1 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11 and non BLKmode operands.
13 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
19 2018-03-23 Richard Biener <rguenther@suse.de>
22 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23 we are going to emit early debug for LTO.
25 2018-03-23 Jakub Jelinek <jakub@redhat.com>
28 * function.c (match_asm_constraints_1): Don't optimize if input
29 doesn't satisfy general_operand predicate for output's mode.
32 * alias.c (write_dependence_p): Don't require for x_canonicalized
33 non-VOIDmode if x has VOIDmode.
36 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
37 just don't try to optimize it rather than assert it never happens.
39 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
41 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
42 macro expansions for definition of ST_INTERNAL_<mode> and
43 LD_INTERNAL_<mode> builtins.
44 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
46 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
48 (altivec_expand_st_builtin): Likewise.
49 (altivec_expand_builtin): Remove calls to deleted functions.
50 (rs6000_address_for_altivec): Delete this function.
51 * config/rs6000/vector.md: Remove expands for
52 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
54 2018-03-22 Sudakshina Das <sudi.das@arm.com>
57 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
58 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
59 re-computing once computed.
60 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
61 (arm_init_machine_status): Initialize
62 machine->static_chain_stack_bytes.
64 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
67 * doc/extend.texi: Add four new prototypes for vec_ld.
68 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
69 definitions for more logical presentation.
70 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
71 entries for V1TI variants of __builtin_altivec_ld builtin.
72 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
73 handling of V1TI variant of LVX icode pattern.
74 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
75 (rs6000_gimple_fold_builtin): Likewise.
76 (altivec_init_builtins): Add code to define
77 __builtin_altivec_lvx_v1ti function.
79 2018-03-22 Jakub Jelinek <jakub@redhat.com>
82 * function.c (match_asm_constraints_1): Don't do the optimization
83 if input isn't a REG, SUBREG, MEM or constant.
85 2018-03-22 Tom de Vries <tom@codesourcery.com>
87 PR tree-optimization/84956
88 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
91 2018-03-22 Jakub Jelinek <jakub@redhat.com>
94 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
95 DECL_INITIAL (decl) to decl at the end.
96 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
99 2018-03-21 Joseph Myers <joseph@codesourcery.com>
101 * doc/extend.texi (__builtin_tgmath): Document when complex
102 integer types are treated as _Complex _Float64.
104 2018-03-21 Tom de Vries <tom@codesourcery.com>
106 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
108 2018-03-21 Jakub Jelinek <jakub@redhat.com>
110 PR tree-optimization/84960
111 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
112 if it is ENTRY block, move them into single succ of ENTRY in that case.
114 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
116 PR tree-optimization/84811
117 * poly-int.h (poly_span_traits): Remove the T3 parameter and
118 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
119 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
120 (known_subrange_p): Update accordingly. Cast each value involved
121 in the size comparison, rather than casting the result of the
124 2018-03-21 Jakub Jelinek <jakub@redhat.com>
126 PR tree-optimization/84982
127 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
128 by flipping the least significant bit rather than all bits from
129 bitpos to bitpos + bitsize - 1.
131 2018-03-21 Nathan Sidwell <nathan@acm.org>
133 * doc/extend.texi (Deprecated Features): Remove mention of
134 long-deleted deprecations.
136 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
139 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
140 * configure: Regenerate.
142 2018-03-21 Tom de Vries <tom@codesourcery.com>
144 PR tree-optimization/83126
145 * tree-parloops.c (num_phis): New function.
146 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
148 2018-03-21 Nathan Sidwell <nathan@acm.org>
150 * doc/extend.texi (Deprecated Features): Update deprecated flags,
151 mention anon-struct/union members and trailing attributes.
153 2018-03-21 Bin Cheng <bin.cheng@arm.com>
155 PR tree-optimization/84969
156 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
157 builtin memset partitions if they set different rhs values.
159 2018-03-21 Jakub Jelinek <jakub@redhat.com>
161 PR rtl-optimization/84989
162 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
163 VEC_DUPLICATE with scalar result mode.
165 2018-03-21 Martin Liska <mliska@suse.cz>
168 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
169 not intended return statement.
171 2018-03-21 Martin Liska <mliska@suse.cz>
174 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
175 (chkp_find_bound_slots_1): Limit number of iterations.
177 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
180 * Minor grammar fixes for x86 options.
182 2018-03-20 Jakub Jelinek <jakub@redhat.com>
185 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
186 holding REG_CFA_RESTORE notes, instead turn them into a USE.
188 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
191 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
192 (altivec_lvx_<mode>_1op): Likewise.
193 (altivec_stvx_<mode>_2op): Likewise.
194 (altivec_stvx_<mode>_1op): Likewise.
195 (altivec_lvx_<VM2:mode>): New define_expand.
196 (altivec_stvx_<VM2:mode>): Likewise.
197 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
198 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
199 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
200 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
201 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
202 (rs6000_gen_lvx): Likewise.
203 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
204 (altivec_expand_stv_builtin): Likewise.
205 (altivec_expand_builtin): Likewise.
206 * config/rs6000/vector.md: Likewise.
208 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
211 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
214 2018-03-20 Richard Biener <rguenther@suse.de>
217 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
218 sign-conversions as zero, fall back to standard scalar_stmt
221 2018-03-20 Martin Liska <mliska@suse.cz>
224 * predict.c (rebuild_frequencies): Handle case when we have
225 PROFILE_ABSENT, but flag_guess_branch_prob is false.
227 2018-03-20 Jakub Jelinek <jakub@redhat.com>
230 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
231 flag_section_anchors.
232 * varasm.c (use_blocks_for_decl_p): Remove hack for
236 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
238 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
239 be created, use lowpart_subreg of operands[0] rather than operands[0]
241 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
242 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
243 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
244 and n constraint instead of aarch64_shift_imm_di and Usd.
245 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
246 (*aarch64_<optab>_reg_minus<mode>3): ... this.
248 2018-03-20 Sudakshina Das <sudi.das@arm.com>
251 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
252 to favor GPR over NEON registers.
253 (<shift>di3_neon): Likewise.
255 2018-03-20 Tom de Vries <tom@codesourcery.com>
258 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
259 (nvptx_process_pars): Emit bar.sync asap and alap.
261 2018-03-20 Tom de Vries <tom@codesourcery.com>
264 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
265 seen_label if seen_label is already set.
267 2018-03-20 Jakub Jelinek <jakub@redhat.com>
270 * config/i386/i386.c (fold_builtin_cpu): For features above 31
271 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
272 Use 1U instead of 1. Formatting fixes.
275 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
276 instead of TREE_TYPE (s1) for the return value.
278 2018-03-19 Jakub Jelinek <jakub@redhat.com>
280 PR tree-optimization/84946
281 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
282 bitsize + bitsize in poly_uint64 rather than poly_int64.
285 * dwarf2asm.c: Include fold-const.c.
286 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
287 of decl rather than decl itself.
289 PR rtl-optimization/84643
290 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
292 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
295 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
296 calling assemble_variable.
298 2018-03-19 Sudakshina Das <sudi.das@arm.com>
301 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
302 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
304 2018-03-19 Jim Wilson <jimw@sifive.com>
307 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
308 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
309 (riscv_first_stack_step): Likewise.
310 (riscv_option_override): Use STACK_BOUNDARY instead of
312 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
314 (BIGGEST_ALIGNMENT): Set to 128.
315 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
316 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
319 2018-03-19 Richard Biener <rguenther@suse.de>
321 PR tree-optimization/84933
322 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
323 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
325 2018-03-19 Richard Biener <rguenther@suse.de>
327 PR tree-optimization/84859
328 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
329 (cond_if_else_store_replacement): Perform sinking operation on
330 single-store BBs regardless of MAX_STORES_TO_SINK setting.
331 Generalize what a BB with a single eligible store is.
333 2018-03-19 Richard Biener <rguenther@suse.de>
335 PR tree-optimization/84929
336 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
337 chrec_is_positive against non-chrec arg.
339 2018-03-19 Tamar Christina <tamar.christina@arm.com>
342 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
344 2018-03-18 Martin Liska <mliska@suse.cz>
346 PR rtl-optimization/84635
347 * regrename.c (build_def_use): Use matches_mode only when
350 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
352 PR tree-optimization/84913
353 * tree-vect-loop.c (vectorizable_reduction): Don't try to
354 vectorize chains of COND_EXPRs.
356 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
358 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
360 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
362 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
364 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
368 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
369 Kito Cheng <kito.cheng@gmail.com>
371 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
372 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
373 (nds32_adjust_reg_alloc_order): New function.
374 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
376 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
378 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
379 nds32_print_operand, nds32_print_operand_address): Use
380 HOST_WIDE_INT_PRINT_DEC instead.
382 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
384 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
386 2018-03-17 Jakub Jelinek <jakub@redhat.com>
389 * config/i386/i386.c (initial_ix86_tune_features,
390 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
392 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
393 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
394 rather than 1u << ix86_tune. Formatting fix.
395 (ix86_option_override_internal): Change ix86_arch_mask from
396 unsigned int to unsigned HOST_WIDE_INT, initialize to
397 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
398 (ix86_function_specific_restore): Likewise.
400 2018-03-16 Jakub Jelinek <jakub@redhat.com>
403 * postreload.c (reload_combine_recognize_pattern): Perform
404 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
405 truncate_int_for_mode the result for the destination's mode.
408 * hsa-gen.c (mem_type_for_type): Fix comment typo.
409 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
411 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
414 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
417 * lra-assigns.c (lra_split_hard_reg_for): Don't use
418 regno_allocno_class_array and sorted_pseudos.
419 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
420 insns where regno is used.
422 2018-03-16 Martin Liska <mliska@suse.cz>
425 * multiple_target.c (create_dispatcher_calls): Redirect
426 reference in the symbol table.
428 2018-03-16 Martin Liska <mliska@suse.cz>
431 * multiple_target.c (create_dispatcher_calls): Redirect also
434 2018-03-16 Jakub Jelinek <jakub@redhat.com>
438 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
439 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
440 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
442 2018-03-16 Julia Koval <julia.koval@intel.com>
444 * doc/invoke.texi (Skylake Server): Add CLWB.
445 Cannonlake): Remove CLWB.
447 2018-03-16 Jakub Jelinek <jakub@redhat.com>
449 PR tree-optimization/84841
450 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
452 (FLOAT_ONE_CONST_TYPE): Define.
453 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
454 (sort_by_operand_rank): Put entries with higher constant_type last
455 rather than first to match comments.
457 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
459 * config/nios2/nios2.md (movsi_internal): Fix thinko in
462 2018-03-15 Jakub Jelinek <jakub@redhat.com>
465 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
466 check and use address of target always.
468 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
471 * config/i386/i386.c (indirect_thunk_needed): Update comments.
472 (indirect_thunk_bnd_needed): Likewise.
473 (indirect_thunks_used): Likewise.
474 (indirect_thunks_bnd_used): Likewise.
475 (indirect_return_needed): New.
476 (indirect_return_bnd_needed): Likewise.
477 (output_indirect_thunk_function): Add a bool argument for
479 (output_indirect_thunk_function): Don't generate alias for
480 function return thunk.
481 (ix86_code_end): Call output_indirect_thunk_function to generate
482 function return thunks.
483 (ix86_output_function_return): Set indirect_return_bnd_needed
484 and indirect_return_needed instead of indirect_thunk_bnd_needed
485 and indirect_thunk_needed.
487 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
489 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
490 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
491 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
493 2018-03-15 David Malcolm <dmalcolm@redhat.com>
494 Paul Hua <paul.hua.gm@gmail.com>
497 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
499 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
501 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
502 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
505 2018-03-15 Tamar Christina <tamar.christina@arm.com>
508 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
509 instead of GET_MODE_SIZE when comparing Units.
511 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
514 * varasm.c (hash_section): Return an unchangeble hash value
515 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
516 Return !aarch64_can_use_per_function_literal_pools_p ().
518 2018-03-15 Jakub Jelinek <jakub@redhat.com>
521 * optabs.c (emit_conditional_move): Pass address of cmode's copy
522 rather than address of cmode as last argument to prepare_cmp_insn.
524 2018-03-15 Julia Koval <julia.koval@intel.com>
526 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
527 F_AVX512VNNI, F_AVX512BITALG): New.
529 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
532 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
533 insn for floating-point loads and stores.
535 2018-03-14 Carl Love <cel@us.ibm.com>
537 * config/rs6000/rs6000-c.c: Add macro definitions for
538 ALTIVEC_BUILTIN_VEC_PERMXOR.
539 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
540 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
541 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
542 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
544 * config/doc/extend.texi: Add prototypes for vec_permxor.
546 2018-03-14 David Malcolm <dmalcolm@redhat.com>
549 * diagnostic-show-locus.c (class layout_point): Convert m_line
550 from int to linenum_type.
551 (line_span::comparator): Use linenum "compare" function when
552 comparing line numbers.
553 (test_line_span): New function.
554 (layout_range::contains_point): Convert param "row" from int to
556 (layout_range::intersects_line_p): Likewise.
557 (layout::will_show_line_p): Likewise.
558 (layout::print_source_line): Likewise.
559 (layout::should_print_annotation_line_p): Likewise.
560 (layout::print_annotation_line): Likewise.
561 (layout::print_leading_fixits): Likewise.
562 (layout::annotation_line_showed_range_p): Likewise.
563 (struct line_corrections): Likewise for field m_row.
564 (line_corrections::line_corrections): Likewise for param "row".
565 (layout::print_trailing_fixits): Likewise.
566 (layout::get_state_at_point): Likewise.
567 (layout::get_x_bound_for_row): Likewise.
568 (layout::print_line): Likewise.
569 (diagnostic_show_locus): Likewise for locals "last_line" and
571 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
572 * input.c (selftest::test_linenum_comparisons): New function.
573 (selftest::input_c_tests): Call it.
574 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
575 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
576 * selftest.h (ASSERT_GT): New macro.
577 (ASSERT_GT_AT): New macro.
578 (ASSERT_LT): New macro.
579 (ASSERT_LT_AT): New macro.
581 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
583 PR rtl-optimization/84780
584 * combine.c (distribute_links): Don't make a link based on pc_rtx.
586 2018-03-14 Martin Liska <mliska@suse.cz>
588 * tree.c (record_node_allocation_statistics): Use
590 (get_stats_node_kind): New function extracted from
591 record_node_allocation_statistics.
592 (free_node): Use get_stats_node_kind.
594 2018-03-14 Richard Biener <rguenther@suse.de>
596 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
597 that the value-set of ANTIC_IN doesn't grow.
600 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
602 (BB_VISITED_WITH_VISITED_SUCCS): New define.
603 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
605 2018-03-14 Julia Koval <julia.koval@intel.com>
607 * config.gcc (icelake-client, icelake-server): New.
609 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
610 (initial_ix86_arch_features): Ditto.
611 (PTA_SKYLAKE): Add SGX.
612 (PTA_ICELAKE): Remove.
613 (PTA_ICELAKE_CLIENT): New.
614 (PTA_ICELAKE_SERVER): New.
615 (ix86_option_override_internal): Split up icelake on icelake client and
617 (get_builtin_code_for_version): Ditto.
618 (fold_builtin_cpu): Ditto.
619 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
620 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
621 * config/i386/i386.h (processor_type): Ditto.
622 * doc/invoke.texi: Ditto.
624 2018-03-14 Jakub Jelinek <jakub@redhat.com>
627 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
628 INTEGER_CST offset, add it together with bitpos / 8 and
629 sign extend based on POINTER_SIZE.
633 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
636 * config/i386/constraints.md (Yc): New register constraint.
637 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
638 Use Yc constraint for alternative 2 of operand 0. Remove
639 preferred_for_speed attribute.
641 2018-03-14 Richard Biener <rguenther@suse.de>
643 PR tree-optimization/84830
644 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
645 with the old one to avoid oscillations.
647 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
650 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
652 (assign_by_spills): Return a flag of reload assignment failure.
653 Do not process the reload assignment failures. Do not spill other
654 reload pseudos if they has the same reg class. Update n if
656 (lra_assign): Add a return arg. Set up from the result of
657 assign_by_spills call.
658 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
659 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
660 usage_insns if it is not NULL.
661 (spill_hard_reg_in_range): New function.
662 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
663 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
665 (lra_assign): Change prototype.
666 * lra.c (lra): Add code to deal with fails by splitting hard reg
669 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
671 * config/riscv/riscv.opt (mrelax): New option.
672 * config/riscv/riscv.c (riscv_file_start): Emit ".option
673 "norelax" when riscv_mrelax is disabled.
674 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
676 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
679 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
680 reassociation for int modes.
682 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
684 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
685 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
687 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
688 * config/aarch64/aarch64-sve.md
689 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
690 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
691 (*extend<mode><Vwide>2): Rename to...
692 (aarch64_sve_extend<mode><Vwide>2): ...this.
693 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
694 renaming the old pattern to...
695 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
697 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
698 define_expand, renaming the old pattern to...
699 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
700 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
701 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
702 account when deciding which SVE instruction the optab should use.
703 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
705 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
707 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
708 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
709 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
710 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
711 (tlsdesc_small_<mode>): Turn a define_expand and use
712 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
713 (tlsdesc_small_advsimd_<mode>): ...this.
714 (tlsdesc_small_sve_<mode>): New pattern.
716 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
718 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
719 (UNSPEC_UMUL_HIGHPART): New constants.
720 (MUL_HIGHPART): New int iteraor.
721 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
722 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
724 (*<su>mul<mode>3_highpart): New define_insn.
726 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
729 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
732 2018-03-13 Martin Liska <mliska@suse.cz>
735 * (sem_item_optimizer::sem_item_optimizer): Initialize new
737 (sem_item_optimizer::~sem_item_optimizer): Release it.
738 (sem_item_optimizer::merge_classes): Register variable aliases.
739 (sem_item_optimizer::fixup_pt_set): New function.
740 (sem_item_optimizer::fixup_points_to_sets): Likewise.
741 * ipa-icf.h: Declare new variables and functions.
743 2018-03-13 Jakub Jelinek <jakub@redhat.com>
746 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
747 integer_pow2p@2 and test integer_pow2p in condition.
748 (A < 0 ? C : 0): Similarly for @1.
751 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
752 characters starting at p contain '\0' character, don't look beyond
756 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
757 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
760 * reg-stack.c (change_stack): Change update_end var from int to
761 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
762 also call set_block_for_insn on the newly added insns and rescan.
765 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
769 * tree.c (stabilize_reference_1): Return save_expr (e) for
770 STATEMENT_LIST even if it doesn't have side-effects.
772 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
774 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
776 2018-03-12 Renlin Li <renlin.li@arm.com>
778 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
779 aarch64_output_scalar_simd_mov_immediate.
781 2018-03-12 Martin Sebor <msebor@redhat.com>
783 PR tree-optimization/83456
784 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
785 for perfectly overlapping calls to memcpy.
786 (gimple_fold_builtin_memory_chk): Same.
787 (gimple_fold_builtin_strcpy): Handle no-warning.
788 (gimple_fold_builtin_stxcpy_chk): Same.
789 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
791 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
793 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
794 parameter. Use it for SFmode.
795 (rs6000_function_arg_advance_1): Adjust.
796 (rs6000_function_arg): Adjust.
797 (rs6000_gimplify_va_arg): Pass false for that new parameter.
799 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
801 PR rtl-optimization/84169
802 PR rtl-optimization/84780
803 * combine.c (can_combine_p): Check for a 2-insn combination whether
804 the destination register is used between the two insns, too.
806 2018-03-12 Richard Biener <rguenther@suse.de>
808 PR tree-optimization/84803
809 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
810 for refs DR analysis didn't process.
812 2018-03-12 Richard Biener <rguenther@suse.de>
814 PR tree-optimization/84777
815 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
816 force-vectorize loops ignore whether we are optimizing for size.
818 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
820 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
821 (TARGET_MD_ASM_ADJUST): Define.
823 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
824 Kito Cheng <kito.cheng@gmail.com>
825 Chung-Ju Wu <jasonwucj@gmail.com>
827 * config/nds32/nds32.c (nds32_compute_stack_frame,
828 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
829 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
830 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
831 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
832 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
833 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
834 * config/nds32/nds32.md (prologue, epilogue): Use macro
835 NDS32_V3PUSH_AVAILABLE_P to do checking.
837 2018-03-11 Jakub Jelinek <jakub@redhat.com>
840 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
841 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
842 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
843 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
844 addition of most attributes on !orig_type_die or the attribute not
845 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
847 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
848 Chung-Ju Wu <jasonwucj@gmail.com>
850 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
852 * config/nds32/nds32.opt (mvh): New option.
854 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
855 Chung-Ju Wu <jasonwucj@gmail.com>
857 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
859 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
860 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
863 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
864 Chung-Ju Wu <jasonwucj@gmail.com>
866 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
868 * config/nds32/nds32-multiple.md (strlensi): New pattern.
869 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
871 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
872 Kito Cheng <kito.cheng@gmail.com>
873 Chung-Ju Wu <jasonwucj@gmail.com>
875 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
876 UNSPEC_FFMISM and UNSPEC_FLMISM.
877 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
878 for ffb, ffmism and flmism.
879 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
880 (unspec_ffmism): Ditto.
881 (unspec_flmism): Ditto.
882 (nds32_expand_builtin_impl): Check if string extension is available.
883 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
884 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
886 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
889 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
892 * lra-assigns.c (assign_by_spills): Return a flag of reload
893 assignment failure. Do not process the reload assignment
894 failures. Do not spill other reload pseudos if they has the same
896 (lra_assign): Add a return arg. Set up from the result of
897 assign_by_spills call.
898 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
899 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
900 usage_insns if it is not NULL.
901 (spill_hard_reg_in_range): New function.
902 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
903 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
905 (lra_assign): Change prototype.
906 * lra.c (lra): Add code to deal with fails by splitting hard reg
909 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
912 * config/i386/i386.opt: Replace Enforcment with Enforcement.
914 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
917 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
918 (dw_val_node): Add val_symbolic_view.
919 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
920 (symview_upper_bound): New.
921 (new_line_info_table): Initialize symviews_since_reset.
922 (dwarf2out_source_line): Count symviews_since_reset and set
924 (dw_val_equal_p): Handle symview.
925 (add_AT_symview): New.
926 (print_dw_val): Handle symview.
927 (attr_checksum, attr_checksum_ordered): Likewise.
928 (same_dw_val_p, size_of_die): Likewise.
929 (value_format, output_die): Likewise.
930 (add_high_low_attributes): Use add_AT_symview for entry_view.
931 (dwarf2out_finish): Reset symview_upper_bound, clear
934 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
937 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
938 Add strict argument and use it.
939 (rs6000_split_multireg_move): Update for new strict argument.
940 (mem_operand_gpr): Disallow all non-offsettable addresses.
941 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
943 2018-03-09 Jakub Jelinek <jakub@redhat.com>
946 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
947 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
948 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
951 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
952 decl, use remap_type if we want to use the type.
954 2018-03-09 Martin Sebor <msebor@redhat.com>
956 PR tree-optimization/84526
957 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
959 (builtin_access::generic_overlap): Be prepared to handle non-array
962 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
964 PR rtl-optimization/84682
965 * lra-constraints.c (process_address_1): Check is_address flag
966 for address constraints.
967 (process_alt_operands): Likewise.
968 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
969 preprocess_constraints.
970 * recog.h (preprocess_constraints): Add oploc parameter.
972 * recog.c (preprocess_constraints): Test address_operand for
973 CT_ADDRESS constraints.
975 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
978 * lra-assigns.c (assign_by_spills): Return a flag of reload
979 assignment failure. Do not process the reload assignment
980 failures. Do not spill other reload pseudos if they has the same
982 (lra_assign): Add a return arg. Set up from the result of
983 assign_by_spills call.
984 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
985 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
986 usage_insns if it is not NULL.
987 (spill_hard_reg_in_range): New function.
988 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
989 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
991 (lra_assign): Change prototype.
992 * lra.c (lra): Add code to deal with fails by splitting hard reg
995 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
998 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
999 Accept complain bool parameter. Only emit errors if it is true.
1000 (arm_parse_cpu_option_name): Likewise.
1001 (arm_target_thumb_only): Adjust callers of the above.
1002 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1003 prototype to take a default true bool parameter.
1004 (arm_parse_arch_option_name): Likewise.
1006 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1007 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1011 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1012 * configure: Regenerate.
1013 * configure.ac ("linker --version-script option"): New.
1014 ("linker soname option"): New.
1016 2018-03-09 Richard Biener <rguenther@suse.de>
1018 PR tree-optimization/84775
1019 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1020 immediate uses of predicate stmts and mark them modified.
1023 PR tree-optimization/84178
1024 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1026 (version_loop_for_if_conversion): Delay update_ssa call.
1027 (tree_if_conversion): Delay update_ssa until after predicate
1030 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1033 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1034 when the function accesses prior frames.
1036 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1039 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1040 gen_llsym, otherwise call maybe_gen_llsym.
1043 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1044 has ',' character inside of it.
1046 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1049 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1050 as clobbering CC_REGNUM.
1052 2018-03-08 Richard Biener <rguenther@suse.de>
1055 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1056 (follow_copies_to_constant): Do not follow SSA names registered
1059 2018-03-08 Richard Biener <rguenther@suse.de>
1061 PR tree-optimization/84178
1062 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1064 (version_loop_for_if_conversion): Delay update_ssa call.
1065 (tree_if_conversion): Delay update_ssa until after predicate
1068 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1070 PR tree-optimization/84178
1071 * tree-if-conv.c (release_bb_predicate): Remove the
1072 the assertion that the stmts have NULL use_ops.
1073 Discard the statements, asserting that they haven't
1074 yet been added to a BB.
1076 2018-03-08 Richard Biener <rguenther@suse.de>
1078 PR tree-optimization/84746
1079 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1080 (phi_translate): Pass in destination ANTIC_OUT set.
1081 (phi_translate_1): Likewise. For a simplified result lookup
1082 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1083 (phi_translate_set): Adjust.
1084 (do_pre_regular_insertion): Likewise.
1085 (do_pre_partial_partial_insertion): Likewise.
1087 2018-03-08 Martin Liska <mliska@suse.cz>
1089 PR gcov-profile/84735
1090 * doc/gcov.texi: Document usage of profile files.
1091 * gcov-io.h: Document changes in the format.
1093 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1097 * dwarf2out.c (struct dw_line_info_table): Update comments for
1099 (FORCE_RESET_NEXT_VIEW): New.
1100 (FORCE_RESETTING_VIEW_P): New.
1101 (RESETTING_VIEW_P): Check for -1 too.
1102 (ZERO_VIEW_P): Likewise.
1103 (new_line_info_table): Force-reset next view.
1104 (dwarf2out_begin_function): Likewise.
1105 (dwarf2out_source_line): Simplify zero_view_p initialization.
1106 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1107 view directly. Omit view when omitting .loc at line 0.
1109 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1111 PR tree-optimization/84740
1112 * tree-switch-conversion.c (process_switch): Call build_constructors
1113 only if info.phi_count is non-zero.
1115 PR tree-optimization/84739
1116 * tree-tailcall.c (find_tail_calls): Check call arguments against
1117 DECL_ARGUMENTS (current_function_decl) rather than
1118 DECL_ARGUMENTS (func) when checking for tail recursion.
1120 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1122 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1123 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1124 Volker Reichelt's entry and add entries for people that perform
1125 GCC fuzzy testing and report numerous bugs.
1127 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1130 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1131 readonly data in sdata, if that is disabled.
1132 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1133 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1134 -mreadonly-in-sdata option.
1136 2018-03-07 Martin Sebor <msebor@redhat.com>
1138 PR tree-optimization/84468
1139 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1140 basic block when looking for nul assignment.
1142 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1145 * except.h (output_function_exception_table): Adjust prototype.
1146 * except.c (output_function_exception_table): Remove FNNAME parameter
1147 and add SECTION parameter. Ouput one part of the table at a time.
1148 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1149 the first part of the exception table and emit unwind directives.
1150 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1151 (i386_pe_seh_cold_init): Likewise.
1152 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1153 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1154 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1155 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1156 * config/i386/winnt.c: Include except.h.
1157 (struct seh_frame_state): Add reg_offset, after_prologue and
1158 in_cold_section fields.
1159 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1160 (i386_pe_seh_cold_init): New function.
1161 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1162 to seh->in_cold_section.
1163 (seh_emit_push): Record the offset of the push.
1164 (seh_emit_save): Record the offet of the save.
1165 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1166 Test seh->after_prologue to disregard the epilogue.
1167 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1168 (i386_pe_end_cold_function): New function.
1170 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1173 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1174 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1177 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1179 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1180 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1183 * multiple_target.c: Include tree-inline.h and intl.h.
1184 (expand_target_clones): Diagnose and fail if node->definition and
1185 !tree_versionable_function_p (node->decl).
1187 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1189 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1191 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1192 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1193 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1195 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1198 * combine.c (try_combine): Use reg_or_subregno instead of handling
1199 just paradoxical SUBREGs and REGs.
1201 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1203 * config/arc/arc.c (arc_finalize_pic): Remove function.
1204 (arc_must_save_register): We use single base PIC register, remove
1205 checks to save/restore the PIC register.
1206 (arc_expand_prologue): Likewise.
1207 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1209 (arc_verify_short): Likewise.
1210 (arc_attr_type): Likewise.
1211 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1212 (walk_stores): Likewise.
1213 (arc_address_cost): Make it static.
1214 (arc_verify_short): Likewise.
1215 (branch_dest): Likewise.
1216 (arc_attr_type): Likewise.
1217 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1218 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1219 (arc_final_prescan_insn): Remove inserting the nops due to
1220 hardware hazards. It is done in reorg step.
1221 (insn_length_variant_t): Remove.
1222 (insn_length_parameters_t): Likewise.
1223 (arc_insn_length_parameters): Likewise.
1224 (arc_get_insn_variants): Likewise.
1225 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1227 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1230 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1233 PR tree-optimization/84687
1234 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1236 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1238 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1240 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1241 Rename to ppc_speculation_barrier.
1242 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1243 __builtin_ppc_speculation_barrier.
1245 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1248 * combine.c (combine_simplify_rtx): Don't try to simplify if
1249 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1252 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1254 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1255 to 32 bytes when compiling for POWER9.
1257 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1260 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1261 regparm >= 3 with no arg reg available also for calls with
1262 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1265 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1267 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1269 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1272 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1274 2018-03-05 Richard Biener <rguenther@suse.de>
1276 PR tree-optimization/84486
1277 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1278 When inserting a __builtin_assume_aligned call set the LHS
1279 SSA name alignment info accordingly.
1281 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1283 PR tree-optimization/84114
1284 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1285 Avoid reassociation of FLOAT_MODE addition.
1287 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1289 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1290 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1291 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1292 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1293 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1294 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1295 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1297 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1298 __builtin_ia32_wbinvd): New builtins.
1299 (SPECIAL_ARGS2): New.
1300 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1301 (SPECIAL_ARGS2): New.
1302 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1303 (ix86_valid_target_attribute_inner_p): Ditto.
1304 (ix86_init_mmx_sse_builtins): Add special_args2.
1305 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1306 TARGET_WBNOINVD_P): New.
1307 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1308 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1309 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1310 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1311 * config/i386/pconfigintrin.h: New file.
1312 * config/i386/wbnoinvdintrin.h: Ditto.
1313 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1314 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1316 2018-03-05 Richard Biener <rguenther@suse.de>
1318 PR tree-optimization/84670
1319 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1321 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1322 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1323 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1324 doesn't grow if all successors (recursively) were visited at least
1327 2018-03-05 Richard Biener <rguenther@suse.de>
1329 PR tree-optimization/84650
1330 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1331 if executed in the loop pipeline.
1333 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1335 * doc/configfiles.texi (Configuration Files): Move info about
1336 conditionalizing $target-protos.h to...
1337 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1338 differs from $target-protos.h.
