf).
7342 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7343 with the new names of the branch decrement patterns, and added the
7344 names of the branch decrement conditional patterns.
7346 2018-01-08 Richard Biener <rguenther@suse.de>
7348 PR tree-optimization/83563
7349 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7352 2018-01-08 Richard Biener <rguenther@suse.de>
7355 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7357 2018-01-08 Richard Biener <rguenther@suse.de>
7359 PR tree-optimization/83685
7360 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7361 references to abnormals.
7363 2018-01-08 Richard Biener <rguenther@suse.de>
7366 * dwarf2out.c (output_indirect_strings): Handle empty
7367 skeleton_debug_str_hash.
7368 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7370 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7372 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7373 (emit_store_direct): Likewise.
7374 (arc_trampoline_adjust_address): Likewise.
7375 (arc_asm_trampoline_template): New function.
7376 (arc_initialize_trampoline): Use asm_trampoline_template.
7377 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7378 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7379 * config/arc/arc.md (flush_icache): Delete pattern.
7381 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7383 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7384 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7387 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7390 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7391 by not USED_FOR_TARGET.
7392 (make_pass_resolve_sw_modes): Likewise.
7394 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7396 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7399 2018-01-08 Richard Biener <rguenther@suse.de>
7402 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7404 2018-01-08 Richard Biener <rguenther@suse.de>
7407 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7409 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7412 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7413 basic blocks with a small number of successors.
7414 (convert_control_dep_chain_into_preds): Improve handling of
7416 (dump_predicates): Split apart into...
7417 (dump_pred_chain): ...here...
7418 (dump_pred_info): ...and here.
7419 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7420 (can_chain_union_be_invalidated_p): Improve check for invalidation
7422 (uninit_uses_cannot_happen): Avoid unnecessary if
7423 convert_control_dep_chain_into_preds yielded nothing.
7425 2018-01-06 Martin Sebor <msebor@redhat.com>
7427 PR tree-optimization/83640
7428 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7429 subtracting negative offset from size.
7430 (builtin_access::overlap): Adjust offset bounds of the access to fall
7431 within the size of the object if possible.
7433 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7435 PR rtl-optimization/83699
7436 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7437 extract_bit_field_as_subreg to cases in which the extracted
7438 value is also a vector.
7440 * lra-constraints.c (process_alt_operands): Test for the equivalence
7441 substitutions when detecting a possible reload cycle.
7443 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7446 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7447 by default if flag_selective_schedling{,2}. Formatting fixes.
7449 PR rtl-optimization/83682
7450 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7451 if it has non-VECTOR_MODE element mode.
7452 (vec_duplicate_p): Likewise.
7455 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7456 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7458 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7461 * config/i386/i386-builtin.def
7462 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7463 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7464 Require also OPTION_MASK_ISA_AVX512F in addition to
7465 OPTION_MASK_ISA_GFNI.
7466 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7467 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7468 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7469 to OPTION_MASK_ISA_GFNI.
7470 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7471 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7472 OPTION_MASK_ISA_AVX512BW.
7473 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7474 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7475 addition to OPTION_MASK_ISA_GFNI.
7476 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7477 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7478 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7479 to OPTION_MASK_ISA_GFNI.
7480 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7481 a requirement for all ISAs rather than any of them with a few
7483 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7485 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7486 bitmasks to be enabled with 3 exceptions, instead of requiring any
7487 enabled ISA with lots of exceptions.
7488 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7489 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7490 Change avx512bw in isa attribute to avx512f.
7491 * config/i386/sgxintrin.h: Add license boilerplate.
7492 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7493 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7494 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7495 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7497 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7498 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7499 temporarily sse2 rather than sse if not enabled already.
7502 * config/i386/sse.md (VI248_VLBW): Rename to ...
7503 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7504 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7505 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7506 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7507 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7508 mode iterator instead of VI248_VLBW.
7510 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7512 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7513 (record_modified): Skip clobbers; add debug output.
7514 (param_change_prob): Use sreal frequencies.
7516 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7518 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7519 punt for user-aligned variables.
7521 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7523 * tree-chrec.c (chrec_contains_symbols): Return true for
7526 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7529 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7530 of (x|y) == x for BICS pattern.
7532 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7534 PR tree-optimization/83605
7535 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7536 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7539 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7541 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7542 * config/epiphany/rtems.h: New file.
7544 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7545 Uros Bizjak <ubizjak@gmail.com>
7548 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7549 QIreg_operand instead of register_operand predicate.
7550 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7551 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7552 comments instead of -fmitigate[-_]rop.
7554 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7557 * cgraphunit.c (symbol_table::compile): Switch to text_section
7558 before calling assembly_start debug hook.
7559 * run-rtl-passes.c (run_rtl_passes): Likewise.
7562 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7564 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7565 range_int_cst_p rather than !symbolic_range_p before calling
7566 extract_range_from_multiplicative_op_1.
7568 2018-01-04 Jeff Law <law@redhat.com>
7570 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7571 redundant test in assertion.
7573 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7575 * doc/rtl.texi: Document machine_mode wrapper classes.
7577 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7579 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7582 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7584 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7585 the VEC_PERM_EXPR fold to fail.
7587 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7590 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7591 to switched_sections.
7593 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7596 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7599 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7602 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7603 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7605 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7608 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7609 is BLKmode and bitpos not zero or mode change is needed.
7611 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7614 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7617 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7620 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7621 instead of MULT rtx. Update all corresponding splitters.
7623 (*ssub<modesuffix>): Ditto.
7625 (*cmp_sadd_di): Update split patterns.
7626 (*cmp_sadd_si): Ditto.
7627 (*cmp_sadd_sidi): Ditto.
7628 (*cmp_ssub_di): Ditto.
7629 (*cmp_ssub_si): Ditto.
7630 (*cmp_ssub_sidi): Ditto.
7631 * config/alpha/predicates.md (const23_operand): New predicate.
7632 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7633 Look for ASHIFT, not MULT inner operand.
7634 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7636 2018-01-04 Martin Liska <mliska@suse.cz>
7638 PR gcov-profile/83669
7639 * gcov.c (output_intermediate_file): Add version to intermediate
7641 * doc/gcov.texi: Document new field 'version' in intermediate
7642 file format. Fix location of '-k' option of gcov command.
7644 2018-01-04 Martin Liska <mliska@suse.cz>
7647 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7649 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7651 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7653 2018-01-03 Martin Sebor <msebor@redhat.com>
7655 PR tree-optimization/83655
7656 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7657 checking calls with invalid arguments.
7659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7661 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7662 (vectorizable_mask_load_store): Delete.
7663 (vectorizable_call): Return false for masked loads and stores.
7664 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7665 instead of gimple_assign_rhs1.
7666 (vectorizable_load): Handle IFN_MASK_LOAD.
7667 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7671 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7673 (vectorizable_mask_load_store): ...here.
7674 (vectorizable_load): ...and here.
7676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7678 * tree-vect-stmts.c (vect_build_all_ones_mask)
7679 (vect_build_zero_merge_argument): New functions, split out from...
7680 (vectorizable_load): ...here.
7682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7684 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7686 (vectorizable_mask_load_store): ...here.
