1 2018-03-23 Richard Biener <rguenther@suse.de>
4 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5 we are going to emit early debug for LTO.
7 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10 * function.c (match_asm_constraints_1): Don't optimize if input
11 doesn't satisfy general_operand predicate for output's mode.
14 * alias.c (write_dependence_p): Don't require for x_canonicalized
15 non-VOIDmode if x has VOIDmode.
18 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
19 just don't try to optimize it rather than assert it never happens.
21 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
23 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
24 macro expansions for definition of ST_INTERNAL_<mode> and
25 LD_INTERNAL_<mode> builtins.
26 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
28 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
30 (altivec_expand_st_builtin): Likewise.
31 (altivec_expand_builtin): Remove calls to deleted functions.
32 (rs6000_address_for_altivec): Delete this function.
33 * config/rs6000/vector.md: Remove expands for
34 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
36 2018-03-22 Sudakshina Das <sudi.das@arm.com>
39 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
40 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
41 re-computing once computed.
42 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
43 (arm_init_machine_status): Initialize
44 machine->static_chain_stack_bytes.
46 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
49 * doc/extend.texi: Add four new prototypes for vec_ld.
50 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
51 definitions for more logical presentation.
52 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
53 entries for V1TI variants of __builtin_altivec_ld builtin.
54 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
55 handling of V1TI variant of LVX icode pattern.
56 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
57 (rs6000_gimple_fold_builtin): Likewise.
58 (altivec_init_builtins): Add code to define
59 __builtin_altivec_lvx_v1ti function.
61 2018-03-22 Jakub Jelinek <jakub@redhat.com>
64 * function.c (match_asm_constraints_1): Don't do the optimization
65 if input isn't a REG, SUBREG, MEM or constant.
67 2018-03-22 Tom de Vries <tom@codesourcery.com>
69 PR tree-optimization/84956
70 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
73 2018-03-22 Jakub Jelinek <jakub@redhat.com>
76 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
77 DECL_INITIAL (decl) to decl at the end.
78 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
81 2018-03-21 Joseph Myers <joseph@codesourcery.com>
83 * doc/extend.texi (__builtin_tgmath): Document when complex
84 integer types are treated as _Complex _Float64.
86 2018-03-21 Tom de Vries <tom@codesourcery.com>
88 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
90 2018-03-21 Jakub Jelinek <jakub@redhat.com>
92 PR tree-optimization/84960
93 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
94 if it is ENTRY block, move them into single succ of ENTRY in that case.
96 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
98 PR tree-optimization/84811
99 * poly-int.h (poly_span_traits): Remove the T3 parameter and
100 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
101 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
102 (known_subrange_p): Update accordingly. Cast each value involved
103 in the size comparison, rather than casting the result of the
106 2018-03-21 Jakub Jelinek <jakub@redhat.com>
108 PR tree-optimization/84982
109 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
110 by flipping the least significant bit rather than all bits from
111 bitpos to bitpos + bitsize - 1.
113 2018-03-21 Nathan Sidwell <nathan@acm.org>
115 * doc/extend.texi (Deprecated Features): Remove mention of
116 long-deleted deprecations.
118 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
121 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
122 * configure: Regenerate.
124 2018-03-21 Tom de Vries <tom@codesourcery.com>
126 PR tree-optimization/83126
127 * tree-parloops.c (num_phis): New function.
128 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
130 2018-03-21 Nathan Sidwell <nathan@acm.org>
132 * doc/extend.texi (Deprecated Features): Update deprecated flags,
133 mention anon-struct/union members and trailing attributes.
135 2018-03-21 Bin Cheng <bin.cheng@arm.com>
137 PR tree-optimization/84969
138 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
139 builtin memset partitions if they set different rhs values.
141 2018-03-21 Jakub Jelinek <jakub@redhat.com>
143 PR rtl-optimization/84989
144 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
145 VEC_DUPLICATE with scalar result mode.
147 2018-03-21 Martin Liska <mliska@suse.cz>
150 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
151 not intended return statement.
153 2018-03-21 Martin Liska <mliska@suse.cz>
156 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
157 (chkp_find_bound_slots_1): Limit number of iterations.
159 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
162 * Minor grammar fixes for x86 options.
164 2018-03-20 Jakub Jelinek <jakub@redhat.com>
167 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
168 holding REG_CFA_RESTORE notes, instead turn them into a USE.
170 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
173 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
174 (altivec_lvx_<mode>_1op): Likewise.
175 (altivec_stvx_<mode>_2op): Likewise.
176 (altivec_stvx_<mode>_1op): Likewise.
177 (altivec_lvx_<VM2:mode>): New define_expand.
178 (altivec_stvx_<VM2:mode>): Likewise.
179 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
180 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
181 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
182 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
183 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
184 (rs6000_gen_lvx): Likewise.
185 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
186 (altivec_expand_stv_builtin): Likewise.
187 (altivec_expand_builtin): Likewise.
188 * config/rs6000/vector.md: Likewise.
190 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
193 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
196 2018-03-20 Richard Biener <rguenther@suse.de>
199 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
200 sign-conversions as zero, fall back to standard scalar_stmt
203 2018-03-20 Martin Liska <mliska@suse.cz>
206 * predict.c (rebuild_frequencies): Handle case when we have
207 PROFILE_ABSENT, but flag_guess_branch_prob is false.
209 2018-03-20 Jakub Jelinek <jakub@redhat.com>
212 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
213 flag_section_anchors.
214 * varasm.c (use_blocks_for_decl_p): Remove hack for
218 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
220 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
221 be created, use lowpart_subreg of operands[0] rather than operands[0]
223 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
224 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
225 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
226 and n constraint instead of aarch64_shift_imm_di and Usd.
227 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
228 (*aarch64_<optab>_reg_minus<mode>3): ... this.
230 2018-03-20 Sudakshina Das <sudi.das@arm.com>
233 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
234 to favor GPR over NEON registers.
235 (<shift>di3_neon): Likewise.
237 2018-03-20 Tom de Vries <tom@codesourcery.com>
240 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
241 (nvptx_process_pars): Emit bar.sync asap and alap.
243 2018-03-20 Tom de Vries <tom@codesourcery.com>
246 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
247 seen_label if seen_label is already set.
249 2018-03-20 Jakub Jelinek <jakub@redhat.com>
252 * config/i386/i386.c (fold_builtin_cpu): For features above 31
253 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
254 Use 1U instead of 1. Formatting fixes.
257 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
258 instead of TREE_TYPE (s1) for the return value.
260 2018-03-19 Jakub Jelinek <jakub@redhat.com>
262 PR tree-optimization/84946
263 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
264 bitsize + bitsize in poly_uint64 rather than poly_int64.
267 * dwarf2asm.c: Include fold-const.c.
268 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
269 of decl rather than decl itself.
271 PR rtl-optimization/84643
272 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
274 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
277 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
278 calling assemble_variable.
280 2018-03-19 Sudakshina Das <sudi.das@arm.com>
283 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
284 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
286 2018-03-19 Jim Wilson <jimw@sifive.com>
289 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
290 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
291 (riscv_first_stack_step): Likewise.
292 (riscv_option_override): Use STACK_BOUNDARY instead of
294 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
296 (BIGGEST_ALIGNMENT): Set to 128.
297 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
298 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
301 2018-03-19 Richard Biener <rguenther@suse.de>
303 PR tree-optimization/84933
304 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
305 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
307 2018-03-19 Richard Biener <rguenther@suse.de>
309 PR tree-optimization/84859
310 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
311 (cond_if_else_store_replacement): Perform sinking operation on
312 single-store BBs regardless of MAX_STORES_TO_SINK setting.
313 Generalize what a BB with a single eligible store is.
315 2018-03-19 Richard Biener <rguenther@suse.de>
317 PR tree-optimization/84929
318 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
319 chrec_is_positive against non-chrec arg.
321 2018-03-19 Tamar Christina <tamar.christina@arm.com>
324 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
326 2018-03-18 Martin Liska <mliska@suse.cz>
328 PR rtl-optimization/84635
329 * regrename.c (build_def_use): Use matches_mode only when
332 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
334 PR tree-optimization/84913
335 * tree-vect-loop.c (vectorizable_reduction): Don't try to
336 vectorize chains of COND_EXPRs.
338 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
340 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
342 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
344 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
346 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
348 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
350 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
351 Kito Cheng <kito.cheng@gmail.com>
353 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
354 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
355 (nds32_adjust_reg_alloc_order): New function.
356 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
358 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
360 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
361 nds32_print_operand, nds32_print_operand_address): Use
362 HOST_WIDE_INT_PRINT_DEC instead.
364 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
368 2018-03-17 Jakub Jelinek <jakub@redhat.com>
371 * config/i386/i386.c (initial_ix86_tune_features,
372 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
374 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
375 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
376 rather than 1u << ix86_tune. Formatting fix.