1340 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1341 Chung-Ju Wu <jasonwucj@gmail.com>
1343 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1344 * config/nds32/nds32-multiple.md (setmemsi): Define.
1345 * config/nds32/nds32-memory-manipulation.c
1346 (nds32_gen_dup_4_byte_to_word_value): New.
1347 (emit_setmem_word_loop): New.
1348 (emit_setmem_byte_loop): New.
1349 (nds32_expand_setmem_loop): New.
1350 (nds32_expand_setmem_loop_v3m): New.
1351 (nds32_expand_setmem_unroll): New.
1352 (nds32_expand_setmem): New.
1354 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1355 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * config/nds32/nds32-memory-manipulation.c
1358 (nds32_emit_load_store): New.
1359 (nds32_emit_post_inc_load_store): New.
1360 (nds32_emit_mem_move): New.
1361 (nds32_emit_mem_move_block): New.
1362 (nds32_expand_movmemsi_loop_unknown_size): New.
1363 (nds32_expand_movmemsi_loop_known_size): New.
1364 (nds32_expand_movmemsi_loop): New.
1365 (nds32_expand_movmemsi_unroll): New.
1366 (nds32_expand_movmemqi): Rename ...
1367 (nds32_expand_movmemsi): ... to this.
1368 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1369 (movmemsi): ... to this.
1370 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1371 (nds32_expand_movmemsi): ... to this.
1373 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1374 Monk Chiang <sh.chiang04@gmail.com>
1375 Chung-Ju Wu <jasonwucj@gmail.com>
1377 * config/nds32/nds32-protos.h
1378 (nds32_expand_load_multiple): New arguments.
1379 (nds32_expand_store_multiple): Ditto.
1380 (nds32_valid_multiple_load_store): Rename ...
1381 (nds32_valid_multiple_load_store_p): ... to this.
1382 * config/nds32/nds32-memory-manipulation.c
1383 (nds32_expand_load_multiple): Refine implementation.
1384 (nds32_expand_store_multiple): Ditto.
1385 * config/nds32/nds32-multiple.md
1386 (load_multiple): Update nds32_expand_load_multiple interface.
1387 (store_multiple): Update nds32_expand_store_multiple interface.
1388 * config/nds32/nds32-predicates.c
1389 (nds32_valid_multiple_load_store): Rename ...
1390 (nds32_valid_multiple_load_store_p): ... to this and refine
1392 * config/nds32/predicates.md
1393 (nds32_load_multiple_and_update_address_operation): New predicate.
1394 (nds32_store_multiple_and_update_address_operation): New predicate.
1396 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1397 Chung-Ju Wu <jasonwucj@gmail.com>
1399 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1400 (combo): New attribute.
1401 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1403 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1405 * config/nds32/nds32.opt: Change -mcmodel= default value.
1407 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1408 Monk Chiang <sh.chiang04@gmail.com>
1409 Chung-Ju Wu <jasonwucj@gmail.com>
1411 * config/nds32/constants.md (unspec_element): New enum.
1412 * config/nds32/constraints.md (Umw): New constraint.
1413 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1414 * config/nds32/nds32-intrinsic.md: Likewise.
1415 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1416 (nds32_valid_smw_lwm_base_p): New.
1417 (nds32_output_smw_single_word): New.
1418 (nds32_output_lmw_single_word): New.
1419 (nds32_expand_unaligned_load): New.
1420 (nds32_expand_unaligned_store): New.
1421 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1422 (nds32_output_smw_single_word): Declare.
1423 (nds32_output_lmw_single_word): Declare.
1424 (nds32_expand_unaligned_load): Declare.
1425 (nds32_expand_unaligned_store): Declare.
1426 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1427 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1428 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1429 NDS32_BUILTIN_UASTORE_DW.
1430 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1433 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1434 Kito Cheng <kito.cheng@gmail.com>
1435 Chung-Ju Wu <jasonwucj@gmail.com>
1437 * config/nds32/nds32-intrinsic.c
1438 (nds32_expand_builtin_null_ftype_reg): Delete.
1439 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1440 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1441 (nds32_read_argument): New.
1442 (nds32_legitimize_target): Ditto.
1443 (nds32_legitimize_argument): Ditto.
1444 (nds32_check_constant_argument): Ditto.
1445 (nds32_expand_unop_builtin): Ditto.
1446 (nds32_expand_unopimm_builtin): Ditto.
1447 (nds32_expand_binop_builtin): Ditto.
1448 (nds32_builtin_decl_impl): Ditto.
1449 (builtin_description): Ditto.
1450 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1451 (nds32_init_builtins_impl): Ditto.
1452 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1453 (nds32_builtin_decl): New.
1454 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1455 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1457 2018-03-02 Jeff Law <law@redhat.com>
1459 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1460 (redundant_insn, fill_simple_delay_slots): Likewise.
1461 (fill_slots_from_thread): Likewise.
1462 * resource.c (mark_referenced_resources): Likewise.
1463 (mark_set_resources, find_dead_or_set_registers): Likewise.
1465 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1467 * substring-locations.h (format_warning_va): Formatting fix for
1469 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1471 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1472 * substring-locations.c: Include intl.h.
1473 (format_warning_va): Turned into small wrapper around
1474 format_warning_n_va, renamed to ...
1475 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1476 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1478 (format_warning_at_substring_n): New function.
1479 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1480 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1481 format_warning_at_substring with just a shorter name instead of
1482 const function pointer.
1483 (fmtwarn_n): New function.
1484 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1485 appropriate, get rid of all the fmtstr temporaries, move conditionals
1486 with G_() wrapped string literals directly into fmtwarn arguments,
1487 cast dir.len to (int), formatting fixes.
1489 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1491 * doc/invoke.texi: Remove "Cilk Plus" references.
1493 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1494 Richard Biener <rguenther@suse.de>
1497 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1498 for error or warning attributes if CALL_FROM_THUNK_P is set.
1501 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1504 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1505 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1508 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1509 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1511 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1512 instead of -1U in last predictors element's probability member.
1514 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1517 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1518 arguments if they are comparable.
1520 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1522 PR tree-optimization/84634
1523 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1524 masks and masked_loop_p with a single loop_masks, making sure it's
1525 null for bb vectorization.
1527 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1529 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1530 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1531 loop->force_vectorize to check whether there is no alias.
1533 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1536 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1538 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1540 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1541 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1542 instead of a loop around prev_real_insn.
1543 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1547 * config/i386/i386.c (ix86_print_operand): Use conditional
1548 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1551 2018-03-02 Richard Biener <rguenther@suse.de>
1553 PR tree-optimization/84427
1554 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1555 (bitmap_set_subtract_values): Rewrite to handle multiple
1558 (prune_clobbered_mems): Likewise.
1559 (phi_translate): Take edge instead of pred/phiblock.
1560 (phi_translate_1): Likewise.
1561 (phi_translate_set): Likewise. Insert all translated
1562 exprs for a value into the set, keeping possibly multiple
1563 expressions per value.
1564 (compute_antic_aux): Adjust for phi_translate changes.
1565 When intersecting union the expressions and prune those
1566 not in the final value set, keeping possibly multiple
1567 expressions per value. Do not use value-insertion
1568 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1569 all expressions. Add verification that the value-sets
1570 only shrink during iteration.
1571 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1572 (do_pre_regular_insertion): Likewise.
1573 (do_pre_partial_partial_insertion): Likewise.
1575 2018-03-02 Richard Biener <rguenther@suse.de>
1578 * config/darwin.c (saved_debug_info_level): New static global.
1579 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1580 (darwin_asm_lto_end): Restore debug info generation settings.
1582 2018-03-01 Martin Liska <mliska@suse.cz>
1585 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1588 2018-03-01 Richard Biener <rguenther@suse.de>
1591 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1592 type annotation with LTO when debug was disabled at compile-time.
1594 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1596 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1598 (mips_final_postscan_insn): Likewise.
1600 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1602 PR rtl-optimization/84528
1603 * alias.c (init_alias_target): Add commentary.
1604 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1605 a unique base value if the frame pointer is not eliminated
1606 to the stack pointer.
1608 2018-03-01 Tom de Vries <tom@codesourcery.com>
1610 PR rtl-optimization/83327
1611 * lra-int.h (hard_regs_spilled_into): Declare.
1612 * lra.c (hard_regs_spilled_into): Define.
1613 (init_reg_info): Init hard_regs_spilled_into.
1614 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1615 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1616 (process_bb_lives): Handle hard_regs_spilled_into.
1617 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1618 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1620 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1622 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1623 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1624 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1625 * config/rs6000/aix72.h: New file.
1627 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1629 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1630 instead of warning_at with conditional singular and plural messages
1634 * stor-layout.c (update_alignment_for_field): For
1635 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1636 && !DECL_PACKED (field), do the alignment update, just use
1637 only desired_align instead of MAX (type_align, desired_align)
1639 (place_field): Don't do known_align < desired_align handling
1640 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1641 is non-NULL, instead do it after rli->prev_field handling and
1642 only if not within a bitfield word. For DECL_PACKED (field)
1643 use type_align of BITS_PER_UNIT.
1645 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1647 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1648 superfluous parentheses and trailing spaces.
1650 2018-02-28 Richard Biener <rguenther@suse.de>
1652 PR tree-optimization/84584
1653 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1654 SCoPs with fake exit edge.
1656 2018-02-28 Martin Liska <mliska@suse.cz>
1659 * timevar.c (timer::print): Fix format to properly print 100%
1662 2018-02-28 Richard Biener <rguenther@suse.de>
1665 * genmatch.c (capture_info::walk_match): Do not mark
1666 captured expressions without operands as expr_p given
1667 they act more like predicates and should be subject to
1668 "lost tail" side-effect preserving.
1670 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1672 PR rtl-optimization/81611
1673 * auto-inc-dec.c (attempt_change): Move dead note from
1674 mem_insn if it's the next use of regno
1675 (find_address): Take address use of reg holding
1676 non-incremented value. Add parm to limit search to the named
1678 (merge_in_block): Attempt to use a mem insn that is the next
1679 use of the original regno.
1681 2018-02-27 Martin Sebor <msebor@redhat.com>
1684 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1685 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1687 2018-02-27 Martin Sebor <msebor@redhat.com>
1689 PR translation/84207
1690 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1691 n argument to unsigned HOST_WIDE_INT.
1692 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1693 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1694 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1695 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1697 2018-02-27 Richard Biener <rguenther@suse.de>
1699 PR tree-optimization/84512
1700 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1701 Do not use the estimate returned from record_stmt_cost for
1702 the scalar iteration cost but sum properly using add_stmt_cost.
1704 2018-02-27 Richard Biener <rguenther@suse.de>
1706 PR tree-optimization/84466
1707 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1708 Adjust last change to less strictly validate use operands.
1710 2018-02-27 Martin Liska <mliska@suse.cz>
1712 PR gcov-profile/84548
1713 * gcov.c (process_file): Allow partial overlap and consider it
1714 also as group functions.
1715 (output_lines): Properly calculate range of lines for a group.
1717 2018-02-27 Martin Liska <mliska@suse.cz>
1719 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1720 'ggc' suffixes. Change first column width.
1721 (timer::print): Fix formatting of the column.
1723 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1725 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1726 preserve inline entry blocks for the sake of debug inline
1727 entry point markers alone.
1728 (remove_unused_locals): Suggest in comments a better place to
1729 force the preservation of inline entry blocks that are
1730 otherwise unused, but do not preserve them.
1732 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1734 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1736 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1739 * config/i386/constraints.md (Bs): Replace
1740 ix86_indirect_branch_register with
1741 TARGET_INDIRECT_BRANCH_REGISTER.
1743 * config/i386/i386.md (indirect_jump): Likewise.
1744 (tablejump): Likewise.
1745 (*sibcall_memory): Likewise.
1746 (*sibcall_value_memory): Likewise.
1747 Peepholes of indirect call and jump via memory: Likewise.
1748 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1749 (*sibcall_value_GOT_32): Likewise.
1750 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1751 (GOT_memory_operand): Likewise.
1752 (call_insn_operand): Likewise.
1753 (sibcall_insn_operand): Likewise.
1754 (GOT32_symbol_operand): Likewise.
1755 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1757 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1759 PR rtl-optimization/83496
1760 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1761 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1762 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1763 redundant insn, if any.
1764 (relax_delay_slots): Likewise.
1765 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1767 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1769 PR tree-optimization/83965
1770 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1771 that grouped statements are part of a reduction chain. Return
1772 true if the statement is not marked as a reduction itself but
1774 (vect_recog_dot_prod_pattern): Don't check whether the statement
1775 is part of a group here.
1776 (vect_recog_sad_pattern): Likewise.
1777 (vect_recog_widen_sum_pattern): Likewise.
1779 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1782 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1784 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1787 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1789 (ix86_output_indirect_function_return): New prototype.
1790 (ix86_split_simple_return_pop_internal): Likewise.
1791 * config/i386/i386.c (indirect_return_via_cx): New.
1792 (indirect_return_via_cx_bnd): Likewise.
1793 (indirect_thunk_name): Handle return va CX_REG.
1794 (output_indirect_thunk_function): Create alias for
1795 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1796 (ix86_output_indirect_jmp): Remove the bool argument.
1797 (ix86_output_indirect_function_return): New function.
1798 (ix86_split_simple_return_pop_internal): Likewise.
1799 * config/i386/i386.md (*indirect_jump): Don't pass false
1800 to ix86_output_indirect_jmp.
1801 (*tablejump_1): Likewise.
1802 (simple_return_pop_internal): Change it to define_insn_and_split.
1803 Call ix86_split_simple_return_pop_internal to split it for
1805 (simple_return_indirect_internal): Call
1806 ix86_output_indirect_function_return instead of
1807 ix86_output_indirect_jmp.
1809 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1812 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1813 memset and value initialization afterwards.
1815 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1817 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1819 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1822 * common/config/aarch64/aarch64-common.c
1823 (aarch_option_optimization_table[]): Switch
1824 off fomit-frame-pointer
1826 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1827 Chung-Ju Wu <jasonwucj@gmail.com>
1829 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1831 (store_multiple): Ditto.
1833 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1835 * config.gcc: Add --with-cpu support for nds32 target.
1836 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1837 * config/nds32/nds32.opt: Add -mcpu= option.
1839 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1841 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1842 isel=yes): Warn for these deprecated options.
1844 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1846 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1847 ISA_2_5_MASKS_EMBEDDED.
1849 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1851 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1852 p->max as pointers rather than using iterative_hash_expr.
1854 2018-02-23 Carl Love <cel@us.ibm.com>
1856 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1857 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1859 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1860 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1861 P8V_BUILTIN_VEC_VUNSIGNED2.
1863 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1866 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1867 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1869 * lra-constraints.c (curr_insn_transform): Set up
1870 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1872 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1873 * lra-eliminations.c (spill_pseudos): Ditto.
1874 (process_insn_for_elimination): Ditto.
1875 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1876 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1877 LRA_NON_CLOBBERED_ALT.
1879 2018-02-22 Martin Sebor <msebor@redhat.com>
1881 PR tree-optimization/84480
1882 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1883 to maybe_diag_stxncpy_trunc. Call it.
1884 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1885 from gimple_fold_builtin_strcpy. Print inlining stack.
1886 (handle_builtin_stxncpy): Print inlining stack.
1887 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1889 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1892 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1893 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1894 and -fcheck-pointer-bounds are used together.
1895 (indirect_thunk_prefix): New enum.
1896 (indirect_thunk_need_prefix): New function.
1897 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1898 "_nt" instead of "_bnd" for NOTRACK prefix.
1899 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1900 (output_indirect_thunk_function): Likewise.
1902 (ix86_code_end): Update output_indirect_thunk_function calls.
1903 (ix86_output_indirect_branch_via_reg): Replace
1904 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1905 (ix86_output_indirect_branch_via_push): Likewise.
1906 (ix86_output_function_return): Likewise.
1907 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1908 incompatible with -fcf-protection=branch and
1909 -fcheck-pointer-bounds.
1911 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1914 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1915 Change gcc_assert call to output_operand_lossage.
1917 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1919 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1921 2018-02-22 DJ Delorie <dj@redhat.com>
1922 Sebastian Perta <sebastian.perta@renesas.com>
1923 Oleg Endo <olegendo@gcc.gnu.org>
1925 * config/rx/rx.c (rx_rtx_costs): New function.
1926 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1928 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1930 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1932 2018-02-22 Martin Liska <mliska@suse.cz>
1935 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1936 Add "native" as a possible value.
1938 2018-02-22 Martin Liska <mliska@suse.cz>
1941 * config/i386/i386.c (ix86_option_override_internal):
1942 Add "native" as a possible value for -march and -mtune.
1944 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1947 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1948 to all type variants.
1950 PR tree-optimization/84503
1951 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1952 width as info->bitpos + info->bitsize - start.
1953 (merged_store_group::merge_overlapping): Simplify width computation.
1954 (check_no_overlap): New function.
1955 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1956 start + width and last_order of the group, fail if check_no_overlap
1958 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1959 to group if check_no_overlap fails.
1961 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1963 * config/rs6000/altivec.md: Delete contraint arguments to
1964 define_expand, define_split, and define_peephole2, and in
1965 define_insn_and_split if always unused.
1966 * config/rs6000/darwin.md: Ditto.
1967 * config/rs6000/dfp.md: Ditto.
1968 * config/rs6000/rs6000.md: Ditto.
1969 * config/rs6000/sync.md: Ditto.
1970 * config/rs6000/vector.md: Ditto.
1971 * config/rs6000/vsx.md: Ditto.
1973 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1975 * config/rs6000/altivec.md: Write output control strings as braced
1976 blocks instead of double-quoted strings.
1977 * config/rs6000/darwin.md: Ditto.
1978 * config/rs6000/rs6000.md: Ditto.
1979 * config/rs6000/vector.md: Ditto.
1980 * config/rs6000/vsx.md: Ditto.
1982 2018-02-21 Jason Merrill <jason@redhat.com>
1984 PR c++/84314 - ICE with templates and fastcall attribute.
1985 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1987 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1989 * ipa-cp.c (determine_versionability): Fix comment typos.
1991 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1994 * ipa-cp.c (determine_versionability): Do not version functions caling
1997 2018-02-21 Martin Liska <mliska@suse.cz>
2000 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2001 Add "native" as a possible value.
2002 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2003 the macro when native cpu detection is available.
2005 2018-02-21 Martin Liska <mliska@suse.cz>
2008 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2009 Add "native" as a possible value.
2010 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2011 when native cpu detection is available.
2013 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2014 Martin Sebor <msebor@redhat.com>
2016 PR tree-optimization/84478
2017 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2019 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2020 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2021 support which is conservatively correct, for 2 only stay conservative
2022 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2023 argument to the 2 argument get_range_strlen, adjust 6 arg
2024 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2026 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2027 (gimple_fold_builtin_strlen): Pass true as last argument to
2030 2018-02-20 Martin Sebor <msebor@redhat.com>
2033 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2034 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2035 (builtin_memref::builtin_memref): Factor out parts into
2036 set_base_and_offset and call it.
2038 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2041 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2042 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2043 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2044 search at the associated MODE_INT.
2046 2018-02-20 Jeff Law <law@redhat.com>
2049 PR tree-optimization/81592
2051 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2052 for range data rather than using global data.
2053 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2054 range data rather than using global data.
2055 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2056 pass it to children as needed.
2057 (struct directive::fmtresult): Similarly.
2058 (struct directive::set_width): Similarly.
2059 (struct directive::set_precision): Similarly.
2060 (format_integer, format_directive, parse_directive): Similarly.
2061 (format_none): Accept unnamed vr_values parameter.
2062 (format_percent, format_floating, format_character): Similarly.
2063 (format_string, format_plain): Similarly.
2064 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2065 the EVRP range analyzer for range data rather than using global data.
2066 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2067 gimple-ssa-evrp-analyze.h
2068 (class sprintf_dom_walker): Add after_dom_children member function.
2069 Add evrp_range_analyzer member.
2070 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2071 range analyzer as needed.
2072 (sprintf_dom_walker::after_dom_children): New member function.
2073 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2075 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2076 (evrp_range_analyzer::pop_to_marker): Likewise.
2078 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2080 PR tree-optimization/84419
2081 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2082 with the required type if its current type is compatible but
2085 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2088 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2089 after vectorization.
2091 2018-02-20 Martin Liska <mliska@suse.cz>
2094 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2095 possible values if we don't have a hint.
2097 2018-02-20 Martin Liska <mliska@suse.cz>
2101 * final.c (shorten_branches): Build align_tab array with one
2103 * opts.c (finish_options): Add alignment option limit check.
2104 (MAX_CODE_ALIGN): Likewise.
2105 (MAX_CODE_ALIGN_VALUE): Likewise.
2106 * doc/invoke.texi: Document maximum allowed option value for
2107 all -falign-* options.
2109 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2112 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2113 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2114 * var-tracking.c (emit_note_insn_var_location): Remove all references
2115 to NOTE_INSN_CALL_ARG_LOCATION.
2116 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2117 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2118 Use copy_rtx_if_shared.
2119 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2120 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2121 (dwarf2out_var_location): Remove handling of
2122 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2124 * final.c (final_scan_insn): Remove all references to
2125 NOTE_INSN_CALL_ARG_LOCATION.
2126 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2127 before dumping final insns.
2128 * except.c (emit_note_eh_region_end): Remove all references to
2129 NOTE_INSN_CALL_ARG_LOCATION.
2130 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2131 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2132 * config/arc/arc.c (hwloop_optimize): Likewise.
2133 * config/arm/arm.c (create_fix_barrier): Likewise.
2134 * config/s390/s390.c (s390_chunkify_start): Likewise.
2135 * config/sh/sh.c (find_barrier): Likewise.
2136 * config/i386/i386.c (rest_of_insert_endbranch,
2137 ix86_seh_fixup_eh_fallthru): Likewise.
2138 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2139 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2140 * config/frv/frv.c (frv_function_prologue): Likewise.
2141 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2143 (note_outside_basic_block_p): Remove all references to
2144 NOTE_INSN_CALL_ARG_LOCATION.
2145 * gengtype.c (adjust_field_rtx_def): Likewise.
2146 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2148 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2149 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2152 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2155 PR tree-optimization/84452
2156 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2157 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2160 2018-02-19 Martin Liska <mliska@suse.cz>
2163 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2165 2018-02-19 Martin Liska <mliska@suse.cz>
2166 Richard Sandiford <richard.sandiford@linaro.org>
2168 PR tree-optimization/82491
2169 * gimple-fold.c (get_base_constructor): Make earlier bail out
2172 2018-02-19 Carl Love <cel@us.ibm.com>
2174 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2175 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2177 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2178 P8V_BUILTIN_VEC_NEG.
2180 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2182 * config/rl78/rl78.md (movdf): New define expand.
2184 2018-02-19 Martin Liska <mliska@suse.cz>
2187 * doc/invoke.texi: Fix typo.
2188 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2190 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2192 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2193 handle rs6000_single_float and rs6000_double_float specially for
2196 2018-02-16 Jeff Law <law@redhat.com>
2198 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2201 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2204 * ipa-inline.c (inline_small_functions): Fix a typo.
2206 2018-02-16 Nathan Sidwell <nathan@acm.org>
2208 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2210 2018-02-16 Carl Love <cel@us.ibm.com>
2212 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2213 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2214 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2215 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2216 expansion to P8V_BUILTIN_VEC_FLOAT2.
2218 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2220 PR rtl-optimization/70023
2221 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2222 src_regno into account.
2224 2018-02-16 Carl Love <cel@us.ibm.com>
2226 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2227 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2228 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2229 * config/rs6000/rs6000.c: Remove case statements for
2230 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2231 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2232 and P9V_BUILTIN_VEC_VINSERT4B.
2233 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2234 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2235 * config/rs6000/vsx.md:
2236 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2239 2018-02-16 Carl Love <cel@us.ibm.com>
2241 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2243 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2245 * config/rs6000/rs6000-c.c: Add the definitions for
2246 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2247 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2248 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2249 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2250 definition for insert4b and define insn *insert3b_internal.
2251 * doc/extend.texi: Add documentation for vec_extract4b.
2253 2018-02-16 Nathan Sidwell <nathan@acm.org>
2255 * doc/extend.texi (Backwards Compatibility): Mention friend
2256 injection. Note for-scope is deprecated.
2257 * doc/invoke.texi (-ffriend-injection): Deprecate.
2259 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2261 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2262 that moved to I2, also allow destinations that are a paradoxical
2263 subreg (instead of a normal reg).
2265 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2268 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2271 2018-02-16 Richard Biener <rguenther@suse.de>
2273 PR tree-optimization/84037
2274 PR tree-optimization/84016
2276 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2277 Adjust vec_construct for the fact we need additional higher latency
2278 128bit inserts for AVX256 and AVX512 vector builds.
2279 (ix86_add_stmt_cost): Scale vector construction cost for
2282 2018-02-16 Richard Biener <rguenther@suse.de>
2284 PR tree-optimization/84417
2285 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2286 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2287 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2289 2018-02-16 Martin Liska <mliska@suse.cz>
2292 * internal-fn.def (ASAN_CHECK): Set proper flags.
2293 (ASAN_MARK): Likewise.
2295 2018-02-16 Julia Koval <julia.koval@intel.com>
2297 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2298 from PTA_CANNONLAKE.
2300 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2303 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2304 Use ++iter rather than iter++ for std::list iterators.
2305 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2306 defer deleting them until all nodes in the forest are processed. Do
2307 free even leaf nodes. Change to_process into auto_vec.
2310 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2311 * vec.h (vec_default_construct): Use memset instead of placement new
2312 if BROKEN_VALUE_INITIALIZATION is defined.
2313 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2314 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2317 PR rtl-optimization/83723
2318 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2319 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2320 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2322 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2324 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2326 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2328 PR rtl-optimization/81443
2329 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2330 from inner REGs to paradoxical SUBREGs.
2332 2018-02-16 Richard Biener <rguenther@suse.de>
2334 PR tree-optimization/84399
2335 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2336 For operands we can analyze at their definition make sure we can
2337 analyze them at each use as well.
2339 2018-02-16 Richard Biener <rguenther@suse.de>
2341 PR tree-optimization/84190
2342 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2343 volatile accesses if the decl isn't volatile.
2345 2018-02-15 Jason Merrill <jason@redhat.com>
2347 PR c++/84314 - ICE with templates and fastcall attribute.
2348 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2349 TYPE_CANONICAL on an existing type.
2351 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2353 PR tree-optimization/84383
2354 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2355 dstoff nor call operand_equal_p if dstbase is NULL.
2357 PR tree-optimization/84334
2358 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2359 also a CONSTANT_CLASS_P, punt.
2361 2018-02-14 Jim Wilson <jimw@sifive.com>
2363 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2364 first SMALL_OPERAND check. New local min_second_step. Move assert
2365 to where locals are set. Add TARGET_RVC support.
2366 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2368 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2370 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2372 2018-02-14 Martin Sebor <msebor@redhat.com>
2374 PR tree-optimization/83698
2375 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2376 arrays constrain the offset range to their bounds.
2377 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2378 (builtin_access::overlap): Avoid setting the size of overlap if it's
2380 (maybe_diag_overlap): Also consider arrays when deciding what values
2381 of offsets to include in diagnostics.
2383 2018-02-14 Martin Sebor <msebor@redhat.com>
2386 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2387 that correspond to the kind of a declaration.
2389 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2392 * config/pa/pa.md: Load address of PIC label using the linkage table
2393 if the label is nonlocal.
2395 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2397 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2398 warning message if user requests -maltivec=be.
2399 * doc/invoke.texi: Document deprecation of -maltivec=be.
2401 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2404 * config/rs6000/rs6000-c.c: Update definitions for
2405 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2406 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2408 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2411 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2412 add _get_ssp intrinsics. Remove argument from
2413 __builtin_ia32_rdssp[d|q].
2414 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2415 * config/i386/i386-builtin.def: Remove argument from
2416 __builtin_ia32_rdssp[d|q].
2417 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2418 ix86_expand_special_args_builtin for _rdssp[d|q].
2419 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2420 Clear register before usage.
2421 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2422 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2424 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2426 PR tree-optimization/84357
2427 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2428 operand 1 of an ARRAY_REF too.
2430 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2433 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2434 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2436 (set_of_reg): New struct.
2437 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2438 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2439 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2441 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2442 Split into bitclr, bitset, bitinvert patterns if appropriate.
2443 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2444 use rx_fuse_in_memory_bitop.
2445 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2446 to named insn, correct maximum insn length.
2448 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2451 * machmode.def: Define a complex mode for PARTIAL_INT.
2452 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2454 * doc/rtl.texi: Document CSPImode.
2455 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2457 (msp430_hard_regno_nregs_with_padding): Likewise.
2459 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2462 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2464 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2466 PR rtl-optimization/84169
2467 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2468 we generated a parallel as new i3 and we split that to new i2 and i3
2469 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2470 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2471 those to i2, not i1. Partially rewrite this scan code.
2473 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2476 * stor-layout.c (place_field): For variable length fields, adjust
2477 offset_align afterwards not just based on the field's alignment,
2478 but also on the size.
2481 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2482 of exps and logs in the use_exp2 case.
2484 2018-02-13 Jeff Law <law@redhat.com>
2486 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2489 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2492 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2496 * common.opt (gas-loc-support, gas-locview-support): New.
2497 (ginline-points, ginternal-reset-location-views): New.
2498 * doc/invoke.texi: Document them. Use @itemx where intended.
2499 (gvariable-location-views): Adjust.
2500 * target.def (reset_location_view): New.
2501 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2502 (TARGET_RESET_LOCATION_VIEW): New.
2503 * doc/tm.texi: Rebuilt.
2504 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2505 (dwarf2out_default_as_locview_support): New.
2506 (output_asm_line_debug_info): Use option variables.
2507 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2508 (output_loc_list): Likewise.
2509 (add_high_low_attributes): Check option variables.
2510 Don't output entry view attribute in strict mode.
2511 (gen_inlined_subroutine_die): Check option variables.
2512 (dwarf2out_inline_entry): Likewise.
2513 (init_sections_and_labels): Likewise.
2514 (dwarf2out_early_finish): Likewise.
2515 (maybe_reset_location_view): New, from...
2516 (dwarf2out_var_location): ... here. Call it.
2517 * debug.h (dwarf2out_default_as_loc_support): Declare.
2518 (dwarf2out_default_as_locview_support): Declare.
2519 * hooks.c (hook_int_rtx_insn_0): New.
2520 * hooks.h (hook_int_rtx_insn_0): Declare.
2521 * toplev.c (process_options): Take -gas-loc-support and
2522 -gas-locview-support from dwarf2out. Enable
2523 -gvariable-location-views by default only with locview
2524 assembler support. Enable -ginternal-reset-location-views by
2525 default only if the target defines the corresponding hook.
2526 Enable -ginline-points by default if location views are
2527 enabled; force it disabled if statement frontiers are
2529 * tree-inline.c (expand_call_inline): Check option variables.
2530 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2532 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2534 PR tree-optimization/84321
2535 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2536 handling. Also check whether the anti-range contains any values
2537 that satisfy the mask; switch to a VR_RANGE if not.
2539 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2542 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2544 2018-02-13 Martin Jambor <mjambor@suse.cz>
2547 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2548 of call statements, also set location of a load to a temporary.
2550 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2552 * config/rl78/rl78.c (add_vector_labels): New function.
2553 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2554 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2555 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2556 which checks that no arguments are passed.
2557 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2558 * doc/extend.texi: Documentation for the new attribute.
2560 2018-02-13 Andreas Schwab <schwab@suse.de>
2562 * config/riscv/linux.h (CPP_SPEC): Define.
2564 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2567 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2568 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2569 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2570 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2571 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2572 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2573 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2574 temporarily for AES and PCLMUL builtins.
2576 PR tree-optimization/84339
2577 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2578 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2582 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2583 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2584 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2586 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2588 * omp-simd-clone.h: New file.
2589 * omp-simd-clone.c: Include omp-simd-clone.h.