7687 (vectorizable_store): ...and here.
7689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7691 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7693 (vectorizable_mask_load_store): ...here.
7695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7697 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7698 (vect_model_store_cost): Take a vec_load_store_type instead of a
7700 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7701 (vect_model_store_cost): Take a vec_load_store_type instead of a
7703 (vectorizable_mask_load_store): Update accordingly.
7704 (vectorizable_store): Likewise.
7705 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7709 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7710 IFN_MASK_LOAD calls here rather than...
7711 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7714 Alan Hayward <alan.hayward@arm.com>
7715 David Sherwood <david.sherwood@arm.com>
7717 * expmed.c (extract_bit_field_1): For vector extracts,
7718 fall back to extract_bit_field_as_subreg if vec_extract
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7726 they are variable or constant sized.
7727 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7728 slots for constant-sized data.
7730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7731 Alan Hayward <alan.hayward@arm.com>
7732 David Sherwood <david.sherwood@arm.com>
7734 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7735 handling COND_EXPRs with boolean comparisons, try to find a better
7736 basis for the mask type than the boolean itself.
7738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7740 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7741 is calculated and how it can be overridden.
7742 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7743 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7745 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7749 Alan Hayward <alan.hayward@arm.com>
7750 David Sherwood <david.sherwood@arm.com>
7752 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7753 Remove the mode argument.
7754 (aarch64_simd_valid_immediate): Remove the mode and inverse
7756 * config/aarch64/iterators.md (bitsize): New iterator.
7757 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7758 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7759 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7760 aarch64_simd_valid_immediate.
7761 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7762 (aarch64_reg_or_bic_imm): Likewise.
7763 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7764 with an insn_type enum and msl with a modifier_type enum.
7765 Replace element_width with a scalar_mode. Change the shift
7766 to unsigned int. Add constructors for scalar_float_mode and
7767 scalar_int_mode elements.
7768 (aarch64_vect_float_const_representable_p): Delete.
7769 (aarch64_can_const_movi_rtx_p)
7770 (aarch64_simd_scalar_immediate_valid_for_move)
7771 (aarch64_simd_make_constant): Update call to
7772 aarch64_simd_valid_immediate.
7773 (aarch64_advsimd_valid_immediate_hs): New function.
7774 (aarch64_advsimd_valid_immediate): Likewise.
7775 (aarch64_simd_valid_immediate): Remove mode and inverse
7776 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7777 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7778 and aarch64_float_const_representable_p on the result.
7779 (aarch64_output_simd_mov_immediate): Remove mode argument.
7780 Update call to aarch64_simd_valid_immediate and use of
7781 simd_immediate_info.
7782 (aarch64_output_scalar_simd_mov_immediate): Update call
7785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7786 Alan Hayward <alan.hayward@arm.com>
7787 David Sherwood <david.sherwood@arm.com>
7789 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7790 (mode_nunits): Likewise CONST_MODE_NUNITS.
7791 * machmode.def (ADJUST_NUNITS): Document.
7792 * genmodes.c (mode_data::need_nunits_adj): New field.
7793 (blank_mode): Update accordingly.
7794 (adj_nunits): New variable.
7795 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7797 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7798 listed in adj_nunits.
7799 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7800 listed in adj_nunits. Don't emit case statements for such modes.
7801 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7802 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7803 nothing if adj_nunits is nonnull.
7804 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7805 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7806 (emit_mode_fbit): Update use of print_maybe_const_decl.
7807 (emit_move_size): Likewise. Treat the array as non-const
7809 (emit_mode_adjustments): Handle adj_nunits.
7811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7813 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7814 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7815 (VECTOR_MODES): Use it.
7816 (make_vector_modes): Take the prefix as an argument.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7819 Alan Hayward <alan.hayward@arm.com>
7820 David Sherwood <david.sherwood@arm.com>
7822 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7823 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7824 for MODE_VECTOR_BOOL.
7825 * machmode.def (VECTOR_BOOL_MODE): Document.
7826 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7827 (make_vector_bool_mode): New function.
7828 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7830 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7831 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7833 * stor-layout.c (int_mode_for_mode): Likewise.
7834 * tree.c (build_vector_type_for_mode): Likewise.
7835 * varasm.c (output_constant_pool_2): Likewise.
7836 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7837 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7838 for MODE_VECTOR_BOOL.
7839 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7840 of mode class checks.
7841 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7842 instead of a list of mode class checks.
7843 (expand_vector_scalar_condition): Likewise.
7844 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7847 Alan Hayward <alan.hayward@arm.com>
7848 David Sherwood <david.sherwood@arm.com>
7850 * machmode.h (mode_size): Change from unsigned short to
7852 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7853 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7854 or if measurement_type is not polynomial.
7855 (fixed_size_mode::includes_p): Check for constant-sized modes.
7856 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7857 return a poly_uint16 rather than an unsigned short.
7858 (emit_mode_size): Change the type of mode_size from unsigned short
7859 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7860 (emit_mode_adjustments): Cope with polynomial vector sizes.
7861 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7863 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7865 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7866 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7867 * caller-save.c (setup_save_areas): Likewise.
7868 (replace_reg_with_saved_mem): Likewise.
7869 * calls.c (emit_library_call_value_1): Likewise.
7870 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7871 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7872 (gen_lowpart_for_combine): Likewise.
7873 * convert.c (convert_to_integer_1): Likewise.
7874 * cse.c (equiv_constant, cse_insn): Likewise.
7875 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7876 (cselib_subst_to_values): Likewise.
7877 * dce.c (word_dce_process_block): Likewise.
7878 * df-problems.c (df_word_lr_mark_ref): Likewise.
7879 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7880 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7881 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7882 (rtl_for_decl_location): Likewise.
7883 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7884 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7885 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7886 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7887 (expand_expr_real_1): Likewise.
7888 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7889 (pad_below): Likewise.
7890 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7891 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7892 * ira.c (get_subreg_tracking_sizes): Likewise.
7893 * ira-build.c (ira_create_allocno_objects): Likewise.
7894 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7895 (ira_sort_regnos_for_alter_reg): Likewise.
7896 * ira-costs.c (record_operand_costs): Likewise.
7897 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7898 (resolve_simple_move): Likewise.
7899 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7900 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7901 (lra_constraints): Likewise.
7902 (CONST_POOL_OK_P): Reject variable-sized modes.
7903 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7904 (add_pseudo_to_slot, lra_spill): Likewise.
7905 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7906 * optabs-query.c (get_best_extraction_insn): Likewise.
7907 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7908 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7909 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7910 * recog.c (offsettable_address_addr_space_p): Likewise.
7911 * regcprop.c (maybe_mode_change): Likewise.
7912 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7913 * regrename.c (build_def_use): Likewise.
7914 * regstat.c (dump_reg_info): Likewise.
7915 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7916 (find_reloads, find_reloads_subreg_address): Likewise.
7917 * reload1.c (eliminate_regs_1): Likewise.
7918 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7919 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7920 (simplify_binary_operation_1, simplify_subreg): Likewise.
7921 * targhooks.c (default_function_arg_padding): Likewise.