377 (ix86_option_override_internal): Change ix86_arch_mask from
378 unsigned int to unsigned HOST_WIDE_INT, initialize to
379 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
380 (ix86_function_specific_restore): Likewise.
382 2018-03-16 Jakub Jelinek <jakub@redhat.com>
385 * postreload.c (reload_combine_recognize_pattern): Perform
386 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
387 truncate_int_for_mode the result for the destination's mode.
390 * hsa-gen.c (mem_type_for_type): Fix comment typo.
391 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
393 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
396 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
399 * lra-assigns.c (lra_split_hard_reg_for): Don't use
400 regno_allocno_class_array and sorted_pseudos.
401 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
402 insns where regno is used.
404 2018-03-16 Martin Liska <mliska@suse.cz>
407 * multiple_target.c (create_dispatcher_calls): Redirect
408 reference in the symbol table.
410 2018-03-16 Martin Liska <mliska@suse.cz>
413 * multiple_target.c (create_dispatcher_calls): Redirect also
416 2018-03-16 Jakub Jelinek <jakub@redhat.com>
420 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
421 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
422 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
424 2018-03-16 Julia Koval <julia.koval@intel.com>
426 * doc/invoke.texi (Skylake Server): Add CLWB.
427 Cannonlake): Remove CLWB.
429 2018-03-16 Jakub Jelinek <jakub@redhat.com>
431 PR tree-optimization/84841
432 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
434 (FLOAT_ONE_CONST_TYPE): Define.
435 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
436 (sort_by_operand_rank): Put entries with higher constant_type last
437 rather than first to match comments.
439 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
441 * config/nios2/nios2.md (movsi_internal): Fix thinko in
444 2018-03-15 Jakub Jelinek <jakub@redhat.com>
447 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
448 check and use address of target always.
450 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
453 * config/i386/i386.c (indirect_thunk_needed): Update comments.
454 (indirect_thunk_bnd_needed): Likewise.
455 (indirect_thunks_used): Likewise.
456 (indirect_thunks_bnd_used): Likewise.
457 (indirect_return_needed): New.
458 (indirect_return_bnd_needed): Likewise.
459 (output_indirect_thunk_function): Add a bool argument for
461 (output_indirect_thunk_function): Don't generate alias for
462 function return thunk.
463 (ix86_code_end): Call output_indirect_thunk_function to generate
464 function return thunks.
465 (ix86_output_function_return): Set indirect_return_bnd_needed
466 and indirect_return_needed instead of indirect_thunk_bnd_needed
467 and indirect_thunk_needed.
469 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
471 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
472 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
473 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
475 2018-03-15 David Malcolm <dmalcolm@redhat.com>
476 Paul Hua <paul.hua.gm@gmail.com>
479 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
481 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
483 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
484 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
487 2018-03-15 Tamar Christina <tamar.christina@arm.com>
490 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
491 instead of GET_MODE_SIZE when comparing Units.
493 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
496 * varasm.c (hash_section): Return an unchangeble hash value
497 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
498 Return !aarch64_can_use_per_function_literal_pools_p ().
500 2018-03-15 Jakub Jelinek <jakub@redhat.com>
503 * optabs.c (emit_conditional_move): Pass address of cmode's copy
504 rather than address of cmode as last argument to prepare_cmp_insn.
506 2018-03-15 Julia Koval <julia.koval@intel.com>
508 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
509 F_AVX512VNNI, F_AVX512BITALG): New.
511 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
514 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
515 insn for floating-point loads and stores.
517 2018-03-14 Carl Love <cel@us.ibm.com>
519 * config/rs6000/rs6000-c.c: Add macro definitions for
520 ALTIVEC_BUILTIN_VEC_PERMXOR.
521 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
522 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
523 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
524 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
526 * config/doc/extend.texi: Add prototypes for vec_permxor.
528 2018-03-14 David Malcolm <dmalcolm@redhat.com>
531 * diagnostic-show-locus.c (class layout_point): Convert m_line
532 from int to linenum_type.
533 (line_span::comparator): Use linenum "compare" function when
534 comparing line numbers.
535 (test_line_span): New function.
536 (layout_range::contains_point): Convert param "row" from int to
538 (layout_range::intersects_line_p): Likewise.
539 (layout::will_show_line_p): Likewise.
540 (layout::print_source_line): Likewise.
541 (layout::should_print_annotation_line_p): Likewise.
542 (layout::print_annotation_line): Likewise.
543 (layout::print_leading_fixits): Likewise.
544 (layout::annotation_line_showed_range_p): Likewise.
545 (struct line_corrections): Likewise for field m_row.
546 (line_corrections::line_corrections): Likewise for param "row".
547 (layout::print_trailing_fixits): Likewise.
548 (layout::get_state_at_point): Likewise.
549 (layout::get_x_bound_for_row): Likewise.
550 (layout::print_line): Likewise.
551 (diagnostic_show_locus): Likewise for locals "last_line" and
553 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
554 * input.c (selftest::test_linenum_comparisons): New function.
555 (selftest::input_c_tests): Call it.
556 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
557 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
558 * selftest.h (ASSERT_GT): New macro.
559 (ASSERT_GT_AT): New macro.
560 (ASSERT_LT): New macro.
561 (ASSERT_LT_AT): New macro.
563 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
565 PR rtl-optimization/84780
566 * combine.c (distribute_links): Don't make a link based on pc_rtx.
568 2018-03-14 Martin Liska <mliska@suse.cz>
570 * tree.c (record_node_allocation_statistics): Use
572 (get_stats_node_kind): New function extracted from
573 record_node_allocation_statistics.
574 (free_node): Use get_stats_node_kind.
576 2018-03-14 Richard Biener <rguenther@suse.de>
578 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
579 that the value-set of ANTIC_IN doesn't grow.
582 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
584 (BB_VISITED_WITH_VISITED_SUCCS): New define.
585 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
587 2018-03-14 Julia Koval <julia.koval@intel.com>
589 * config.gcc (icelake-client, icelake-server): New.
591 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
592 (initial_ix86_arch_features): Ditto.
593 (PTA_SKYLAKE): Add SGX.
594 (PTA_ICELAKE): Remove.
595 (PTA_ICELAKE_CLIENT): New.
596 (PTA_ICELAKE_SERVER): New.
597 (ix86_option_override_internal): Split up icelake on icelake client and
599 (get_builtin_code_for_version): Ditto.
600 (fold_builtin_cpu): Ditto.
601 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
602 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
603 * config/i386/i386.h (processor_type): Ditto.
604 * doc/invoke.texi: Ditto.
606 2018-03-14 Jakub Jelinek <jakub@redhat.com>
609 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
610 INTEGER_CST offset, add it together with bitpos / 8 and
611 sign extend based on POINTER_SIZE.
615 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
618 * config/i386/constraints.md (Yc): New register constraint.
619 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
620 Use Yc constraint for alternative 2 of operand 0. Remove
621 preferred_for_speed attribute.
623 2018-03-14 Richard Biener <rguenther@suse.de>
625 PR tree-optimization/84830
626 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
627 with the old one to avoid oscillations.
629 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
632 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
634 (assign_by_spills): Return a flag of reload assignment failure.
635 Do not process the reload assignment failures. Do not spill other
636 reload pseudos if they has the same reg class. Update n if
638 (lra_assign): Add a return arg. Set up from the result of
639 assign_by_spills call.
640 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
641 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
642 usage_insns if it is not NULL.
643 (spill_hard_reg_in_range): New function.
644 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
645 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
647 (lra_assign): Change prototype.
648 * lra.c (lra): Add code to deal with fails by splitting hard reg
651 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
653 * config/riscv/riscv.opt (mrelax): New option.
654 * config/riscv/riscv.c (riscv_file_start): Emit ".option
655 "norelax" when riscv_mrelax is disabled.
656 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
658 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
661 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
662 reassociation for int modes.
664 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
666 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
667 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
669 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
670 * config/aarch64/aarch64-sve.md
671 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
672 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
673 (*extend<mode><Vwide>2): Rename to...
674 (aarch64_sve_extend<mode><Vwide>2): ...this.
675 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
676 renaming the old pattern to...
677 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
679 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
680 define_expand, renaming the old pattern to...
681 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
682 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
683 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
684 account when deciding which SVE instruction the optab should use.
685 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
687 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
689 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
690 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
691 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
692 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
693 (tlsdesc_small_<mode>): Turn a define_expand and use
694 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
695 (tlsdesc_small_advsimd_<mode>): ...this.
696 (tlsdesc_small_sve_<mode>): New pattern.
698 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
700 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
701 (UNSPEC_UMUL_HIGHPART): New constants.
702 (MUL_HIGHPART): New int iteraor.