2590 (expand_simd_clones): No longer static.
2591 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2592 cgraph.h and omp-simd-clone.h.
2593 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2594 (vect_recog_widen_shift_pattern): Formatting fix.
2595 (vect_pattern_recog_1): Don't check optab for calls.
2598 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2599 operands[2] into a REG before using gen_lowpart on it.
2601 2018-02-12 Jeff Law <law@redhat.com>
2604 * config/sh/sh.c (find_barrier): Consider a sibling call
2607 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2608 successfully back substituting a reg.
2610 2018-02-12 Richard Biener <rguenther@suse.de>
2612 PR tree-optimization/84037
2613 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2614 parameter, move visited init to caller.
2615 (vect_slp_analyze_operations): Separate cost from validity
2616 check, initialize visited once for all instances.
2617 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2619 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2621 (vect_model_promotion_demotion_cost): Likewise.
2622 (vectorizable_bswap): Guard cost modeling with !slp_node
2623 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2625 (vectorizable_call): Likewise.
2626 (vectorizable_conversion): Likewise.
2627 (vectorizable_assignment): Likewise.
2628 (vectorizable_shift): Likewise.
2629 (vectorizable_operation): Likewise.
2630 (vectorizable_store): Likewise.
2631 (vectorizable_load): Likewise.
2632 (vectorizable_condition): Likewise.
2633 (vectorizable_comparison): Likewise.
2635 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2638 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2639 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2640 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2642 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2645 * params.def (inline-min-speedup): Increase from 8 to 15.
2646 (max-inline-insns-auto): Decrease from 40 to 30.
2647 * ipa-split.c (consider_split): Add some buffer for function to
2648 be considered inlining candidate.
2649 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2652 2018-02-12 Richard Biener <rguenther@suse.de>
2654 PR tree-optimization/84037
2655 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2656 matched stmts if we cannot swap the non-matched ones.
2658 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2660 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2661 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2662 _mm_maskz_scalef_round_ss): New intrinsics.
2663 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2664 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2665 __builtin_ia32_scalefss_round): Remove.
2666 (__builtin_ia32_scalefsd_mask_round,
2667 __builtin_ia32_scalefss_mask_round): New intrinsics.
2668 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2669 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2670 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2671 "<round_constraint>")): Changed to ...
2672 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2673 "<round_scalar_constraint>")): ... this.
2674 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2675 %0, %1, %2<round_op3>}"): Changed to ...
2676 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2677 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2678 %2<round_scalar_mask_op3>}"): ... this.
2679 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2681 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2683 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2684 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2685 (_mm_maskz_sqrt_round_ss): New intrinsics.
2686 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2687 (__builtin_ia32_sqrtsd_mask_round)
2688 (__builtin_ia32_sqrtss_mask_round): New builtins.
2689 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2690 (__builtin_ia32_sqrtss_round): Remove.
2691 (__builtin_ia32_sqrtsd_mask_round)
2692 (__builtin_ia32_sqrtss_mask_round): New builtins.
2693 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2694 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2695 ((match_operand:VF_128 1 "vector_operand"
2696 "xBm,<round_constraint>")): Changed to ...
2697 ((match_operand:VF_128 1 "vector_operand"
2698 "xBm,<round_scalar_constraint>")): ... this.
2699 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2700 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2701 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2702 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2703 %<iptr>1<round_scalar_mask_op3>}): ... this.
2704 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2705 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2707 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2710 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2711 Cast vec_cmpeq result to correct type.
2712 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2713 Cast vec_cmpgt result to correct type.
2715 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2717 * final.c (final_scan_insn_1): Renamed from...
2718 (final_scan_insn): ... this. New wrapper, to recover
2719 seen from the outermost call in recursive ones.
2720 * config/sparc/sparc.c (output_return): Drop seen from call.
2721 (output_sibcall): Likewise.
2722 * config/visium/visium.c (output_branch): Likewise.
2724 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2726 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2729 2018-02-10 Alan Modra <amodra@gmail.com>
2732 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2733 Specify LR as an input.
2735 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2738 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2739 remove_member_access_dummy_vars): New functions.
2740 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2741 lower_omp_1, execute_lower_omp): Use them.
2743 PR rtl-optimization/84308
2744 * shrink-wrap.c (spread_components): Release todo vector.
2746 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2748 PR rtl-optimization/57193
2749 * ira-color.c (struct allocno_color_data): Add member
2750 conflict_allocno_hard_prefs.
2751 (update_conflict_allocno_hard_prefs): New.
2752 (bucket_allocno_compare_func): Add a preference based on
2753 conflict_allocno_hard_prefs.
2754 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2755 (color_allocnos): Remove a dead code. Initiate
2756 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2758 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2761 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2762 constraint from =wa to wa. Avoid a subreg on the output operand,
2763 instead use a pseudo and subreg it in a move.
2764 (p9_xxbrd_<mode>): Changed to ...
2765 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2766 (p9_xxbrd_v2df): New expander.
2767 (p9_xxbrw_<mode>): Changed to ...
2768 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2769 (p9_xxbrw_v4sf): New expander.
2771 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2773 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2775 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2778 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2779 multiply in 32-bit mode.
2780 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2781 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2784 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2786 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2787 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2788 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2789 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2791 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2794 * dwarf2out.c (is_trivial_indirect_ref): New function.
2795 (dwarf2out_late_global_decl): Do not generate a location
2796 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2797 and that are not defined in the current unit.
2799 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2801 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2802 instead of a libcall for UNORDERED.
2804 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2807 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2808 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2810 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2813 * config/s390/s390.c (s390_set_current_function): Invoke
2814 s390_indirect_branch_settings also if fndecl didn't change.
2816 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2818 * config/rs6000/rs6000.md (blockage): Set length to zero.
2820 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2822 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2824 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2827 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2828 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2832 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2833 PARALLEL incoming that failed vt_get_decl_and_offset check.
2836 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2837 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2839 (get_variable_section): For decls in named .bss* sections pass true as
2840 second argument to bss_initializer_p.
2842 2018-02-09 Marek Polacek <polacek@redhat.com>
2843 Jakub Jelinek <jakub@redhat.com>
2846 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2847 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2848 Sync some changes from cxx_fold_indirect_ref.
2850 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2852 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2854 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2855 (BLOCK_INLINE_ENTRY_LABEL): New.
2856 (dwarf2out_var_location): Disregard inline entry markers.
2857 (inline_entry_data): New struct.
2858 (inline_entry_data_hasher): New hashtable type.
2859 (inline_entry_data_hasher::hash): New.
2860 (inline_entry_data_hasher::equal): New.
2861 (inline_entry_data_table): New variable.
2862 (add_high_low_attributes): Add DW_AT_entry_pc and
2863 DW_AT_GNU_entry_view attributes if a pending entry is found
2864 in inline_entry_data_table. Add old entry_pc attribute only
2865 if debug nonbinding markers are disabled.
2866 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2867 markers are enabled.
2868 (block_within_block_p, dwarf2out_inline_entry): New.
2869 (dwarf2out_finish): Check that no entries remained in
2870 inline_entry_data_table.
2871 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2872 (final_scan_insn, notice_source_line): Likewise.
2873 (rest_of_clean_state): Skip inline entry markers.
2874 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2876 * gimple.c (gimple_build_debug_inline_entry): New.
2877 * gimple.h (enum gimple_debug_subcode): Add
2878 GIMPLE_DEBUG_INLINE_ENTRY.
2879 (gimple_build_debug_inline_entry): Declare.
2880 (gimple_debug_inline_entry_p): New.
2881 (gimple_debug_nonbind_marker_p): Adjust.
2882 * insn-notes.def (INLINE_ENTRY): New.
2883 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2884 inline entry marker notes.
2885 (print_insn): Likewise.
2886 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2887 (INSN_DEBUG_MARKER_KIND): Likewise.
2888 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2889 * tree-inline.c (expand_call_inline): Build and insert
2890 debug_inline_entry stmt.
2891 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2892 inline entry blocks early, if nonbind markers are enabled.
2893 (dump_scope_block): Dump fragment info.
2894 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2895 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2896 (gimple_build_debug_inline_entry): New.
2897 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2898 Enable/disable inline entry points too.
2899 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2900 (DEBUG_INSN): Describe inline entry markers.
2902 * common.opt (gvariable-location-views): New.
2903 (gvariable-location-views=incompat5): New.
2904 * config.in: Rebuilt.
2905 * configure: Rebuilt.
2906 * configure.ac: Test assembler for view support.
2907 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2908 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2909 * dwarf2out.c (var_loc_view): New typedef.
2910 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2911 (dwarf2out_locviews_in_attribute): New.
2912 (dwarf2out_locviews_in_loclist): New.
2913 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2914 (enum dw_line_info_opcode): Add LI_adv_address.
2915 (struct dw_line_info_table): Add view.
2916 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2917 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2918 (zero_view_p): New variable.
2919 (ZERO_VIEW_P): New macro.
2920 (output_asm_line_debug_info): New.
2921 (struct var_loc_node): Add view.
2922 (add_AT_view_list, AT_loc_list): New.
2923 (add_var_loc_to_decl): Add view param. Test it against last.
2924 (new_loc_list): Add view params. Record them.
2925 (AT_loc_list_ptr): Handle loc and view lists.
2926 (view_list_to_loc_list_val_node): New.
2927 (print_dw_val): Handle dw_val_class_view_list.
2928 (size_of_die): Likewise.
2929 (value_format): Likewise.
2930 (loc_list_has_views): New.
2931 (gen_llsym): Set vl_symbol too.
2932 (maybe_gen_llsym, skip_loc_list_entry): New.
2933 (dwarf2out_maybe_output_loclist_view_pair): New.
2934 (output_loc_list): Output view list or entries too.
2935 (output_view_list_offset): New.
2936 (output_die): Handle dw_val_class_view_list.
2937 (output_dwarf_version): New.
2938 (output_compilation_unit_header): Use it.
2939 (output_skeleton_debug_sections): Likewise.
2940 (output_rnglists, output_line_info): Likewise.
2941 (output_pubnames, output_aranges): Update version comments.
2942 (output_one_line_info_table): Output view numbers in asm comments.
2943 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2944 Call maybe_gen_llsym.
2945 (loc_list_from_tree_1): Adjust.
2946 (add_AT_location_description): Create view list attribute if
2947 needed, check it's absent otherwise.
2948 (convert_cfa_to_fb_loc_list): Adjust.
2949 (maybe_emit_file): Call output_asm_line_debug_info for test.
2950 (dwarf2out_var_location): Reset views as needed. Precompute
2951 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2952 attribute. Set view.
2953 (new_line_info_table): Reset next view.
2954 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2955 (dwarf2out_source_line): Likewise. Output view resets and labels to
2956 the assembler, or select appropriate line info opcodes.
2957 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2958 (optimize_string_length): Catch it. Adjust.
2959 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2960 dw_val_class_view_list, and remove it if no longer needed.
2961 (hash_loc_list): Hash view numbers.
2962 (loc_list_hasher::equal): Compare them.
2963 (optimize_location_lists): Check whether a view list symbol is
2964 needed, and whether the locview attribute is present, and
2965 whether they match. Remove the locview attribute if no longer
2967 (index_location_lists): Call skip_loc_list_entry for test.
2968 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2969 Use output_dwarf_version.
2970 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2971 (struct dw_val_node): Add val_view_list.
2972 * final.c (SEEN_NEXT_VIEW): New.
2973 (set_next_view_needed): New.
2974 (clear_next_view_needed): New.
2975 (maybe_output_next_view): New.
2976 (final_start_function): Rename to...
2977 (final_start_function_1): ... this. Take pointer to FIRST,
2978 add SEEN parameter. Emit param bindings in the initial view.
2979 (final_start_function): Reintroduce SEEN-less interface.
2980 (final): Rename to...
2981 (final_1): ... this. Take SEEN parameter. Output final pending
2982 next view at the end.
2983 (final): Reintroduce seen-less interface.
2984 (final_scan_insn): Output pending next view before switching
2985 sections or ending a block. Mark the next view as needed when
2986 outputting variable locations. Notify debug backend of section
2987 changes, and of location view changes.
2988 (rest_of_handle_final): Adjust.
2989 * toplev.c (process_options): Autodetect value for debug variable
2990 location views option. Warn on incompat5 without -gdwarf-5.
2991 * doc/invoke.texi (gvariable-location-views): New.
2992 (gvariable-location-views=incompat5): New.
2993 (gno-variable-location-views): New.
2995 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2997 PR tree-optimization/84136
2998 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2999 that the result of find_edge is non-NULL.
3001 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3004 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3005 storing integer register in SImode. Fix cost of 256 and 512
3006 byte aligned SSE register store.
3008 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3010 * config/i386/i386.c (ix86_multiplication_cost): Fix
3011 multiplication cost for TARGET_AVX512DQ.
3013 2018-02-08 Marek Polacek <polacek@redhat.com>
3015 PR tree-optimization/84238
3016 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3019 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3021 PR tree-optimization/84265
3022 * tree-vect-stmts.c (vectorizable_store): Don't treat
3023 VMAT_CONTIGUOUS accesses as grouped.
3024 (vectorizable_load): Likewise.
3026 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3028 PR tree-optimization/81635
3029 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3030 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3031 (test_round_for_mask): New functions.
3032 (wide_int_cc_tests): Call test_round_for_mask.
3033 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3034 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3035 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3036 range returned by get_range_info.
3038 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3041 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3042 * symtab.c: Include builtins.h
3043 (symtab_node::output_to_lto_symbol_table_p): Move here
3044 from lto-streamer-out.c:output_symbol_p.
3045 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3046 (output_symbol_p): Move all logic to symtab.c
3047 (produce_symtab): Update.
3049 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3051 * config/s390/s390-opts.h (enum indirect_branch): Define.
3052 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3053 (s390_indirect_branch_via_thunk)
3054 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3055 (enum s390_indirect_branch_type): Define.
3056 * config/s390/s390.c (struct s390_frame_layout, struct
3057 machine_function): Remove.
3058 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3059 (indirect_branch_table_label_no, indirect_branch_table_name):
3061 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3062 (enum s390_indirect_branch_option): Define.
3063 (s390_return_addr_from_memory): New function.
3064 (s390_handle_string_attribute): New function.
3065 (s390_attribute_table): Add new attribute handler.
3066 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3067 (s390_indirect_branch_via_thunk): New function.
3068 (s390_indirect_branch_via_inline_thunk): New function.
3069 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3070 sibling call optimization for non z10 compiles.
3071 (s390_emit_call): Force indirect branch target to be a single
3072 register. Add r1 clobber for non-z10 compiles.
3073 (s390_emit_epilogue): Emit return jump via return_use expander.
3074 (s390_reorg): Handle JUMP_INSNs as execute targets.
3075 (s390_option_override_internal): Perform validity checks for the
3076 new command line options.
3077 (s390_indirect_branch_attrvalue): New function.
3078 (s390_indirect_branch_settings): New function.
3079 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3080 (s390_output_indirect_thunk_function): New function.
3081 (s390_code_end): Implement target hook.
3082 (s390_case_values_threshold): Implement target hook.
3083 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3085 * config/s390/s390.h (struct s390_frame_layout)
3086 (struct machine_function): Move here from s390.c.
3087 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3088 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3089 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3090 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3091 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3092 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3093 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3094 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3095 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3096 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3097 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3098 (mnemonic attribute): Add values which aren't recognized
3100 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3101 pattern for branch conversion. Fix mnemonic attribute.
3102 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3103 indirect branch via thunk if requested.
3104 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3105 ("*indirect_jump"): Disable for branch conversion using out of
3107 ("indirect_jump_via_thunk<mode>_z10")
3108 ("indirect_jump_via_thunk<mode>")
3109 ("indirect_jump_via_inlinethunk<mode>_z10")
3110 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3111 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3112 ("casesi_jump_via_inlinethunk<mode>_z10")
3113 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3114 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3115 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3116 ("*indirect2_jump"): Disable for branch conversion.
3117 ("casesi_jump"): Turn into expander and expand patterns for branch
3119 ("return_use"): New expander.
3120 ("*return"): Emit return via thunk and rename it to ...
3121 ("*return<mode>"): ... this one.
3122 * config/s390/s390.opt: Add new options and and enum for the
3125 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3127 * lra-constraints.c (match_reload): Unconditionally use
3128 gen_lowpart_SUBREG, rather than selecting between that
3129 and equivalent gen_rtx_SUBREG code.
3131 2018-02-08 Richard Biener <rguenther@suse.de>
3133 PR tree-optimization/84233
3134 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3135 changed flag instead of boguously re-using phi_inserted.
3137 2018-02-08 Martin Jambor <mjambor@suse.cz>
3139 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3140 static local variables.
3142 2018-02-08 Richard Biener <rguenther@suse.de>
3144 PR tree-optimization/84278
3145 * tree-vect-stmts.c (vectorizable_store): When looking for
3146 smaller vector types to perform grouped strided loads/stores
3147 make sure the mode is supported by the target.
3148 (vectorizable_load): Likewise.
3150 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3152 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3153 Increase LDP/STP opportunities by adding adjacent callee-saves.
3155 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3157 PR rtl-optimization/84068
3158 PR rtl-optimization/83459
3159 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3161 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3163 PR tree-optimization/84224
3164 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3165 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3168 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3171 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3172 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3173 count, remove one USE.
3175 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3177 * doc/install.texi (Configuration): Document the
3178 --with-long-double-format={ibm,ieee} PowerPC configuration
3182 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3183 Convert from define_expand to be define_insn_and_split. Rework
3184 float/double/_Float128 conversions to QI/HI/SImode to work with
3185 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3186 conversions to QI/HImode types did a store and then a load to
3187 truncate the value. For conversions to VSX registers, don't split
3188 the insn, instead emit the code directly. Use the code iterator
3189 any_fix to combine signed and unsigned conversions.
3190 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3191 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3192 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3193 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3194 (fix_<mode>di2_hw): Likewise.
3195 (fixuns_<mode>di2_hw): Likewise.
3196 (fix_<mode>si2_hw): Likewise.
3197 (fixuns_<mode>si2_hw): Likewise.
3198 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3199 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3200 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3201 fix<uns>_trunc<SFDF:mode>si2_p8.
3202 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3204 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3205 (fix<uns>_<mode>_mem): Likewise.
3206 (fctiw<u>z_<mode>_mem): Likewise.
3207 (fix<uns>_<mode>_mem): Likewise.
3208 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3209 the register allocator from doing a direct move to the GPRs to do
3210 a store, and instead use the ISA 3.0 store byte/half-word from
3211 vector register instruction. For IEEE 128-bit floating point,
3212 also optimize stores of 32-bit ints.
3213 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3215 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3217 * genextract.c (push_pathstr_operand): New function to support
3219 (walk_rtx): Call push_pathstr_operand.
3220 (print_path): Support [a-zA-Z].
3222 2018-02-07 Richard Biener <rguenther@suse.de>
3224 PR tree-optimization/84037
3225 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3226 (cse_and_gimplify_to_preheader): Declare.
3227 (vect_get_place_in_interleaving_chain): Likewise.
3228 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3230 (_loop_vec_info::~_loop_vec_info): Delete it.
3231 (cse_and_gimplify_to_preheader): New function.
3232 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3233 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3234 (vectorizable_load): Likewise. For grouped stores always base
3235 the IV on the first element.
3236 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3237 condition before gimplifying.
3239 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3241 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3242 *DIV_EXPR and *MOD_EXPR.
3244 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3247 * config/i386/i386.c (ix86_option_override_internal): Mask out
3248 the CF_SET bit when checking -fcf-protection.
3250 2018-02-07 Tom de Vries <tom@codesourcery.com>
3253 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3256 2018-02-07 Richard Biener <rguenther@suse.de>
3258 PR tree-optimization/84204
3259 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3262 PR tree-optimization/84205
3263 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3264 special-case isl_ast_op_zdiv_r.
3266 PR tree-optimization/84223
3267 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3268 Only add conditions from within the region.
3269 (gather_bbs::after_dom_children): Adjust.
3271 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3274 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3275 * config/avr/avr.md: Only post-reload split REG-REG moves if
3276 either register is GENERAL_REG_P.
3278 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3280 PR tree-optimization/84235
3281 * tree-ssa-scopedtables.c
3282 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3283 if the subtraction is performed in floating point type where NaNs are
3284 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3285 build 1. Formatting fix.
3287 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3290 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3291 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3292 and skip it regardless of bb boundaries. Use CALL_P macro,
3293 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3294 unnecessarily, formatting fix.
3296 2018-02-06 Michael Collison <michael.collison@arm.com>
3298 * config/arm/thumb2.md:
3299 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3300 (*thumb_mov_notscc): Ditto.
3302 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3305 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3306 unsigned_fix, not "s".
3308 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3310 * configure.ac (gcc_fn_eh_frame_ro): New function.
3311 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3312 correct .eh_frame permissions.
3313 * configure: Regenerate.
3315 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3317 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3320 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3322 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3323 Display warning message for -mno-speculate-indirect-jumps.
3325 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3327 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3329 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3331 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3333 PR tree-optimization/84225
3334 * tree-eh.c (find_trapping_overflow): Only call
3335 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3337 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3340 * config/i386/i386.c: Reimplement the check of possible options
3341 -mibt/-mshstk conbination. Change error messages.
3342 * doc/invoke.texi: Fix a typo: remove extra '='.
3344 2018-02-06 Marek Polacek <polacek@redhat.com>
3346 PR tree-optimization/84228
3347 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3349 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3352 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3353 emitted arch directives.
3354 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3355 __ARM_FEATURE_COPROC before changing architectures.
3357 2018-02-06 Richard Biener <rguenther@suse.de>
3359 * config/i386/i386.c (print_reg): Fix typo.
3360 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3362 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3364 * configure: Regenerate.
3366 2018-02-05 Martin Sebor <msebor@redhat.com>
3368 PR tree-optimization/83369
3369 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3372 2018-02-05 Martin Liska <mliska@suse.cz>
3374 * doc/invoke.texi: Cherry-pick upstream r323995.
3376 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3378 * ira.c (ira_init_register_move_cost): Adjust comment.
3380 2018-02-05 Martin Liska <mliska@suse.cz>
3382 PR gcov-profile/84137
3383 * doc/gcov.texi: Fix typo in documentation.
3385 2018-02-05 Martin Liska <mliska@suse.cz>
3387 PR gcov-profile/83879
3388 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3389 using dlopen functionality.
3391 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3393 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3394 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3395 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3396 _mm_maskz_range_round_ss): New intrinsics.
3397 (__builtin_ia32_rangesd128_round)
3398 (__builtin_ia32_rangess128_round): Remove.
3399 (__builtin_ia32_rangesd128_mask_round,
3400 __builtin_ia32_rangess128_mask_round): New builtins.
3401 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3402 __builtin_ia32_rangess128_round): Remove.
3403 (__builtin_ia32_rangesd128_mask_round,
3404 __builtin_ia32_rangess128_mask_round): New builtins.
3405 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3406 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3407 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3408 "<round_saeonly_constraint>")): Changed to ...
3409 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3410 "<round_saeonly_scalar_constraint>")): ... this.
3411 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3412 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3413 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3414 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3415 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3417 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3419 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3421 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3422 Remove all values except native, 8540 and 8548.
3424 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3426 * config/i386/i386.c (ix86_output_function_return): Pass
3427 INVALID_REGNUM, instead of -1, as invalid register number to
3428 indirect_thunk_name and output_indirect_thunk.
3430 2018-02-02 Julia Koval <julia.koval@intel.com>
3432 * config.gcc: Add -march=icelake.
3433 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3434 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3435 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3436 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3437 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3438 (processor_target_table): Add icelake.
3439 (ix86_option_override_internal): Handle new PTAs.
3440 (get_builtin_code_for_version): Handle icelake.
3441 (M_INTEL_COREI7_ICELAKE): New.
3442 (fold_builtin_cpu): Handle icelake.
3443 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3444 * doc/invoke.texi: Add -march=icelake.
3446 2018-02-02 Julia Koval <julia.koval@intel.com>
3448 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3449 to wide_int_bitmask.
3450 * wide-int-bitmask.h: New.
3452 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3455 * config/i386/i386.md: Replace Pmode with word_mode in
3456 builtin_setjmp_setup and builtin_longjmp to support x32.
3458 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3462 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3464 (rs6000_supported_cpu_names): New static variable.
3465 (linux_cpu_translation_table): Likewise.
3466 (elf_platform) <cpu>: Define new static variable and use it.
3467 Translate kernel AT_PLATFORM name to canonical name if needed.
3468 Error if platform name is unknown.
3470 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3473 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3475 2018-02-01 Jeff Law <law@redhat.com>
3478 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3479 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3480 the scratch if RELEASE_VIA_POP is false.
3481 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3482 If we have to save a temporary register, decrement SIZE appropriately.
3483 Pass new arguments to release_scratch_register_on_entry.
3484 (ix86_adjust_stack_and_probe): Likewise.
3485 (ix86_emit_probe_stack_range): Pass new arguments to
3486 release_scratch_register_on_entry.
3488 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3490 PR rtl-optimization/84157
3491 * combine.c (change_zero_ext): Use REG_P predicate in
3492 front of HARD_REGISTER_P predicate.
3494 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3496 * config/avr/avr.c (avr_option_override): Move disabling of
3497 -fdelete-null-pointer-checks to...
3498 * common/config/avr/avr-common.c (avr_option_optimization_table):
3501 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3503 PR tree-optimization/81635
3504 * tree-data-ref.c (split_constant_offset_1): For types that
3505 wrap on overflow, try to use range info to prove that wrapping
3508 2018-02-01 Renlin Li <renlin.li@arm.com>
3511 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3513 (aarch64_register_move_cost): Likewise.
3514 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3516 (REG_CLASS_NAMES): Likewise.
3517 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3518 TAILCALL_ADDR_REGS. Remove IP registers.
3519 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3521 2018-02-01 Richard Biener <rguenther@suse.de>
3523 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3524 for specifying RPO order and allow NULL for that.
3525 * domwalk.c (dom_walker::dom_walker): Likewise.
3526 (dom_walker::walk): Handle NULL RPO order.
3527 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3529 (rewrite_update_dom_walker): Likewise.
3530 (mark_def_dom_walker): Likewise.
3532 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3534 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3535 (aarch64_maybe_expand_sve_subreg_move): Declare.
3536 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3537 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3539 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3540 that are semantically a reverse operation.
3541 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3542 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3543 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3545 (aarch64_can_change_mode_class): For big-endian, forbid changes
3546 between two SVE modes if they have different element sizes.
3548 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3550 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3551 the TImode handling for big-endian targets.
3553 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3555 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3556 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3558 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3559 Remove BSWAP handing for big-endian targets and use the form of
3560 LD1RQ appropariate for the mode.
3562 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3564 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3565 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3568 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3571 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3572 check for operands that need to go through aarch64_sve_reload_be.
3574 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3576 PR tree-optimization/81661
3577 PR tree-optimization/84117
3578 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3579 * tree-eh.c: Include gimplify.h.
3580 (find_trapping_overflow, replace_trapping_overflow,
3581 rewrite_to_non_trapping_overflow): New functions.
3582 * tree-vect-loop.c: Include tree-eh.h.
3583 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3584 * tree-data-ref.c: Include tree-eh.h.
3585 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3587 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3589 PR rtl-optimization/84123
3590 * combine.c (change_zero_ext): Check if hard register satisfies
3591 can_change_dest_mode before calling gen_lowpart_SUBREG.
3593 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3596 * ira.c (ira_init_register_move_cost): Remove assert.
3598 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3600 PR rtl-optimization/84071
3601 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3602 * doc/tm.texi: Regenerate.
3604 2018-01-31 Richard Biener <rguenther@suse.de>
3606 PR tree-optimization/84132
3607 * tree-data-ref.c (analyze_miv_subscript): Properly
3608 check whether evolution_function_is_affine_multivariate_p
3609 before calling gcd_of_steps_may_divide_p.
3611 2018-01-31 Julia Koval <julia.koval@intel.com>
3614 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3615 * config/i386/i386.md (rdpid_rex64) New.
3616 (rdpid): Make 32bit only.
3618 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3621 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3622 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3624 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3627 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3629 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3631 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3633 PR rtl-optimization/84071
3634 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3635 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3637 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3639 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3640 (arc_attribute_table): Add 'aux' attribute.
3641 (arc_in_small_data_p): Consider aux like variables.
3642 (arc_is_aux_reg_p): New function.
3643 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3644 (arc_get_aux_arg): New function.
3645 (prepare_move_operands): Handle aux-register access.
3646 (arc_handle_aux_attribute): New function.
3647 * doc/extend.texi (ARC Variable attributes): Add subsection.
3649 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3651 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3652 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3653 (arc_attribute_table): Add 'uncached' attribute.
3654 (arc_print_operand): Print '.di' flag for uncached memory
3656 (arc_in_small_data_p): Do not consider for small data the uncached
3658 (arc_is_uncached_mem_p): New function.
3659 * config/arc/predicates.md (compact_store_memory_operand): Check
3660 for uncached memory accesses.
3661 (nonvol_nonimm_operand): Likewise.
3662 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3664 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3667 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3668 falign-loops=): Add Optimization flag.
3670 2018-01-30 Jeff Law <law@redhat.com>
3673 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3674 INT_REGISTERS_SAVED. Check it prior to calling
3675 get_scratch_register_on_entry.
3676 (ix86_adjust_stack_and_probe): Similarly.
3677 (ix86_emit_probe_stack_range): Similarly.
3678 (ix86_expand_prologue): Corresponding changes.
3680 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3683 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3684 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3686 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3689 * lra-constraints.c (curr_insn_transform): Process AND in the
3692 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3694 PR rtl-optimization/83986
3695 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3696 dependence against last_pending_memory_flush in addition to
3699 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3701 PR tree-optimization/81611
3702 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3705 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3708 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3711 2018-01-30 Richard Biener <rguenther@suse.de>
3712 Jakub Jelinek <jakub@redhat.com>
3714 PR tree-optimization/84111
3715 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3716 inner loops added during recursion, as they don't have up-to-date
3719 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3722 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3723 (can_inline_edge_by_limits_p): ... here.
3724 (can_early_inline_edge_p, check_callers,
3725 update_caller_keys, update_callee_keys, recursive_inlining,
3726 add_new_edges_to_heap, speculation_useful_p,
3727 inline_small_functions,
3728 inline_small_functions, flatten_function,
3729 inline_to_all_callers_1): Update.
3731 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3733 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3736 2018-01-30 Richard Biener <rguenther@suse.de>
3738 PR tree-optimization/83008
3739 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3740 invariant and constant vector uses in stmts when they need
3743 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3746 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3747 * configure: Regenerate.
3749 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3751 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3753 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3754 Use gen_rtx_REG rather than gen_lowpart.
3756 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3758 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3759 rather than 0 when creating partial subregs.
3761 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3763 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3766 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3769 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3770 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3771 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3772 flags. This restores the settings used before the 2017-07-24.
3773 Turning off pre increment/decrement/modify allows IVOPTS to
3774 optimize DF/SF loops where the index is an int.
3776 2018-01-29 Richard Biener <rguenther@suse.de>
3777 Kelvin Nilsen <kelvin@gcc.gnu.org>
3780 * tree-vect-stmts.c (vectorizable_call): Don't call
3781 targetm.vectorize_builtin_md_vectorized_function if callee is
3784 2018-01-22 Carl Love <cel@us.ibm.com>
3786 * doc/extend.tex: Fix typo in second arg in
3787 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3789 2018-01-29 Richard Biener <rguenther@suse.de>
3791 PR tree-optimization/84086
3792 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3793 (flush_ssaname_freelist): When SSA names were released reset
3794 the SCEV hash table.
3796 2018-01-29 Richard Biener <rguenther@suse.de>
3798 PR tree-optimization/84057
3799 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3800 removed paths when removing edges.
3802 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3804 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3805 -mfunction-return=@var{choice}.
3807 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3810 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3812 (layout::print_source_line): Likewise.