7922 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7923 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7924 (verify_gimple_assign_ternary): Likewise.
7925 * tree-inline.c (estimate_move_cost): Likewise.
7926 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7927 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7928 (get_address_cost_ainc): Likewise.
7929 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7930 (vect_supportable_dr_alignment): Likewise.
7931 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7932 (vectorizable_reduction): Likewise.
7933 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7934 (vectorizable_operation, vectorizable_load): Likewise.
7935 * tree.c (build_same_sized_truth_vector_type): Likewise.
7936 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7937 * var-tracking.c (emit_note_insn_var_location): Likewise.
7938 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7939 (ADDR_VEC_ALIGN): Likewise.
7941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7942 Alan Hayward <alan.hayward@arm.com>
7943 David Sherwood <david.sherwood@arm.com>
7945 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7947 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7948 or if measurement_type is polynomial.
7949 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7950 * combine.c (make_extraction): Likewise.
7951 * dse.c (find_shift_sequence): Likewise.
7952 * dwarf2out.c (mem_loc_descriptor): Likewise.
7953 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7954 (extract_bit_field, extract_low_bits): Likewise.
7955 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7956 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7957 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7958 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7959 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7960 * reload.c (find_reloads): Likewise.
7961 * reload1.c (alter_reg): Likewise.
7962 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7963 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7964 * tree-if-conv.c (predicate_mem_writes): Likewise.
7965 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7966 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7967 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7968 * valtrack.c (dead_debug_insert_temp): Likewise.
7969 * varasm.c (mergeable_constant_section): Likewise.
7970 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7973 Alan Hayward <alan.hayward@arm.com>
7974 David Sherwood <david.sherwood@arm.com>
7976 * expr.c (expand_assignment): Cope with polynomial mode sizes
7977 when assigning to a CONCAT.
7979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7980 Alan Hayward <alan.hayward@arm.com>
7981 David Sherwood <david.sherwood@arm.com>
7983 * machmode.h (mode_precision): Change from unsigned short to
7985 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7987 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7988 or if measurement_type is not polynomial.
7989 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7990 in which the mode is already known to be a scalar_int_mode.
7991 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7992 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7994 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7995 for GET_MODE_PRECISION.
7996 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7997 for GET_MODE_PRECISION.
7998 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8000 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8001 (expand_field_assignment, make_extraction): Likewise.
8002 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8003 (get_last_value): Likewise.
8004 * convert.c (convert_to_integer_1): Likewise.
8005 * cse.c (cse_insn): Likewise.
8006 * expr.c (expand_expr_real_1): Likewise.
8007 * lra-constraints.c (simplify_operand_subreg): Likewise.
8008 * optabs-query.c (can_atomic_load_p): Likewise.
8009 * optabs.c (expand_atomic_load): Likewise.
8010 (expand_atomic_store): Likewise.
8011 * ree.c (combine_reaching_defs): Likewise.
8012 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8013 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8014 * tree.h (type_has_mode_precision_p): Likewise.
8015 * ubsan.c (instrument_si_overflow): Likewise.
8017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8018 Alan Hayward <alan.hayward@arm.com>
8019 David Sherwood <david.sherwood@arm.com>
8021 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8022 polynomial numbers of units.
8023 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8024 (valid_vector_subparts_p): New function.
8025 (build_vector_type): Remove temporary shim and take the number
8026 of units as a poly_uint64 rather than an int.
8027 (build_opaque_vector_type): Take the number of units as a
8028 poly_uint64 rather than an int.
8029 * tree.c (build_vector_from_ctor): Handle polynomial
8030 TYPE_VECTOR_SUBPARTS.
8031 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8032 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8033 (build_vector_from_val): If the number of units is variable,
8034 use build_vec_duplicate_cst for constant operands and
8035 VEC_DUPLICATE_EXPR otherwise.
8036 (make_vector_type): Remove temporary is_constant ().
8037 (build_vector_type, build_opaque_vector_type): Take the number of
8038 units as a poly_uint64 rather than an int.
8039 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8041 * cfgexpand.c (expand_debug_expr): Likewise.
8042 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8043 (store_constructor, expand_expr_real_1): Likewise.
8044 (const_scalar_mask_from_tree): Likewise.
8045 * fold-const-call.c (fold_const_reduction): Likewise.
8046 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8047 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8048 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8049 (fold_relational_const): Likewise.
8050 (native_interpret_vector): Likewise. Change the size from an
8051 int to an unsigned int.
8052 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8053 TYPE_VECTOR_SUBPARTS.
8054 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8055 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8056 duplicating a non-constant operand into a variable-length vector.
8057 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8058 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8059 * ipa-icf.c (sem_variable::equals): Likewise.
8060 * match.pd: Likewise.
8061 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8062 * print-tree.c (print_node): Likewise.
8063 * stor-layout.c (layout_type): Likewise.
8064 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8065 * tree-cfg.c (verify_gimple_comparison): Likewise.
8066 (verify_gimple_assign_binary): Likewise.
8067 (verify_gimple_assign_ternary): Likewise.
8068 (verify_gimple_assign_single): Likewise.
8069 * tree-pretty-print.c (dump_generic_node): Likewise.
8070 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8071 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8072 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8073 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8074 (vect_shift_permute_load_chain): Likewise.
8075 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8076 (expand_vector_condition, optimize_vector_constructor): Likewise.
8077 (lower_vec_perm, get_compute_type): Likewise.
8078 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8079 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8080 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8081 (vect_recog_mask_conversion_pattern): Likewise.
8082 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8083 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8084 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8085 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8086 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8087 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8088 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8089 (supportable_widening_operation): Likewise.
8090 (supportable_narrowing_operation): Likewise.
8091 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8093 * varasm.c (output_constant): Likewise.
8095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8096 Alan Hayward <alan.hayward@arm.com>
8097 David Sherwood <david.sherwood@arm.com>
8099 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8100 so that both the length == 3 and length != 3 cases set up their
8101 own permute vectors. Add comments explaining why we know the
8102 number of elements is constant.
8103 (vect_permute_load_chain): Likewise.
8105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8106 Alan Hayward <alan.hayward@arm.com>
8107 David Sherwood <david.sherwood@arm.com>
8109 * machmode.h (mode_nunits): Change from unsigned char to
8111 (ONLY_FIXED_SIZE_MODES): New macro.
8112 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8113 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8114 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8116 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8117 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8118 or if measurement_type is not polynomial.
8119 * genmodes.c (ZERO_COEFFS): New macro.
8120 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8122 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8123 Use ZERO_COEFFS when emitting initializers.
8124 * data-streamer.h (bp_pack_poly_value): New function.
8125 (bp_unpack_poly_value): Likewise.
8126 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8127 for GET_MODE_NUNITS.
8128 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8129 for GET_MODE_NUNITS.
8130 * tree.c (make_vector_type): Remove temporary shim and make
8131 the real function take the number of units as a poly_uint64
8133 (build_vector_type_for_mode): Handle polynomial nunits.
8134 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8135 * emit-rtl.c (const_vec_series_p_1): Likewise.
8136 (gen_rtx_CONST_VECTOR): Likewise.
8137 * fold-const.c (test_vec_duplicate_folding): Likewise.