703 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
704 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
706 (*<su>mul<mode>3_highpart): New define_insn.
708 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
711 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
714 2018-03-13 Martin Liska <mliska@suse.cz>
717 * (sem_item_optimizer::sem_item_optimizer): Initialize new
719 (sem_item_optimizer::~sem_item_optimizer): Release it.
720 (sem_item_optimizer::merge_classes): Register variable aliases.
721 (sem_item_optimizer::fixup_pt_set): New function.
722 (sem_item_optimizer::fixup_points_to_sets): Likewise.
723 * ipa-icf.h: Declare new variables and functions.
725 2018-03-13 Jakub Jelinek <jakub@redhat.com>
728 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
729 integer_pow2p@2 and test integer_pow2p in condition.
730 (A < 0 ? C : 0): Similarly for @1.
733 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
734 characters starting at p contain '\0' character, don't look beyond
738 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
739 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
742 * reg-stack.c (change_stack): Change update_end var from int to
743 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
744 also call set_block_for_insn on the newly added insns and rescan.
747 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
751 * tree.c (stabilize_reference_1): Return save_expr (e) for
752 STATEMENT_LIST even if it doesn't have side-effects.
754 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
756 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
758 2018-03-12 Renlin Li <renlin.li@arm.com>
760 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
761 aarch64_output_scalar_simd_mov_immediate.
763 2018-03-12 Martin Sebor <msebor@redhat.com>
765 PR tree-optimization/83456
766 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
767 for perfectly overlapping calls to memcpy.
768 (gimple_fold_builtin_memory_chk): Same.
769 (gimple_fold_builtin_strcpy): Handle no-warning.
770 (gimple_fold_builtin_stxcpy_chk): Same.
771 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
773 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
775 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
776 parameter. Use it for SFmode.
777 (rs6000_function_arg_advance_1): Adjust.
778 (rs6000_function_arg): Adjust.
779 (rs6000_gimplify_va_arg): Pass false for that new parameter.
781 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
783 PR rtl-optimization/84169
784 PR rtl-optimization/84780
785 * combine.c (can_combine_p): Check for a 2-insn combination whether
786 the destination register is used between the two insns, too.
788 2018-03-12 Richard Biener <rguenther@suse.de>
790 PR tree-optimization/84803
791 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
792 for refs DR analysis didn't process.
794 2018-03-12 Richard Biener <rguenther@suse.de>
796 PR tree-optimization/84777
797 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
798 force-vectorize loops ignore whether we are optimizing for size.
800 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
802 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
803 (TARGET_MD_ASM_ADJUST): Define.
805 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
806 Kito Cheng <kito.cheng@gmail.com>
807 Chung-Ju Wu <jasonwucj@gmail.com>
809 * config/nds32/nds32.c (nds32_compute_stack_frame,
810 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
811 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
812 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
813 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
814 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
815 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
816 * config/nds32/nds32.md (prologue, epilogue): Use macro
817 NDS32_V3PUSH_AVAILABLE_P to do checking.
819 2018-03-11 Jakub Jelinek <jakub@redhat.com>
822 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
823 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
824 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
825 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
826 addition of most attributes on !orig_type_die or the attribute not
827 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
829 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
830 Chung-Ju Wu <jasonwucj@gmail.com>
832 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
834 * config/nds32/nds32.opt (mvh): New option.
836 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
837 Chung-Ju Wu <jasonwucj@gmail.com>
839 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
841 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
842 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
845 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
846 Chung-Ju Wu <jasonwucj@gmail.com>
848 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
850 * config/nds32/nds32-multiple.md (strlensi): New pattern.
851 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
853 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
854 Kito Cheng <kito.cheng@gmail.com>
855 Chung-Ju Wu <jasonwucj@gmail.com>
857 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
858 UNSPEC_FFMISM and UNSPEC_FLMISM.
859 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
860 for ffb, ffmism and flmism.
861 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
862 (unspec_ffmism): Ditto.
863 (unspec_flmism): Ditto.
864 (nds32_expand_builtin_impl): Check if string extension is available.
865 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
866 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
868 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
871 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
874 * lra-assigns.c (assign_by_spills): Return a flag of reload
875 assignment failure. Do not process the reload assignment
876 failures. Do not spill other reload pseudos if they has the same
878 (lra_assign): Add a return arg. Set up from the result of
879 assign_by_spills call.
880 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
881 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
882 usage_insns if it is not NULL.
883 (spill_hard_reg_in_range): New function.
884 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
885 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
887 (lra_assign): Change prototype.
888 * lra.c (lra): Add code to deal with fails by splitting hard reg
891 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
894 * config/i386/i386.opt: Replace Enforcment with Enforcement.
896 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
899 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
900 (dw_val_node): Add val_symbolic_view.
901 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
902 (symview_upper_bound): New.
903 (new_line_info_table): Initialize symviews_since_reset.
904 (dwarf2out_source_line): Count symviews_since_reset and set
906 (dw_val_equal_p): Handle symview.
907 (add_AT_symview): New.
908 (print_dw_val): Handle symview.
909 (attr_checksum, attr_checksum_ordered): Likewise.
910 (same_dw_val_p, size_of_die): Likewise.
911 (value_format, output_die): Likewise.
912 (add_high_low_attributes): Use add_AT_symview for entry_view.
913 (dwarf2out_finish): Reset symview_upper_bound, clear
916 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
919 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
920 Add strict argument and use it.
921 (rs6000_split_multireg_move): Update for new strict argument.
922 (mem_operand_gpr): Disallow all non-offsettable addresses.
923 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
925 2018-03-09 Jakub Jelinek <jakub@redhat.com>
928 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
929 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
930 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
933 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
934 decl, use remap_type if we want to use the type.
936 2018-03-09 Martin Sebor <msebor@redhat.com>
938 PR tree-optimization/84526
939 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
941 (builtin_access::generic_overlap): Be prepared to handle non-array
944 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
946 PR rtl-optimization/84682
947 * lra-constraints.c (process_address_1): Check is_address flag
948 for address constraints.
949 (process_alt_operands): Likewise.
950 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
951 preprocess_constraints.
952 * recog.h (preprocess_constraints): Add oploc parameter.
954 * recog.c (preprocess_constraints): Test address_operand for
955 CT_ADDRESS constraints.
957 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
960 * lra-assigns.c (assign_by_spills): Return a flag of reload
961 assignment failure. Do not process the reload assignment
962 failures. Do not spill other reload pseudos if they has the same
964 (lra_assign): Add a return arg. Set up from the result of
965 assign_by_spills call.
966 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
967 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
968 usage_insns if it is not NULL.
969 (spill_hard_reg_in_range): New function.
970 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
971 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
973 (lra_assign): Change prototype.
974 * lra.c (lra): Add code to deal with fails by splitting hard reg
977 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
980 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
981 Accept complain bool parameter. Only emit errors if it is true.
982 (arm_parse_cpu_option_name): Likewise.
983 (arm_target_thumb_only): Adjust callers of the above.
984 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
985 prototype to take a default true bool parameter.
986 (arm_parse_arch_option_name): Likewise.
988 2018-03-09 David Malcolm <dmalcolm@redhat.com>
989 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
993 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
994 * configure: Regenerate.
995 * configure.ac ("linker --version-script option"): New.
996 ("linker soname option"): New.
998 2018-03-09 Richard Biener <rguenther@suse.de>
1000 PR tree-optimization/84775
1001 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1002 immediate uses of predicate stmts and mark them modified.
1005 PR tree-optimization/84178
1006 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1008 (version_loop_for_if_conversion): Delay update_ssa call.
1009 (tree_if_conversion): Delay update_ssa until after predicate
1012 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1015 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1016 when the function accesses prior frames.
1018 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1021 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1022 gen_llsym, otherwise call maybe_gen_llsym.
1025 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1026 has ',' character inside of it.
1028 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1031 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1032 as clobbering CC_REGNUM.
1034 2018-03-08 Richard Biener <rguenther@suse.de>
1037 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1038 (follow_copies_to_constant): Do not follow SSA names registered
1041 2018-03-08 Richard Biener <rguenther@suse.de>
1043 PR tree-optimization/84178
1044 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1046 (version_loop_for_if_conversion): Delay update_ssa call.
1047 (tree_if_conversion): Delay update_ssa until after predicate
1050 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1052 PR tree-optimization/84178
1053 * tree-if-conv.c (release_bb_predicate): Remove the
1054 the assertion that the stmts have NULL use_ops.
1055 Discard the statements, asserting that they haven't
1056 yet been added to a BB.
1058 2018-03-08 Richard Biener <rguenther@suse.de>
1060 PR tree-optimization/84746
1061 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1062 (phi_translate): Pass in destination ANTIC_OUT set.
1063 (phi_translate_1): Likewise. For a simplified result lookup
1064 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1065 (phi_translate_set): Adjust.