3813 (test_get_line_width_without_trailing_whitespace): Add test cases.
3815 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3818 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3821 2018-01-26 Jim Wilson <jimw@sifive.com>
3823 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3825 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3828 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3830 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3831 and CMP + SUB-immediate -> SUBS.
3833 2018-01-26 Martin Sebor <msebor@redhat.com>
3835 PR tree-optimization/83896
3836 * tree-ssa-strlen.c (get_string_len): Rename...
3837 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3838 Avoid assuming length is constant.
3839 (handle_char_store): Use HOST_WIDE_INT for string length.
3841 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3844 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3845 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3847 2018-01-26 Richard Biener <rguenther@suse.de>
3849 PR rtl-optimization/84003
3850 * dse.c (record_store): Only record redundant stores when
3851 the earlier store aliases at least all accesses the later one does.
3853 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3855 PR rtl-optimization/83985
3856 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3857 REG_CFA_RESTORE insns.
3858 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3859 REG_CFA_RESTORE insns here.
3862 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3863 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3865 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3867 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3868 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3869 (arc_init): Likewise.
3870 (arc_override_options): Likewise.
3871 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3873 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3875 * config/arc/arc.h (TARGET_DBNZ): Define.
3876 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3877 properly set the tune attribute.
3878 (dbnz): Use TARGET_DBNZ guard.
3879 * config/arc/arc.opt (mtune): Add core3 option.
3881 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3883 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3884 recognize new pic like addresses.
3885 (arc_delegitimize_address): Clean up.
3887 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3889 * config/arc/arc-arches.def: Option mrf16 valid for all
3891 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3892 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3893 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3894 * config/arc/arc-tables.opt: Regenerate.
3895 * config/arc/arc.c (arc_conditional_register_usage): Handle
3896 reduced register file case.
3897 (arc_file_start): Set must have build attributes.
3898 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3900 * config/arc/arc.opt (mrf16): Add new option.
3901 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3902 * config/arc/genmultilib.awk: Handle new mrf16 option.
3903 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3904 * config/arc/t-multilib: Regenerate.
3905 * doc/invoke.texi (ARC Options): Document mrf16 option.
3907 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3909 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3910 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3911 (arc_attribute_table): Add 'secure_call' attribute.
3912 (arc_print_operand): Print secure call operand.
3913 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3915 (arc_is_secure_call_p): New function. * config/arc/arc.md
3916 (call_i): Add support for sjli instruction.
3917 (call_value_i): Likewise.
3918 * config/arc/constraints.md (Csc): New constraint.
3920 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3921 John Eric Martin <John.Martin@emmicro-us.com>
3923 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3924 * config/arc/arc.c (_arc_jli_section): New struct.
3925 (arc_jli_section): New type.
3926 (rc_jli_sections): New static variable.
3927 (arc_handle_jli_attribute): New function.
3928 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3929 (arc_file_end): New function.
3930 (TARGET_ASM_FILE_END): Define.
3931 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3932 (arc_add_jli_section): New function.
3933 (jli_call_scan): Likewise.
3934 (arc_reorg): Call jli_call_scan.
3935 (arc_output_addsi): Remove 'S' from printing asm operand.
3936 (arc_is_jli_call_p): New function.
3937 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3939 (movhi_insn): Likewise.
3940 (movsi_insn): Likewise.
3941 (movsi_set_cc_insn): Likewise.
3942 (loadqi_update): Likewise.
3943 (load_zeroextendqisi_update): Likewise.
3944 (load_signextendqisi_update): Likewise.
3945 (loadhi_update): Likewise.
3946 (load_zeroextendhisi_update): Likewise.
3947 (load_signextendhisi_update): Likewise.
3948 (loadsi_update): Likewise.
3949 (loadsf_update): Likewise.
3950 (movsicc_insn): Likewise.
3951 (bset_insn): Likewise.
3952 (bxor_insn): Likewise.
3953 (bclr_insn): Likewise.
3954 (bmsk_insn): Likewise.
3955 (bicsi3_insn): Likewise.
3956 (cmpsi_cc_c_insn): Likewise.
3957 (movsi_ne): Likewise.
3958 (movsi_cond_exec): Likewise.
3959 (clrsbsi2): Likewise.
3970 (call_i): Remove 'S' asm letter, add jli instruction.
3971 (call_value_i): Likewise.
3972 * config/arc/arc.op (mjli-always): New option.
3973 * config/arc/constraints.md (Cji): New constraint.
3974 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3976 (subsf3_fpx): Likewise.
3977 (mulsf3_fpx): Likewise.
3978 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3980 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3982 * doc/invoke.texi (ARC): Document mjli-always option.
3984 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3986 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3987 avoid addition with 0 and use incw and decw where possible.
3989 2018-01-26 Richard Biener <rguenther@suse.de>
3991 PR tree-optimization/81082
3992 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3993 association if it requires casting to unsigned.
3994 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3995 from fold_plusminus_mult_expr to catch important cases late when
3996 range info is available.
3998 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4000 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4001 * configure.ac (hidden_linkonce): New test.
4002 * configure: Regenerate.
4003 * config.in: Regenerate.
4005 2018-01-26 Julia Koval <julia.koval@intel.com>
4007 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4008 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4009 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4010 _mm_mask_bitshuffle_epi64_mask): Fix type.
4011 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4012 USI_FTYPE_V4DI_V4DI_USI): Remove.
4013 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4014 __builtin_ia32_vpshufbitqmb256_mask,
4015 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4016 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4017 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4019 2018-01-26 Alan Modra <amodra@gmail.com>
4022 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4023 UNSPEC_VBPERMQ. Sort other unspecs.
4025 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4027 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4029 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4032 * predict.c (drop_profile): Do not push/pop cfun; update also
4034 (handle_missing_profiles): Fix logic looking for zero profiles.
4036 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4039 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4040 on functions with #pragma omp declare simd or functions with simd
4042 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4043 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4044 Remove trailing \n from warning_at calls.
4046 2018-01-25 Tom de Vries <tom@codesourcery.com>
4049 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4050 for neutered workers.
4052 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4055 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4056 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4058 2018-01-24 Jeff Law <law@redhat.com>
4061 * i386.c (get_probe_interval): Move to earlier point.
4062 (ix86_compute_frame_layout): If -fstack-clash-protection and
4063 the frame is larger than the probe interval, then use pushes
4064 to save registers rather than reg->mem moves.
4065 (ix86_expand_prologue): Remove conditional for int_registers_saved
4068 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4071 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4072 min/max for never referenced object.
4074 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4077 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4079 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4080 attributes from DECL_ATTRIBUTES (decl) without affecting
4081 DECL_ATTRIBUTES (current_function_decl).
4082 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4083 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4085 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4087 PR tree-optimization/83979
4088 * fold-const.c (fold_comparison): Use constant_boolean_node
4089 instead of boolean_{true,false}_node.
4091 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4093 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4096 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4098 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4099 Simplify the clause that sets the length attribute.
4100 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4101 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4102 clause that sets the length attribute.
4103 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4105 2018-01-24 Tom de Vries <tom@codesourcery.com>
4108 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4109 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4110 Add strict parameter.
4111 (prevent_branch_around_nothing): Insert dummy insn between branch to
4112 label and label with no ptx insn inbetween.
4113 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4115 2018-01-24 Tom de Vries <tom@codesourcery.com>
4118 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4119 for neutered threads in warp.
4120 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4122 2018-01-24 Richard Biener <rguenther@suse.de>
4124 PR tree-optimization/83176
4125 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4128 2018-01-24 Richard Biener <rguenther@suse.de>
4130 PR tree-optimization/82819
4131 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4132 code generating pluses that are no-ops in the target precision.
4134 2018-01-24 Richard Biener <rguenther@suse.de>
4137 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4139 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4141 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4142 to merge probabilities.
4143 * predict.c (probably_never_executed): Also mark as cold functions
4144 with global 0 profile and guessed local profile.
4145 * profile-count.c (profile_probability::combine_with_count): New
4147 * profile-count.h (profile_probability::operator*,
4148 profile_probability::operator*=, profile_probability::operator/,
4149 profile_probability::operator/=): Reduce precision to adjusted
4150 and set value to guessed on contradictory divisions.
4151 (profile_probability::combine_with_freq): Remove.
4152 (profile_probability::combine_wiht_count): Declare.
4153 (profile_count::force_nonzero):: Set to adjusted.
4154 (profile_count::probability_in):: Set quality to adjusted.
4155 * tree-ssa-tail-merge.c (replace_block_by): Use
4158 2018-01-23 Andrew Waterman <andrew@sifive.com>
4159 Jim Wilson <jimw@sifive.com>
4161 * config/riscv/riscv.c (riscv_stack_boundary): New.
4162 (riscv_option_override): Set riscv_stack_boundary. Handle
4163 riscv_preferred_stack_boundary_arg.
4164 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4165 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4166 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4167 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4168 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4169 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4171 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4174 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4175 of struct ix86_frame.
4176 (ix86_expand_epilogue): Likewise. Add a local variable for
4177 the reg_save_offset field in struct ix86_frame.
4179 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4181 PR tree-optimization/82604
4182 * tree-loop-distribution.c (enum partition_kind): New enum item
4183 PKIND_PARTIAL_MEMSET.
4184 (partition_builtin_p): Support above new enum item.
4185 (generate_code_for_partition): Ditto.
4186 (compute_access_range): Differentiate cases that equality can be
4187 proven at all loops, the innermost loops or no loops.
4188 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4189 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4190 (finalize_partitions, distribute_loop): Don't fuse partition of
4191 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4192 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4195 2018-01-23 Martin Liska <mliska@suse.cz>
4197 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4198 order to ignore the predictor.
4199 (PRED_POLYMORPHIC_CALL): Likewise.
4200 (PRED_RECURSIVE_CALL): Likewise.
4202 2018-01-23 Martin Liska <mliska@suse.cz>
4204 * tree-profile.c (tree_profiling): Print function header to
4205 aware reader which function we are working on.
4206 * value-prof.c (gimple_find_values_to_profile): Do not print
4207 not interesting value histograms.
4209 2018-01-23 Martin Liska <mliska@suse.cz>
4211 * profile-count.h (enum profile_quality): Add
4212 profile_uninitialized as the first value. Do not number values
4213 as they are zero based.
4214 (profile_count::verify): Update sanity check.
4215 (profile_probability::verify): Likewise.
4217 2018-01-23 Nathan Sidwell <nathan@acm.org>
4219 * doc/invoke.texi (ffor-scope): Deprecate.
4221 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4223 PR tree-optimization/83510
4224 * domwalk.c (set_all_edges_as_executable): New function.
4225 (dom_walker::dom_walker): Convert bool param
4226 "skip_unreachable_blocks" to enum reachability. Move setup of
4227 edge flags to set_all_edges_as_executable and only do it when
4228 reachability is REACHABLE_BLOCKS.
4229 * domwalk.h (enum dom_walker::reachability): New enum.
4230 (dom_walker::dom_walker): Convert bool param
4231 "skip_unreachable_blocks" to enum reachability.
4232 (set_all_edges_as_executable): New decl.
4233 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4234 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4236 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4237 but converting true to REACHABLE_BLOCKS.
4238 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4240 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4241 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4242 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4244 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4245 if check_all_array_refs will be called.
4247 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4249 * tree.c (selftest::test_location_wrappers): Add more test
4252 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4254 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4255 (selftest::test_bit_in_range): Likewise.
4257 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4260 * doc/sourcebuild.texi (vect_float): Say that the selector
4261 only describes the situation when -funsafe-math-optimizations is on.
4262 (vect_float_strict): Document.
4264 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4266 PR tree-optimization/83965
4267 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4268 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4269 instead of checking only for a reduction.
4270 (vect_recog_widen_sum_pattern): Likewise.
4272 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4274 * predict.c (probably_never_executed): Only use precise profile info.
4275 (compute_function_frequency): Skip after inlining hack since we now
4276 have quality checking.
4278 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4280 * profile-count.h (profile_probability::very_unlikely,
4281 profile_probability::unlikely, profile_probability::even): Set
4282 precision to guessed.
4284 2018-01-23 Richard Biener <rguenther@suse.de>
4286 PR tree-optimization/83963
4287 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4288 Properly terminate dominator walk when crossing the exit edge not
4289 when visiting its source block.
4291 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4294 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4295 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4297 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4299 PR tree-optimization/83957
4300 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4301 semicolon after for body surrounded by braces.
4303 PR tree-optimization/83081
4304 * profile-count.h (profile_probability::split): New method.
4305 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4306 Use profile_probability::split.
4307 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4308 when splitting a single conditional jump into 2.
4310 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4312 PR tree-optimization/69452
4313 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4318 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4319 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4320 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4322 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4324 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4325 * config/rl78/rl78.md (movdi): New define_expand.
4326 * config/rl78/rl78.c (rl78_split_movdi): New function.
4328 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4331 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4333 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4334 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4335 128-bit to produce an UNSPEC move to get the double word with the
4336 signbit and then a shift directly to do signbit.
4337 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4338 implementation with a new version that just does either a direct
4339 move or a regular move. Move memory interface to separate insns.
4340 Move insns so they are next to the expander.
4341 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4342 with signbit move. Split big and little endian case.
4343 (signbit<mode>2_dm_mem_le): Likewise.
4344 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4345 (signbit<mode>2_dm2): Likewise.
4347 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4349 * config/rl78/rl78.md (anddi3): New define_expand.
4351 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4353 * config/rl78/rl78.md (umindi3): New define_expand.
4355 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4357 * config/rl78/rl78.md (smindi3): New define_expand.
4359 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4361 * config/rl78/rl78.md (smaxdi3): New define_expand.
4363 2018-01-22 Carl Love <cel@us.ibm.com>
4365 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4366 LVX_V1TI): Add macro expansion.
4367 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4368 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4369 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4370 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4371 Change check to determine if the instruction is a byte reversing
4372 entry. Fix typo in comment.
4373 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4374 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4375 Add def_builtin calls for new builtins.
4376 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4377 Add define_insn expansion.
4379 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4381 * config/rl78/rl78.md (umaxdi3): New define_expand.
4383 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4385 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4386 for non-QImode registers.
4388 2018-01-22 Richard Biener <rguenther@suse.de>
4390 PR tree-optimization/83963
4391 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4392 including the loop exit block.
4393 (scop_detection::merge_sese): Likewise.
4394 (scop_detection::add_scop): Do it here instead.
4396 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4398 * doc/sourcebuild.texi (arm_softfloat): Document.
4400 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4403 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4404 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4405 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4407 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4408 David Edelsohn <dje.gcc@gmail.com>
4411 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4412 Change "crset eq" to "crset 2".
4413 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4414 (*call_indirect_aix<mode>_nospec): Likewise.
4415 (*call_value_indirect_aix<mode>_nospec): Likewise.
4416 (*call_indirect_elfv2<mode>_nospec): Likewise.
4417 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4418 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4419 change assembly output from . to $.
4420 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4421 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4422 (*tablejump<mode>_internal1_nospec): Likewise.
4424 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4427 * config/sh/sh_optimize_sett_clrt.cc:
4428 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4430 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4432 PR tree-optimization/83940
4433 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4434 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4435 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4436 use it to pass back the definition type.
4437 (vect_check_store_rhs): Likewise rhs_dt_out.
4438 (vect_build_gather_load_calls): Add a mask_dt argument and use
4439 it instead of a call to vect_is_simple_use.
4440 (vectorizable_store): Update calls to vect_check_load_store_mask
4441 and vect_check_store_rhs. Use the dt returned by the latter instead
4442 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4443 instead of calls to vect_is_simple_use. Pass the scalar rather
4444 than the vector operand to vect_is_simple_use when handling
4445 second and subsequent copies of an rhs value.
4446 (vectorizable_load): Update calls to vect_check_load_store_mask
4447 and vect_build_gather_load_calls. Use the cached mask_dt and
4448 gs_info.offset_dt instead of calls to vect_is_simple_use.
4450 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4453 * tree-emutls.c: Include gimplify.h.
4454 (lower_emutls_2): New function.
4455 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4456 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4457 it before further processing.
4460 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4461 UINTVAL (trueop1) instead of INTVAL (op1).
4463 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4467 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4468 INCOMING_FRAME_SP_OFFSET if not defined.
4469 (scan_trace): Add ENTRY argument. If true and
4470 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4471 emit a note to adjust the CFA offset.
4472 (create_cfi_notes): Adjust scan_trace callers.
4473 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4474 INCOMING_FRAME_SP_OFFSET in the CIE.
4475 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4476 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4478 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4479 * doc/tm.texi: Regenerated.
4481 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4483 PR rtl-optimization/83147
4484 * lra-constraints.c (remove_inheritance_pseudos): Use
4485 lra_substitute_pseudo_within_insn.
4487 2018-01-19 Tom de Vries <tom@codesourcery.com>
4488 Cesar Philippidis <cesar@codesourcery.com>
4491 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4493 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4496 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4497 spaces for function labels.
4499 2018-01-19 Martin Liska <mliska@suse.cz>
4501 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4502 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4503 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4504 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4505 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4506 (PRED_CONST_RETURN): Change from 69 to 65.
4507 (PRED_NULL_RETURN): Change from 91 to 71.
4508 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4509 (PRED_LOOP_GUARD): Change from 66 to 73.
4511 2018-01-19 Martin Liska <mliska@suse.cz>
4513 * predict.c (predict_insn_def): Add new assert.
4514 (struct branch_predictor): Change type to signed integer.
4515 (test_prediction_value_range): Amend test to cover
4517 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4518 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4519 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4520 (PRED_LOOP_IV_COMPARE): Likewise.
4521 * predict.h (PROB_UNINITIALIZED): Define new constant.
4523 2018-01-19 Martin Liska <mliska@suse.cz>
4525 * predict.c (dump_prediction): Add new format for
4526 analyze_brprob.py script which is enabled with -details
4528 * profile-count.h (precise_p): New function.
4530 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4532 PR tree-optimization/83922
4533 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4534 there are no statements that need masking.
4535 (vect_active_double_reduction_p): New function.
4536 (vect_analyze_loop_operations): Use it when handling phis that
4537 are not in the loop header.
4539 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4541 PR tree-optimization/83914
4542 * tree-vect-loop.c (vectorizable_induction): Don't convert
4543 init_expr or apply the peeling adjustment for inductions
4544 that are nested within the vectorized loop.
4546 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4548 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4551 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4555 * function.h (gimplify_parameters): Add gimple_seq * argument.
4556 * function.c: Include gimple.h and options.h.
4557 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4558 for the added local temporaries if needed.
4559 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4560 if there are any parameter cleanups, wrap whole body into a
4561 try/finally with the cleanups.
4563 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4566 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4567 Use GET_MODE_CLASS for scalar floating point.
4569 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4573 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4574 Fix call of call_cgraph_insertion_hooks.
4576 2018-01-18 Martin Sebor <msebor@redhat.com>
4578 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4580 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4583 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4586 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4589 * common.opt: (-ffile-prefix-map): New option.
4590 * opts.c (common_handle_option): Defer it.
4591 * opts-global.c (handle_common_deferred_options): Handle it.
4592 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4593 * file-prefix-map.h: New file.
4594 (remap_debug_filename, add_debug_prefix_map): ...here.
4595 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4596 * final.c (debug_prefix_map, add_debug_prefix_map
4597 remap_debug_filename): Move to...
4598 * file-prefix-map.c: New file.
4599 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4600 generalize, get rid of alloca(), use strrchr() instead of strchr().
4601 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4602 Implement in terms of add_prefix_map().
4603 (remap_macro_filename, remap_debug_filename): Implement in term of
4605 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4606 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4607 * dbxout.c: Include file-prefix-map.h.
4608 * varasm.c: Likewise.
4609 * vmsdbgout.c: Likewise.
4610 * xcoffout.c: Likewise.
4611 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4612 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4613 * doc/invoke.texi (-ffile-prefix-map): Document.
4614 (-fdebug-prefix-map): Update description.
4616 2018-01-18 Martin Liska <mliska@suse.cz>
4618 * config/i386/i386.c (indirect_thunk_name): Document that also
4620 (output_indirect_thunk): Document why both instructions
4621 (pause and lfence) are generated.
4623 2018-01-18 Richard Biener <rguenther@suse.de>
4625 PR tree-optimization/83887
4626 * graphite-scop-detection.c
4627 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4628 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4629 (scop_detection::merge_sese): Re-implement with a flood-fill
4630 algorithm that properly finds a SESE region if it exists.
4632 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4635 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4636 pointer_diff optimizations use view_convert instead of convert.
4638 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4640 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4641 Generate different code for -mno-speculate-indirect-jumps.
4642 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4643 (*call_indirect_aix<mode>): Disable for
4644 -mno-speculate-indirect-jumps.
4645 (*call_indirect_aix<mode>_nospec): New define_insn.
4646 (*call_value_indirect_aix<mode>): Disable for
4647 -mno-speculate-indirect-jumps.
4648 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4649 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4650 -mno-speculate-indirect-jumps.
4651 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4653 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4655 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4656 long double type, set the flags for noting the default long double
4657 type, even if we don't pass or return a long double type.
4659 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4662 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4665 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4667 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4668 support for merge[hl].
4669 (fold_mergehl_helper): New helper function.
4670 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4671 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4672 (altivec_vmrglw_direct): Add xxmrglw insn.
4674 2018-01-17 Andrew Waterman <andrew@sifive.com>
4676 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4677 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4679 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4682 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4683 call the lto_location_cache before reading the
4684 DECL_SOURCE_LOCATION of the types.
4686 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4687 Richard Sandiford <richard.sandiford@linaro.org>
4689 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4690 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4691 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4692 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4693 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4695 * config/aarch64/constraints.md (aarch64_movti_operand):
4697 * config/aarch64/predicates.md (Uti): Add new constraint.
4699 2018-01-17 Carl Love <cel@us.ibm.com>
4701 * config/rs6000/vsx.md (define_expand xl_len_r,
4702 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4703 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4705 (define_expand, define_insn): Move the shift left from the
4706 define_insn to the define_expand for lxvl and stxvl instructions.
4707 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4708 and XL_LEN_R definitions to PURE.
4710 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4712 * config/i386/i386.c (indirect_thunk_name): Declare regno
4713 as unsigned int. Compare regno with INVALID_REGNUM.
4714 (output_indirect_thunk): Ditto.
4715 (output_indirect_thunk_function): Ditto.
4716 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4717 in the call to output_indirect_thunk_function.
4719 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4722 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4723 rather than the size of inner_type to determine the stack slot size
4724 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4726 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4729 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4732 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4734 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4735 endian Linux systems to optionally enable multilibs for selecting
4736 the long double type if the user configured an explicit type.
4737 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4738 have no long double multilibs if not defined.
4739 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4740 warn if the user used -mabi={ieee,ibm}longdouble and we built
4741 multilibs for long double.
4742 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4743 appropriate multilib option.
4744 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4746 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4747 for building long double multilibs.
4748 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4750 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4752 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4755 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4757 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4760 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4763 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4766 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4768 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4769 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4771 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4773 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4774 different rtl trees depending on TARGET_64BIT.
4775 (rs6000_gen_lvx): Likewise.
4777 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4779 * config/visium/visium.md (nop): Tweak comment.
4780 (hazard_nop): Likewise.
4782 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4784 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4785 -mspeculate-indirect-jumps.
4786 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4787 for -mno-speculate-indirect-jumps.
4788 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4789 (*call_value_indirect_elfv2<mode>): Disable for
4790 -mno-speculate-indirect-jumps.
4791 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4792 (indirect_jump): Emit different RTL for
4793 -mno-speculate-indirect-jumps.
4794 (*indirect_jump<mode>): Disable for
4795 -mno-speculate-indirect-jumps.
4796 (*indirect_jump<mode>_nospec): New define_insn.
4797 (tablejump): Emit different RTL for
4798 -mno-speculate-indirect-jumps.
4799 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4800 (tablejumpsi_nospec): New define_expand.
4801 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4802 (tablejumpdi_nospec): New define_expand.
4803 (*tablejump<mode>_internal1): Disable for
4804 -mno-speculate-indirect-jumps.
4805 (*tablejump<mode>_internal1_nospec): New define_insn.
4806 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4809 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4811 * caller-save.c (insert_save): Drop unnecessary parameter. All
4814 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4815 Richard Biener <rguenth@suse.de>
4818 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4819 return early, inline manually is_gimple_sizepos. Make sure if we
4820 call gimplify_expr we don't end up with a gimple constant.
4821 * tree.c (variably_modified_type_p): Don't return true for
4822 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4823 * gimplify.h (is_gimple_sizepos): Remove.
4825 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4827 PR tree-optimization/83857
4828 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4829 vectorizable_live_operation for pure SLP statements.
4830 (vectorizable_live_operation): Handle PHIs.
4832 2018-01-16 Richard Biener <rguenther@suse.de>
4834 PR tree-optimization/83867
4835 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4836 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4838 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4841 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4842 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4843 If off is not INTEGER_CST, issue a may not be aligned warning
4844 rather than isn't aligned. Use isn%'t rather than isn't.
4845 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4847 <case MULT_EXPR>: Improve the case when bottom and one of the
4848 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4849 operand, in that case check if the other operand is multiple of
4850 bottom divided by the INTEGER_CST operand.
4852 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4855 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4856 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4857 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4858 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4859 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4860 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4861 (pa_function_arg_size): New function.
4863 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4865 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4866 in a separate statement.
4868 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4870 PR tree-optimization/83847
4871 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4872 group gathers and scatters.
4874 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4876 PR rtl-optimization/86620
4877 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4879 PR rtl-optimization/83213
4880 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4881 to last if both are JUMP_INSNs.
4883 PR tree-optimization/83843
4884 * gimple-ssa-store-merging.c
4885 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4886 store_immediate_info for bswap/nop orig_stores.
4888 2018-01-15 Andrew Waterman <andrew@sifive.com>
4890 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4892 <UDIV>: Increase cost if !TARGET_DIV.
4894 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4896 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4897 (define_attr "cr_logical_3op"): New.
4898 (cceq_ior_compare): Adjust.
4899 (cceq_ior_compare_complement): Adjust.
4900 (*cceq_rev_compare): Adjust.
4901 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4902 (is_cracked_insn): Adjust.
4903 (insn_must_be_first_in_group): Adjust.
4904 * config/rs6000/40x.md: Adjust.
4905 * config/rs6000/440.md: Adjust.
4906 * config/rs6000/476.md: Adjust.
4907 * config/rs6000/601.md: Adjust.
4908 * config/rs6000/603.md: Adjust.
4909 * config/rs6000/6xx.md: Adjust.
4910 * config/rs6000/7450.md: Adjust.
4911 * config/rs6000/7xx.md: Adjust.
4912 * config/rs6000/8540.md: Adjust.
4913 * config/rs6000/cell.md: Adjust.
4914 * config/rs6000/e300c2c3.md: Adjust.
4915 * config/rs6000/e500mc.md: Adjust.
4916 * config/rs6000/e500mc64.md: Adjust.
4917 * config/rs6000/e5500.md: Adjust.
4918 * config/rs6000/e6500.md: Adjust.
4919 * config/rs6000/mpc.md: Adjust.
4920 * config/rs6000/power4.md: Adjust.
4921 * config/rs6000/power5.md: Adjust.
4922 * config/rs6000/power6.md: Adjust.
4923 * config/rs6000/power7.md: Adjust.
4924 * config/rs6000/power8.md: Adjust.
4925 * config/rs6000/power9.md: Adjust.
4926 * config/rs6000/rs64.md: Adjust.
4927 * config/rs6000/titan.md: Adjust.
4929 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4931 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4932 ix86_indirect_branch_register logic.
4934 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4936 * config/i386/constraints.md (Bs): Update
4937 ix86_indirect_branch_register check. Don't check
4938 ix86_indirect_branch_register with GOT_memory_operand.
4940 * config/i386/predicates.md (GOT_memory_operand): Don't check
4941 ix86_indirect_branch_register here.
4942 (GOT32_symbol_operand): Likewise.
4944 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4946 * config/i386/predicates.md (constant_call_address_operand):
4947 Rewrite ix86_indirect_branch_register logic.
4948 (sibcall_insn_operand): Likewise.
4950 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4952 * config/i386/constraints.md (Bs): Replace
4953 ix86_indirect_branch_thunk_register with
4954 ix86_indirect_branch_register.
4956 * config/i386/i386.md (indirect_jump): Likewise.
4957 (tablejump): Likewise.
4958 (*sibcall_memory): Likewise.
4959 (*sibcall_value_memory): Likewise.
4960 Peepholes of indirect call and jump via memory: Likewise.
4961 * config/i386/i386.opt: Likewise.
4962 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4963 (GOT_memory_operand): Likewise.
4964 (call_insn_operand): Likewise.
4965 (sibcall_insn_operand): Likewise.
4966 (GOT32_symbol_operand): Likewise.
4968 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4971 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4972 type rather than type addr's type points to.
4973 (expand_omp_atomic_mutex): Likewise.
4974 (expand_omp_atomic): Likewise.
4976 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4979 * config/i386/i386.c (output_indirect_thunk_function): Use
4980 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4981 for __x86_return_thunk.
4983 2018-01-15 Richard Biener <rguenther@suse.de>
4986 * expmed.c (extract_bit_field_1): Fix typo.
4988 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4991 * config/arm/iterators.md (VF): New mode iterator.
4992 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4993 Remove integer-related logic from pattern.
4994 (neon_vabd<mode>_3): Likewise.
4996 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4999 * common.opt (fstrict-overflow): No longer an alias.
5000 (fwrapv-pointer): New option.
5001 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5002 also for pointer types based on flag_wrapv_pointer.
5003 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5004 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5005 opts->x_flag_wrapv got set.
5006 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5007 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5008 POINTER_TYPE_OVERFLOW_UNDEFINED.
5009 * match.pd: Likewise in address comparison pattern.
5010 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5012 2018-01-15 Richard Biener <rguenther@suse.de>
5015 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5016 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5017 Reset type names to their identifier if their TYPE_DECL doesn't
5018 have linkage (and thus is used for ODR and devirt).
5019 (save_debug_info_for_decl): Remove.
5020 (save_debug_info_for_type): Likewise.
5021 (add_tree_to_fld_list): Adjust.
5022 * tree-pretty-print.c (dump_generic_node): Make dumping of
5023 type names more robust.
5025 2018-01-15 Richard Biener <rguenther@suse.de>
5027 * BASE-VER: Bump to 8.0.1.
5029 2018-01-14 Martin Sebor <msebor@redhat.com>
5032 * builtins.c (check_access): Avoid warning when the no-warning bit
5035 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5037 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5038 * ira-color (allocno_hard_regs_compare): Likewise.
5040 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5043 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5044 Use .pushsection/.popsection.
5046 2018-01-14 Martin Sebor <msebor@redhat.com>
5049 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5051 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5053 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5054 entry from extra_headers.
5055 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5056 extra_headers, make the list bitwise identical to the i?86-*-* one.
5058 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5060 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5061 -mcmodel=large with -mindirect-branch=thunk,
5062 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5063 -mfunction-return=thunk-extern.
5064 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5065 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5066 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5068 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5070 * config/i386/i386.c (print_reg): Print the name of the full
5071 integer register without '%'.
5072 (ix86_print_operand): Handle 'V'.
5073 * doc/extend.texi: Document 'V' modifier.
5075 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5077 * config/i386/constraints.md (Bs): Disallow memory operand for
5078 -mindirect-branch-register.
5080 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5081 (GOT_memory_operand): Likewise.
5082 (call_insn_operand): Likewise.
5083 (sibcall_insn_operand): Likewise.
5084 (GOT32_symbol_operand): Likewise.
5085 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5086 for -mindirect-branch-register.
5087 (tablejump): Likewise.
5088 (*sibcall_memory): Likewise.
5089 (*sibcall_value_memory): Likewise.
5090 Disallow peepholes of indirect call and jump via memory for
5091 -mindirect-branch-register.