8138 * genrecog.c (validate_pattern): Likewise.
8139 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8140 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8141 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8142 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8143 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8144 * rtlanal.c (subreg_get_info): Likewise.
8145 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8146 (vect_grouped_load_supported): Likewise.
8147 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8148 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8149 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8150 (simplify_const_unary_operation, simplify_binary_operation_1)
8151 (simplify_const_binary_operation, simplify_ternary_operation)
8152 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8153 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8154 instead of CONST_VECTOR_NUNITS.
8155 * varasm.c (output_constant_pool_2): Likewise.
8156 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8157 explicit-encoded elements in the XVEC for variable-length vectors.
8159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8161 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8164 Alan Hayward <alan.hayward@arm.com>
8165 David Sherwood <david.sherwood@arm.com>
8167 * coretypes.h (fixed_size_mode): Declare.
8168 (fixed_size_mode_pod): New typedef.
8169 * builtins.h (target_builtins::x_apply_args_mode)
8170 (target_builtins::x_apply_result_mode): Change type to
8171 fixed_size_mode_pod.
8172 * builtins.c (apply_args_size, apply_result_size, result_vector)
8173 (expand_builtin_apply_args_1, expand_builtin_apply)
8174 (expand_builtin_return): Update accordingly.
8176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8178 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8179 * cselib.c (cselib_hash_rtx): Likewise.
8180 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8181 CONST_VECTOR encoding.
8183 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8184 Jeff Law <law@redhat.com>
8187 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8188 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8189 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8190 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8193 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8194 explicitly probe *sp in a noreturn function if there were any callee
8195 register saves or frame pointer is needed.
8197 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8200 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8201 BLKmode for ternary, binary or unary expressions.
8204 * var-tracking.c (delete_vta_debug_insn): New inline function.
8205 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8206 insns from get_insns () to NULL instead of each bb separately.
8207 Use delete_vta_debug_insn. No longer static.
8208 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8209 delete_vta_debug_insns callers.
8210 * rtl.h (delete_vta_debug_insns): Declare.
8211 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8212 instead of variable_tracking_main.
8214 2018-01-03 Martin Sebor <msebor@redhat.com>
8216 PR tree-optimization/83603
8217 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8218 arguments past the endof the argument list in functions declared
8219 without a prototype.
8220 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8221 Avoid checking when arguments are null.
8223 2018-01-03 Martin Sebor <msebor@redhat.com>
8226 * doc/extend.texi (attribute const): Fix a typo.
8227 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8228 issuing -Wsuggest-attribute for void functions.
8230 2018-01-03 Martin Sebor <msebor@redhat.com>
8232 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8233 offset_int::from instead of wide_int::to_shwi.
8234 (maybe_diag_overlap): Remove assertion.
8235 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8236 * gimple-ssa-sprintf.c (format_directive): Same.
8237 (parse_directive): Same.
8238 (sprintf_dom_walker::compute_format_length): Same.
8239 (try_substitute_return_value): Same.
8241 2018-01-03 Jeff Law <law@redhat.com>
8244 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8245 non-constant residual for zero at runtime and avoid probing in
8246 that case. Reorganize code for trailing problem to mirror handling
8249 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8251 PR tree-optimization/83501
8252 * tree-ssa-strlen.c (get_string_cst): New.
8253 (handle_char_store): Call get_string_cst.
8255 2018-01-03 Martin Liska <mliska@suse.cz>
8257 PR tree-optimization/83593
8258 * tree-ssa-strlen.c: Include tree-cfg.h.
8259 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8260 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8261 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8263 (strlen_dom_walker::before_dom_children): Call
8264 gimple_purge_dead_eh_edges. Dump tranformation with details
8266 (strlen_dom_walker::before_dom_children): Update call by adding
8267 new argument cleanup_eh.
8268 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8270 2018-01-03 Martin Liska <mliska@suse.cz>
8273 * cif-code.def (VARIADIC_THUNK): New enum value.
8274 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8277 2018-01-03 Jan Beulich <jbeulich@suse.com>
8279 * sse.md (mov<mode>_internal): Tighten condition for when to use
8280 vmovdqu<ssescalarsize> for TI and OI modes.
8282 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8284 Update copyright years.
8286 2018-01-03 Martin Liska <mliska@suse.cz>
8289 * ipa-visibility.c (function_and_variable_visibility): Skip
8290 functions with noipa attribure.
8292 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8294 * gcc.c (process_command): Update copyright notice dates.
8295 * gcov-dump.c (print_version): Ditto.
8296 * gcov.c (print_version): Ditto.
8297 * gcov-tool.c (print_version): Ditto.
8298 * gengtype.c (create_file): Ditto.
8299 * doc/cpp.texi: Bump @copying's copyright year.
8300 * doc/cppinternals.texi: Ditto.
8301 * doc/gcc.texi: Ditto.
8302 * doc/gccint.texi: Ditto.
8303 * doc/gcov.texi: Ditto.
8304 * doc/install.texi: Ditto.
8305 * doc/invoke.texi: Ditto.
8307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8309 * vector-builder.h (vector_builder::m_full_nelts): Change from
8310 unsigned int to poly_uint64.
8311 (vector_builder::full_nelts): Update prototype accordingly.
8312 (vector_builder::new_vector): Likewise.
8313 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8314 (vector_builder::operator ==): Likewise.
8315 (vector_builder::finalize): Likewise.
8316 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8317 Take the number of elements as a poly_uint64 rather than an
8319 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8320 from unsigned int to poly_uint64.
8321 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8322 (vec_perm_indices::new_vector): Likewise.
8323 (vec_perm_indices::length): Likewise.
8324 (vec_perm_indices::nelts_per_input): Likewise.
8325 (vec_perm_indices::input_nelts): Likewise.
8326 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8327 number of elements per input as a poly_uint64 rather than an
8328 unsigned int. Use the original encoding for variable-length
8329 vectors, rather than clamping each individual element.
8330 For the second and subsequent elements in each pattern,
8331 clamp the step and base before clamping their sum.
8332 (vec_perm_indices::series_p): Handle polynomial element counts.
8333 (vec_perm_indices::all_in_range_p): Likewise.
8334 (vec_perm_indices_to_tree): Likewise.
8335 (vec_perm_indices_to_rtx): Likewise.
8336 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8337 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8338 (tree_vector_builder::new_binary_operation): Handle polynomial
8339 element counts. Return false if we need to know the number
8340 of elements at compile time.
8341 * fold-const.c (fold_vec_perm): Punt if the number of elements
8342 isn't known at compile time.
8344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8346 * vec-perm-indices.h (vec_perm_builder): Change element type
8347 from HOST_WIDE_INT to poly_int64.
8348 (vec_perm_indices::element_type): Update accordingly.
8349 (vec_perm_indices::clamp): Handle polynomial element_types.
8350 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8351 (vec_perm_indices::all_in_range_p): Likewise.
8352 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8354 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8355 polynomial vec_perm_indices element types.
8356 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8357 * fold-const.c (fold_vec_perm): Likewise.
8358 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8359 * tree-vect-generic.c (lower_vec_perm): Likewise.