1066 (do_pre_regular_insertion): Likewise.
1067 (do_pre_partial_partial_insertion): Likewise.
1069 2018-03-08 Martin Liska <mliska@suse.cz>
1071 PR gcov-profile/84735
1072 * doc/gcov.texi: Document usage of profile files.
1073 * gcov-io.h: Document changes in the format.
1075 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1079 * dwarf2out.c (struct dw_line_info_table): Update comments for
1081 (FORCE_RESET_NEXT_VIEW): New.
1082 (FORCE_RESETTING_VIEW_P): New.
1083 (RESETTING_VIEW_P): Check for -1 too.
1084 (ZERO_VIEW_P): Likewise.
1085 (new_line_info_table): Force-reset next view.
1086 (dwarf2out_begin_function): Likewise.
1087 (dwarf2out_source_line): Simplify zero_view_p initialization.
1088 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1089 view directly. Omit view when omitting .loc at line 0.
1091 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1093 PR tree-optimization/84740
1094 * tree-switch-conversion.c (process_switch): Call build_constructors
1095 only if info.phi_count is non-zero.
1097 PR tree-optimization/84739
1098 * tree-tailcall.c (find_tail_calls): Check call arguments against
1099 DECL_ARGUMENTS (current_function_decl) rather than
1100 DECL_ARGUMENTS (func) when checking for tail recursion.
1102 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1104 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1105 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1106 Volker Reichelt's entry and add entries for people that perform
1107 GCC fuzzy testing and report numerous bugs.
1109 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1112 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1113 readonly data in sdata, if that is disabled.
1114 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1115 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1116 -mreadonly-in-sdata option.
1118 2018-03-07 Martin Sebor <msebor@redhat.com>
1120 PR tree-optimization/84468
1121 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1122 basic block when looking for nul assignment.
1124 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1127 * except.h (output_function_exception_table): Adjust prototype.
1128 * except.c (output_function_exception_table): Remove FNNAME parameter
1129 and add SECTION parameter. Ouput one part of the table at a time.
1130 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1131 the first part of the exception table and emit unwind directives.
1132 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1133 (i386_pe_seh_cold_init): Likewise.
1134 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1135 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1136 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1137 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1138 * config/i386/winnt.c: Include except.h.
1139 (struct seh_frame_state): Add reg_offset, after_prologue and
1140 in_cold_section fields.
1141 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1142 (i386_pe_seh_cold_init): New function.
1143 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1144 to seh->in_cold_section.
1145 (seh_emit_push): Record the offset of the push.
1146 (seh_emit_save): Record the offet of the save.
1147 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1148 Test seh->after_prologue to disregard the epilogue.
1149 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1150 (i386_pe_end_cold_function): New function.
1152 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1155 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1156 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1159 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1161 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1162 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1165 * multiple_target.c: Include tree-inline.h and intl.h.
1166 (expand_target_clones): Diagnose and fail if node->definition and
1167 !tree_versionable_function_p (node->decl).
1169 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1171 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1173 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1174 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1175 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1177 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1180 * combine.c (try_combine): Use reg_or_subregno instead of handling
1181 just paradoxical SUBREGs and REGs.
1183 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1185 * config/arc/arc.c (arc_finalize_pic): Remove function.
1186 (arc_must_save_register): We use single base PIC register, remove
1187 checks to save/restore the PIC register.
1188 (arc_expand_prologue): Likewise.
1189 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1191 (arc_verify_short): Likewise.
1192 (arc_attr_type): Likewise.
1193 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1194 (walk_stores): Likewise.
1195 (arc_address_cost): Make it static.
1196 (arc_verify_short): Likewise.
1197 (branch_dest): Likewise.
1198 (arc_attr_type): Likewise.
1199 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1200 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1201 (arc_final_prescan_insn): Remove inserting the nops due to
1202 hardware hazards. It is done in reorg step.
1203 (insn_length_variant_t): Remove.
1204 (insn_length_parameters_t): Likewise.
1205 (arc_insn_length_parameters): Likewise.
1206 (arc_get_insn_variants): Likewise.
1207 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1209 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1212 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1215 PR tree-optimization/84687
1216 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1218 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1220 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1222 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1223 Rename to ppc_speculation_barrier.
1224 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1225 __builtin_ppc_speculation_barrier.
1227 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1230 * combine.c (combine_simplify_rtx): Don't try to simplify if
1231 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1234 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1236 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1237 to 32 bytes when compiling for POWER9.
1239 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1242 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1243 regparm >= 3 with no arg reg available also for calls with
1244 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1247 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1249 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1251 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1254 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1256 2018-03-05 Richard Biener <rguenther@suse.de>
1258 PR tree-optimization/84486
1259 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1260 When inserting a __builtin_assume_aligned call set the LHS
1261 SSA name alignment info accordingly.
1263 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1265 PR tree-optimization/84114
1266 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1267 Avoid reassociation of FLOAT_MODE addition.
1269 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1271 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1272 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1273 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1274 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1275 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1276 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1277 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1279 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1280 __builtin_ia32_wbinvd): New builtins.
1281 (SPECIAL_ARGS2): New.
1282 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1283 (SPECIAL_ARGS2): New.
1284 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1285 (ix86_valid_target_attribute_inner_p): Ditto.
1286 (ix86_init_mmx_sse_builtins): Add special_args2.
1287 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1288 TARGET_WBNOINVD_P): New.
1289 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1290 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1291 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1292 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1293 * config/i386/pconfigintrin.h: New file.
1294 * config/i386/wbnoinvdintrin.h: Ditto.
1295 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1296 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1298 2018-03-05 Richard Biener <rguenther@suse.de>
1300 PR tree-optimization/84670
1301 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1303 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1304 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1305 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1306 doesn't grow if all successors (recursively) were visited at least
1309 2018-03-05 Richard Biener <rguenther@suse.de>
1311 PR tree-optimization/84650
1312 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1313 if executed in the loop pipeline.
1315 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1317 * doc/configfiles.texi (Configuration Files): Move info about
1318 conditionalizing $target-protos.h to...
1319 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1320 differs from $target-protos.h.
1322 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1323 Chung-Ju Wu <jasonwucj@gmail.com>
1325 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1326 * config/nds32/nds32-multiple.md (setmemsi): Define.
1327 * config/nds32/nds32-memory-manipulation.c
1328 (nds32_gen_dup_4_byte_to_word_value): New.
1329 (emit_setmem_word_loop): New.
1330 (emit_setmem_byte_loop): New.
1331 (nds32_expand_setmem_loop): New.
1332 (nds32_expand_setmem_loop_v3m): New.
1333 (nds32_expand_setmem_unroll): New.
1334 (nds32_expand_setmem): New.
1336 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1337 Chung-Ju Wu <jasonwucj@gmail.com>
1339 * config/nds32/nds32-memory-manipulation.c
1340 (nds32_emit_load_store): New.
1341 (nds32_emit_post_inc_load_store): New.
1342 (nds32_emit_mem_move): New.
1343 (nds32_emit_mem_move_block): New.
1344 (nds32_expand_movmemsi_loop_unknown_size): New.
1345 (nds32_expand_movmemsi_loop_known_size): New.
1346 (nds32_expand_movmemsi_loop): New.
1347 (nds32_expand_movmemsi_unroll): New.
1348 (nds32_expand_movmemqi): Rename ...
1349 (nds32_expand_movmemsi): ... to this.
1350 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1351 (movmemsi): ... to this.
1352 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1353 (nds32_expand_movmemsi): ... to this.
1355 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1356 Monk Chiang <sh.chiang04@gmail.com>
1357 Chung-Ju Wu <jasonwucj@gmail.com>
1359 * config/nds32/nds32-protos.h
1360 (nds32_expand_load_multiple): New arguments.
1361 (nds32_expand_store_multiple): Ditto.
1362 (nds32_valid_multiple_load_store): Rename ...
1363 (nds32_valid_multiple_load_store_p): ... to this.
1364 * config/nds32/nds32-memory-manipulation.c
1365 (nds32_expand_load_multiple): Refine implementation.
1366 (nds32_expand_store_multiple): Ditto.
1367 * config/nds32/nds32-multiple.md
1368 (load_multiple): Update nds32_expand_load_multiple interface.
1369 (store_multiple): Update nds32_expand_store_multiple interface.
1370 * config/nds32/nds32-predicates.c
1371 (nds32_valid_multiple_load_store): Rename ...
1372 (nds32_valid_multiple_load_store_p): ... to this and refine
1374 * config/nds32/predicates.md
1375 (nds32_load_multiple_and_update_address_operation): New predicate.
1376 (nds32_store_multiple_and_update_address_operation): New predicate.
1378 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1379 Chung-Ju Wu <jasonwucj@gmail.com>
1381 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1382 (combo): New attribute.