5092 (*call_pop): Replace m with Bw.
5093 (*call_value_pop): Likewise.
5094 (*sibcall_pop_memory): Replace m with Bs.
5095 * config/i386/i386.opt (mindirect-branch-register): New option.
5096 * doc/invoke.texi: Document -mindirect-branch-register option.
5098 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5100 * config/i386/i386-protos.h (ix86_output_function_return): New.
5101 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5102 set function_return_type.
5103 (indirect_thunk_name): Add ret_p to indicate thunk for function
5105 (output_indirect_thunk_function): Pass false to
5106 indirect_thunk_name.
5107 (ix86_output_indirect_branch_via_reg): Likewise.
5108 (ix86_output_indirect_branch_via_push): Likewise.
5109 (output_indirect_thunk_function): Create alias for function
5110 return thunk if regno < 0.
5111 (ix86_output_function_return): New function.
5112 (ix86_handle_fndecl_attribute): Handle function_return.
5113 (ix86_attribute_table): Add function_return.
5114 * config/i386/i386.h (machine_function): Add
5115 function_return_type.
5116 * config/i386/i386.md (simple_return_internal): Use
5117 ix86_output_function_return.
5118 (simple_return_internal_long): Likewise.
5119 * config/i386/i386.opt (mfunction-return=): New option.
5120 (indirect_branch): Mention -mfunction-return=.
5121 * doc/extend.texi: Document function_return function attribute.
5122 * doc/invoke.texi: Document -mfunction-return= option.
5124 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5126 * config/i386/i386-opts.h (indirect_branch): New.
5127 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5128 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5129 with local indirect jump when converting indirect call and jump.
5130 (ix86_set_indirect_branch_type): New.
5131 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5132 (indirectlabelno): New.
5133 (indirect_thunk_needed): Likewise.
5134 (indirect_thunk_bnd_needed): Likewise.
5135 (indirect_thunks_used): Likewise.
5136 (indirect_thunks_bnd_used): Likewise.
5137 (INDIRECT_LABEL): Likewise.
5138 (indirect_thunk_name): Likewise.
5139 (output_indirect_thunk): Likewise.
5140 (output_indirect_thunk_function): Likewise.
5141 (ix86_output_indirect_branch_via_reg): Likewise.
5142 (ix86_output_indirect_branch_via_push): Likewise.
5143 (ix86_output_indirect_branch): Likewise.
5144 (ix86_output_indirect_jmp): Likewise.
5145 (ix86_code_end): Call output_indirect_thunk_function if needed.
5146 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5148 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5149 (ix86_attribute_table): Add indirect_branch.
5150 * config/i386/i386.h (machine_function): Add indirect_branch_type
5151 and has_local_indirect_jump.
5152 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5154 (tablejump): Likewise.
5155 (*indirect_jump): Use ix86_output_indirect_jmp.
5156 (*tablejump_1): Likewise.
5157 (simple_return_indirect_internal): Likewise.
5158 * config/i386/i386.opt (mindirect-branch=): New option.
5159 (indirect_branch): New.
5162 (thunk-inline): Likewise.
5163 (thunk-extern): Likewise.
5164 * doc/extend.texi: Document indirect_branch function attribute.
5165 * doc/invoke.texi: Document -mindirect-branch= option.
5167 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5170 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5172 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5174 * ipa-inline.c (want_inline_small_function_p): Return false if
5175 inlining has already failed with CIF_FINAL_ERROR.
5176 (update_caller_keys): Call want_inline_small_function_p before
5178 (update_callee_keys): Likewise.
5180 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5182 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5184 (rs6000_quadword_masked_address_p): Likewise.
5185 (quad_aligned_load_p): Likewise.
5186 (quad_aligned_store_p): Likewise.
5187 (const_load_sequence_p): Add comment to describe the outer-most loop.
5188 (mimic_memory_attributes_and_flags): New function.
5189 (rs6000_gen_stvx): Likewise.
5190 (replace_swapped_aligned_store): Likewise.
5191 (rs6000_gen_lvx): Likewise.
5192 (replace_swapped_aligned_load): Likewise.
5193 (replace_swapped_load_constant): Capitalize argument name in
5194 comment describing this function.
5195 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5196 and stores that access quad-word aligned addresses and replace
5197 with stvx or lvx instructions when appropriate.
5198 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5199 New function prototype.
5200 (rs6000_quadword_masked_address_p): Likewise.
5201 (rs6000_gen_lvx): Likewise.
5202 (rs6000_gen_stvx): Likewise.
5203 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5204 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5205 when memory address is aligned.
5206 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5207 this split to select lvx instruction when memory address is aligned.
5208 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5209 instruction when memory address is aligned.
5210 (*vsx_le_perm_load_v16qi): Likewise.
5211 (four unnamed splitters): Modify to select the stvx instruction
5212 when memory is aligned.
5214 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5216 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5217 which appears in the queue multiple times.
5219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5220 Alan Hayward <alan.hayward@arm.com>
5221 David Sherwood <david.sherwood@arm.com>
5223 * tree-vectorizer.h (vec_lower_bound): New structure.
5224 (_loop_vec_info): Add check_nonzero and lower_bounds.
5225 (LOOP_VINFO_CHECK_NONZERO): New macro.
5226 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5227 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5228 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5229 fields. Make seg_len the distance travelled, not including the
5231 (dr_direction_indicator): Declare.
5232 (dr_zero_step_indicator): Likewise.
5233 (dr_known_forward_stride_p): Likewise.
5234 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5236 (runtime_alias_check_p): Allow runtime alias checks with
5238 (operator ==): Compare access_size and align.
5239 (prune_runtime_alias_test_list): Rework for new distinction between
5240 the access_size and seg_len.
5241 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5243 (get_segment_min_max): New function.
5244 (create_intersect_range_checks): Use it.
5245 (dr_step_indicator): New function.
5246 (dr_direction_indicator): Likewise.
5247 (dr_zero_step_indicator): Likewise.
5248 (dr_known_forward_stride_p): Likewise.
5249 * tree-loop-distribution.c (data_ref_segment_size): Return
5250 DR_STEP * (niters - 1).
5251 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5253 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5254 (vect_preserves_scalar_order_p): New function, split out from...
5255 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5256 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5257 (vect_vfa_access_size): New function.
5258 (vect_vfa_align): Likewise.
5259 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5260 (dump_lower_bound): New function.
5261 (vect_check_lower_bound): Likewise.
5262 (vect_small_gap_p): Likewise.
5263 (vectorizable_with_step_bound_p): Likewise.
5264 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5265 depencies if the vectorization factor is 1. Convert the checks
5266 for nonzero steps into checks on the bounds of DR_STEP. Try using
5267 a bunds check for variable steps if the minimum required step is
5268 relatively small. Update calls to the dr_with_seg_len
5269 constructor and to vect_compile_time_alias.
5270 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5272 (vect_loop_versioning): Call it.
5273 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5275 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5278 Alan Hayward <alan.hayward@arm.com>
5279 David Sherwood <david.sherwood@arm.com>
5281 * doc/sourcebuild.texi (vect_scatter_store): Document.
5282 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5284 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5286 * genopinit.c (main): Add supports_vec_scatter_store and
5287 supports_vec_scatter_store_cached to target_optabs.
5288 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5289 IFN_MASK_SCATTER_STORE.
5290 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5292 * internal-fn.h (internal_store_fn_p): Declare.
5293 (internal_fn_stored_value_index): Likewise.
5294 * internal-fn.c (scatter_store_direct): New macro.
5295 (expand_scatter_store_optab_fn): New function.
5296 (direct_scatter_store_optab_supported_p): New macro.
5297 (internal_store_fn_p): New function.
5298 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5299 IFN_MASK_SCATTER_STORE.
5300 (internal_fn_mask_index): Likewise.
5301 (internal_fn_stored_value_index): New function.
5302 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5304 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5305 * optabs-query.c (supports_vec_scatter_store_p): New function.
5306 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5307 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5308 true for scatter stores.
5309 (vect_gather_scatter_fn_p): Handle scatter stores too.
5310 (vect_check_gather_scatter): Consider using scatter stores if
5311 supports_vec_scatter_store_p.
5312 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5314 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5315 internal_fn_stored_value_index.
5316 (check_load_store_masking): Handle scatter stores too.
5317 (vect_get_store_rhs): Make public.
5318 (vectorizable_call): Use internal_store_fn_p.
5319 (vectorizable_store): Handle scatter store internal functions.
5320 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5321 when deciding whether the end of the group has been reached.
5322 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5323 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5324 (mask_scatter_store<mode>): New insns.
5326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5327 Alan Hayward <alan.hayward@arm.com>
5328 David Sherwood <david.sherwood@arm.com>
5330 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5331 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5332 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5334 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5335 Use vect_truncate_gather_scatter_offset if we can't treat the
5336 operation as a normal gather load or scatter store.
5337 (get_group_load_store_type): Take the gather_scatter_info
5338 as argument. Try using a gather load or scatter store for
5339 single-element groups.
5340 (get_load_store_type): Update calls to get_group_load_store_type
5341 and vect_use_strided_gather_scatters_p.
5343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5344 Alan Hayward <alan.hayward@arm.com>
5345 David Sherwood <david.sherwood@arm.com>
5347 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5348 optional tree argument.
5349 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5351 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5352 but continue to use the current value as a fallback.
5353 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5354 to compare the updates.
5355 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5356 (get_load_store_type): Use it when handling a strided access.
5357 (vect_get_strided_load_store_ops): New function.
5358 (vect_get_data_ptr_increment): Likewise.
5359 (vectorizable_load): Handle strided gather loads. Always pass
5360 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5363 Alan Hayward <alan.hayward@arm.com>
5364 David Sherwood <david.sherwood@arm.com>
5366 * doc/md.texi (gather_load@var{m}): Document.
5367 (mask_gather_load@var{m}): Likewise.
5368 * genopinit.c (main): Add supports_vec_gather_load and
5369 supports_vec_gather_load_cached to target_optabs.
5370 * optabs-tree.c (init_tree_optimization_optabs): Use
5371 ggc_cleared_alloc to allocate target_optabs.
5372 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5373 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5375 * internal-fn.h (internal_load_fn_p): Declare.
5376 (internal_gather_scatter_fn_p): Likewise.
5377 (internal_fn_mask_index): Likewise.
5378 (internal_gather_scatter_fn_supported_p): Likewise.
5379 * internal-fn.c (gather_load_direct): New macro.
5380 (expand_gather_load_optab_fn): New function.
5381 (direct_gather_load_optab_supported_p): New macro.
5382 (direct_internal_fn_optab): New function.
5383 (internal_load_fn_p): Likewise.
5384 (internal_gather_scatter_fn_p): Likewise.
5385 (internal_fn_mask_index): Likewise.
5386 (internal_gather_scatter_fn_supported_p): Likewise.
5387 * optabs-query.c (supports_at_least_one_mode_p): New function.
5388 (supports_vec_gather_load_p): Likewise.
5389 * optabs-query.h (supports_vec_gather_load_p): Declare.
5390 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5391 and memory_type field.
5392 (NUM_PATTERNS): Bump to 15.
5393 * tree-vect-data-refs.c: Include internal-fn.h.
5394 (vect_gather_scatter_fn_p): New function.
5395 (vect_describe_gather_scatter_call): Likewise.
5396 (vect_check_gather_scatter): Try using internal functions for
5397 gather loads. Recognize existing calls to a gather load function.
5398 (vect_analyze_data_refs): Consider using gather loads if
5399 supports_vec_gather_load_p.
5400 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5401 (vect_get_gather_scatter_offset_type): Likewise.
5402 (vect_convert_mask_for_vectype): Likewise.
5403 (vect_add_conversion_to_patterm): Likewise.
5404 (vect_try_gather_scatter_pattern): Likewise.
5405 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5406 (vect_vect_recog_func_ptrs): Add it.
5407 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5408 internal_fn_mask_index and internal_gather_scatter_fn_p.
5409 (check_load_store_masking): Take the gather_scatter_info as an
5410 argument and handle gather loads.
5411 (vect_get_gather_scatter_ops): New function.
5412 (vectorizable_call): Check internal_load_fn_p.
5413 (vectorizable_load): Likewise. Handle gather load internal
5415 (vectorizable_store): Update call to check_load_store_masking.
5416 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5417 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5418 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5419 (aarch64_gather_scale_operand_d): New predicates.
5420 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5421 (mask_gather_load<mode>): New insns.
5423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5424 Alan Hayward <alan.hayward@arm.com>
5425 David Sherwood <david.sherwood@arm.com>
5427 * optabs.def (fold_left_plus_optab): New optab.
5428 * doc/md.texi (fold_left_plus_@var{m}): Document.
5429 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5430 * internal-fn.c (fold_left_direct): Define.
5431 (expand_fold_left_optab_fn): Likewise.
5432 (direct_fold_left_optab_supported_p): Likewise.
5433 * fold-const-call.c (fold_const_fold_left): New function.
5434 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5435 * tree-parloops.c (valid_reduction_p): New function.
5436 (gather_scalar_reductions): Use it.
5437 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5438 (vect_finish_replace_stmt): Declare.
5439 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5440 (needs_fold_left_reduction_p): New function, split out from...
5441 (vect_is_simple_reduction): ...here. Accept reductions that
5442 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5443 (vect_force_simple_reduction): Also store the reduction type in
5444 the assignment's STMT_VINFO_REDUC_TYPE.
5445 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5446 (merge_with_identity): New function.
5447 (vect_expand_fold_left): Likewise.
5448 (vectorize_fold_left_reduction): Likewise.
5449 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5450 scalar phi in place for it. Check for target support and reject
5451 cases that would reassociate the operation. Defer the transform
5452 phase to vectorize_fold_left_reduction.
5453 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5454 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5455 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5459 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5460 call to ifc_temp_var.
5462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5463 Alan Hayward <alan.hayward@arm.com>
5464 David Sherwood <david.sherwood@arm.com>
5466 * target.def (legitimize_address_displacement): Take the original
5467 offset as a poly_int.
5468 * targhooks.h (default_legitimize_address_displacement): Update
5470 * targhooks.c (default_legitimize_address_displacement): Likewise.
5471 * doc/tm.texi: Regenerate.
5472 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5473 as an argument, moving assert of ad->disp == ad->disp_term to...
5474 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5475 Try calling targetm.legitimize_address_displacement before expanding
5476 the address rather than afterwards, and adjust for the new interface.
5477 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5478 Match the new hook interface. Handle SVE addresses.
5479 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5484 * Makefile.in (OBJS): Add early-remat.o.
5485 * target.def (select_early_remat_modes): New hook.
5486 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5487 * doc/tm.texi: Regenerate.
5488 * targhooks.h (default_select_early_remat_modes): Declare.
5489 * targhooks.c (default_select_early_remat_modes): New function.
5490 * timevar.def (TV_EARLY_REMAT): New timevar.
5491 * passes.def (pass_early_remat): New pass.
5492 * tree-pass.h (make_pass_early_remat): Declare.
5493 * early-remat.c: New file.
5494 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5496 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5499 Alan Hayward <alan.hayward@arm.com>
5500 David Sherwood <david.sherwood@arm.com>
5502 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5503 vfm1 with a bound_epilog parameter.
5504 (vect_do_peeling): Update calls accordingly, and move the prologue
5505 call earlier in the function. Treat the base bound_epilog as 0 for
5506 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5507 this base when peeling for gaps.
5508 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5509 with fully-masked loops.
5510 (vect_estimate_min_profitable_iters): Handle the single peeled
5511 iteration in that case.
5513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5514 Alan Hayward <alan.hayward@arm.com>
5515 David Sherwood <david.sherwood@arm.com>
5517 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5518 single-element interleaving even if the size is not a power of 2.
5519 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5520 accesses for single-element interleaving if the group size is
5523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5524 Alan Hayward <alan.hayward@arm.com>
5525 David Sherwood <david.sherwood@arm.com>
5527 * doc/md.texi (fold_extract_last_@var{m}): Document.
5528 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5529 * optabs.def (fold_extract_last_optab): New optab.
5530 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5531 * internal-fn.c (fold_extract_direct): New macro.
5532 (expand_fold_extract_optab_fn): Likewise.
5533 (direct_fold_extract_optab_supported_p): Likewise.
5534 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5535 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5536 EXTRACT_LAST_REDUCTION.
5537 (get_initial_def_for_reduction): Do not create an initial vector
5538 for EXTRACT_LAST_REDUCTION reductions.
5539 (vectorizable_reduction): Leave the scalar phi in place for
5540 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5541 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5542 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5543 transform phase to vectorizable_condition.
5544 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5546 (vect_finish_stmt_generation): ...here.
5547 (vect_finish_replace_stmt): New function.
5548 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5549 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5551 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5554 Alan Hayward <alan.hayward@arm.com>
5555 David Sherwood <david.sherwood@arm.com>
5557 * doc/md.texi (extract_last_@var{m}): Document.
5558 * optabs.def (extract_last_optab): New optab.
5559 * internal-fn.def (EXTRACT_LAST): New internal function.
5560 * internal-fn.c (cond_unary_direct): New macro.
5561 (expand_cond_unary_optab_fn): Likewise.
5562 (direct_cond_unary_optab_supported_p): Likewise.
5563 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5564 loops using EXTRACT_LAST.
5565 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5566 (extract_last_<mode>): ...this optab.
5567 (vec_extract<mode><Vel>): Update accordingly.
5569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5570 Alan Hayward <alan.hayward@arm.com>
5571 David Sherwood <david.sherwood@arm.com>
5573 * target.def (empty_mask_is_expensive): New hook.
5574 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5575 * doc/tm.texi: Regenerate.
5576 * targhooks.h (default_empty_mask_is_expensive): Declare.
5577 * targhooks.c (default_empty_mask_is_expensive): New function.
5578 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5579 if the target says that empty masks are expensive.
5580 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5582 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5589 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5590 (vect_use_loop_mask_for_alignment_p): New function.
5591 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5592 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5593 niters_skip argument. Make sure that the first niters_skip elements
5594 of the first iteration are inactive.
5595 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5596 Update call to vect_set_loop_masks_directly.
5597 (get_misalign_in_elems): New function, split out from...
5598 (vect_gen_prolog_loop_niters): ...here.
5599 (vect_update_init_of_dr): Take a code argument that specifies whether
5600 the adjustment should be added or subtracted.
5601 (vect_update_init_of_drs): Likewise.
5602 (vect_prepare_for_masked_peels): New function.
5603 (vect_do_peeling): Skip prologue peeling if we're using a mask
5604 instead. Update call to vect_update_inits_of_drs.
5605 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5607 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5608 alignment. Do not include the number of peeled iterations in
5609 the minimum threshold in that case.
5610 (vectorizable_induction): Adjust the start value down by
5611 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5612 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5613 Take the number of skipped iterations into account when calculating
5615 * tree-vect-stmts.c (vect_gen_while_not): New function.
5617 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5618 Alan Hayward <alan.hayward@arm.com>
5619 David Sherwood <david.sherwood@arm.com>
5621 * doc/sourcebuild.texi (vect_fully_masked): Document.
5622 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5624 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5626 (vect_analyze_loop_2): ...here. Don't check the vectorization
5627 factor against the number of loop iterations if the loop is
5630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5631 Alan Hayward <alan.hayward@arm.com>
5632 David Sherwood <david.sherwood@arm.com>
5634 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5635 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5636 (dump_groups): Update accordingly.
5637 (iv_use::mem_type): New member variable.
5638 (address_p): New function.
5639 (record_use): Add a mem_type argument and initialize the new
5641 (record_group_use): Add a mem_type argument. Use address_p.
5642 Remove obsolete null checks of base_object. Update call to record_use.
5643 (find_interesting_uses_op): Update call to record_group_use.
5644 (find_interesting_uses_cond): Likewise.
5645 (find_interesting_uses_address): Likewise.
5646 (get_mem_type_for_internal_fn): New function.
5647 (find_address_like_use): Likewise.
5648 (find_interesting_uses_stmt): Try find_address_like_use before
5649 calling find_interesting_uses_op.
5650 (addr_offset_valid_p): Use the iv mem_type field as the type
5651 of the addressed memory.
5652 (add_autoinc_candidates): Likewise.
5653 (get_address_cost): Likewise.
5654 (split_small_address_groups_p): Use address_p.
5655 (split_address_groups): Likewise.
5656 (add_iv_candidate_for_use): Likewise.
5657 (autoinc_possible_for_pair): Likewise.
5658 (rewrite_groups): Likewise.
5659 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5660 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5661 (get_alias_ptr_type_for_ptr_address): New function.
5662 (rewrite_use_address): Rewrite address uses in calls that were
5663 identified by find_address_like_use.
5665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5666 Alan Hayward <alan.hayward@arm.com>
5667 David Sherwood <david.sherwood@arm.com>
5669 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5671 * gimple-expr.h (is_gimple_addressable: Likewise.
5672 * gimple-expr.c (is_gimple_address): Likewise.
5673 * internal-fn.c (expand_call_mem_ref): New function.
5674 (expand_mask_load_optab_fn): Use it.
5675 (expand_mask_store_optab_fn): Likewise.
5677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5678 Alan Hayward <alan.hayward@arm.com>
5679 David Sherwood <david.sherwood@arm.com>
5681 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5682 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5683 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5684 (cond_umax@var{mode}): Document.
5685 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5686 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5687 (cond_umin_optab, cond_umax_optab): New optabs.
5688 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5689 (COND_IOR, COND_XOR): New internal functions.
5690 * internal-fn.h (get_conditional_internal_fn): Declare.
5691 * internal-fn.c (cond_binary_direct): New macro.
5692 (expand_cond_binary_optab_fn): Likewise.
5693 (direct_cond_binary_optab_supported_p): Likewise.
5694 (get_conditional_internal_fn): New function.
5695 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5696 Cope with reduction statements that are vectorized as calls rather
5698 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5699 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5700 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5701 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5702 (UNSPEC_COND_EOR): New unspecs.
5703 (optab): Add mappings for them.
5704 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5705 (sve_int_op, sve_fp_op): New int attributes.
5707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5708 Alan Hayward <alan.hayward@arm.com>
5709 David Sherwood <david.sherwood@arm.com>
5711 * optabs.def (while_ult_optab): New optab.
5712 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5713 * internal-fn.def (WHILE_ULT): New internal function.
5714 * internal-fn.h (direct_internal_fn_supported_p): New override
5715 that takes two types as argument.
5716 * internal-fn.c (while_direct): New macro.
5717 (expand_while_optab_fn): New function.
5718 (convert_optab_supported_p): Likewise.
5719 (direct_while_optab_supported_p): New macro.
5720 * wide-int.h (wi::udiv_ceil): New function.
5721 * tree-vectorizer.h (rgroup_masks): New structure.
5722 (vec_loop_masks): New typedef.
5723 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5725 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5726 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5727 (vect_max_vf): New function.
5728 (slpeel_make_loop_iterate_ntimes): Delete.
5729 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5730 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5731 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5732 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5733 internal-fn.h, stor-layout.h and optabs-query.h.
5734 (vect_set_loop_mask): New function.
5735 (add_preheader_seq): Likewise.
5736 (add_header_seq): Likewise.
5737 (interleave_supported_p): Likewise.
5738 (vect_maybe_permute_loop_masks): Likewise.
5739 (vect_set_loop_masks_directly): Likewise.
5740 (vect_set_loop_condition_masked): Likewise.
5741 (vect_set_loop_condition_unmasked): New function, split out from
5742 slpeel_make_loop_iterate_ntimes.
5743 (slpeel_make_loop_iterate_ntimes): Rename to..
5744 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5745 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5746 (vect_do_peeling): Update call accordingly.
5747 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5749 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5750 mask_compare_type, can_fully_mask_p and fully_masked_p.
5751 (release_vec_loop_masks): New function.
5752 (_loop_vec_info): Use it to free the loop masks.
5753 (can_produce_all_loop_masks_p): New function.
5754 (vect_get_max_nscalars_per_iter): Likewise.
5755 (vect_verify_full_masking): Likewise.
5756 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5757 retries, and free the mask rgroups before retrying. Check loop-wide
5758 reasons for disallowing fully-masked loops. Make the final decision
5759 about whether use a fully-masked loop or not.
5760 (vect_estimate_min_profitable_iters): Do not assume that peeling
5761 for the number of iterations will be needed for fully-masked loops.
5762 (vectorizable_reduction): Disable fully-masked loops.
5763 (vectorizable_live_operation): Likewise.
5764 (vect_halve_mask_nunits): New function.
5765 (vect_double_mask_nunits): Likewise.
5766 (vect_record_loop_mask): Likewise.
5767 (vect_get_loop_mask): Likewise.
5768 (vect_transform_loop): Handle the case in which the final loop
5769 iteration might handle a partial vector. Call vect_set_loop_condition
5770 instead of slpeel_make_loop_iterate_ntimes.
5771 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5772 (check_load_store_masking): New function.
5773 (prepare_load_store_mask): Likewise.
5774 (vectorizable_store): Handle fully-masked loops.
5775 (vectorizable_load): Likewise.
5776 (supportable_widening_operation): Use vect_halve_mask_nunits for
5778 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5779 (vect_gen_while): New function.
5780 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5781 (aarch64_uqdec<mode>): New insn.
5783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5784 Alan Hayward <alan.hayward@arm.com>
5785 David Sherwood <david.sherwood@arm.com>
5787 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5788 (reduc_xor_scal_optab): New optabs.
5789 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5790 (reduc_xor_scal_@var{m}): Document.
5791 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5792 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5794 * fold-const-call.c (fold_const_call): Handle them.
5795 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5796 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5797 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5798 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5799 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5800 (UNSPEC_XORV): New unspecs.
5801 (optab): Add entries for them.
5802 (BITWISEV): New int iterator.
5803 (bit_reduc_op): New int attributes.
5805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5806 Alan Hayward <alan.hayward@arm.com>
5807 David Sherwood <david.sherwood@arm.com>
5809 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5810 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5811 * optabs.def (vec_shl_insert_optab): New optab.
5812 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5813 (duplicate_and_interleave): Likewise.
5814 * tree-vect-loop.c: Include internal-fn.h.
5815 (neutral_op_for_slp_reduction): New function, split out from
5816 get_initial_defs_for_reduction.
5817 (get_initial_def_for_reduction): Handle option 2 for variable-length
5818 vectors by loading the neutral value into a vector and then shifting
5819 the initial value into element 0.
5820 (get_initial_defs_for_reduction): Replace the code argument with
5821 the neutral value calculated by neutral_op_for_slp_reduction.
5822 Use gimple_build_vector for constant-length vectors.
5823 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5824 but the first group_size elements have a neutral value.
5825 Use duplicate_and_interleave otherwise.
5826 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5827 Update call to get_initial_defs_for_reduction. Handle SLP
5828 reductions for variable-length vectors by creating one vector
5829 result for each scalar result, with the elements associated
5830 with other scalar results stubbed out with the neutral value.
5831 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5832 Require IFN_VEC_SHL_INSERT for double reductions on
5833 variable-length vectors, or SLP reductions that have
5834 a neutral value. Require can_duplicate_and_interleave_p
5835 support for variable-length unchained SLP reductions if there
5836 is no neutral value, such as for MIN/MAX reductions. Also require
5837 the number of vector elements to be a multiple of the number of
5838 SLP statements when doing variable-length unchained SLP reductions.
5839 Update call to vect_create_epilog_for_reduction.
5840 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5841 and remove initial values.
5842 (duplicate_and_interleave): Make public.
5843 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5844 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5847 Alan Hayward <alan.hayward@arm.com>
5848 David Sherwood <david.sherwood@arm.com>
5850 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5851 (can_duplicate_and_interleave_p): New function.
5852 (vect_get_and_check_slp_defs): Take the vector of statements
5853 rather than just the current one. Remove excess parentheses.
5854 Restriction rejectinon of vect_constant_def and vect_external_def
5855 for variable-length vectors to boolean types, or types for which
5856 can_duplicate_and_interleave_p is false.
5857 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5858 (duplicate_and_interleave): New function.
5859 (vect_get_constant_vectors): Use gimple_build_vector for
5860 constant-length vectors and suitable variable-length constant
5861 vectors. Use duplicate_and_interleave for other variable-length
5862 vectors. Don't defer the update when inserting new statements.
5864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5865 Alan Hayward <alan.hayward@arm.com>
5866 David Sherwood <david.sherwood@arm.com>
5868 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5869 min_profitable_iters doesn't go negative.
5871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5872 Alan Hayward <alan.hayward@arm.com>
5873 David Sherwood <david.sherwood@arm.com>
5875 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5876 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5877 * optabs.def (vec_mask_load_lanes_optab): New optab.
5878 (vec_mask_store_lanes_optab): Likewise.
5879 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5880 (MASK_STORE_LANES): Likewise.
5881 * internal-fn.c (mask_load_lanes_direct): New macro.
5882 (mask_store_lanes_direct): Likewise.
5883 (expand_mask_load_optab_fn): Handle masked operations.
5884 (expand_mask_load_lanes_optab_fn): New macro.
5885 (expand_mask_store_optab_fn): Handle masked operations.
5886 (expand_mask_store_lanes_optab_fn): New macro.
5887 (direct_mask_load_lanes_optab_supported_p): Likewise.
5888 (direct_mask_store_lanes_optab_supported_p): Likewise.
5889 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5891 (vect_load_lanes_supported): Likewise.
5892 * tree-vect-data-refs.c (strip_conversion): New function.
5893 (can_group_stmts_p): Likewise.
5894 (vect_analyze_data_ref_accesses): Use it instead of checking
5895 for a pair of assignments.
5896 (vect_store_lanes_supported): Take a masked_p parameter.
5897 (vect_load_lanes_supported): Likewise.
5898 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5899 vect_store_lanes_supported and vect_load_lanes_supported.
5900 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5901 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5902 parameter. Don't allow gaps for masked accesses.
5903 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5904 and vect_load_lanes_supported.
5905 (get_load_store_type): Take a masked_p parameter and update
5906 call to get_group_load_store_type.
5907 (vectorizable_store): Update call to get_load_store_type.
5908 Handle IFN_MASK_STORE_LANES.
5909 (vectorizable_load): Update call to get_load_store_type.
5910 Handle IFN_MASK_LOAD_LANES.
5912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5913 Alan Hayward <alan.hayward@arm.com>
5914 David Sherwood <david.sherwood@arm.com>
5916 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5918 * config/aarch64/aarch64-protos.h
5919 (aarch64_sve_struct_memory_operand_p): Declare.
5920 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5921 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5922 (VPRED, vpred): Handle SVE structure modes.
5923 * config/aarch64/constraints.md (Utx): New constraint.
5924 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5925 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5926 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5927 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5928 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5929 structure modes. Split into pieces after RA.
5930 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5931 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5933 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5934 SVE structure modes.
5935 (aarch64_classify_address): Likewise.
5936 (sizetochar): Move earlier in file.
5937 (aarch64_print_operand): Handle SVE register lists.
5938 (aarch64_array_mode): New function.
5939 (aarch64_sve_struct_memory_operand_p): Likewise.
5940 (TARGET_ARRAY_MODE): Redefine.
5942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5943 Alan Hayward <alan.hayward@arm.com>
5944 David Sherwood <david.sherwood@arm.com>
5946 * target.def (array_mode): New target hook.
5947 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5948 * doc/tm.texi: Regenerate.
5949 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5950 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5951 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5953 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5957 Alan Hayward <alan.hayward@arm.com>
5958 David Sherwood <david.sherwood@arm.com>
5960 * fold-const.c (fold_binary_loc): Check the argument types
5961 rather than the result type when testing for a vector operation.
5963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5965 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5966 * doc/tm.texi: Regenerate.
5968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5969 Alan Hayward <alan.hayward@arm.com>
5970 David Sherwood <david.sherwood@arm.com>
5972 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5973 (sve): Document new AArch64 extension.
5974 * doc/md.texi (w): Extend the description of the AArch64
5975 constraint to include SVE vectors.
5976 (Upl, Upa): Document new AArch64 predicate constraints.