8360 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8361 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8362 element type to HOST_WIDE_INT.
8364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8365 Alan Hayward <alan.hayward@arm.com>
8366 David Sherwood <david.sherwood@arm.com>
8368 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8369 rather than an int. Use plus_constant.
8370 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8371 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8374 Alan Hayward <alan.hayward@arm.com>
8375 David Sherwood <david.sherwood@arm.com>
8377 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8378 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
8385 mode sizes. Require a constant size for BLKmode parameters
8386 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8387 forces a parameter to be padded at the lsb end in order to
8388 fill a complete number of words, require the parameter size
8389 to be ordered wrt UNITS_PER_WORD.
8391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8392 Alan Hayward <alan.hayward@arm.com>
8393 David Sherwood <david.sherwood@arm.com>
8395 * reload1.c (spill_stack_slot_width): Change element type
8396 from unsigned int to poly_uint64_pod.
8397 (alter_reg): Treat mode sizes as polynomial.
8399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8400 Alan Hayward <alan.hayward@arm.com>
8401 David Sherwood <david.sherwood@arm.com>
8403 * reload.c (complex_word_subreg_p): New function.
8404 (reload_inner_reg_of_subreg, push_reload): Use it.
8406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8407 Alan Hayward <alan.hayward@arm.com>
8408 David Sherwood <david.sherwood@arm.com>
8410 * lra-constraints.c (process_alt_operands): Reject matched
8411 operands whose sizes aren't ordered.
8412 (match_reload): Refer to this check here.
8414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8415 Alan Hayward <alan.hayward@arm.com>
8416 David Sherwood <david.sherwood@arm.com>
8418 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8419 that the mode size is in the set {1, 2, 4, 8, 16}.
8421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8422 Alan Hayward <alan.hayward@arm.com>
8423 David Sherwood <david.sherwood@arm.com>
8425 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8426 Use plus_constant instead of gen_rtx_PLUS.
8428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8429 Alan Hayward <alan.hayward@arm.com>
8430 David Sherwood <david.sherwood@arm.com>
8432 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8433 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8434 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8435 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8436 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8437 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8438 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8439 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8440 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8441 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8443 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8444 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8445 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8446 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8447 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8448 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8449 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8450 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8451 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8452 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8454 * expr.c (emit_move_resolve_push): Treat the input and result
8455 of PUSH_ROUNDING as a poly_int64.
8456 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8457 (emit_push_insn): Likewise.
8458 * lra-eliminations.c (mark_not_eliminable): Likewise.
8459 * recog.c (push_operand): Likewise.
8460 * reload1.c (elimination_effects): Likewise.
8461 * rtlanal.c (nonzero_bits1): Likewise.
8462 * calls.c (store_one_arg): Likewise. Require the padding to be
8463 known at compile time.
8465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8466 Alan Hayward <alan.hayward@arm.com>
8467 David Sherwood <david.sherwood@arm.com>
8469 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8470 Use plus_constant instead of gen_rtx_PLUS.
8472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8473 Alan Hayward <alan.hayward@arm.com>
8474 David Sherwood <david.sherwood@arm.com>
8476 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8480 Alan Hayward <alan.hayward@arm.com>
8481 David Sherwood <david.sherwood@arm.com>
8483 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8484 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8485 via stack temporaries. Treat the mode size as polynomial too.
8487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8488 Alan Hayward <alan.hayward@arm.com>
8489 David Sherwood <david.sherwood@arm.com>
8491 * expr.c (expand_expr_real_2): When handling conversions involving
8492 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8493 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8494 as a poly_uint64 too.
8496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8497 Alan Hayward <alan.hayward@arm.com>
8498 David Sherwood <david.sherwood@arm.com>
8500 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8503 Alan Hayward <alan.hayward@arm.com>
8504 David Sherwood <david.sherwood@arm.com>
8506 * combine.c (can_change_dest_mode): Handle polynomial
8507 REGMODE_NATURAL_SIZE.
8508 * expmed.c (store_bit_field_1): Likewise.
8509 * expr.c (store_constructor): Likewise.
8510 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8511 and polynomial REGMODE_NATURAL_SIZE.
8512 (gen_lowpart_common): Likewise.
8513 * reginfo.c (record_subregs_of_mode): Likewise.
8514 * rtlanal.c (read_modify_subreg_p): Likewise.
8516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8517 Alan Hayward <alan.hayward@arm.com>
8518 David Sherwood <david.sherwood@arm.com>
8520 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8521 numbers of elements.
8523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8524 Alan Hayward <alan.hayward@arm.com>
8525 David Sherwood <david.sherwood@arm.com>
8527 * match.pd: Cope with polynomial numbers of vector elements.
8529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8530 Alan Hayward <alan.hayward@arm.com>
8531 David Sherwood <david.sherwood@arm.com>
8533 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8534 in a POINTER_PLUS_EXPR.
8536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8537 Alan Hayward <alan.hayward@arm.com>
8538 David Sherwood <david.sherwood@arm.com>
8540 * omp-simd-clone.c (simd_clone_subparts): New function.
8541 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8542 (ipa_simd_modify_function_body): Likewise.
8544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8545 Alan Hayward <alan.hayward@arm.com>
8546 David Sherwood <david.sherwood@arm.com>
8548 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8549 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8550 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8551 (expand_vector_condition, vector_element): Likewise.
8552 (subparts_gt): New function.
8553 (get_compute_type): Use subparts_gt.
8554 (count_type_subparts): Delete.
8555 (expand_vector_operations_1): Use subparts_gt instead of
8556 count_type_subparts.
8558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8559 Alan Hayward <alan.hayward@arm.com>
8560 David Sherwood <david.sherwood@arm.com>
8562 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8563 (vect_compile_time_alias): ...this new function. Do the calculation
8564 on poly_ints rather than trees.
8565 (vect_prune_runtime_alias_test_list): Update call accordingly.
8567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8568 Alan Hayward <alan.hayward@arm.com>
8569 David Sherwood <david.sherwood@arm.com>
8571 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8573 (vect_schedule_slp_instance): Likewise.
8575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8576 Alan Hayward <alan.hayward@arm.com>
8577 David Sherwood <david.sherwood@arm.com>
8579 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8580 constant and extern definitions for variable-length vectors.
8581 (vect_get_constant_vectors): Note that the number of units
8582 is known to be constant.
8584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8585 Alan Hayward <alan.hayward@arm.com>
8586 David Sherwood <david.sherwood@arm.com>
8588 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8589 of units as polynomial. Choose between WIDE and NARROW based
8592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8593 Alan Hayward <alan.hayward@arm.com>
8594 David Sherwood <david.sherwood@arm.com>
8596 * tree-vect-stmts.c (simd_clone_subparts): New function.
8597 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8600 Alan Hayward <alan.hayward@arm.com>
8601 David Sherwood <david.sherwood@arm.com>
8603 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8604 vectors as polynomial. Use build_index_vector for
8607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8608 Alan Hayward <alan.hayward@arm.com>
8609 David Sherwood <david.sherwood@arm.com>
8611 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8612 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8613 for variable-length vectors.
8614 (vectorizable_mask_load_store): Treat the number of units as
8615 polynomial, asserting that it is constant if the condition has
8616 already been enforced.