1383 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1385 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1387 * config/nds32/nds32.opt: Change -mcmodel= default value.
1389 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1390 Monk Chiang <sh.chiang04@gmail.com>
1391 Chung-Ju Wu <jasonwucj@gmail.com>
1393 * config/nds32/constants.md (unspec_element): New enum.
1394 * config/nds32/constraints.md (Umw): New constraint.
1395 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1396 * config/nds32/nds32-intrinsic.md: Likewise.
1397 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1398 (nds32_valid_smw_lwm_base_p): New.
1399 (nds32_output_smw_single_word): New.
1400 (nds32_output_lmw_single_word): New.
1401 (nds32_expand_unaligned_load): New.
1402 (nds32_expand_unaligned_store): New.
1403 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1404 (nds32_output_smw_single_word): Declare.
1405 (nds32_output_lmw_single_word): Declare.
1406 (nds32_expand_unaligned_load): Declare.
1407 (nds32_expand_unaligned_store): Declare.
1408 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1409 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1410 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1411 NDS32_BUILTIN_UASTORE_DW.
1412 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1415 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1416 Kito Cheng <kito.cheng@gmail.com>
1417 Chung-Ju Wu <jasonwucj@gmail.com>
1419 * config/nds32/nds32-intrinsic.c
1420 (nds32_expand_builtin_null_ftype_reg): Delete.
1421 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1422 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1423 (nds32_read_argument): New.
1424 (nds32_legitimize_target): Ditto.
1425 (nds32_legitimize_argument): Ditto.
1426 (nds32_check_constant_argument): Ditto.
1427 (nds32_expand_unop_builtin): Ditto.
1428 (nds32_expand_unopimm_builtin): Ditto.
1429 (nds32_expand_binop_builtin): Ditto.
1430 (nds32_builtin_decl_impl): Ditto.
1431 (builtin_description): Ditto.
1432 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1433 (nds32_init_builtins_impl): Ditto.
1434 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1435 (nds32_builtin_decl): New.
1436 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1437 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1439 2018-03-02 Jeff Law <law@redhat.com>
1441 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1442 (redundant_insn, fill_simple_delay_slots): Likewise.
1443 (fill_slots_from_thread): Likewise.
1444 * resource.c (mark_referenced_resources): Likewise.
1445 (mark_set_resources, find_dead_or_set_registers): Likewise.
1447 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1449 * substring-locations.h (format_warning_va): Formatting fix for
1451 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1453 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1454 * substring-locations.c: Include intl.h.
1455 (format_warning_va): Turned into small wrapper around
1456 format_warning_n_va, renamed to ...
1457 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1458 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1460 (format_warning_at_substring_n): New function.
1461 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1462 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1463 format_warning_at_substring with just a shorter name instead of
1464 const function pointer.
1465 (fmtwarn_n): New function.
1466 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1467 appropriate, get rid of all the fmtstr temporaries, move conditionals
1468 with G_() wrapped string literals directly into fmtwarn arguments,
1469 cast dir.len to (int), formatting fixes.
1471 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1473 * doc/invoke.texi: Remove "Cilk Plus" references.
1475 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1476 Richard Biener <rguenther@suse.de>
1479 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1480 for error or warning attributes if CALL_FROM_THUNK_P is set.
1483 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1486 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1487 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1490 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1491 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1493 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1494 instead of -1U in last predictors element's probability member.
1496 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1499 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1500 arguments if they are comparable.
1502 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1504 PR tree-optimization/84634
1505 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1506 masks and masked_loop_p with a single loop_masks, making sure it's
1507 null for bb vectorization.
1509 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1511 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1512 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1513 loop->force_vectorize to check whether there is no alias.
1515 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1518 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1520 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1522 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1523 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1524 instead of a loop around prev_real_insn.
1525 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1529 * config/i386/i386.c (ix86_print_operand): Use conditional
1530 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1533 2018-03-02 Richard Biener <rguenther@suse.de>
1535 PR tree-optimization/84427
1536 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1537 (bitmap_set_subtract_values): Rewrite to handle multiple
1540 (prune_clobbered_mems): Likewise.
1541 (phi_translate): Take edge instead of pred/phiblock.
1542 (phi_translate_1): Likewise.
1543 (phi_translate_set): Likewise. Insert all translated
1544 exprs for a value into the set, keeping possibly multiple
1545 expressions per value.
1546 (compute_antic_aux): Adjust for phi_translate changes.
1547 When intersecting union the expressions and prune those
1548 not in the final value set, keeping possibly multiple
1549 expressions per value. Do not use value-insertion
1550 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1551 all expressions. Add verification that the value-sets
1552 only shrink during iteration.
1553 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1554 (do_pre_regular_insertion): Likewise.
1555 (do_pre_partial_partial_insertion): Likewise.
1557 2018-03-02 Richard Biener <rguenther@suse.de>
1560 * config/darwin.c (saved_debug_info_level): New static global.
1561 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1562 (darwin_asm_lto_end): Restore debug info generation settings.
1564 2018-03-01 Martin Liska <mliska@suse.cz>
1567 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1570 2018-03-01 Richard Biener <rguenther@suse.de>
1573 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1574 type annotation with LTO when debug was disabled at compile-time.
1576 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1578 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1580 (mips_final_postscan_insn): Likewise.
1582 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1584 PR rtl-optimization/84528
1585 * alias.c (init_alias_target): Add commentary.
1586 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1587 a unique base value if the frame pointer is not eliminated
1588 to the stack pointer.
1590 2018-03-01 Tom de Vries <tom@codesourcery.com>
1592 PR rtl-optimization/83327
1593 * lra-int.h (hard_regs_spilled_into): Declare.
1594 * lra.c (hard_regs_spilled_into): Define.
1595 (init_reg_info): Init hard_regs_spilled_into.
1596 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1597 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1598 (process_bb_lives): Handle hard_regs_spilled_into.
1599 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1600 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1602 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1604 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1605 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1606 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1607 * config/rs6000/aix72.h: New file.
1609 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1611 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1612 instead of warning_at with conditional singular and plural messages
1616 * stor-layout.c (update_alignment_for_field): For
1617 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1618 && !DECL_PACKED (field), do the alignment update, just use
1619 only desired_align instead of MAX (type_align, desired_align)
1621 (place_field): Don't do known_align < desired_align handling
1622 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1623 is non-NULL, instead do it after rli->prev_field handling and
1624 only if not within a bitfield word. For DECL_PACKED (field)
1625 use type_align of BITS_PER_UNIT.
1627 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1629 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1630 superfluous parentheses and trailing spaces.
1632 2018-02-28 Richard Biener <rguenther@suse.de>
1634 PR tree-optimization/84584
1635 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1636 SCoPs with fake exit edge.
1638 2018-02-28 Martin Liska <mliska@suse.cz>
1641 * timevar.c (timer::print): Fix format to properly print 100%
1644 2018-02-28 Richard Biener <rguenther@suse.de>
1647 * genmatch.c (capture_info::walk_match): Do not mark
1648 captured expressions without operands as expr_p given
1649 they act more like predicates and should be subject to
1650 "lost tail" side-effect preserving.
1652 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1654 PR rtl-optimization/81611
1655 * auto-inc-dec.c (attempt_change): Move dead note from
1656 mem_insn if it's the next use of regno
1657 (find_address): Take address use of reg holding
1658 non-incremented value. Add parm to limit search to the named
1660 (merge_in_block): Attempt to use a mem insn that is the next
1661 use of the original regno.
1663 2018-02-27 Martin Sebor <msebor@redhat.com>
1666 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1667 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1669 2018-02-27 Martin Sebor <msebor@redhat.com>
1671 PR translation/84207
1672 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1673 n argument to unsigned HOST_WIDE_INT.
1674 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1675 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1676 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1677 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1679 2018-02-27 Richard Biener <rguenther@suse.de>
1681 PR tree-optimization/84512
1682 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1683 Do not use the estimate returned from record_stmt_cost for
1684 the scalar iteration cost but sum properly using add_stmt_cost.
1686 2018-02-27 Richard Biener <rguenther@suse.de>
1688 PR tree-optimization/84466
1689 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1690 Adjust last change to less strictly validate use operands.
1692 2018-02-27 Martin Liska <mliska@suse.cz>
1694 PR gcov-profile/84548
1695 * gcov.c (process_file): Allow partial overlap and consider it
1696 also as group functions.
1697 (output_lines): Properly calculate range of lines for a group.
1699 2018-02-27 Martin Liska <mliska@suse.cz>
1701 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1702 'ggc' suffixes. Change first column width.
1703 (timer::print): Fix formatting of the column.
1705 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1707 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1708 preserve inline entry blocks for the sake of debug inline
1709 entry point markers alone.
1710 (remove_unused_locals): Suggest in comments a better place to
1711 force the preservation of inline entry blocks that are
1712 otherwise unused, but do not preserve them.