5977 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5979 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5980 (msve-vector-bits=): New option.
5981 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5982 SVE when these are disabled.
5983 (sve): New extension.
5984 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5985 modes. Adjust their number of units based on aarch64_sve_vg.
5986 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5987 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5988 aarch64_addr_query_type.
5989 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5990 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5991 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5992 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5993 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5994 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5995 (aarch64_simd_imm_zero_p): Delete.
5996 (aarch64_check_zero_based_sve_index_immediate): Declare.
5997 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5998 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5999 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6000 (aarch64_sve_float_mul_immediate_p): Likewise.
6001 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6003 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6004 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6005 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6006 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6007 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6008 (aarch64_regmode_natural_size): Likewise.
6009 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6010 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6012 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6013 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6014 for VG and the SVE predicate registers.
6015 (V_ALIASES): Add a "z"-prefixed alias.
6016 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6017 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6018 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6019 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6020 (REG_CLASS_NAMES): Add entries for them.
6021 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6022 and the predicate registers.
6023 (aarch64_sve_vg): Declare.
6024 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6025 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6026 (REGMODE_NATURAL_SIZE): Define.
6027 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6029 * config/aarch64/aarch64.c: Include cfgrtl.h.
6030 (simd_immediate_info): Add a constructor for series vectors,
6031 and an associated step field.
6032 (aarch64_sve_vg): New variable.
6033 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6034 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6035 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6036 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6037 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6038 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6039 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6040 (aarch64_get_mask_mode): New functions.
6041 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6042 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6043 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6044 predicate modes and predicate registers. Explicitly restrict
6045 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6046 to store a vector mode if it is recognized by
6047 aarch64_classify_vector_mode.
6048 (aarch64_regmode_natural_size): New function.
6049 (aarch64_hard_regno_caller_save_mode): Return the original mode
6051 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6052 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6053 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6054 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6056 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6057 does not overlap dest if the function is frame-related. Handle
6059 (aarch64_split_add_offset): New function.
6060 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6061 them aarch64_add_offset.
6062 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6063 and update call to aarch64_sub_sp.
6064 (aarch64_add_cfa_expression): New function.
6065 (aarch64_expand_prologue): Pass extra temporary registers to the
6066 functions above. Handle the case in which we need to emit new
6067 DW_CFA_expressions for registers that were originally saved
6068 relative to the stack pointer, but now have to be expressed
6069 relative to the frame pointer.
6070 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6072 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6073 IP0 and IP1 values for SVE frames.
6074 (aarch64_expand_vec_series): New function.
6075 (aarch64_expand_sve_widened_duplicate): Likewise.
6076 (aarch64_expand_sve_const_vector): Likewise.
6077 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6078 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6079 into the register, rather than emitting a SET directly.
6080 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6081 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6082 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6083 (offset_9bit_signed_scaled_p): New functions.
6084 (aarch64_replicate_bitmask_imm): New function.
6085 (aarch64_bitmask_imm): Use it.
6086 (aarch64_cannot_force_const_mem): Reject expressions involving
6087 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6088 (aarch64_classify_index): Handle SVE indices, by requiring
6089 a plain register index with a scale that matches the element size.
6090 (aarch64_classify_address): Handle SVE addresses. Assert that
6091 the mode of the address is VOIDmode or an integer mode.
6092 Update call to aarch64_classify_symbol.
6093 (aarch64_classify_symbolic_expression): Update call to
6094 aarch64_classify_symbol.
6095 (aarch64_const_vec_all_in_range_p): New function.
6096 (aarch64_print_vector_float_operand): Likewise.
6097 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6098 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6099 and the FP immediates 1.0 and 0.5.
6100 (aarch64_print_address_internal): Handle SVE addresses.
6101 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6102 (aarch64_regno_regclass): Handle predicate registers.
6103 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6105 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6106 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6107 (aarch64_convert_sve_vector_bits): New function.
6108 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6109 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6111 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6112 Handle SVE vector and predicate modes. Accept VL-based constants
6113 that need only one temporary register, and VL offsets that require
6114 no temporary registers.
6115 (aarch64_conditional_register_usage): Mark the predicate registers
6116 as fixed if SVE isn't available.
6117 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6118 Return true for SVE vector and predicate modes.
6119 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6120 rather than an unsigned int. Handle SVE modes.
6121 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6123 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6125 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6126 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6127 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6128 (aarch64_sve_float_mul_immediate_p): New functions.
6129 (aarch64_sve_valid_immediate): New function.
6130 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6131 Explicitly reject structure modes. Check for INDEX constants.
6132 Handle PTRUE and PFALSE constants.
6133 (aarch64_check_zero_based_sve_index_immediate): New function.
6134 (aarch64_simd_imm_zero_p): Delete.
6135 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6136 vector modes. Accept constants in the range of CNT[BHWD].
6137 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6138 ask for an Advanced SIMD mode.
6139 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6140 (aarch64_simd_vector_alignment): Handle SVE predicates.
6141 (aarch64_vectorize_preferred_vector_alignment): New function.
6142 (aarch64_simd_vector_alignment_reachable): Use it instead of
6144 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6145 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6147 (MAX_VECT_LEN): Delete.
6148 (expand_vec_perm_d): Add a vec_flags field.
6149 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6150 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6151 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6153 (aarch64_evpc_rev): Rename to...
6154 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6155 (aarch64_evpc_rev_global): New function.
6156 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6157 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6159 (aarch64_evpc_sve_tbl): New function.
6160 (aarch64_expand_vec_perm_const_1): Update after rename of
6161 aarch64_evpc_rev. Handle SVE permutes too, trying
6162 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6163 than aarch64_evpc_tbl.
6164 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6165 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6166 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6167 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6168 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6169 (aarch64_expand_sve_vcond): New functions.
6170 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6171 of aarch64_vector_mode_p.
6172 (aarch64_dwarf_poly_indeterminate_value): New function.
6173 (aarch64_compute_pressure_classes): Likewise.
6174 (aarch64_can_change_mode_class): Likewise.
6175 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6176 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6177 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6178 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6179 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6180 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6181 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6182 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6184 (Dn, Dl, Dr): Accept const as well as const_vector.
6185 (Dz): Likewise. Compare against CONST0_RTX.
6186 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6187 of "vector" where appropriate.
6188 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6189 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6190 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6191 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6192 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6193 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6194 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6195 (v_int_equiv): Extend to SVE modes.
6196 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6198 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6199 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6200 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6201 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6202 (SVE_COND_FP_CMP): New int iterators.
6203 (perm_hilo): Handle the new unpack unspecs.
6204 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6206 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6207 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6208 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6209 (aarch64_equality_operator, aarch64_constant_vector_operand)
6210 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6211 (aarch64_sve_nonimmediate_operand): Likewise.
6212 (aarch64_sve_general_operand): Likewise.
6213 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6214 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6215 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6216 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6217 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6218 (aarch64_sve_float_arith_immediate): Likewise.
6219 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6220 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6221 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6222 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6223 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6224 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6225 (aarch64_sve_float_arith_operand): Likewise.
6226 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6227 (aarch64_sve_float_mul_operand): Likewise.
6228 (aarch64_sve_vec_perm_operand): Likewise.
6229 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6230 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6231 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6232 as well as const_vector.
6233 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6234 in file. Use CONST0_RTX and CONSTM1_RTX.
6235 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6236 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6237 Use aarch64_simd_imm_zero.
6238 * config/aarch64/aarch64-sve.md: New file.
6239 * config/aarch64/aarch64.md: Include it.
6240 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6241 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6242 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6243 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6244 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6245 (sve): New attribute.
6246 (enabled): Disable instructions with the sve attribute unless
6248 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6249 aarch64_expand_mov_immediate.
6250 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6251 CNT[BHSD] immediates.
6252 (movti): Split CONST_POLY_INT moves into two halves.
6253 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6254 Split additions that need a temporary here if the destination
6255 is the stack pointer.
6256 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6257 (*add<mode>3_poly_1): New instruction.
6258 (set_clobber_cc): New expander.
6260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6262 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6263 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6264 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6265 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6266 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6267 Change innermode from fixed_mode_size to machine_mode.
6268 (simplify_subreg): Update call accordingly. Handle a constant-sized
6269 subreg of a variable-length CONST_VECTOR.
6271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6272 Alan Hayward <alan.hayward@arm.com>
6273 David Sherwood <david.sherwood@arm.com>
6275 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6276 (add_offset_to_base): New function, split out from...
6277 (create_mem_ref): ...here. When handling a scale other than 1,
6278 check first whether the address is valid without the offset.
6279 Add it into the base if so, leaving the index and scale as-is.
6281 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6284 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6285 fold_for_warn before checking if arg2 is INTEGER_CST.
6287 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6289 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6290 (store_multiple_operation): Delete.
6291 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6292 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6293 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6294 guarded by TARGET_STRING.
6295 (rs6000_output_load_multiple): Delete.
6296 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6297 OPTION_MASK_STRING / TARGET_STRING handling.
6298 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6299 (const rs6000_opt_masks) <"string">: Change mask to 0.
6300 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6301 (MASK_STRING): Delete.
6302 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6304 (load_multiple): Delete.
6311 (store_multiple): Delete.
6318 (movmemsi_8reg): Delete.
6319 (corresponding unnamed define_insn): Delete.
6320 (movmemsi_6reg): Delete.
6321 (corresponding unnamed define_insn): Delete.
6322 (movmemsi_4reg): Delete.
6323 (corresponding unnamed define_insn): Delete.
6324 (movmemsi_2reg): Delete.
6325 (corresponding unnamed define_insn): Delete.
6326 (movmemsi_1reg): Delete.
6327 (corresponding unnamed define_insn): Delete.
6328 * config/rs6000/rs6000.opt (mno-string): New.
6329 (mstring): Replace by deprecation warning stub.
6330 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6332 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6334 * regrename.c (regrename_do_replace): If replacing the same
6335 reg multiple times, try to reuse last created gen_raw_REG.
6338 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6339 main to workaround a bug in GDB.
6341 2018-01-12 Tom de Vries <tom@codesourcery.com>
6344 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6346 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6348 PR rtl-optimization/80481
6349 * ira-color.c (get_cap_member): New function.
6350 (allocnos_conflict_by_live_ranges_p): Use it.
6351 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6352 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6354 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6357 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6358 (*saddl_se_1): Ditto.
6360 (*ssubl_se_1): Ditto.
6362 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6364 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6365 rather than wi::to_widest for DR_INITs.
6366 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6367 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6368 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6370 (vect_analyze_group_access_1): Note that here.
6372 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6374 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6375 polynomial type sizes.
6377 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6379 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6380 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6381 (gimple_add_tmp_var): Likewise.
6383 2018-01-12 Martin Liska <mliska@suse.cz>
6385 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6386 (gimple_alloc_sizes): Likewise.
6387 (dump_gimple_statistics): Use PRIu64 in printf format.
6388 * gimple.h: Change uint64_t to int.
6390 2018-01-12 Martin Liska <mliska@suse.cz>
6392 * tree-core.h: Use uint64_t instead of int.
6393 * tree.c (tree_node_counts): Likewise.
6394 (tree_node_sizes): Likewise.
6395 (dump_tree_statistics): Use PRIu64 in printf format.
6397 2018-01-12 Martin Liska <mliska@suse.cz>
6399 * Makefile.in: As qsort_chk is implemented in vec.c, add
6400 vec.o to linkage of gencfn-macros.
6401 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6402 passing the info to record_node_allocation_statistics.
6403 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6405 * ggc-common.c (struct ggc_usage): Add operator== and use
6406 it in operator< and compare function.
6407 * mem-stats.h (struct mem_usage): Likewise.
6408 * vec.c (struct vec_usage): Remove operator< and compare
6409 function. Can be simply inherited.
6411 2018-01-12 Martin Jambor <mjambor@suse.cz>
6414 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6415 * tree-ssa-math-opts.c: Include domwalk.h.
6416 (convert_mult_to_fma_1): New function.
6417 (fma_transformation_info): New type.
6418 (fma_deferring_state): Likewise.
6419 (cancel_fma_deferring): New function.
6420 (result_of_phi): Likewise.
6421 (last_fma_candidate_feeds_initial_phi): Likewise.
6422 (convert_mult_to_fma): Added deferring logic, split actual
6423 transformation to convert_mult_to_fma_1.
6424 (math_opts_dom_walker): New type.
6425 (math_opts_dom_walker::after_dom_children): New method, body moved
6426 here from pass_optimize_widening_mul::execute, added deferring logic
6428 (pass_optimize_widening_mul::execute): Moved most of code to
6429 math_opts_dom_walker::after_dom_children.
6430 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6431 * config/i386/i386.c (ix86_option_override_internal): Added
6432 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6434 2018-01-12 Richard Biener <rguenther@suse.de>
6437 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6438 inline instance vars.
6440 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6443 * config/rx/rx.c (rx_is_restricted_memory_address):
6446 2018-01-12 Richard Biener <rguenther@suse.de>
6448 PR tree-optimization/80846
6449 * target.def (split_reduction): New target hook.
6450 * targhooks.c (default_split_reduction): New function.
6451 * targhooks.h (default_split_reduction): Declare.
6452 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6453 target requests first reduce vectors by combining low and high
6455 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6456 (get_vectype_for_scalar_type_and_size): Export.
6457 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6458 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6459 * doc/tm.texi: Regenerate.
6460 * config/i386/i386.c (ix86_split_reduction): Implement
6461 TARGET_VECTORIZE_SPLIT_REDUCTION.
6463 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6466 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6467 in PIC mode except for TARGET_VXWORKS_RTP.
6468 * config/sparc/sparc.c: Include cfgrtl.h.
6469 (TARGET_INIT_PIC_REG): Define.
6470 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6471 (sparc_pic_register_p): New predicate.
6472 (sparc_legitimate_address_p): Use it.
6473 (sparc_legitimize_pic_address): Likewise.
6474 (sparc_delegitimize_address): Likewise.
6475 (sparc_mode_dependent_address_p): Likewise.
6476 (gen_load_pcrel_sym): Remove 4th parameter.
6477 (load_got_register): Adjust call to above. Remove obsolete stuff.
6478 (sparc_expand_prologue): Do not call load_got_register here.
6479 (sparc_flat_expand_prologue): Likewise.
6480 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6481 (sparc_use_pseudo_pic_reg): New function.
6482 (sparc_init_pic_reg): Likewise.
6483 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6484 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6486 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6488 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6489 Add item for branch_cost.
6491 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6493 PR rtl-optimization/83565
6494 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6495 not extend the result to a larger mode for rotate operations.
6496 (num_sign_bit_copies1): Likewise.
6498 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6501 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6503 Use values-Xc.o for -pedantic.
6504 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6506 2018-01-12 Martin Liska <mliska@suse.cz>
6509 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6511 (possible_polymorphic_call_targets): Use it.
6512 (ipa_devirt): Likewise.
6514 2018-01-12 Martin Liska <mliska@suse.cz>
6516 * profile-count.h (enum profile_quality): Use 0 as invalid
6517 enum value of profile_quality.
6519 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6521 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6522 -mext-string options.
6524 2018-01-12 Richard Biener <rguenther@suse.de>
6526 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6527 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6528 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6530 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6532 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6534 * configure.ac (--with-long-double-format): Add support for the
6535 configuration option to change the default long double format on
6537 * config.gcc (powerpc*-linux*-*): Likewise.
6538 * configure: Regenerate.
6539 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6540 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6541 used without modification.
6543 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6545 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6546 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6547 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6548 MISC_BUILTIN_SPEC_BARRIER.
6549 (rs6000_init_builtins): Likewise.
6550 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6552 (speculation_barrier): New define_insn.
6553 * doc/extend.texi: Document __builtin_speculation_barrier.
6555 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6558 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6559 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6560 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6562 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6563 integral modes instead of "ss" and "sd".
6564 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6565 vectors with 32-bit and 64-bit elements.
6566 (vecdupssescalarmodesuffix): New mode attribute.
6567 (vec_dup<mode>): Use it.
6569 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6572 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6573 frame if argument is passed on stack.
6575 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6578 * ree.c (combine_reaching_defs): Optimize also
6579 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6580 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6582 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6585 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6587 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6590 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6591 after they are computed.
6593 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6595 PR tree-optimization/83695
6596 * gimple-loop-linterchange.cc
6597 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6598 reset cached scev information after interchange.
6599 (pass_linterchange::execute): Remove call to scev_reset_htab.
6601 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6603 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6604 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6605 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6606 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6607 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6608 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6609 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6610 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6611 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6612 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6613 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6614 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6615 (V_lane_reg): Likewise.
6616 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6618 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6619 (vfmal_lane_low<mode>_intrinsic,
6620 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6621 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6622 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6623 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6624 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6625 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6627 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6629 * config/arm/arm-cpus.in (fp16fml): New feature.
6630 (ALL_SIMD): Add fp16fml.
6631 (armv8.2-a): Add fp16fml as an option.
6632 (armv8.3-a): Likewise.
6633 (armv8.4-a): Add fp16fml as part of fp16.
6634 * config/arm/arm.h (TARGET_FP16FML): Define.
6635 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6637 * config/arm/arm-modes.def (V2HF): Define.
6638 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6639 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6640 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6641 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6642 vfmsl_low, vfmsl_high): New set of builtins.
6643 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6644 (vfml_op): New code attribute.
6645 (VFMLHALVES): New int iterator.
6646 (VFML, VFMLSEL): New mode attributes.
6647 (V_reg): Define mapping for V2HF.
6648 (V_hi, V_lo): New mode attributes.
6649 (VF_constraint): Likewise.
6650 (vfml_half, vfml_half_selector): New int attributes.
6651 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6653 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6654 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6656 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6657 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6658 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6659 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6661 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6662 Document new effective target and option set.
6664 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6666 * config/arm/arm-cpus.in (armv8_4): New feature.
6667 (ARMv8_4a): New fgroup.
6668 (armv8.4-a): New arch.
6669 * config/arm/arm-tables.opt: Regenerate.
6670 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6671 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6672 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6673 Add matching rules for -march=armv8.4-a and extensions.
6674 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6676 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6679 * config/rx/rx.md (BW): New mode attribute.
6680 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6682 2018-01-11 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/83435
6685 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6686 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6687 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6689 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6690 Alan Hayward <alan.hayward@arm.com>
6691 David Sherwood <david.sherwood@arm.com>
6693 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6695 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6696 (aarch64_print_address_internal): Use it to check for a zero offset.
6698 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6699 Alan Hayward <alan.hayward@arm.com>
6700 David Sherwood <david.sherwood@arm.com>
6702 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6703 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6704 Return a poly_int64 rather than a HOST_WIDE_INT.
6705 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6706 rather than a HOST_WIDE_INT.
6707 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6708 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6709 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6710 final_offset from HOST_WIDE_INT to poly_int64.
6711 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6712 to_constant when getting the number of units in an Advanced SIMD
6714 (aarch64_builtin_vectorized_function): Check for a constant number
6716 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6718 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6719 attribute instead of GET_MODE_NUNITS.
6720 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6721 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6722 GET_MODE_SIZE for fixed-size registers.
6723 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6724 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6725 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6726 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6727 (aarch64_print_operand, aarch64_print_address_internal)
6728 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6729 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6730 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6731 Handle polynomial GET_MODE_SIZE.
6732 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6733 wider than SImode without modification.
6734 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6735 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6736 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6737 passing and returning SVE modes.
6738 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6739 rather than GEN_INT.
6740 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6741 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6742 (aarch64_allocate_and_probe_stack_space): Likewise.
6743 (aarch64_layout_frame): Cope with polynomial offsets.
6744 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6745 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6747 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6748 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6749 poly_int64 rather than a HOST_WIDE_INT.
6750 (aarch64_get_separate_components, aarch64_process_components)
6751 (aarch64_expand_prologue, aarch64_expand_epilogue)
6752 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6753 (aarch64_anchor_offset): New function, split out from...
6754 (aarch64_legitimize_address): ...here.
6755 (aarch64_builtin_vectorization_cost): Handle polynomial
6756 TYPE_VECTOR_SUBPARTS.
6757 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6759 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6760 number of elements from the PARALLEL rather than the mode.
6761 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6762 rather than GET_MODE_BITSIZE.
6763 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6764 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6765 (aarch64_expand_vec_perm_const_1): Handle polynomial
6766 d->perm.length () and d->perm elements.
6767 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6768 Apply to_constant to d->perm elements.
6769 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6770 polynomial CONST_VECTOR_NUNITS.
6771 (aarch64_move_pointer): Take amount as a poly_int64 rather
6773 (aarch64_progress_pointer): Avoid temporary variable.
6774 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6775 the mode attribute instead of GET_MODE.
6777 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6778 Alan Hayward <alan.hayward@arm.com>
6779 David Sherwood <david.sherwood@arm.com>
6781 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6782 x exists before using it.
6783 (aarch64_add_constant_internal): Rename to...
6784 (aarch64_add_offset_1): ...this. Replace regnum with separate
6785 src and dest rtxes. Handle the case in which they're different,
6786 including when the offset is zero. Replace scratchreg with an rtx.
6787 Use 2 additions if there is no spare register into which we can
6788 move a 16-bit constant.
6789 (aarch64_add_constant): Delete.
6790 (aarch64_add_offset): Replace reg with separate src and dest
6791 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6792 Use aarch64_add_offset_1.
6793 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6794 an rtx rather than an int. Take the delta as a poly_int64
6795 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6796 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6797 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6798 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6799 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6801 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6802 aarch64_add_constant.
6804 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6806 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6807 Use scalar_float_mode.
6809 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6811 * config/aarch64/aarch64-simd.md
6812 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6813 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6814 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6815 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6816 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6817 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6818 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6819 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6820 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6821 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6823 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6826 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6827 targ_options->x_arm_arch_string is non NULL.
6829 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6831 * config/aarch64/aarch64.h
6832 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6834 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6837 * expmed.c (emit_store_flag_force): Swap if const op0
6838 and change VOIDmode to mode of op0.
6840 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6842 PR rtl-optimization/83761
6843 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6844 than bytes to mode_for_size.
6846 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6849 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6850 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6853 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6856 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6857 when in layout mode.
6858 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6859 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6862 2018-01-10 Michael Collison <michael.collison@arm.com>
6864 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6865 * config/aarch64/aarch64-option-extension.def: Add
6866 AARCH64_OPT_EXTENSION of 'fp16fml'.
6867 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6868 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6869 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6870 * config/aarch64/constraints.md (Ui7): New constraint.
6871 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6872 (VFMLA_SEL_W): Ditto.
6875 (VFMLA16_LOW): New int iterator.
6876 (VFMLA16_HIGH): Ditto.
6877 (UNSPEC_FMLAL): New unspec.
6878 (UNSPEC_FMLSL): Ditto.
6879 (UNSPEC_FMLAL2): Ditto.
6880 (UNSPEC_FMLSL2): Ditto.
6881 (f16mac): New code attribute.
6882 * config/aarch64/aarch64-simd-builtins.def
6883 (aarch64_fmlal_lowv2sf): Ditto.
6884 (aarch64_fmlsl_lowv2sf): Ditto.
6885 (aarch64_fmlalq_lowv4sf): Ditto.
6886 (aarch64_fmlslq_lowv4sf): Ditto.
6887 (aarch64_fmlal_highv2sf): Ditto.
6888 (aarch64_fmlsl_highv2sf): Ditto.
6889 (aarch64_fmlalq_highv4sf): Ditto.
6890 (aarch64_fmlslq_highv4sf): Ditto.
6891 (aarch64_fmlal_lane_lowv2sf): Ditto.
6892 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6893 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6894 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6895 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6896 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6897 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6898 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6899 (aarch64_fmlal_lane_highv2sf): Ditto.
6900 (aarch64_fmlsl_lane_highv2sf): Ditto.
6901 (aarch64_fmlal_laneq_highv2sf): Ditto.
6902 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6903 (aarch64_fmlalq_lane_highv4sf): Ditto.
6904 (aarch64_fmlsl_lane_highv4sf): Ditto.
6905 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6906 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6907 * config/aarch64/aarch64-simd.md:
6908 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6909 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6910 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6911 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6912 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6913 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6914 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6915 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6916 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6917 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6918 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6919 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6920 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6921 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6922 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6923 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6924 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6925 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6926 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6927 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6928 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6929 (vfmlsl_low_u32): Ditto.
6930 (vfmlalq_low_u32): Ditto.
6931 (vfmlslq_low_u32): Ditto.
6932 (vfmlal_high_u32): Ditto.
6933 (vfmlsl_high_u32): Ditto.
6934 (vfmlalq_high_u32): Ditto.
6935 (vfmlslq_high_u32): Ditto.
6936 (vfmlal_lane_low_u32): Ditto.
6937 (vfmlsl_lane_low_u32): Ditto.
6938 (vfmlal_laneq_low_u32): Ditto.
6939 (vfmlsl_laneq_low_u32): Ditto.
6940 (vfmlalq_lane_low_u32): Ditto.
6941 (vfmlslq_lane_low_u32): Ditto.
6942 (vfmlalq_laneq_low_u32): Ditto.
6943 (vfmlslq_laneq_low_u32): Ditto.
6944 (vfmlal_lane_high_u32): Ditto.
6945 (vfmlsl_lane_high_u32): Ditto.
6946 (vfmlal_laneq_high_u32): Ditto.
6947 (vfmlsl_laneq_high_u32): Ditto.
6948 (vfmlalq_lane_high_u32): Ditto.
6949 (vfmlslq_lane_high_u32): Ditto.
6950 (vfmlalq_laneq_high_u32): Ditto.
6951 (vfmlslq_laneq_high_u32): Ditto.
6952 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6953 (AARCH64_FL_FOR_ARCH8_4): New.
6954 (AARCH64_ISA_F16FML): New ISA flag.
6955 (TARGET_F16FML): New feature flag for fp16fml.
6956 (doc/invoke.texi): Document new fp16fml option.
6958 2018-01-10 Michael Collison <michael.collison@arm.com>
6960 * config/aarch64/aarch64-builtins.c:
6961 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6962 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6963 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6964 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6965 (AARCH64_ISA_SHA3): New ISA flag.
6966 (TARGET_SHA3): New feature flag for sha3.
6967 * config/aarch64/iterators.md (sha512_op): New int attribute.
6968 (CRYPTO_SHA512): New int iterator.
6969 (UNSPEC_SHA512H): New unspec.
6970 (UNSPEC_SHA512H2): Ditto.
6971 (UNSPEC_SHA512SU0): Ditto.
6972 (UNSPEC_SHA512SU1): Ditto.
6973 * config/aarch64/aarch64-simd-builtins.def
6974 (aarch64_crypto_sha512hqv2di): New builtin.
6975 (aarch64_crypto_sha512h2qv2di): Ditto.
6976 (aarch64_crypto_sha512su0qv2di): Ditto.
6977 (aarch64_crypto_sha512su1qv2di): Ditto.
6978 (aarch64_eor3qv8hi): Ditto.
6979 (aarch64_rax1qv2di): Ditto.
6980 (aarch64_xarqv2di): Ditto.
6981 (aarch64_bcaxqv8hi): Ditto.
6982 * config/aarch64/aarch64-simd.md:
6983 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6984 (aarch64_crypto_sha512su0qv2di): Ditto.
6985 (aarch64_crypto_sha512su1qv2di): Ditto.
6986 (aarch64_eor3qv8hi): Ditto.
6987 (aarch64_rax1qv2di): Ditto.
6988 (aarch64_xarqv2di): Ditto.
6989 (aarch64_bcaxqv8hi): Ditto.
6990 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6991 (vsha512h2q_u64): Ditto.
6992 (vsha512su0q_u64): Ditto.
6993 (vsha512su1q_u64): Ditto.
6994 (veor3q_u16): Ditto.
6995 (vrax1q_u64): Ditto.
6997 (vbcaxq_u16): Ditto.
6998 * config/arm/types.md (crypto_sha512): New type attribute.
6999 (crypto_sha3): Ditto.
7000 (doc/invoke.texi): Document new sha3 option.
7002 2018-01-10 Michael Collison <michael.collison@arm.com>
7004 * config/aarch64/aarch64-builtins.c:
7005 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7006 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7007 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7008 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7009 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7010 (AARCH64_ISA_SM4): New ISA flag.
7011 (TARGET_SM4): New feature flag for sm4.
7012 * config/aarch64/aarch64-simd-builtins.def
7013 (aarch64_sm3ss1qv4si): Ditto.
7014 (aarch64_sm3tt1aq4si): Ditto.
7015 (aarch64_sm3tt1bq4si): Ditto.
7016 (aarch64_sm3tt2aq4si): Ditto.
7017 (aarch64_sm3tt2bq4si): Ditto.
7018 (aarch64_sm3partw1qv4si): Ditto.
7019 (aarch64_sm3partw2qv4si): Ditto.
7020 (aarch64_sm4eqv4si): Ditto.
7021 (aarch64_sm4ekeyqv4si): Ditto.
7022 * config/aarch64/aarch64-simd.md:
7023 (aarch64_sm3ss1qv4si): Ditto.
7024 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7025 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7026 (aarch64_sm4eqv4si): Ditto.
7027 (aarch64_sm4ekeyqv4si): Ditto.
7028 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7029 (sm3part_op): Ditto.
7030 (CRYPTO_SM3TT): Ditto.
7031 (CRYPTO_SM3PART): Ditto.
7032 (UNSPEC_SM3SS1): New unspec.
7033 (UNSPEC_SM3TT1A): Ditto.
7034 (UNSPEC_SM3TT1B): Ditto.
7035 (UNSPEC_SM3TT2A): Ditto.
7036 (UNSPEC_SM3TT2B): Ditto.
7037 (UNSPEC_SM3PARTW1): Ditto.
7038 (UNSPEC_SM3PARTW2): Ditto.
7039 (UNSPEC_SM4E): Ditto.
7040 (UNSPEC_SM4EKEY): Ditto.
7041 * config/aarch64/constraints.md (Ui2): New constraint.
7042 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7043 * config/arm/types.md (crypto_sm3): New type attribute.
7044 (crypto_sm4): Ditto.
7045 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7046 (vsm3tt1aq_u32): Ditto.
7047 (vsm3tt1bq_u32): Ditto.
7048 (vsm3tt2aq_u32): Ditto.
7049 (vsm3tt2bq_u32): Ditto.
7050 (vsm3partw1q_u32): Ditto.
7051 (vsm3partw2q_u32): Ditto.
7052 (vsm4eq_u32): Ditto.
7053 (vsm4ekeyq_u32): Ditto.
7054 (doc/invoke.texi): Document new sm4 option.
7056 2018-01-10 Michael Collison <michael.collison@arm.com>
7058 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7059 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7060 (AARCH64_FL_FOR_ARCH8_4): New.
7061 (AARCH64_FL_V8_4): New flag.
7062 (doc/invoke.texi): Document new armv8.4-a option.
7064 2018-01-10 Michael Collison <michael.collison@arm.com>
7066 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7067 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7068 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7069 * config/aarch64/aarch64-option-extension.def: Add
7070 AARCH64_OPT_EXTENSION of 'sha2'.
7071 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7072 (crypto): Disable sha2 and aes if crypto disabled.
7073 (crypto): Enable aes and sha2 if enabled.
7074 (simd): Disable sha2 and aes if simd disabled.
7075 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7077 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7078 (TARGET_SHA2): New feature flag for sha2.
7079 (TARGET_AES): New feature flag for aes.
7080 * config/aarch64/aarch64-simd.md:
7081 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7082 conditional on TARGET_AES.
7083 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7084 (aarch64_crypto_sha1hsi): Make pattern conditional
7086 (aarch64_crypto_sha1hv4si): Ditto.
7087 (aarch64_be_crypto_sha1hv4si): Ditto.
7088 (aarch64_crypto_sha1su1v4si): Ditto.
7089 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7090 (aarch64_crypto_sha1su0v4si): Ditto.
7091 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7092 (aarch64_crypto_sha256su0v4si): Ditto.