8617 (vectorizable_store, vectorizable_load): Likewise.
8619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8620 Alan Hayward <alan.hayward@arm.com>
8621 David Sherwood <david.sherwood@arm.com>
8623 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8624 of units as polynomial. Punt if we can't tell at compile time
8625 which vector contains the final result.
8627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8628 Alan Hayward <alan.hayward@arm.com>
8629 David Sherwood <david.sherwood@arm.com>
8631 * tree-vect-loop.c (vectorizable_induction): Treat the number
8632 of units as polynomial. Punt on SLP inductions. Use an integer
8633 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8634 cast of such a series for variable-length floating-point
8637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8638 Alan Hayward <alan.hayward@arm.com>
8639 David Sherwood <david.sherwood@arm.com>
8641 * tree.h (build_index_vector): Declare.
8642 * tree.c (build_index_vector): New function.
8643 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8644 of units as polynomial, forcibly converting it to a constant if
8645 vectorizable_reduction has already enforced the condition.
8646 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8647 to create a {1,2,3,...} vector.
8648 (vectorizable_reduction): Treat the number of units as polynomial.
8649 Choose vectype_in based on the largest scalar element size rather
8650 than the smallest number of units. Enforce the restrictions
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8658 number of units as polynomial.
8660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8661 Alan Hayward <alan.hayward@arm.com>
8662 David Sherwood <david.sherwood@arm.com>
8664 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8665 * target.def (autovectorize_vector_sizes): Return the vector sizes
8666 by pointer, using vector_sizes rather than a bitmask.
8667 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8668 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8669 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8671 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8672 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8673 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8674 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8675 * omp-general.c (omp_max_vf): Likewise.
8676 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8677 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8678 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8679 * tree-vect-slp.c (vect_slp_bb): Likewise.
8680 * doc/tm.texi: Regenerate.
8681 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8683 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8684 the vector size as a poly_uint64 rather than an unsigned int.
8685 (current_vector_size): Change from an unsigned int to a poly_uint64.
8686 (get_vectype_for_scalar_type): Update accordingly.
8687 * tree.h (build_truth_vector_type): Take the size and number of
8688 units as a poly_uint64 rather than an unsigned int.
8689 (build_vector_type): Add a temporary overload that takes
8690 the number of units as a poly_uint64 rather than an unsigned int.
8691 * tree.c (make_vector_type): Likewise.
8692 (build_truth_vector_type): Take the number of units as a poly_uint64
8693 rather than an unsigned int.
8695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8696 Alan Hayward <alan.hayward@arm.com>
8697 David Sherwood <david.sherwood@arm.com>
8699 * target.def (get_mask_mode): Take the number of units and length
8700 as poly_uint64s rather than unsigned ints.
8701 * targhooks.h (default_get_mask_mode): Update accordingly.
8702 * targhooks.c (default_get_mask_mode): Likewise.
8703 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8704 * doc/tm.texi: Regenerate.
8706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8707 Alan Hayward <alan.hayward@arm.com>
8708 David Sherwood <david.sherwood@arm.com>
8710 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8711 * omp-general.c (omp_max_vf): Likewise.
8712 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8713 (expand_omp_simd): Handle polynomial safelen.
8714 * omp-low.c (omplow_simd_context): Add a default constructor.
8715 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8716 (lower_rec_simd_input_clauses): Update accordingly.
8717 (lower_rec_input_clauses): Likewise.
8719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8720 Alan Hayward <alan.hayward@arm.com>
8721 David Sherwood <david.sherwood@arm.com>
8723 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8724 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8725 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8726 (vect_analyze_slp_cost): Likewise.
8727 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8728 (vect_model_load_cost): Likewise.
8730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8731 Alan Hayward <alan.hayward@arm.com>
8732 David Sherwood <david.sherwood@arm.com>
8734 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8735 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8736 from an unsigned int * to a poly_uint64_pod *.
8737 (calculate_unrolling_factor): New function.
8738 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8741 Alan Hayward <alan.hayward@arm.com>
8742 David Sherwood <david.sherwood@arm.com>
8744 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8745 from an unsigned int to a poly_uint64.
8746 (_loop_vec_info::slp_unrolling_factor): Likewise.
8747 (_loop_vec_info::vectorization_factor): Change from an int
8749 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8750 (vect_get_num_vectors): New function.
8751 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8752 (vect_get_num_copies): Use vect_get_num_vectors.
8753 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8754 to an unsigned int *.
8755 (vect_analyze_data_refs): Change min_vf from an int * to a
8757 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8758 than an unsigned HOST_WIDE_INT.
8759 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8760 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8761 to an unsigned int *.
8762 (vect_analyze_data_ref_dependences): Likewise.
8763 (vect_compute_data_ref_alignment): Handle polynomial vf.
8764 (vect_enhance_data_refs_alignment): Likewise.
8765 (vect_prune_runtime_alias_test_list): Likewise.
8766 (vect_shift_permute_load_chain): Likewise.
8767 (vect_supportable_dr_alignment): Likewise.
8768 (dependence_distance_ge_vf): Take the vectorization factor as a
8769 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8770 (vect_analyze_data_refs): Change min_vf from an int * to a
8772 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8773 vfm1 as a poly_uint64 rather than an int. Make the same change
8774 for the returned bound_scalar.
8775 (vect_gen_vector_loop_niters): Handle polynomial vf.
8776 (vect_do_peeling): Likewise. Update call to
8777 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8778 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8780 * tree-vect-loop.c (vect_determine_vectorization_factor)
8781 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8782 (vect_get_known_peeling_cost): Likewise.
8783 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8784 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8785 (vect_transform_loop): Likewise. Use the lowest possible VF when
8786 updating the upper bounds of the loop.
8787 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8789 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8790 polynomial unroll factors.
8791 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8792 (vect_make_slp_decision): Likewise.
8793 (vect_supported_load_permutation_p): Likewise, and polynomial
8795 (vect_analyze_slp_cost): Handle polynomial vf.
8796 (vect_slp_analyze_node_operations): Likewise.
8797 (vect_slp_analyze_bb_1): Likewise.
8798 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8799 than an unsigned HOST_WIDE_INT.
8800 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8801 (vectorizable_load): Handle polynomial vf.
8802 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8804 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8807 Alan Hayward <alan.hayward@arm.com>
8808 David Sherwood <david.sherwood@arm.com>
8810 * match.pd: Handle bit operations involving three constants
8811 and try to fold one pair.
8813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8815 * tree-vect-loop-manip.c: Include gimple-fold.h.
8816 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8817 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8818 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8819 Add a path that uses a step of VF instead of 1, but disable it
8821 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8822 and niters_no_overflow parameters. Update calls to
8823 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8824 Create a new SSA name if the latter choses to use a ste other
8825 than zero, and return it via niters_vector_mult_vf_var.
8826 * tree-vect-loop.c (vect_transform_loop): Update calls to
8827 vect_do_peeling, vect_gen_vector_loop_niters and
8828 slpeel_make_loop_iterate_ntimes.
8829 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8830 (vect_gen_vector_loop_niters): Update declarations after above changes.