1714 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1716 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1718 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1721 * config/i386/constraints.md (Bs): Replace
1722 ix86_indirect_branch_register with
1723 TARGET_INDIRECT_BRANCH_REGISTER.
1725 * config/i386/i386.md (indirect_jump): Likewise.
1726 (tablejump): Likewise.
1727 (*sibcall_memory): Likewise.
1728 (*sibcall_value_memory): Likewise.
1729 Peepholes of indirect call and jump via memory: Likewise.
1730 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1731 (*sibcall_value_GOT_32): Likewise.
1732 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1733 (GOT_memory_operand): Likewise.
1734 (call_insn_operand): Likewise.
1735 (sibcall_insn_operand): Likewise.
1736 (GOT32_symbol_operand): Likewise.
1737 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1739 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1741 PR rtl-optimization/83496
1742 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1743 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1744 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1745 redundant insn, if any.
1746 (relax_delay_slots): Likewise.
1747 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1749 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1751 PR tree-optimization/83965
1752 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1753 that grouped statements are part of a reduction chain. Return
1754 true if the statement is not marked as a reduction itself but
1756 (vect_recog_dot_prod_pattern): Don't check whether the statement
1757 is part of a group here.
1758 (vect_recog_sad_pattern): Likewise.
1759 (vect_recog_widen_sum_pattern): Likewise.
1761 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1764 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1766 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1769 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1771 (ix86_output_indirect_function_return): New prototype.
1772 (ix86_split_simple_return_pop_internal): Likewise.
1773 * config/i386/i386.c (indirect_return_via_cx): New.
1774 (indirect_return_via_cx_bnd): Likewise.
1775 (indirect_thunk_name): Handle return va CX_REG.
1776 (output_indirect_thunk_function): Create alias for
1777 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1778 (ix86_output_indirect_jmp): Remove the bool argument.
1779 (ix86_output_indirect_function_return): New function.
1780 (ix86_split_simple_return_pop_internal): Likewise.
1781 * config/i386/i386.md (*indirect_jump): Don't pass false
1782 to ix86_output_indirect_jmp.
1783 (*tablejump_1): Likewise.
1784 (simple_return_pop_internal): Change it to define_insn_and_split.
1785 Call ix86_split_simple_return_pop_internal to split it for
1787 (simple_return_indirect_internal): Call
1788 ix86_output_indirect_function_return instead of
1789 ix86_output_indirect_jmp.
1791 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1794 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1795 memset and value initialization afterwards.
1797 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1799 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1801 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1804 * common/config/aarch64/aarch64-common.c
1805 (aarch_option_optimization_table[]): Switch
1806 off fomit-frame-pointer
1808 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1809 Chung-Ju Wu <jasonwucj@gmail.com>
1811 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1813 (store_multiple): Ditto.
1815 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1817 * config.gcc: Add --with-cpu support for nds32 target.
1818 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1819 * config/nds32/nds32.opt: Add -mcpu= option.
1821 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1823 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1824 isel=yes): Warn for these deprecated options.
1826 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1828 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1829 ISA_2_5_MASKS_EMBEDDED.
1831 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1833 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1834 p->max as pointers rather than using iterative_hash_expr.
1836 2018-02-23 Carl Love <cel@us.ibm.com>
1838 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1839 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1841 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1842 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1843 P8V_BUILTIN_VEC_VUNSIGNED2.
1845 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1848 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1849 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1851 * lra-constraints.c (curr_insn_transform): Set up
1852 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1854 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1855 * lra-eliminations.c (spill_pseudos): Ditto.
1856 (process_insn_for_elimination): Ditto.
1857 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1858 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1859 LRA_NON_CLOBBERED_ALT.
1861 2018-02-22 Martin Sebor <msebor@redhat.com>
1863 PR tree-optimization/84480
1864 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1865 to maybe_diag_stxncpy_trunc. Call it.
1866 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1867 from gimple_fold_builtin_strcpy. Print inlining stack.
1868 (handle_builtin_stxncpy): Print inlining stack.
1869 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1871 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1874 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1875 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1876 and -fcheck-pointer-bounds are used together.
1877 (indirect_thunk_prefix): New enum.
1878 (indirect_thunk_need_prefix): New function.
1879 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1880 "_nt" instead of "_bnd" for NOTRACK prefix.
1881 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1882 (output_indirect_thunk_function): Likewise.
1884 (ix86_code_end): Update output_indirect_thunk_function calls.
1885 (ix86_output_indirect_branch_via_reg): Replace
1886 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1887 (ix86_output_indirect_branch_via_push): Likewise.
1888 (ix86_output_function_return): Likewise.
1889 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1890 incompatible with -fcf-protection=branch and
1891 -fcheck-pointer-bounds.
1893 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1896 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1897 Change gcc_assert call to output_operand_lossage.
1899 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1901 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1903 2018-02-22 DJ Delorie <dj@redhat.com>
1904 Sebastian Perta <sebastian.perta@renesas.com>
1905 Oleg Endo <olegendo@gcc.gnu.org>
1907 * config/rx/rx.c (rx_rtx_costs): New function.
1908 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1910 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1912 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1914 2018-02-22 Martin Liska <mliska@suse.cz>
1917 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1918 Add "native" as a possible value.
1920 2018-02-22 Martin Liska <mliska@suse.cz>
1923 * config/i386/i386.c (ix86_option_override_internal):
1924 Add "native" as a possible value for -march and -mtune.
1926 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1929 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1930 to all type variants.
1932 PR tree-optimization/84503
1933 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1934 width as info->bitpos + info->bitsize - start.
1935 (merged_store_group::merge_overlapping): Simplify width computation.
1936 (check_no_overlap): New function.
1937 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1938 start + width and last_order of the group, fail if check_no_overlap
1940 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1941 to group if check_no_overlap fails.
1943 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1945 * config/rs6000/altivec.md: Delete contraint arguments to
1946 define_expand, define_split, and define_peephole2, and in
1947 define_insn_and_split if always unused.
1948 * config/rs6000/darwin.md: Ditto.
1949 * config/rs6000/dfp.md: Ditto.
1950 * config/rs6000/rs6000.md: Ditto.
1951 * config/rs6000/sync.md: Ditto.
1952 * config/rs6000/vector.md: Ditto.
1953 * config/rs6000/vsx.md: Ditto.
1955 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1957 * config/rs6000/altivec.md: Write output control strings as braced
1958 blocks instead of double-quoted strings.
1959 * config/rs6000/darwin.md: Ditto.
1960 * config/rs6000/rs6000.md: Ditto.
1961 * config/rs6000/vector.md: Ditto.
1962 * config/rs6000/vsx.md: Ditto.
1964 2018-02-21 Jason Merrill <jason@redhat.com>
1966 PR c++/84314 - ICE with templates and fastcall attribute.
1967 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1969 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1971 * ipa-cp.c (determine_versionability): Fix comment typos.
1973 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1976 * ipa-cp.c (determine_versionability): Do not version functions caling
1979 2018-02-21 Martin Liska <mliska@suse.cz>
1982 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1983 Add "native" as a possible value.
1984 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1985 the macro when native cpu detection is available.
1987 2018-02-21 Martin Liska <mliska@suse.cz>
1990 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1991 Add "native" as a possible value.
1992 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1993 when native cpu detection is available.
1995 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1996 Martin Sebor <msebor@redhat.com>
1998 PR tree-optimization/84478
1999 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2001 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2002 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2003 support which is conservatively correct, for 2 only stay conservative
2004 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2005 argument to the 2 argument get_range_strlen, adjust 6 arg
2006 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2008 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2009 (gimple_fold_builtin_strlen): Pass true as last argument to
2012 2018-02-20 Martin Sebor <msebor@redhat.com>
2015 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2016 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2017 (builtin_memref::builtin_memref): Factor out parts into
2018 set_base_and_offset and call it.
2020 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2023 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2024 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2025 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2026 search at the associated MODE_INT.
2028 2018-02-20 Jeff Law <law@redhat.com>
2031 PR tree-optimization/81592
2033 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2034 for range data rather than using global data.
2035 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2036 range data rather than using global data.
2037 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2038 pass it to children as needed.
2039 (struct directive::fmtresult): Similarly.
2040 (struct directive::set_width): Similarly.
2041 (struct directive::set_precision): Similarly.
2042 (format_integer, format_directive, parse_directive): Similarly.
2043 (format_none): Accept unnamed vr_values parameter.
2044 (format_percent, format_floating, format_character): Similarly.
2045 (format_string, format_plain): Similarly.
2046 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2047 the EVRP range analyzer for range data rather than using global data.
2048 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2049 gimple-ssa-evrp-analyze.h
2050 (class sprintf_dom_walker): Add after_dom_children member function.