7093 (aarch64_crypto_sha256su1v4si): Ditto.
7094 (doc/invoke.texi): Document new aes and sha2 options.
7096 2018-01-10 Martin Sebor <msebor@redhat.com>
7098 PR tree-optimization/83781
7099 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7102 2018-01-11 Martin Sebor <msebor@gmail.com>
7103 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7105 PR tree-optimization/83501
7106 PR tree-optimization/81703
7108 * tree-ssa-strlen.c (get_string_cst): Rename...
7109 (get_string_len): ...to this. Handle global constants.
7110 (handle_char_store): Adjust.
7112 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7113 Jim Wilson <jimw@sifive.com>
7115 * config/riscv/riscv-protos.h (riscv_output_return): New.
7116 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7117 (riscv_attribute_table, riscv_output_return),
7118 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7119 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7120 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7122 (riscv_expand_prologue): Add early return for naked function.
7123 (riscv_expand_epilogue): Likewise.
7124 (riscv_function_ok_for_sibcall): Return false for naked function.
7125 (riscv_set_current_function): New.
7126 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7127 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7128 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7129 * doc/extend.texi (RISC-V Function Attributes): New.
7131 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7133 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7134 check for 128-bit long double before checking TCmode.
7135 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7136 128-bit long doubles before checking TFmode or TCmode.
7137 (FLOAT128_IBM_P): Likewise.
7139 2018-01-10 Martin Sebor <msebor@redhat.com>
7141 PR tree-optimization/83671
7142 * builtins.c (c_strlen): Unconditionally return zero for the empty
7144 Use -Warray-bounds for warnings.
7145 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7146 for non-constant array indices with COMPONENT_REF, arrays of
7147 arrays, and pointers to arrays.
7148 (gimple_fold_builtin_strlen): Determine and set length range for
7149 non-constant character arrays.
7151 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7154 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7157 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7159 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7161 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7164 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7165 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7166 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7167 indexed_or_indirect_operand predicate.
7168 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7169 (*vsx_le_perm_load_v8hi): Likewise.
7170 (*vsx_le_perm_load_v16qi): Likewise.
7171 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7172 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7173 (*vsx_le_perm_store_v8hi): Likewise.
7174 (*vsx_le_perm_store_v16qi): Likewise.
7175 (eight unnamed splitters): Likewise.
7177 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7179 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7180 * config/rs6000/emmintrin.h: Likewise.
7181 * config/rs6000/mmintrin.h: Likewise.
7182 * config/rs6000/xmmintrin.h: Likewise.
7184 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7187 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7189 * tree.c (tree_nop_conversion): Return true for location wrapper
7191 (maybe_wrap_with_location): New function.
7192 (selftest::check_strip_nops): New function.
7193 (selftest::test_location_wrappers): New function.
7194 (selftest::tree_c_tests): Call it.
7195 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7196 (maybe_wrap_with_location): New decl.
7197 (EXPR_LOCATION_WRAPPER_P): New macro.
7198 (location_wrapper_p): New inline function.
7199 (tree_strip_any_location_wrapper): New inline function.
7201 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7204 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7205 stack_realign_offset for the largest alignment of stack slot
7207 (ix86_find_max_used_stack_alignment): New function.
7208 (ix86_finalize_stack_frame_flags): Use it. Set
7209 max_used_stack_alignment if we don't realign stack.
7210 * config/i386/i386.h (machine_function): Add
7211 max_used_stack_alignment.
7213 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7215 * config/arm/arm.opt (-mbranch-cost): New option.
7216 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7219 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7222 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7223 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7225 2018-01-10 Richard Biener <rguenther@suse.de>
7228 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7229 early out so it also covers the case where we have a non-NULL
7232 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7234 PR tree-optimization/83753
7235 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7236 for non-strided grouped accesses if the number of elements is 1.
7238 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7241 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7242 * i386.h (TARGET_USE_GATHER): Define.
7243 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7245 2018-01-10 Martin Liska <mliska@suse.cz>
7248 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7249 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7251 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7252 CLEANUP_NO_PARTITIONING is not set.
7254 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7256 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7257 for vectors, as a partial revert of r254296.
7258 * rtl.h (const_vec_p): Delete.
7259 (const_vec_duplicate_p): Don't test for vector CONSTs.
7260 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7261 * expmed.c (make_tree): Likewise.
7264 * common.md (E, F): Use CONSTANT_P instead of checking for
7266 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7267 checking for CONST_VECTOR.
7269 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7272 * predict.c (force_edge_cold): Handle in more sane way edges
7275 2018-01-09 Carl Love <cel@us.ibm.com>
7277 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7279 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7280 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7281 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7282 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7283 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7284 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7285 * config/rs6000/rs6000-protos.h: Add extern defition for
7286 rs6000_generate_float2_double_code.
7287 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7289 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7290 (float2_v2df): Add define_expand.
7292 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7295 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7296 op_mode in the force_to_mode call.
7298 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7300 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7301 instead of checking each element individually.
7302 (aarch64_evpc_uzp): Likewise.
7303 (aarch64_evpc_zip): Likewise.
7304 (aarch64_evpc_ext): Likewise.
7305 (aarch64_evpc_rev): Likewise.
7306 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7307 instead of checking each element individually. Return true without
7309 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7310 whether all selected elements come from the same input, instead of
7311 checking each element individually. Remove calls to gen_rtx_REG,
7312 start_sequence and end_sequence and instead assert that no rtl is
7315 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7317 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7318 order of HIGH and CONST checks.
7320 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7322 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7323 if the destination isn't an SSA_NAME.
7325 2018-01-09 Richard Biener <rguenther@suse.de>
7327 PR tree-optimization/83668
7328 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7330 (canonicalize_loop_form): ... here, renamed from ...
7331 (canonicalize_loop_closed_ssa_form): ... this and amended to
7332 swap successor edges for loop exit blocks to make us use
7333 the RPO order we need for initial schedule generation.
7335 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7337 PR tree-optimization/64811
7338 * match.pd: When optimizing comparisons with Inf, avoid
7339 introducing or losing exceptions from comparisons with NaN.
7341 2018-01-09 Martin Liska <mliska@suse.cz>
7344 * asan.c (shadow_mem_size): Add gcc_assert.
7346 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7348 Don't save registers in main().
7351 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7352 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7353 * config/avr/avr.c (avr_set_current_function): Don't error if
7354 naked, OS_task or OS_main are specified at the same time.
7355 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7357 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7359 * common/config/avr/avr-common.c (avr_option_optimization_table):
7360 Switch on -mmain-is-OS_task for optimizing compilations.
7362 2018-01-09 Richard Biener <rguenther@suse.de>
7364 PR tree-optimization/83572
7365 * graphite.c: Include cfganal.h.
7366 (graphite_transform_loops): Connect infinite loops to exit
7367 and remove fake edges at the end.
7369 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7371 * ipa-inline.c (edge_badness): Revert accidental checkin.
7373 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7376 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7377 symbols; not inline clones.
7379 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7382 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7383 hard registers. Formatting fixes.
7385 PR preprocessor/83722
7386 * gcc.c (try_generate_repro): Pass
7387 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7388 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7391 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7392 Kito Cheng <kito.cheng@gmail.com>
7394 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7395 (riscv_leaf_function_p): Delete.
7396 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7398 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7400 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7402 (do_ifelse): New function.
7403 (do_isel): New function.
7404 (do_sub3): New function.
7405 (do_add3): New function.
7406 (do_load_mask_compare): New function.
7407 (do_overlap_load_compare): New function.
7408 (expand_compare_loop): New function.
7409 (expand_block_compare): Call expand_compare_loop() when appropriate.
7410 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7412 (-mblock-compare-inline-loop-limit): New option.
7414 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7417 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7418 Reverse order of second and third operands in first alternative.
7419 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7420 of first and second elements in UNSPEC_VPERMR vector.
7421 (altivec_expand_vec_perm_le): Likewise.
7423 2018-01-08 Jeff Law <law@redhat.com>
7425 PR rtl-optimizatin/81308
7426 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7427 (process_switch): If group_case_labels makes a change, then set
7429 (pass_convert_switch::execute): If a switch is converted, then
7430 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7432 PR rtl-optimization/81308
7433 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7436 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7438 PR target/83663 - Revert r255946
7439 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7440 generation for cases where splatting a value is not useful.
7441 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7442 across a vec_duplicate and a paradoxical subreg forming a vector
7443 mode to a vec_concat.
7445 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7447 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7448 -march=armv8.3-a variants.
7449 * config/arm/t-multilib: Likewise.
7450 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7452 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7454 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7456 (cceq_ior_compare_complement): Give it a name so I can use it, and
7457 change boolean_or_operator predicate to boolean_operator so it can
7458 be used to generate a crand.
7459 (eqne): New code iterator.
7460 (bd/bd_neg): New code_attrs.
7461 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7462 a single define_insn.
7463 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7464 decrement (bdnzt/bdnzf/bdzt/bdzf).
7465 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7466 with the new names of the branch decrement patterns, and added the
7467 names of the branch decrement conditional patterns.
7469 2018-01-08 Richard Biener <rguenther@suse.de>
7471 PR tree-optimization/83563
7472 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7475 2018-01-08 Richard Biener <rguenther@suse.de>
7478 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7480 2018-01-08 Richard Biener <rguenther@suse.de>
7482 PR tree-optimization/83685
7483 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7484 references to abnormals.
7486 2018-01-08 Richard Biener <rguenther@suse.de>
7489 * dwarf2out.c (output_indirect_strings): Handle empty
7490 skeleton_debug_str_hash.
7491 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7493 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7495 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7496 (emit_store_direct): Likewise.
7497 (arc_trampoline_adjust_address): Likewise.
7498 (arc_asm_trampoline_template): New function.
7499 (arc_initialize_trampoline): Use asm_trampoline_template.
7500 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7501 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7502 * config/arc/arc.md (flush_icache): Delete pattern.
7504 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7506 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7507 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7510 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7513 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7514 by not USED_FOR_TARGET.
7515 (make_pass_resolve_sw_modes): Likewise.
7517 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7519 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7522 2018-01-08 Richard Biener <rguenther@suse.de>
7525 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7527 2018-01-08 Richard Biener <rguenther@suse.de>
7530 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7532 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7535 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7536 basic blocks with a small number of successors.
7537 (convert_control_dep_chain_into_preds): Improve handling of
7539 (dump_predicates): Split apart into...
7540 (dump_pred_chain): ...here...
7541 (dump_pred_info): ...and here.
7542 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7543 (can_chain_union_be_invalidated_p): Improve check for invalidation
7545 (uninit_uses_cannot_happen): Avoid unnecessary if
7546 convert_control_dep_chain_into_preds yielded nothing.
7548 2018-01-06 Martin Sebor <msebor@redhat.com>
7550 PR tree-optimization/83640
7551 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7552 subtracting negative offset from size.
7553 (builtin_access::overlap): Adjust offset bounds of the access to fall
7554 within the size of the object if possible.
7556 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7558 PR rtl-optimization/83699
7559 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7560 extract_bit_field_as_subreg to cases in which the extracted
7561 value is also a vector.
7563 * lra-constraints.c (process_alt_operands): Test for the equivalence
7564 substitutions when detecting a possible reload cycle.
7566 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7569 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7570 by default if flag_selective_schedling{,2}. Formatting fixes.
7572 PR rtl-optimization/83682
7573 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7574 if it has non-VECTOR_MODE element mode.
7575 (vec_duplicate_p): Likewise.
7578 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7579 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7581 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7584 * config/i386/i386-builtin.def
7585 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7586 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7587 Require also OPTION_MASK_ISA_AVX512F in addition to
7588 OPTION_MASK_ISA_GFNI.
7589 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7590 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7591 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7592 to OPTION_MASK_ISA_GFNI.
7593 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7594 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7595 OPTION_MASK_ISA_AVX512BW.
7596 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7597 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7598 addition to OPTION_MASK_ISA_GFNI.
7599 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7600 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7601 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7602 to OPTION_MASK_ISA_GFNI.
7603 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7604 a requirement for all ISAs rather than any of them with a few
7606 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7608 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7609 bitmasks to be enabled with 3 exceptions, instead of requiring any
7610 enabled ISA with lots of exceptions.
7611 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7612 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7613 Change avx512bw in isa attribute to avx512f.
7614 * config/i386/sgxintrin.h: Add license boilerplate.
7615 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7616 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7617 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7618 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7620 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7621 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7622 temporarily sse2 rather than sse if not enabled already.
7625 * config/i386/sse.md (VI248_VLBW): Rename to ...
7626 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7627 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7628 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7629 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7630 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7631 mode iterator instead of VI248_VLBW.
7633 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7635 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7636 (record_modified): Skip clobbers; add debug output.
7637 (param_change_prob): Use sreal frequencies.
7639 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7641 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7642 punt for user-aligned variables.
7644 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7646 * tree-chrec.c (chrec_contains_symbols): Return true for
7649 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7652 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7653 of (x|y) == x for BICS pattern.
7655 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7657 PR tree-optimization/83605
7658 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7659 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7662 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7664 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7665 * config/epiphany/rtems.h: New file.
7667 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7668 Uros Bizjak <ubizjak@gmail.com>
7671 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7672 QIreg_operand instead of register_operand predicate.
7673 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7674 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7675 comments instead of -fmitigate[-_]rop.
7677 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7680 * cgraphunit.c (symbol_table::compile): Switch to text_section
7681 before calling assembly_start debug hook.
7682 * run-rtl-passes.c (run_rtl_passes): Likewise.
7685 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7687 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7688 range_int_cst_p rather than !symbolic_range_p before calling
7689 extract_range_from_multiplicative_op_1.
7691 2018-01-04 Jeff Law <law@redhat.com>
7693 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7694 redundant test in assertion.
7696 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7698 * doc/rtl.texi: Document machine_mode wrapper classes.
7700 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7702 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7705 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7707 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7708 the VEC_PERM_EXPR fold to fail.
7710 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7713 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7714 to switched_sections.
7716 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7719 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7722 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7725 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7726 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7728 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7731 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7732 is BLKmode and bitpos not zero or mode change is needed.
7734 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7737 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7740 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7743 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7744 instead of MULT rtx. Update all corresponding splitters.
7746 (*ssub<modesuffix>): Ditto.
7748 (*cmp_sadd_di): Update split patterns.
7749 (*cmp_sadd_si): Ditto.
7750 (*cmp_sadd_sidi): Ditto.
7751 (*cmp_ssub_di): Ditto.
7752 (*cmp_ssub_si): Ditto.
7753 (*cmp_ssub_sidi): Ditto.
7754 * config/alpha/predicates.md (const23_operand): New predicate.
7755 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7756 Look for ASHIFT, not MULT inner operand.
7757 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7759 2018-01-04 Martin Liska <mliska@suse.cz>
7761 PR gcov-profile/83669
7762 * gcov.c (output_intermediate_file): Add version to intermediate
7764 * doc/gcov.texi: Document new field 'version' in intermediate
7765 file format. Fix location of '-k' option of gcov command.
7767 2018-01-04 Martin Liska <mliska@suse.cz>
7770 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7772 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7774 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7776 2018-01-03 Martin Sebor <msebor@redhat.com>
7778 PR tree-optimization/83655
7779 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7780 checking calls with invalid arguments.
7782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7784 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7785 (vectorizable_mask_load_store): Delete.
7786 (vectorizable_call): Return false for masked loads and stores.
7787 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7788 instead of gimple_assign_rhs1.
7789 (vectorizable_load): Handle IFN_MASK_LOAD.
7790 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7794 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7796 (vectorizable_mask_load_store): ...here.
7797 (vectorizable_load): ...and here.
7799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7801 * tree-vect-stmts.c (vect_build_all_ones_mask)
7802 (vect_build_zero_merge_argument): New functions, split out from...
7803 (vectorizable_load): ...here.
7805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7807 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7809 (vectorizable_mask_load_store): ...here.
7810 (vectorizable_store): ...and here.
7812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7814 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7816 (vectorizable_mask_load_store): ...here.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7820 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7821 (vect_model_store_cost): Take a vec_load_store_type instead of a
7823 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7824 (vect_model_store_cost): Take a vec_load_store_type instead of a
7826 (vectorizable_mask_load_store): Update accordingly.
7827 (vectorizable_store): Likewise.
7828 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7832 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7833 IFN_MASK_LOAD calls here rather than...
7834 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7837 Alan Hayward <alan.hayward@arm.com>
7838 David Sherwood <david.sherwood@arm.com>
7840 * expmed.c (extract_bit_field_1): For vector extracts,
7841 fall back to extract_bit_field_as_subreg if vec_extract
7844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7845 Alan Hayward <alan.hayward@arm.com>
7846 David Sherwood <david.sherwood@arm.com>
7848 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7849 they are variable or constant sized.
7850 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7851 slots for constant-sized data.
7853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7854 Alan Hayward <alan.hayward@arm.com>
7855 David Sherwood <david.sherwood@arm.com>
7857 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7858 handling COND_EXPRs with boolean comparisons, try to find a better
7859 basis for the mask type than the boolean itself.
7861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7863 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7864 is calculated and how it can be overridden.
7865 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7866 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7868 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7872 Alan Hayward <alan.hayward@arm.com>
7873 David Sherwood <david.sherwood@arm.com>
7875 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7876 Remove the mode argument.
7877 (aarch64_simd_valid_immediate): Remove the mode and inverse
7879 * config/aarch64/iterators.md (bitsize): New iterator.
7880 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7881 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7882 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7883 aarch64_simd_valid_immediate.
7884 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7885 (aarch64_reg_or_bic_imm): Likewise.
7886 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7887 with an insn_type enum and msl with a modifier_type enum.
7888 Replace element_width with a scalar_mode. Change the shift
7889 to unsigned int. Add constructors for scalar_float_mode and
7890 scalar_int_mode elements.
7891 (aarch64_vect_float_const_representable_p): Delete.
7892 (aarch64_can_const_movi_rtx_p)
7893 (aarch64_simd_scalar_immediate_valid_for_move)
7894 (aarch64_simd_make_constant): Update call to
7895 aarch64_simd_valid_immediate.
7896 (aarch64_advsimd_valid_immediate_hs): New function.
7897 (aarch64_advsimd_valid_immediate): Likewise.
7898 (aarch64_simd_valid_immediate): Remove mode and inverse
7899 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7900 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7901 and aarch64_float_const_representable_p on the result.
7902 (aarch64_output_simd_mov_immediate): Remove mode argument.
7903 Update call to aarch64_simd_valid_immediate and use of
7904 simd_immediate_info.
7905 (aarch64_output_scalar_simd_mov_immediate): Update call
7908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7909 Alan Hayward <alan.hayward@arm.com>
7910 David Sherwood <david.sherwood@arm.com>
7912 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7913 (mode_nunits): Likewise CONST_MODE_NUNITS.
7914 * machmode.def (ADJUST_NUNITS): Document.
7915 * genmodes.c (mode_data::need_nunits_adj): New field.
7916 (blank_mode): Update accordingly.
7917 (adj_nunits): New variable.
7918 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7920 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7921 listed in adj_nunits.
7922 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7923 listed in adj_nunits. Don't emit case statements for such modes.
7924 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7925 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7926 nothing if adj_nunits is nonnull.
7927 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7928 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7929 (emit_mode_fbit): Update use of print_maybe_const_decl.
7930 (emit_move_size): Likewise. Treat the array as non-const
7932 (emit_mode_adjustments): Handle adj_nunits.
7934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7936 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7937 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7938 (VECTOR_MODES): Use it.
7939 (make_vector_modes): Take the prefix as an argument.
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 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7946 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7947 for MODE_VECTOR_BOOL.
7948 * machmode.def (VECTOR_BOOL_MODE): Document.
7949 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7950 (make_vector_bool_mode): New function.
7951 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7953 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7954 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7956 * stor-layout.c (int_mode_for_mode): Likewise.
7957 * tree.c (build_vector_type_for_mode): Likewise.
7958 * varasm.c (output_constant_pool_2): Likewise.
7959 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7960 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7961 for MODE_VECTOR_BOOL.
7962 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7963 of mode class checks.
7964 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7965 instead of a list of mode class checks.
7966 (expand_vector_scalar_condition): Likewise.
7967 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7970 Alan Hayward <alan.hayward@arm.com>
7971 David Sherwood <david.sherwood@arm.com>
7973 * machmode.h (mode_size): Change from unsigned short to
7975 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7976 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7977 or if measurement_type is not polynomial.
7978 (fixed_size_mode::includes_p): Check for constant-sized modes.
7979 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7980 return a poly_uint16 rather than an unsigned short.
7981 (emit_mode_size): Change the type of mode_size from unsigned short
7982 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7983 (emit_mode_adjustments): Cope with polynomial vector sizes.
7984 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7986 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7988 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7989 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7990 * caller-save.c (setup_save_areas): Likewise.
7991 (replace_reg_with_saved_mem): Likewise.
7992 * calls.c (emit_library_call_value_1): Likewise.
7993 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7994 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7995 (gen_lowpart_for_combine): Likewise.
7996 * convert.c (convert_to_integer_1): Likewise.
7997 * cse.c (equiv_constant, cse_insn): Likewise.
7998 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7999 (cselib_subst_to_values): Likewise.
8000 * dce.c (word_dce_process_block): Likewise.
8001 * df-problems.c (df_word_lr_mark_ref): Likewise.
8002 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8003 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8004 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8005 (rtl_for_decl_location): Likewise.
8006 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8007 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8008 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8009 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8010 (expand_expr_real_1): Likewise.
8011 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8012 (pad_below): Likewise.
8013 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8014 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8015 * ira.c (get_subreg_tracking_sizes): Likewise.
8016 * ira-build.c (ira_create_allocno_objects): Likewise.
8017 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8018 (ira_sort_regnos_for_alter_reg): Likewise.
8019 * ira-costs.c (record_operand_costs): Likewise.
8020 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8021 (resolve_simple_move): Likewise.
8022 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8023 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8024 (lra_constraints): Likewise.
8025 (CONST_POOL_OK_P): Reject variable-sized modes.
8026 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8027 (add_pseudo_to_slot, lra_spill): Likewise.
8028 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8029 * optabs-query.c (get_best_extraction_insn): Likewise.
8030 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8031 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8032 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8033 * recog.c (offsettable_address_addr_space_p): Likewise.
8034 * regcprop.c (maybe_mode_change): Likewise.
8035 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8036 * regrename.c (build_def_use): Likewise.
8037 * regstat.c (dump_reg_info): Likewise.
8038 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8039 (find_reloads, find_reloads_subreg_address): Likewise.
8040 * reload1.c (eliminate_regs_1): Likewise.
8041 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8042 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8043 (simplify_binary_operation_1, simplify_subreg): Likewise.
8044 * targhooks.c (default_function_arg_padding): Likewise.
8045 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8046 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8047 (verify_gimple_assign_ternary): Likewise.
8048 * tree-inline.c (estimate_move_cost): Likewise.
8049 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8050 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8051 (get_address_cost_ainc): Likewise.
8052 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8053 (vect_supportable_dr_alignment): Likewise.
8054 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8055 (vectorizable_reduction): Likewise.
8056 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8057 (vectorizable_operation, vectorizable_load): Likewise.
8058 * tree.c (build_same_sized_truth_vector_type): Likewise.
8059 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8060 * var-tracking.c (emit_note_insn_var_location): Likewise.
8061 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8062 (ADDR_VEC_ALIGN): Likewise.
8064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8065 Alan Hayward <alan.hayward@arm.com>
8066 David Sherwood <david.sherwood@arm.com>
8068 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8070 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8071 or if measurement_type is polynomial.
8072 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8073 * combine.c (make_extraction): Likewise.
8074 * dse.c (find_shift_sequence): Likewise.
8075 * dwarf2out.c (mem_loc_descriptor): Likewise.
8076 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8077 (extract_bit_field, extract_low_bits): Likewise.
8078 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8079 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8080 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8081 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8082 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8083 * reload.c (find_reloads): Likewise.
8084 * reload1.c (alter_reg): Likewise.
8085 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8086 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8087 * tree-if-conv.c (predicate_mem_writes): Likewise.
8088 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8089 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8090 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8091 * valtrack.c (dead_debug_insert_temp): Likewise.
8092 * varasm.c (mergeable_constant_section): Likewise.
8093 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
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 * expr.c (expand_assignment): Cope with polynomial mode sizes
8100 when assigning to a CONCAT.
8102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8103 Alan Hayward <alan.hayward@arm.com>
8104 David Sherwood <david.sherwood@arm.com>
8106 * machmode.h (mode_precision): Change from unsigned short to
8108 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8110 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8111 or if measurement_type is not polynomial.
8112 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8113 in which the mode is already known to be a scalar_int_mode.
8114 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8115 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8117 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8118 for GET_MODE_PRECISION.
8119 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8120 for GET_MODE_PRECISION.
8121 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8123 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8124 (expand_field_assignment, make_extraction): Likewise.
8125 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8126 (get_last_value): Likewise.
8127 * convert.c (convert_to_integer_1): Likewise.
8128 * cse.c (cse_insn): Likewise.
8129 * expr.c (expand_expr_real_1): Likewise.
8130 * lra-constraints.c (simplify_operand_subreg): Likewise.
8131 * optabs-query.c (can_atomic_load_p): Likewise.
8132 * optabs.c (expand_atomic_load): Likewise.
8133 (expand_atomic_store): Likewise.
8134 * ree.c (combine_reaching_defs): Likewise.
8135 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8136 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8137 * tree.h (type_has_mode_precision_p): Likewise.
8138 * ubsan.c (instrument_si_overflow): Likewise.
8140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8141 Alan Hayward <alan.hayward@arm.com>
8142 David Sherwood <david.sherwood@arm.com>
8144 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8145 polynomial numbers of units.
8146 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8147 (valid_vector_subparts_p): New function.
8148 (build_vector_type): Remove temporary shim and take the number
8149 of units as a poly_uint64 rather than an int.
8150 (build_opaque_vector_type): Take the number of units as a
8151 poly_uint64 rather than an int.
8152 * tree.c (build_vector_from_ctor): Handle polynomial
8153 TYPE_VECTOR_SUBPARTS.
8154 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8155 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8156 (build_vector_from_val): If the number of units is variable,
8157 use build_vec_duplicate_cst for constant operands and
8158 VEC_DUPLICATE_EXPR otherwise.
8159 (make_vector_type): Remove temporary is_constant ().
8160 (build_vector_type, build_opaque_vector_type): Take the number of
8161 units as a poly_uint64 rather than an int.
8162 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8164 * cfgexpand.c (expand_debug_expr): Likewise.
8165 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8166 (store_constructor, expand_expr_real_1): Likewise.
8167 (const_scalar_mask_from_tree): Likewise.
8168 * fold-const-call.c (fold_const_reduction): Likewise.
8169 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8170 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8171 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8172 (fold_relational_const): Likewise.
8173 (native_interpret_vector): Likewise. Change the size from an
8174 int to an unsigned int.
8175 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8176 TYPE_VECTOR_SUBPARTS.
8177 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8178 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8179 duplicating a non-constant operand into a variable-length vector.
8180 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8181 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8182 * ipa-icf.c (sem_variable::equals): Likewise.
8183 * match.pd: Likewise.
8184 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8185 * print-tree.c (print_node): Likewise.
8186 * stor-layout.c (layout_type): Likewise.
8187 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8188 * tree-cfg.c (verify_gimple_comparison): Likewise.
8189 (verify_gimple_assign_binary): Likewise.
8190 (verify_gimple_assign_ternary): Likewise.
8191 (verify_gimple_assign_single): Likewise.
8192 * tree-pretty-print.c (dump_generic_node): Likewise.
8193 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8194 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8195 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8196 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8197 (vect_shift_permute_load_chain): Likewise.
8198 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8199 (expand_vector_condition, optimize_vector_constructor): Likewise.
8200 (lower_vec_perm, get_compute_type): Likewise.
8201 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8202 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8203 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8204 (vect_recog_mask_conversion_pattern): Likewise.
8205 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8206 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8207 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8208 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8209 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8210 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8211 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8212 (supportable_widening_operation): Likewise.
8213 (supportable_narrowing_operation): Likewise.
8214 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8216 * varasm.c (output_constant): Likewise.
8218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8219 Alan Hayward <alan.hayward@arm.com>
8220 David Sherwood <david.sherwood@arm.com>
8222 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8223 so that both the length == 3 and length != 3 cases set up their
8224 own permute vectors. Add comments explaining why we know the
8225 number of elements is constant.
8226 (vect_permute_load_chain): Likewise.
8228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8229 Alan Hayward <alan.hayward@arm.com>
8230 David Sherwood <david.sherwood@arm.com>
8232 * machmode.h (mode_nunits): Change from unsigned char to
8234 (ONLY_FIXED_SIZE_MODES): New macro.
8235 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8236 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8237 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8239 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8240 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8241 or if measurement_type is not polynomial.
8242 * genmodes.c (ZERO_COEFFS): New macro.
8243 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8245 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8246 Use ZERO_COEFFS when emitting initializers.
8247 * data-streamer.h (bp_pack_poly_value): New function.
8248 (bp_unpack_poly_value): Likewise.
8249 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8250 for GET_MODE_NUNITS.
8251 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8252 for GET_MODE_NUNITS.
8253 * tree.c (make_vector_type): Remove temporary shim and make
8254 the real function take the number of units as a poly_uint64
8256 (build_vector_type_for_mode): Handle polynomial nunits.
8257 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8258 * emit-rtl.c (const_vec_series_p_1): Likewise.
8259 (gen_rtx_CONST_VECTOR): Likewise.
8260 * fold-const.c (test_vec_duplicate_folding): Likewise.
8261 * genrecog.c (validate_pattern): Likewise.
8262 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8263 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8264 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8265 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8266 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8267 * rtlanal.c (subreg_get_info): Likewise.
8268 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8269 (vect_grouped_load_supported): Likewise.
8270 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8271 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8272 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8273 (simplify_const_unary_operation, simplify_binary_operation_1)
8274 (simplify_const_binary_operation, simplify_ternary_operation)
8275 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8276 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8277 instead of CONST_VECTOR_NUNITS.
8278 * varasm.c (output_constant_pool_2): Likewise.
8279 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8280 explicit-encoded elements in the XVEC for variable-length vectors.
8282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8284 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8287 Alan Hayward <alan.hayward@arm.com>
8288 David Sherwood <david.sherwood@arm.com>
8290 * coretypes.h (fixed_size_mode): Declare.
8291 (fixed_size_mode_pod): New typedef.
8292 * builtins.h (target_builtins::x_apply_args_mode)
8293 (target_builtins::x_apply_result_mode): Change type to
8294 fixed_size_mode_pod.
8295 * builtins.c (apply_args_size, apply_result_size, result_vector)
8296 (expand_builtin_apply_args_1, expand_builtin_apply)
8297 (expand_builtin_return): Update accordingly.
8299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8301 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8302 * cselib.c (cselib_hash_rtx): Likewise.
8303 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8304 CONST_VECTOR encoding.
8306 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8307 Jeff Law <law@redhat.com>
8310 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8311 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8312 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8313 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8316 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8317 explicitly probe *sp in a noreturn function if there were any callee
8318 register saves or frame pointer is needed.
8320 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8323 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8324 BLKmode for ternary, binary or unary expressions.
8327 * var-tracking.c (delete_vta_debug_insn): New inline function.
8328 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8329 insns from get_insns () to NULL instead of each bb separately.
8330 Use delete_vta_debug_insn. No longer static.
8331 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8332 delete_vta_debug_insns callers.
8333 * rtl.h (delete_vta_debug_insns): Declare.
8334 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8335 instead of variable_tracking_main.
8337 2018-01-03 Martin Sebor <msebor@redhat.com>
8339 PR tree-optimization/83603
8340 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8341 arguments past the endof the argument list in functions declared
8342 without a prototype.
8343 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8344 Avoid checking when arguments are null.
8346 2018-01-03 Martin Sebor <msebor@redhat.com>
8349 * doc/extend.texi (attribute const): Fix a typo.