8832 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8834 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8835 128-bit round to integer instructions.
8836 (ceil<mode>2): Likewise.
8837 (btrunc<mode>2): Likewise.
8838 (round<mode>2): Likewise.
8840 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8842 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8843 unaligned VSX load/store on P8/P9.
8844 (expand_block_clear): Allow the use of unaligned VSX
8845 load/store on P8/P9.
8847 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8849 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8851 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8852 swap associated with both a load and a store.
8854 2018-01-02 Andrew Waterman <andrew@sifive.com>
8856 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8857 * config/riscv/riscv.md (clear_cache): Use it.
8859 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8861 * web.c: Remove out-of-date comment.
8863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8865 * expr.c (fixup_args_size_notes): Check that any existing
8866 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8867 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8868 (emit_single_push_insn): ...here.
8870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8872 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8873 (const_vector_encoded_nelts): New function.
8874 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8875 (const_vector_int_elt, const_vector_elt): Declare.
8876 * emit-rtl.c (const_vector_int_elt_1): New function.
8877 (const_vector_elt): Likewise.
8878 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8879 of CONST_VECTOR_ELT.
8881 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8883 * expr.c: Include rtx-vector-builder.h.
8884 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8885 directly on the tree encoding.
8886 (const_vector_from_tree): Likewise.
8887 * optabs.c: Include rtx-vector-builder.h.
8888 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8889 sequence of "u" values.
8890 * vec-perm-indices.c: Include rtx-vector-builder.h.
8891 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8892 directly on the vec_perm_indices encoding.
8894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8896 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8897 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8898 * rtx-vector-builder.h: New file.
8899 * rtx-vector-builder.c: Likewise.
8900 * rtl.h (rtx_def::u2): Add a const_vector field.
8901 (CONST_VECTOR_NPATTERNS): New macro.
8902 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8903 (CONST_VECTOR_DUPLICATE_P): Likewise.
8904 (CONST_VECTOR_STEPPED_P): Likewise.
8905 (CONST_VECTOR_ENCODED_ELT): Likewise.
8906 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8907 (unwrap_const_vec_duplicate): Likewise.
8908 (const_vec_series_p): Check for a non-duplicated vector encoding.
8909 Say that the function only returns true for integer vectors.
8910 * emit-rtl.c: Include rtx-vector-builder.h.
8911 (gen_const_vec_duplicate_1): Delete.
8912 (gen_const_vector): Call gen_const_vec_duplicate instead of
8913 gen_const_vec_duplicate_1.
8914 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8915 (gen_const_vec_duplicate): Use rtx_vector_builder.
8916 (gen_const_vec_series): Likewise.
8917 (gen_rtx_CONST_VECTOR): Likewise.
8918 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8919 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8920 Build a new vector rather than modifying a CONST_VECTOR in-place.
8921 (handle_special_swappables): Update call accordingly.
8922 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8923 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8924 Build a new vector rather than modifying a CONST_VECTOR in-place.
8925 (handle_special_swappables): Update call accordingly.
8927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8929 * simplify-rtx.c (simplify_const_binary_operation): Use
8930 CONST_VECTOR_ELT instead of XVECEXP.
8932 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8934 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8935 the selector elements to be different from the data elements
8936 if the selector is a VECTOR_CST.
8937 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8938 ssizetype for the selector.
8940 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8942 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8943 before testing each element individually.
8944 * tree-vect-generic.c (lower_vec_perm): Likewise.
8946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8948 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8949 * selftest-run-tests.c (selftest::run_tests): Call it.
8950 * vector-builder.h (vector_builder::operator ==): New function.
8951 (vector_builder::operator !=): Likewise.
8952 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8953 (vec_perm_indices::all_from_input_p): New function.
8954 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8955 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8956 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8957 instead of reading the VECTOR_CST directly. Detect whether both
8958 vector inputs are the same before constructing the vec_perm_indices,
8959 and update the number of inputs argument accordingly. Use the
8960 utility functions added above. Only construct sel2 if we need to.
8962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8964 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8965 the broadcast of the low byte.
8966 (expand_mult_highpart): Use an explicit encoding for the permutes.
8967 * optabs-query.c (can_mult_highpart_p): Likewise.
8968 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8969 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8970 (vectorizable_bswap): Likewise.
8971 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8972 explicit encoding for the power-of-2 permutes.
8973 (vect_permute_store_chain): Likewise.
8974 (vect_grouped_load_supported): Likewise.
8975 (vect_permute_load_chain): Likewise.
8977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8979 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8980 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8981 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8982 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8983 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8984 (vect_gen_perm_mask_any): Likewise.
8986 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8988 * int-vector-builder.h: New file.
8989 * vec-perm-indices.h: Include int-vector-builder.h.
8990 (vec_perm_indices): Redefine as an int_vector_builder.
8991 (auto_vec_perm_indices): Delete.
8992 (vec_perm_builder): Redefine as a stand-alone class.
8993 (vec_perm_indices::vec_perm_indices): New function.
8994 (vec_perm_indices::clamp): Likewise.
8995 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8996 (vec_perm_indices::new_vector): New function.
8997 (vec_perm_indices::new_expanded_vector): Update for new
8998 vec_perm_indices class.
8999 (vec_perm_indices::rotate_inputs): New function.
9000 (vec_perm_indices::all_in_range_p): Operate directly on the
9001 encoded form, without computing elided elements.
9002 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9003 encoding. Update for new vec_perm_indices class.
9004 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9005 the given vec_perm_builder.
9006 (expand_vec_perm_var): Update vec_perm_builder constructor.
9007 (expand_mult_highpart): Use vec_perm_builder instead of
9008 auto_vec_perm_indices.
9009 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9010 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9011 or double series encoding as appropriate.
9012 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9013 vec_perm_indices instead of auto_vec_perm_indices.
9014 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9015 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9016 (vect_permute_store_chain): Likewise.
9017 (vect_grouped_load_supported): Likewise.
9018 (vect_permute_load_chain): Likewise.
9019 (vect_shift_permute_load_chain): Likewise.
9020 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9021 (vect_transform_slp_perm_load): Likewise.
9022 (vect_schedule_slp_instance): Likewise.
9023 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9024 (vectorizable_mask_load_store): Likewise.
9025 (vectorizable_bswap): Likewise.
9026 (vectorizable_store): Likewise.
9027 (vectorizable_load): Likewise.
9028 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9029 vec_perm_indices instead of auto_vec_perm_indices. Use
9030 tree_to_vec_perm_builder to read the vector from a tree.
9031 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9032 vec_perm_builder instead of a vec_perm_indices.
9033 (have_whole_vector_shift): Use vec_perm_builder and
9034 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9035 truncation to calc_vec_perm_mask_for_shift.
9036 (vect_create_epilog_for_reduction): Likewise.
9037 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9038 from auto_vec_perm_indices to vec_perm_indices.
9039 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9040 instead of changing individual elements.
9041 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9042 the vector in d.perm.
9043 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9044 from auto_vec_perm_indices to vec_perm_indices.
9045 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9046 instead of changing individual elements.
9047 (arm_vectorize_vec_perm_const): Use new_vector to install
9048 the vector in d.perm.