2051 Add evrp_range_analyzer member.
2052 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2053 range analyzer as needed.
2054 (sprintf_dom_walker::after_dom_children): New member function.
2055 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2057 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2058 (evrp_range_analyzer::pop_to_marker): Likewise.
2060 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2062 PR tree-optimization/84419
2063 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2064 with the required type if its current type is compatible but
2067 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2070 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2071 after vectorization.
2073 2018-02-20 Martin Liska <mliska@suse.cz>
2076 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2077 possible values if we don't have a hint.
2079 2018-02-20 Martin Liska <mliska@suse.cz>
2083 * final.c (shorten_branches): Build align_tab array with one
2085 * opts.c (finish_options): Add alignment option limit check.
2086 (MAX_CODE_ALIGN): Likewise.
2087 (MAX_CODE_ALIGN_VALUE): Likewise.
2088 * doc/invoke.texi: Document maximum allowed option value for
2089 all -falign-* options.
2091 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2094 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2095 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2096 * var-tracking.c (emit_note_insn_var_location): Remove all references
2097 to NOTE_INSN_CALL_ARG_LOCATION.
2098 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2099 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2100 Use copy_rtx_if_shared.
2101 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2102 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2103 (dwarf2out_var_location): Remove handling of
2104 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2106 * final.c (final_scan_insn): Remove all references to
2107 NOTE_INSN_CALL_ARG_LOCATION.
2108 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2109 before dumping final insns.
2110 * except.c (emit_note_eh_region_end): Remove all references to
2111 NOTE_INSN_CALL_ARG_LOCATION.
2112 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2113 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2114 * config/arc/arc.c (hwloop_optimize): Likewise.
2115 * config/arm/arm.c (create_fix_barrier): Likewise.
2116 * config/s390/s390.c (s390_chunkify_start): Likewise.
2117 * config/sh/sh.c (find_barrier): Likewise.
2118 * config/i386/i386.c (rest_of_insert_endbranch,
2119 ix86_seh_fixup_eh_fallthru): Likewise.
2120 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2121 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2122 * config/frv/frv.c (frv_function_prologue): Likewise.
2123 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2125 (note_outside_basic_block_p): Remove all references to
2126 NOTE_INSN_CALL_ARG_LOCATION.
2127 * gengtype.c (adjust_field_rtx_def): Likewise.
2128 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2130 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2131 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2134 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2137 PR tree-optimization/84452
2138 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2139 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2142 2018-02-19 Martin Liska <mliska@suse.cz>
2145 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2147 2018-02-19 Martin Liska <mliska@suse.cz>
2148 Richard Sandiford <richard.sandiford@linaro.org>
2150 PR tree-optimization/82491
2151 * gimple-fold.c (get_base_constructor): Make earlier bail out
2154 2018-02-19 Carl Love <cel@us.ibm.com>
2156 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2157 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2159 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2160 P8V_BUILTIN_VEC_NEG.
2162 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2164 * config/rl78/rl78.md (movdf): New define expand.
2166 2018-02-19 Martin Liska <mliska@suse.cz>
2169 * doc/invoke.texi: Fix typo.
2170 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2172 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2174 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2175 handle rs6000_single_float and rs6000_double_float specially for
2178 2018-02-16 Jeff Law <law@redhat.com>
2180 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2183 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2186 * ipa-inline.c (inline_small_functions): Fix a typo.
2188 2018-02-16 Nathan Sidwell <nathan@acm.org>
2190 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2192 2018-02-16 Carl Love <cel@us.ibm.com>
2194 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2195 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2196 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2197 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2198 expansion to P8V_BUILTIN_VEC_FLOAT2.
2200 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2202 PR rtl-optimization/70023
2203 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2204 src_regno into account.
2206 2018-02-16 Carl Love <cel@us.ibm.com>
2208 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2209 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2210 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2211 * config/rs6000/rs6000.c: Remove case statements for
2212 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2213 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2214 and P9V_BUILTIN_VEC_VINSERT4B.
2215 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2216 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2217 * config/rs6000/vsx.md:
2218 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2221 2018-02-16 Carl Love <cel@us.ibm.com>
2223 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2225 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2227 * config/rs6000/rs6000-c.c: Add the definitions for
2228 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2229 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2230 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2231 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2232 definition for insert4b and define insn *insert3b_internal.
2233 * doc/extend.texi: Add documentation for vec_extract4b.
2235 2018-02-16 Nathan Sidwell <nathan@acm.org>
2237 * doc/extend.texi (Backwards Compatibility): Mention friend
2238 injection. Note for-scope is deprecated.
2239 * doc/invoke.texi (-ffriend-injection): Deprecate.
2241 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2243 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2244 that moved to I2, also allow destinations that are a paradoxical
2245 subreg (instead of a normal reg).
2247 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2250 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2253 2018-02-16 Richard Biener <rguenther@suse.de>
2255 PR tree-optimization/84037
2256 PR tree-optimization/84016
2258 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2259 Adjust vec_construct for the fact we need additional higher latency
2260 128bit inserts for AVX256 and AVX512 vector builds.
2261 (ix86_add_stmt_cost): Scale vector construction cost for
2264 2018-02-16 Richard Biener <rguenther@suse.de>
2266 PR tree-optimization/84417
2267 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2268 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2269 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2271 2018-02-16 Martin Liska <mliska@suse.cz>
2274 * internal-fn.def (ASAN_CHECK): Set proper flags.
2275 (ASAN_MARK): Likewise.
2277 2018-02-16 Julia Koval <julia.koval@intel.com>
2279 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2280 from PTA_CANNONLAKE.
2282 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2285 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2286 Use ++iter rather than iter++ for std::list iterators.
2287 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2288 defer deleting them until all nodes in the forest are processed. Do
2289 free even leaf nodes. Change to_process into auto_vec.
2292 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2293 * vec.h (vec_default_construct): Use memset instead of placement new
2294 if BROKEN_VALUE_INITIALIZATION is defined.
2295 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2296 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2299 PR rtl-optimization/83723
2300 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2301 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2302 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2304 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2306 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2308 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2310 PR rtl-optimization/81443
2311 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2312 from inner REGs to paradoxical SUBREGs.
2314 2018-02-16 Richard Biener <rguenther@suse.de>
2316 PR tree-optimization/84399
2317 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2318 For operands we can analyze at their definition make sure we can
2319 analyze them at each use as well.
2321 2018-02-16 Richard Biener <rguenther@suse.de>
2323 PR tree-optimization/84190
2324 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2325 volatile accesses if the decl isn't volatile.
2327 2018-02-15 Jason Merrill <jason@redhat.com>
2329 PR c++/84314 - ICE with templates and fastcall attribute.
2330 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2331 TYPE_CANONICAL on an existing type.
2333 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2335 PR tree-optimization/84383
2336 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2337 dstoff nor call operand_equal_p if dstbase is NULL.
2339 PR tree-optimization/84334
2340 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2341 also a CONSTANT_CLASS_P, punt.
2343 2018-02-14 Jim Wilson <jimw@sifive.com>
2345 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2346 first SMALL_OPERAND check. New local min_second_step. Move assert
2347 to where locals are set. Add TARGET_RVC support.
2348 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2350 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2352 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2354 2018-02-14 Martin Sebor <msebor@redhat.com>
2356 PR tree-optimization/83698
2357 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2358 arrays constrain the offset range to their bounds.
2359 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2360 (builtin_access::overlap): Avoid setting the size of overlap if it's
2362 (maybe_diag_overlap): Also consider arrays when deciding what values
2363 of offsets to include in diagnostics.
2365 2018-02-14 Martin Sebor <msebor@redhat.com>
2368 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2369 that correspond to the kind of a declaration.
2371 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2374 * config/pa/pa.md: Load address of PIC label using the linkage table
2375 if the label is nonlocal.
2377 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2379 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2380 warning message if user requests -maltivec=be.
2381 * doc/invoke.texi: Document deprecation of -maltivec=be.
2383 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2386 * config/rs6000/rs6000-c.c: Update definitions for
2387 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2388 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2390 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2393 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2394 add _get_ssp intrinsics. Remove argument from
2395 __builtin_ia32_rdssp[d|q].
2396 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2397 * config/i386/i386-builtin.def: Remove argument from
2398 __builtin_ia32_rdssp[d|q].
2399 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2400 ix86_expand_special_args_builtin for _rdssp[d|q].
2401 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2402 Clear register before usage.
2403 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2404 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2406 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2408 PR tree-optimization/84357
2409 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2410 operand 1 of an ARRAY_REF too.
2412 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2415 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2416 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2418 (set_of_reg): New struct.
2419 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2420 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2421 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2423 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2424 Split into bitclr, bitset, bitinvert patterns if appropriate.
2425 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2426 use rx_fuse_in_memory_bitop.