8350 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8351 issuing -Wsuggest-attribute for void functions.
8353 2018-01-03 Martin Sebor <msebor@redhat.com>
8355 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8356 offset_int::from instead of wide_int::to_shwi.
8357 (maybe_diag_overlap): Remove assertion.
8358 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8359 * gimple-ssa-sprintf.c (format_directive): Same.
8360 (parse_directive): Same.
8361 (sprintf_dom_walker::compute_format_length): Same.
8362 (try_substitute_return_value): Same.
8364 2018-01-03 Jeff Law <law@redhat.com>
8367 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8368 non-constant residual for zero at runtime and avoid probing in
8369 that case. Reorganize code for trailing problem to mirror handling
8372 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8374 PR tree-optimization/83501
8375 * tree-ssa-strlen.c (get_string_cst): New.
8376 (handle_char_store): Call get_string_cst.
8378 2018-01-03 Martin Liska <mliska@suse.cz>
8380 PR tree-optimization/83593
8381 * tree-ssa-strlen.c: Include tree-cfg.h.
8382 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8383 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8384 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8386 (strlen_dom_walker::before_dom_children): Call
8387 gimple_purge_dead_eh_edges. Dump tranformation with details
8389 (strlen_dom_walker::before_dom_children): Update call by adding
8390 new argument cleanup_eh.
8391 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8393 2018-01-03 Martin Liska <mliska@suse.cz>
8396 * cif-code.def (VARIADIC_THUNK): New enum value.
8397 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8400 2018-01-03 Jan Beulich <jbeulich@suse.com>
8402 * sse.md (mov<mode>_internal): Tighten condition for when to use
8403 vmovdqu<ssescalarsize> for TI and OI modes.
8405 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8407 Update copyright years.
8409 2018-01-03 Martin Liska <mliska@suse.cz>
8412 * ipa-visibility.c (function_and_variable_visibility): Skip
8413 functions with noipa attribure.
8415 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8417 * gcc.c (process_command): Update copyright notice dates.
8418 * gcov-dump.c (print_version): Ditto.
8419 * gcov.c (print_version): Ditto.
8420 * gcov-tool.c (print_version): Ditto.
8421 * gengtype.c (create_file): Ditto.
8422 * doc/cpp.texi: Bump @copying's copyright year.
8423 * doc/cppinternals.texi: Ditto.
8424 * doc/gcc.texi: Ditto.
8425 * doc/gccint.texi: Ditto.
8426 * doc/gcov.texi: Ditto.
8427 * doc/install.texi: Ditto.
8428 * doc/invoke.texi: Ditto.
8430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8432 * vector-builder.h (vector_builder::m_full_nelts): Change from
8433 unsigned int to poly_uint64.
8434 (vector_builder::full_nelts): Update prototype accordingly.
8435 (vector_builder::new_vector): Likewise.
8436 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8437 (vector_builder::operator ==): Likewise.
8438 (vector_builder::finalize): Likewise.
8439 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8440 Take the number of elements as a poly_uint64 rather than an
8442 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8443 from unsigned int to poly_uint64.
8444 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8445 (vec_perm_indices::new_vector): Likewise.
8446 (vec_perm_indices::length): Likewise.
8447 (vec_perm_indices::nelts_per_input): Likewise.
8448 (vec_perm_indices::input_nelts): Likewise.
8449 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8450 number of elements per input as a poly_uint64 rather than an
8451 unsigned int. Use the original encoding for variable-length
8452 vectors, rather than clamping each individual element.
8453 For the second and subsequent elements in each pattern,
8454 clamp the step and base before clamping their sum.
8455 (vec_perm_indices::series_p): Handle polynomial element counts.
8456 (vec_perm_indices::all_in_range_p): Likewise.
8457 (vec_perm_indices_to_tree): Likewise.
8458 (vec_perm_indices_to_rtx): Likewise.
8459 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8460 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8461 (tree_vector_builder::new_binary_operation): Handle polynomial
8462 element counts. Return false if we need to know the number
8463 of elements at compile time.
8464 * fold-const.c (fold_vec_perm): Punt if the number of elements
8465 isn't known at compile time.
8467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8469 * vec-perm-indices.h (vec_perm_builder): Change element type
8470 from HOST_WIDE_INT to poly_int64.
8471 (vec_perm_indices::element_type): Update accordingly.
8472 (vec_perm_indices::clamp): Handle polynomial element_types.
8473 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8474 (vec_perm_indices::all_in_range_p): Likewise.
8475 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8477 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8478 polynomial vec_perm_indices element types.
8479 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8480 * fold-const.c (fold_vec_perm): Likewise.
8481 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8482 * tree-vect-generic.c (lower_vec_perm): Likewise.
8483 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8484 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8485 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8492 rather than an int. Use plus_constant.
8493 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8494 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8501 a HOST_WIDE_INT to a poly_int64.
8503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8504 Alan Hayward <alan.hayward@arm.com>
8505 David Sherwood <david.sherwood@arm.com>
8507 * calls.c (load_register_parameters): Cope with polynomial
8508 mode sizes. Require a constant size for BLKmode parameters
8509 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8510 forces a parameter to be padded at the lsb end in order to
8511 fill a complete number of words, require the parameter size
8512 to be ordered wrt UNITS_PER_WORD.
8514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8515 Alan Hayward <alan.hayward@arm.com>
8516 David Sherwood <david.sherwood@arm.com>
8518 * reload1.c (spill_stack_slot_width): Change element type
8519 from unsigned int to poly_uint64_pod.
8520 (alter_reg): Treat mode sizes as polynomial.
8522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8523 Alan Hayward <alan.hayward@arm.com>
8524 David Sherwood <david.sherwood@arm.com>
8526 * reload.c (complex_word_subreg_p): New function.
8527 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
8534 operands whose sizes aren't ordered.
8535 (match_reload): Refer to this check here.
8537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8538 Alan Hayward <alan.hayward@arm.com>
8539 David Sherwood <david.sherwood@arm.com>
8541 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8542 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8549 Use plus_constant instead of gen_rtx_PLUS.
8551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8552 Alan Hayward <alan.hayward@arm.com>
8553 David Sherwood <david.sherwood@arm.com>
8555 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8556 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8557 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8558 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8559 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8560 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8561 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8562 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8563 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8564 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8566 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8567 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8568 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8569 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8570 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8571 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8572 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8573 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8574 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8575 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8577 * expr.c (emit_move_resolve_push): Treat the input and result
8578 of PUSH_ROUNDING as a poly_int64.
8579 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8580 (emit_push_insn): Likewise.
8581 * lra-eliminations.c (mark_not_eliminable): Likewise.
8582 * recog.c (push_operand): Likewise.
8583 * reload1.c (elimination_effects): Likewise.
8584 * rtlanal.c (nonzero_bits1): Likewise.
8585 * calls.c (store_one_arg): Likewise. Require the padding to be
8586 known at compile time.
8588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8589 Alan Hayward <alan.hayward@arm.com>
8590 David Sherwood <david.sherwood@arm.com>
8592 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8593 Use plus_constant instead of gen_rtx_PLUS.
8595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8596 Alan Hayward <alan.hayward@arm.com>
8597 David Sherwood <david.sherwood@arm.com>
8599 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8603 Alan Hayward <alan.hayward@arm.com>
8604 David Sherwood <david.sherwood@arm.com>
8606 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8607 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8608 via stack temporaries. Treat the mode size as polynomial too.
8610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8611 Alan Hayward <alan.hayward@arm.com>
8612 David Sherwood <david.sherwood@arm.com>
8614 * expr.c (expand_expr_real_2): When handling conversions involving
8615 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8616 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8617 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8626 Alan Hayward <alan.hayward@arm.com>
8627 David Sherwood <david.sherwood@arm.com>
8629 * combine.c (can_change_dest_mode): Handle polynomial
8630 REGMODE_NATURAL_SIZE.
8631 * expmed.c (store_bit_field_1): Likewise.
8632 * expr.c (store_constructor): Likewise.
8633 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8634 and polynomial REGMODE_NATURAL_SIZE.
8635 (gen_lowpart_common): Likewise.
8636 * reginfo.c (record_subregs_of_mode): Likewise.
8637 * rtlanal.c (read_modify_subreg_p): Likewise.
8639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8640 Alan Hayward <alan.hayward@arm.com>
8641 David Sherwood <david.sherwood@arm.com>
8643 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8644 numbers of elements.
8646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8647 Alan Hayward <alan.hayward@arm.com>
8648 David Sherwood <david.sherwood@arm.com>
8650 * match.pd: Cope with polynomial numbers of vector elements.
8652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8653 Alan Hayward <alan.hayward@arm.com>
8654 David Sherwood <david.sherwood@arm.com>
8656 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8657 in a POINTER_PLUS_EXPR.
8659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8660 Alan Hayward <alan.hayward@arm.com>
8661 David Sherwood <david.sherwood@arm.com>
8663 * omp-simd-clone.c (simd_clone_subparts): New function.
8664 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8665 (ipa_simd_modify_function_body): Likewise.
8667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8668 Alan Hayward <alan.hayward@arm.com>
8669 David Sherwood <david.sherwood@arm.com>
8671 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8672 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8673 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8674 (expand_vector_condition, vector_element): Likewise.
8675 (subparts_gt): New function.
8676 (get_compute_type): Use subparts_gt.
8677 (count_type_subparts): Delete.
8678 (expand_vector_operations_1): Use subparts_gt instead of
8679 count_type_subparts.
8681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8682 Alan Hayward <alan.hayward@arm.com>
8683 David Sherwood <david.sherwood@arm.com>
8685 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8686 (vect_compile_time_alias): ...this new function. Do the calculation
8687 on poly_ints rather than trees.
8688 (vect_prune_runtime_alias_test_list): Update call accordingly.
8690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8691 Alan Hayward <alan.hayward@arm.com>
8692 David Sherwood <david.sherwood@arm.com>
8694 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8696 (vect_schedule_slp_instance): Likewise.
8698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8699 Alan Hayward <alan.hayward@arm.com>
8700 David Sherwood <david.sherwood@arm.com>
8702 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8703 constant and extern definitions for variable-length vectors.
8704 (vect_get_constant_vectors): Note that the number of units
8705 is known to be constant.
8707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8708 Alan Hayward <alan.hayward@arm.com>
8709 David Sherwood <david.sherwood@arm.com>
8711 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8712 of units as polynomial. Choose between WIDE and NARROW based
8715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8716 Alan Hayward <alan.hayward@arm.com>
8717 David Sherwood <david.sherwood@arm.com>
8719 * tree-vect-stmts.c (simd_clone_subparts): New function.
8720 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8723 Alan Hayward <alan.hayward@arm.com>
8724 David Sherwood <david.sherwood@arm.com>
8726 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8727 vectors as polynomial. Use build_index_vector for
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-stmts.c (get_load_store_type): Treat the number of
8735 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8736 for variable-length vectors.
8737 (vectorizable_mask_load_store): Treat the number of units as
8738 polynomial, asserting that it is constant if the condition has
8739 already been enforced.
8740 (vectorizable_store, vectorizable_load): Likewise.
8742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8743 Alan Hayward <alan.hayward@arm.com>
8744 David Sherwood <david.sherwood@arm.com>
8746 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8747 of units as polynomial. Punt if we can't tell at compile time
8748 which vector contains the final result.
8750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8751 Alan Hayward <alan.hayward@arm.com>
8752 David Sherwood <david.sherwood@arm.com>
8754 * tree-vect-loop.c (vectorizable_induction): Treat the number
8755 of units as polynomial. Punt on SLP inductions. Use an integer
8756 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8757 cast of such a series for variable-length floating-point
8760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8761 Alan Hayward <alan.hayward@arm.com>
8762 David Sherwood <david.sherwood@arm.com>
8764 * tree.h (build_index_vector): Declare.
8765 * tree.c (build_index_vector): New function.
8766 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8767 of units as polynomial, forcibly converting it to a constant if
8768 vectorizable_reduction has already enforced the condition.
8769 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8770 to create a {1,2,3,...} vector.
8771 (vectorizable_reduction): Treat the number of units as polynomial.
8772 Choose vectype_in based on the largest scalar element size rather
8773 than the smallest number of units. Enforce the restrictions
8776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8777 Alan Hayward <alan.hayward@arm.com>
8778 David Sherwood <david.sherwood@arm.com>
8780 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8781 number of units as polynomial.
8783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8784 Alan Hayward <alan.hayward@arm.com>
8785 David Sherwood <david.sherwood@arm.com>
8787 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8788 * target.def (autovectorize_vector_sizes): Return the vector sizes
8789 by pointer, using vector_sizes rather than a bitmask.
8790 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8791 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8792 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8794 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8795 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8796 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8797 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8798 * omp-general.c (omp_max_vf): Likewise.
8799 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8800 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8801 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8802 * tree-vect-slp.c (vect_slp_bb): Likewise.
8803 * doc/tm.texi: Regenerate.
8804 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8806 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8807 the vector size as a poly_uint64 rather than an unsigned int.
8808 (current_vector_size): Change from an unsigned int to a poly_uint64.
8809 (get_vectype_for_scalar_type): Update accordingly.
8810 * tree.h (build_truth_vector_type): Take the size and number of
8811 units as a poly_uint64 rather than an unsigned int.
8812 (build_vector_type): Add a temporary overload that takes
8813 the number of units as a poly_uint64 rather than an unsigned int.
8814 * tree.c (make_vector_type): Likewise.
8815 (build_truth_vector_type): Take the number of units as a poly_uint64
8816 rather than an unsigned int.
8818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8819 Alan Hayward <alan.hayward@arm.com>
8820 David Sherwood <david.sherwood@arm.com>
8822 * target.def (get_mask_mode): Take the number of units and length
8823 as poly_uint64s rather than unsigned ints.
8824 * targhooks.h (default_get_mask_mode): Update accordingly.
8825 * targhooks.c (default_get_mask_mode): Likewise.
8826 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8827 * doc/tm.texi: Regenerate.
8829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8830 Alan Hayward <alan.hayward@arm.com>
8831 David Sherwood <david.sherwood@arm.com>
8833 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8834 * omp-general.c (omp_max_vf): Likewise.
8835 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8836 (expand_omp_simd): Handle polynomial safelen.
8837 * omp-low.c (omplow_simd_context): Add a default constructor.
8838 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8839 (lower_rec_simd_input_clauses): Update accordingly.
8840 (lower_rec_input_clauses): Likewise.
8842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8843 Alan Hayward <alan.hayward@arm.com>
8844 David Sherwood <david.sherwood@arm.com>
8846 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8847 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8848 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8849 (vect_analyze_slp_cost): Likewise.
8850 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8851 (vect_model_load_cost): Likewise.
8853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8854 Alan Hayward <alan.hayward@arm.com>
8855 David Sherwood <david.sherwood@arm.com>
8857 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8858 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8859 from an unsigned int * to a poly_uint64_pod *.
8860 (calculate_unrolling_factor): New function.
8861 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8864 Alan Hayward <alan.hayward@arm.com>
8865 David Sherwood <david.sherwood@arm.com>
8867 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8868 from an unsigned int to a poly_uint64.
8869 (_loop_vec_info::slp_unrolling_factor): Likewise.
8870 (_loop_vec_info::vectorization_factor): Change from an int
8872 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8873 (vect_get_num_vectors): New function.
8874 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8875 (vect_get_num_copies): Use vect_get_num_vectors.
8876 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8877 to an unsigned int *.
8878 (vect_analyze_data_refs): Change min_vf from an int * to a
8880 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8881 than an unsigned HOST_WIDE_INT.
8882 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8883 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8884 to an unsigned int *.
8885 (vect_analyze_data_ref_dependences): Likewise.
8886 (vect_compute_data_ref_alignment): Handle polynomial vf.
8887 (vect_enhance_data_refs_alignment): Likewise.
8888 (vect_prune_runtime_alias_test_list): Likewise.
8889 (vect_shift_permute_load_chain): Likewise.
8890 (vect_supportable_dr_alignment): Likewise.
8891 (dependence_distance_ge_vf): Take the vectorization factor as a
8892 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8893 (vect_analyze_data_refs): Change min_vf from an int * to a
8895 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8896 vfm1 as a poly_uint64 rather than an int. Make the same change
8897 for the returned bound_scalar.
8898 (vect_gen_vector_loop_niters): Handle polynomial vf.
8899 (vect_do_peeling): Likewise. Update call to
8900 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8901 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8903 * tree-vect-loop.c (vect_determine_vectorization_factor)
8904 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8905 (vect_get_known_peeling_cost): Likewise.
8906 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8907 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8908 (vect_transform_loop): Likewise. Use the lowest possible VF when
8909 updating the upper bounds of the loop.
8910 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8912 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8913 polynomial unroll factors.
8914 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8915 (vect_make_slp_decision): Likewise.
8916 (vect_supported_load_permutation_p): Likewise, and polynomial
8918 (vect_analyze_slp_cost): Handle polynomial vf.
8919 (vect_slp_analyze_node_operations): Likewise.
8920 (vect_slp_analyze_bb_1): Likewise.
8921 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8922 than an unsigned HOST_WIDE_INT.
8923 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8924 (vectorizable_load): Handle polynomial vf.
8925 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8927 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8930 Alan Hayward <alan.hayward@arm.com>
8931 David Sherwood <david.sherwood@arm.com>
8933 * match.pd: Handle bit operations involving three constants
8934 and try to fold one pair.
8936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8938 * tree-vect-loop-manip.c: Include gimple-fold.h.
8939 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8940 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8941 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8942 Add a path that uses a step of VF instead of 1, but disable it
8944 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8945 and niters_no_overflow parameters. Update calls to
8946 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8947 Create a new SSA name if the latter choses to use a ste other
8948 than zero, and return it via niters_vector_mult_vf_var.
8949 * tree-vect-loop.c (vect_transform_loop): Update calls to
8950 vect_do_peeling, vect_gen_vector_loop_niters and
8951 slpeel_make_loop_iterate_ntimes.
8952 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8953 (vect_gen_vector_loop_niters): Update declarations after above changes.
8955 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8957 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8958 128-bit round to integer instructions.
8959 (ceil<mode>2): Likewise.
8960 (btrunc<mode>2): Likewise.
8961 (round<mode>2): Likewise.
8963 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8965 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8966 unaligned VSX load/store on P8/P9.
8967 (expand_block_clear): Allow the use of unaligned VSX
8968 load/store on P8/P9.
8970 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8972 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8974 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8975 swap associated with both a load and a store.
8977 2018-01-02 Andrew Waterman <andrew@sifive.com>
8979 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8980 * config/riscv/riscv.md (clear_cache): Use it.
8982 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8984 * web.c: Remove out-of-date comment.
8986 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8988 * expr.c (fixup_args_size_notes): Check that any existing
8989 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8990 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8991 (emit_single_push_insn): ...here.
8993 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8995 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8996 (const_vector_encoded_nelts): New function.
8997 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8998 (const_vector_int_elt, const_vector_elt): Declare.
8999 * emit-rtl.c (const_vector_int_elt_1): New function.
9000 (const_vector_elt): Likewise.
9001 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9002 of CONST_VECTOR_ELT.
9004 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9006 * expr.c: Include rtx-vector-builder.h.
9007 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9008 directly on the tree encoding.
9009 (const_vector_from_tree): Likewise.
9010 * optabs.c: Include rtx-vector-builder.h.
9011 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9012 sequence of "u" values.
9013 * vec-perm-indices.c: Include rtx-vector-builder.h.
9014 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9015 directly on the vec_perm_indices encoding.
9017 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9019 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9020 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9021 * rtx-vector-builder.h: New file.
9022 * rtx-vector-builder.c: Likewise.
9023 * rtl.h (rtx_def::u2): Add a const_vector field.
9024 (CONST_VECTOR_NPATTERNS): New macro.
9025 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9026 (CONST_VECTOR_DUPLICATE_P): Likewise.
9027 (CONST_VECTOR_STEPPED_P): Likewise.
9028 (CONST_VECTOR_ENCODED_ELT): Likewise.
9029 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9030 (unwrap_const_vec_duplicate): Likewise.
9031 (const_vec_series_p): Check for a non-duplicated vector encoding.
9032 Say that the function only returns true for integer vectors.
9033 * emit-rtl.c: Include rtx-vector-builder.h.
9034 (gen_const_vec_duplicate_1): Delete.
9035 (gen_const_vector): Call gen_const_vec_duplicate instead of
9036 gen_const_vec_duplicate_1.
9037 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9038 (gen_const_vec_duplicate): Use rtx_vector_builder.
9039 (gen_const_vec_series): Likewise.
9040 (gen_rtx_CONST_VECTOR): Likewise.
9041 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9042 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9043 Build a new vector rather than modifying a CONST_VECTOR in-place.
9044 (handle_special_swappables): Update call accordingly.
9045 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9046 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9047 Build a new vector rather than modifying a CONST_VECTOR in-place.
9048 (handle_special_swappables): Update call accordingly.
9050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9052 * simplify-rtx.c (simplify_const_binary_operation): Use
9053 CONST_VECTOR_ELT instead of XVECEXP.
9055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9057 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9058 the selector elements to be different from the data elements
9059 if the selector is a VECTOR_CST.
9060 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9061 ssizetype for the selector.
9063 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9065 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9066 before testing each element individually.
9067 * tree-vect-generic.c (lower_vec_perm): Likewise.
9069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9071 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9072 * selftest-run-tests.c (selftest::run_tests): Call it.
9073 * vector-builder.h (vector_builder::operator ==): New function.
9074 (vector_builder::operator !=): Likewise.
9075 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9076 (vec_perm_indices::all_from_input_p): New function.
9077 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9078 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9079 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9080 instead of reading the VECTOR_CST directly. Detect whether both
9081 vector inputs are the same before constructing the vec_perm_indices,
9082 and update the number of inputs argument accordingly. Use the
9083 utility functions added above. Only construct sel2 if we need to.
9085 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9087 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9088 the broadcast of the low byte.
9089 (expand_mult_highpart): Use an explicit encoding for the permutes.
9090 * optabs-query.c (can_mult_highpart_p): Likewise.
9091 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9092 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9093 (vectorizable_bswap): Likewise.
9094 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9095 explicit encoding for the power-of-2 permutes.
9096 (vect_permute_store_chain): Likewise.
9097 (vect_grouped_load_supported): Likewise.
9098 (vect_permute_load_chain): Likewise.
9100 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9102 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9103 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9104 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9105 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9106 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9107 (vect_gen_perm_mask_any): Likewise.
9109 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9111 * int-vector-builder.h: New file.
9112 * vec-perm-indices.h: Include int-vector-builder.h.
9113 (vec_perm_indices): Redefine as an int_vector_builder.
9114 (auto_vec_perm_indices): Delete.
9115 (vec_perm_builder): Redefine as a stand-alone class.
9116 (vec_perm_indices::vec_perm_indices): New function.
9117 (vec_perm_indices::clamp): Likewise.
9118 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9119 (vec_perm_indices::new_vector): New function.
9120 (vec_perm_indices::new_expanded_vector): Update for new
9121 vec_perm_indices class.
9122 (vec_perm_indices::rotate_inputs): New function.
9123 (vec_perm_indices::all_in_range_p): Operate directly on the
9124 encoded form, without computing elided elements.
9125 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9126 encoding. Update for new vec_perm_indices class.
9127 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9128 the given vec_perm_builder.
9129 (expand_vec_perm_var): Update vec_perm_builder constructor.
9130 (expand_mult_highpart): Use vec_perm_builder instead of
9131 auto_vec_perm_indices.
9132 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9133 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9134 or double series encoding as appropriate.
9135 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9136 vec_perm_indices instead of auto_vec_perm_indices.
9137 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9138 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9139 (vect_permute_store_chain): Likewise.
9140 (vect_grouped_load_supported): Likewise.
9141 (vect_permute_load_chain): Likewise.
9142 (vect_shift_permute_load_chain): Likewise.
9143 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9144 (vect_transform_slp_perm_load): Likewise.
9145 (vect_schedule_slp_instance): Likewise.
9146 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9147 (vectorizable_mask_load_store): Likewise.
9148 (vectorizable_bswap): Likewise.
9149 (vectorizable_store): Likewise.
9150 (vectorizable_load): Likewise.
9151 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9152 vec_perm_indices instead of auto_vec_perm_indices. Use
9153 tree_to_vec_perm_builder to read the vector from a tree.
9154 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9155 vec_perm_builder instead of a vec_perm_indices.
9156 (have_whole_vector_shift): Use vec_perm_builder and
9157 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9158 truncation to calc_vec_perm_mask_for_shift.
9159 (vect_create_epilog_for_reduction): Likewise.
9160 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9161 from auto_vec_perm_indices to vec_perm_indices.
9162 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9163 instead of changing individual elements.
9164 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9165 the vector in d.perm.
9166 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9167 from auto_vec_perm_indices to vec_perm_indices.
9168 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9169 instead of changing individual elements.
9170 (arm_vectorize_vec_perm_const): Use new_vector to install
9171 the vector in d.perm.
9172 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9173 Update vec_perm_builder constructor.
9174 (rs6000_expand_interleave): Likewise.
9175 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9176 (rs6000_expand_interleave): Likewise.
9178 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9180 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9181 to qimode could truncate the indices.
9182 * optabs.c (expand_vec_perm_var): Likewise.
9184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9186 * Makefile.in (OBJS): Add vec-perm-indices.o.
9187 * vec-perm-indices.h: New file.
9188 * vec-perm-indices.c: Likewise.
9189 * target.h (vec_perm_indices): Replace with a forward class
9191 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9192 * optabs.h: Include vec-perm-indices.h.
9193 (expand_vec_perm): Delete.
9194 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9195 (expand_vec_perm_const): Declare.
9196 * target.def (vec_perm_const_ok): Replace with...
9197 (vec_perm_const): ...this new hook.
9198 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9199 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9200 * doc/tm.texi: Regenerate.
9201 * optabs.def (vec_perm_const): Delete.
9202 * doc/md.texi (vec_perm_const): Likewise.
9203 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9204 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9205 expand_vec_perm for constant permutation vectors. Assert that
9206 the mode of variable permutation vectors is the integer equivalent
9207 of the mode that is being permuted.
9208 * optabs-query.h (selector_fits_mode_p): Declare.
9209 * optabs-query.c: Include vec-perm-indices.h.
9210 (selector_fits_mode_p): New function.
9211 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9212 is defined, instead of checking whether the vec_perm_const_optab
9213 exists. Use targetm.vectorize.vec_perm_const instead of
9214 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9215 fit in the vector mode before using a variable permute.
9216 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9217 vec_perm_indices instead of an rtx.
9218 (expand_vec_perm): Replace with...
9219 (expand_vec_perm_const): ...this new function. Take the selector
9220 as a vec_perm_indices rather than an rtx. Also take the mode of
9221 the selector. Update call to shift_amt_for_vec_perm_mask.
9222 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9223 Use vec_perm_indices::new_expanded_vector to expand the original
9224 selector into bytes. Check whether the indices fit in the vector
9225 mode before using a variable permute.
9226 (expand_vec_perm_var): Make global.
9227 (expand_mult_highpart): Use expand_vec_perm_const.
9228 * fold-const.c: Includes vec-perm-indices.h.
9229 * tree-ssa-forwprop.c: Likewise.
9230 * tree-vect-data-refs.c: Likewise.
9231 * tree-vect-generic.c: Likewise.
9232 * tree-vect-loop.c: Likewise.
9233 * tree-vect-slp.c: Likewise.
9234 * tree-vect-stmts.c: Likewise.
9235 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9237 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9238 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9239 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9240 (aarch64_vectorize_vec_perm_const): ...this new function.
9241 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9242 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9243 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9244 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9245 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9246 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9247 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9249 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9250 check for NEON modes.
9251 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9252 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9253 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9254 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9256 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9257 the old VEC_PERM_CONST conditions.
9258 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9259 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9260 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9261 (ia64_vectorize_vec_perm_const_ok): Merge into...
9262 (ia64_vectorize_vec_perm_const): ...this new function.
9263 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9264 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9265 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9266 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9267 * config/mips/mips.c (mips_expand_vec_perm_const)
9268 (mips_vectorize_vec_perm_const_ok): Merge into...
9269 (mips_vectorize_vec_perm_const): ...this new function.
9270 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9271 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9272 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9273 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9274 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9275 (rs6000_expand_vec_perm_const): Delete.
9276 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9278 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9279 (altivec_expand_vec_perm_const_le): Take each operand individually.
9280 Operate on constant selectors rather than rtxes.
9281 (altivec_expand_vec_perm_const): Likewise. Update call to
9282 altivec_expand_vec_perm_const_le.
9283 (rs6000_expand_vec_perm_const): Delete.
9284 (rs6000_vectorize_vec_perm_const_ok): Delete.
9285 (rs6000_vectorize_vec_perm_const): New function.
9286 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9287 an element count and rtx array.
9288 (rs6000_expand_extract_even): Update call accordingly.
9289 (rs6000_expand_interleave): Likewise.
9290 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9291 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9292 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9293 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9294 (rs6000_expand_vec_perm_const): Delete.
9295 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9296 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9297 (altivec_expand_vec_perm_const_le): Take each operand individually.
9298 Operate on constant selectors rather than rtxes.
9299 (altivec_expand_vec_perm_const): Likewise. Update call to
9300 altivec_expand_vec_perm_const_le.
9301 (rs6000_expand_vec_perm_const): Delete.
9302 (rs6000_vectorize_vec_perm_const_ok): Delete.
9303 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9304 reference to the SPE evmerge intructions.
9305 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9306 an element count and rtx array.
9307 (rs6000_expand_extract_even): Update call accordingly.
9308 (rs6000_expand_interleave): Likewise.
9309 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9310 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9312 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9314 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9316 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9317 vector mode and that that mode matches the mode of the data
9319 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9320 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9321 directly using expand_vec_perm_1 when forcing selectors into
9323 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9325 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9327 * optabs-query.h (can_vec_perm_p): Delete.
9328 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9329 * optabs-query.c (can_vec_perm_p): Split into...
9330 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9331 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9332 particular selector is valid.
9333 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9334 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9335 (vect_grouped_load_supported): Likewise.
9336 (vect_shift_permute_load_chain): Likewise.
9337 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9338 (vect_transform_slp_perm_load): Likewise.
9339 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9340 (vectorizable_bswap): Likewise.
9341 (vect_gen_perm_mask_checked): Likewise.
9342 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9343 implementations of variable permutation vectors into account
9344 when deciding which selector to use.
9345 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9346 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9347 with a false third argument.
9348 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9349 to test whether the constant selector is valid and can_vec_perm_var_p
9350 to test whether a variable selector is valid.
9352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9354 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9355 * optabs-query.c (can_vec_perm_p): Likewise.
9356 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9357 instead of vec_perm_indices.
9358 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9359 (vect_gen_perm_mask_checked): Likewise,
9360 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9361 (vect_gen_perm_mask_checked): Likewise,
9363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9365 * optabs-query.h (qimode_for_vec_perm): Declare.
9366 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9367 (qimode_for_vec_perm): ...this new function.
9368 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9370 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9372 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9373 does not have a conditional at the top.
9375 2018-01-02 Richard Biener <rguenther@suse.de>
9377 * ipa-inline.c (big_speedup_p): Fix expression.
9379 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9382 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9385 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9389 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9390 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9391 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9392 cond_taken_branch_cost 3->4.
9394 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9396 PR tree-optimization/83581
9397 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9398 TODO_cleanup_cfg if any changes have been made.
9401 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9402 convert_modes if target mode has the right side, but different mode
9406 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9407 last argument when extracting from CONCAT. If either from_real or
9408 from_imag is NULL, use expansion through memory. If result is not
9409 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9410 the parts directly to inner mode, if even that fails, use expansion
9414 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9415 check for bswap in mode rather than HImode and use that in expand_unop
9418 Copyright (C) 2018 Free Software Foundation, Inc.
9420 Copying and distribu