9049 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9050 Update vec_perm_builder constructor.
9051 (rs6000_expand_interleave): Likewise.
9052 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9053 (rs6000_expand_interleave): Likewise.
9055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9057 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9058 to qimode could truncate the indices.
9059 * optabs.c (expand_vec_perm_var): Likewise.
9061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9063 * Makefile.in (OBJS): Add vec-perm-indices.o.
9064 * vec-perm-indices.h: New file.
9065 * vec-perm-indices.c: Likewise.
9066 * target.h (vec_perm_indices): Replace with a forward class
9068 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9069 * optabs.h: Include vec-perm-indices.h.
9070 (expand_vec_perm): Delete.
9071 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9072 (expand_vec_perm_const): Declare.
9073 * target.def (vec_perm_const_ok): Replace with...
9074 (vec_perm_const): ...this new hook.
9075 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9076 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9077 * doc/tm.texi: Regenerate.
9078 * optabs.def (vec_perm_const): Delete.
9079 * doc/md.texi (vec_perm_const): Likewise.
9080 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9081 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9082 expand_vec_perm for constant permutation vectors. Assert that
9083 the mode of variable permutation vectors is the integer equivalent
9084 of the mode that is being permuted.
9085 * optabs-query.h (selector_fits_mode_p): Declare.
9086 * optabs-query.c: Include vec-perm-indices.h.
9087 (selector_fits_mode_p): New function.
9088 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9089 is defined, instead of checking whether the vec_perm_const_optab
9090 exists. Use targetm.vectorize.vec_perm_const instead of
9091 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9092 fit in the vector mode before using a variable permute.
9093 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9094 vec_perm_indices instead of an rtx.
9095 (expand_vec_perm): Replace with...
9096 (expand_vec_perm_const): ...this new function. Take the selector
9097 as a vec_perm_indices rather than an rtx. Also take the mode of
9098 the selector. Update call to shift_amt_for_vec_perm_mask.
9099 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9100 Use vec_perm_indices::new_expanded_vector to expand the original
9101 selector into bytes. Check whether the indices fit in the vector
9102 mode before using a variable permute.
9103 (expand_vec_perm_var): Make global.
9104 (expand_mult_highpart): Use expand_vec_perm_const.
9105 * fold-const.c: Includes vec-perm-indices.h.
9106 * tree-ssa-forwprop.c: Likewise.
9107 * tree-vect-data-refs.c: Likewise.
9108 * tree-vect-generic.c: Likewise.
9109 * tree-vect-loop.c: Likewise.
9110 * tree-vect-slp.c: Likewise.
9111 * tree-vect-stmts.c: Likewise.
9112 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9114 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9115 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9116 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9117 (aarch64_vectorize_vec_perm_const): ...this new function.
9118 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9119 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9120 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9121 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9122 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9123 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9124 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9126 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9127 check for NEON modes.
9128 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9129 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9130 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9131 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9133 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9134 the old VEC_PERM_CONST conditions.
9135 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9136 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9137 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9138 (ia64_vectorize_vec_perm_const_ok): Merge into...
9139 (ia64_vectorize_vec_perm_const): ...this new function.
9140 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9141 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9142 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9143 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9144 * config/mips/mips.c (mips_expand_vec_perm_const)
9145 (mips_vectorize_vec_perm_const_ok): Merge into...
9146 (mips_vectorize_vec_perm_const): ...this new function.
9147 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9148 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9149 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9150 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9151 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9152 (rs6000_expand_vec_perm_const): Delete.
9153 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9155 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9156 (altivec_expand_vec_perm_const_le): Take each operand individually.
9157 Operate on constant selectors rather than rtxes.
9158 (altivec_expand_vec_perm_const): Likewise. Update call to
9159 altivec_expand_vec_perm_const_le.
9160 (rs6000_expand_vec_perm_const): Delete.
9161 (rs6000_vectorize_vec_perm_const_ok): Delete.
9162 (rs6000_vectorize_vec_perm_const): New function.
9163 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9164 an element count and rtx array.
9165 (rs6000_expand_extract_even): Update call accordingly.
9166 (rs6000_expand_interleave): Likewise.
9167 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9168 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9169 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9170 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9171 (rs6000_expand_vec_perm_const): Delete.
9172 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9173 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9174 (altivec_expand_vec_perm_const_le): Take each operand individually.
9175 Operate on constant selectors rather than rtxes.
9176 (altivec_expand_vec_perm_const): Likewise. Update call to
9177 altivec_expand_vec_perm_const_le.
9178 (rs6000_expand_vec_perm_const): Delete.
9179 (rs6000_vectorize_vec_perm_const_ok): Delete.
9180 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9181 reference to the SPE evmerge intructions.
9182 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9183 an element count and rtx array.
9184 (rs6000_expand_extract_even): Update call accordingly.
9185 (rs6000_expand_interleave): Likewise.
9186 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9187 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9189 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9191 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9193 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9194 vector mode and that that mode matches the mode of the data
9196 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9197 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9198 directly using expand_vec_perm_1 when forcing selectors into
9200 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9202 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9204 * optabs-query.h (can_vec_perm_p): Delete.
9205 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9206 * optabs-query.c (can_vec_perm_p): Split into...
9207 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9208 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9209 particular selector is valid.
9210 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9211 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9212 (vect_grouped_load_supported): Likewise.
9213 (vect_shift_permute_load_chain): Likewise.
9214 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9215 (vect_transform_slp_perm_load): Likewise.
9216 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9217 (vectorizable_bswap): Likewise.
9218 (vect_gen_perm_mask_checked): Likewise.
9219 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9220 implementations of variable permutation vectors into account
9221 when deciding which selector to use.
9222 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9223 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9224 with a false third argument.
9225 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9226 to test whether the constant selector is valid and can_vec_perm_var_p
9227 to test whether a variable selector is valid.
9229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9231 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9232 * optabs-query.c (can_vec_perm_p): Likewise.
9233 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9234 instead of vec_perm_indices.
9235 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9236 (vect_gen_perm_mask_checked): Likewise,
9237 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9238 (vect_gen_perm_mask_checked): Likewise,
9240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9242 * optabs-query.h (qimode_for_vec_perm): Declare.
9243 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9244 (qimode_for_vec_perm): ...this new function.
9245 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9247 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9249 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9250 does not have a conditional at the top.
9252 2018-01-02 Richard Biener <rguenther@suse.de>
9254 * ipa-inline.c (big_speedup_p): Fix expression.
9256 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9259 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9262 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9266 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9267 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9268 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9269 cond_taken_branch_cost 3->4.
9271 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9273 PR tree-optimization/83581
9274 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9275 TODO_cleanup_cfg if any changes have been made.
9278 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9279 convert_modes if target mode has the right side, but different mode
9283 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9284 last argument when extracting from CONCAT. If either from_real or
9285 from_imag is NULL, use expansion through memory. If result is not
9286 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9287 the parts directly to inner mode, if even that fails, use expansion
9291 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9292 check for bswap in mode rather than HImode and use that in expand_unop
9295 Copyright (C) 2018 Free Software Foundation, Inc.
9297 Copying and distribu