2427 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2428 to named insn, correct maximum insn length.
2430 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2433 * machmode.def: Define a complex mode for PARTIAL_INT.
2434 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2436 * doc/rtl.texi: Document CSPImode.
2437 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2439 (msp430_hard_regno_nregs_with_padding): Likewise.
2441 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2444 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2446 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2448 PR rtl-optimization/84169
2449 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2450 we generated a parallel as new i3 and we split that to new i2 and i3
2451 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2452 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2453 those to i2, not i1. Partially rewrite this scan code.
2455 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2458 * stor-layout.c (place_field): For variable length fields, adjust
2459 offset_align afterwards not just based on the field's alignment,
2460 but also on the size.
2463 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2464 of exps and logs in the use_exp2 case.
2466 2018-02-13 Jeff Law <law@redhat.com>
2468 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2471 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2474 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2478 * common.opt (gas-loc-support, gas-locview-support): New.
2479 (ginline-points, ginternal-reset-location-views): New.
2480 * doc/invoke.texi: Document them. Use @itemx where intended.
2481 (gvariable-location-views): Adjust.
2482 * target.def (reset_location_view): New.
2483 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2484 (TARGET_RESET_LOCATION_VIEW): New.
2485 * doc/tm.texi: Rebuilt.
2486 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2487 (dwarf2out_default_as_locview_support): New.
2488 (output_asm_line_debug_info): Use option variables.
2489 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2490 (output_loc_list): Likewise.
2491 (add_high_low_attributes): Check option variables.
2492 Don't output entry view attribute in strict mode.
2493 (gen_inlined_subroutine_die): Check option variables.
2494 (dwarf2out_inline_entry): Likewise.
2495 (init_sections_and_labels): Likewise.
2496 (dwarf2out_early_finish): Likewise.
2497 (maybe_reset_location_view): New, from...
2498 (dwarf2out_var_location): ... here. Call it.
2499 * debug.h (dwarf2out_default_as_loc_support): Declare.
2500 (dwarf2out_default_as_locview_support): Declare.
2501 * hooks.c (hook_int_rtx_insn_0): New.
2502 * hooks.h (hook_int_rtx_insn_0): Declare.
2503 * toplev.c (process_options): Take -gas-loc-support and
2504 -gas-locview-support from dwarf2out. Enable
2505 -gvariable-location-views by default only with locview
2506 assembler support. Enable -ginternal-reset-location-views by
2507 default only if the target defines the corresponding hook.
2508 Enable -ginline-points by default if location views are
2509 enabled; force it disabled if statement frontiers are
2511 * tree-inline.c (expand_call_inline): Check option variables.
2512 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2514 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2516 PR tree-optimization/84321
2517 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2518 handling. Also check whether the anti-range contains any values
2519 that satisfy the mask; switch to a VR_RANGE if not.
2521 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2524 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2526 2018-02-13 Martin Jambor <mjambor@suse.cz>
2529 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2530 of call statements, also set location of a load to a temporary.
2532 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2534 * config/rl78/rl78.c (add_vector_labels): New function.
2535 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2536 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2537 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2538 which checks that no arguments are passed.
2539 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2540 * doc/extend.texi: Documentation for the new attribute.
2542 2018-02-13 Andreas Schwab <schwab@suse.de>
2544 * config/riscv/linux.h (CPP_SPEC): Define.
2546 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2549 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2550 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2551 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2552 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2553 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2554 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2555 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2556 temporarily for AES and PCLMUL builtins.
2558 PR tree-optimization/84339
2559 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2560 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2564 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2565 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2566 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2568 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2570 * omp-simd-clone.h: New file.
2571 * omp-simd-clone.c: Include omp-simd-clone.h.
2572 (expand_simd_clones): No longer static.
2573 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2574 cgraph.h and omp-simd-clone.h.
2575 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2576 (vect_recog_widen_shift_pattern): Formatting fix.
2577 (vect_pattern_recog_1): Don't check optab for calls.
2580 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2581 operands[2] into a REG before using gen_lowpart on it.
2583 2018-02-12 Jeff Law <law@redhat.com>
2586 * config/sh/sh.c (find_barrier): Consider a sibling call
2589 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2590 successfully back substituting a reg.
2592 2018-02-12 Richard Biener <rguenther@suse.de>
2594 PR tree-optimization/84037
2595 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2596 parameter, move visited init to caller.
2597 (vect_slp_analyze_operations): Separate cost from validity
2598 check, initialize visited once for all instances.
2599 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2601 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2603 (vect_model_promotion_demotion_cost): Likewise.
2604 (vectorizable_bswap): Guard cost modeling with !slp_node
2605 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2607 (vectorizable_call): Likewise.
2608 (vectorizable_conversion): Likewise.
2609 (vectorizable_assignment): Likewise.
2610 (vectorizable_shift): Likewise.
2611 (vectorizable_operation): Likewise.
2612 (vectorizable_store): Likewise.
2613 (vectorizable_load): Likewise.
2614 (vectorizable_condition): Likewise.
2615 (vectorizable_comparison): Likewise.
2617 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2620 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2621 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2622 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2624 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2627 * params.def (inline-min-speedup): Increase from 8 to 15.
2628 (max-inline-insns-auto): Decrease from 40 to 30.
2629 * ipa-split.c (consider_split): Add some buffer for function to
2630 be considered inlining candidate.
2631 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2634 2018-02-12 Richard Biener <rguenther@suse.de>
2636 PR tree-optimization/84037
2637 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2638 matched stmts if we cannot swap the non-matched ones.
2640 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2642 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2643 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2644 _mm_maskz_scalef_round_ss): New intrinsics.
2645 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2646 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2647 __builtin_ia32_scalefss_round): Remove.
2648 (__builtin_ia32_scalefsd_mask_round,
2649 __builtin_ia32_scalefss_mask_round): New intrinsics.
2650 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2651 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2652 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2653 "<round_constraint>")): Changed to ...
2654 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2655 "<round_scalar_constraint>")): ... this.
2656 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2657 %0, %1, %2<round_op3>}"): Changed to ...
2658 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2659 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2660 %2<round_scalar_mask_op3>}"): ... this.
2661 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2663 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2665 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2666 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2667 (_mm_maskz_sqrt_round_ss): New intrinsics.
2668 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2669 (__builtin_ia32_sqrtsd_mask_round)
2670 (__builtin_ia32_sqrtss_mask_round): New builtins.
2671 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2672 (__builtin_ia32_sqrtss_round): Remove.
2673 (__builtin_ia32_sqrtsd_mask_round)
2674 (__builtin_ia32_sqrtss_mask_round): New builtins.
2675 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2676 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2677 ((match_operand:VF_128 1 "vector_operand"
2678 "xBm,<round_constraint>")): Changed to ...
2679 ((match_operand:VF_128 1 "vector_operand"
2680 "xBm,<round_scalar_constraint>")): ... this.
2681 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2682 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2683 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2684 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2685 %<iptr>1<round_scalar_mask_op3>}): ... this.
2686 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2687 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2689 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2692 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2693 Cast vec_cmpeq result to correct type.
2694 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2695 Cast vec_cmpgt result to correct type.
2697 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2699 * final.c (final_scan_insn_1): Renamed from...
2700 (final_scan_insn): ... this. New wrapper, to recover
2701 seen from the outermost call in recursive ones.
2702 * config/sparc/sparc.c (output_return): Drop seen from call.
2703 (output_sibcall): Likewise.
2704 * config/visium/visium.c (output_branch): Likewise.
2706 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2708 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2711 2018-02-10 Alan Modra <amodra@gmail.com>
2714 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2715 Specify LR as an input.
2717 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2720 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2721 remove_member_access_dummy_vars): New functions.
2722 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2723 lower_omp_1, execute_lower_omp): Use them.
2725 PR rtl-optimization/84308
2726 * shrink-wrap.c (spread_components): Release todo vector.
2728 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2730 PR rtl-optimization/57193
2731 * ira-color.c (struct allocno_color_data): Add member
2732 conflict_allocno_hard_prefs.
2733 (update_conflict_allocno_hard_prefs): New.
2734 (bucket_allocno_compare_func): Add a preference based on
2735 conflict_allocno_hard_prefs.
2736 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2737 (color_allocnos): Remove a dead code. Initiate
2738 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2740 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2743 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2744 constraint from =wa to wa. Avoid a subreg on the output operand,
2745 instead use a pseudo and subreg it in a move.
2746 (p9_xxbrd_<mode>): Changed to ...
2747 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2748 (p9_xxbrd_v2df): New expander.
2749 (p9_xxbrw_<mode>): Changed to ...
2750 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2751 (p9_xxbrw_v4sf): New expander.
2753 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2755 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2757 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2760 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2761 multiply in 32-bit mode.
2762 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2763 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2766 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2768 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2769 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2770 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2771 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2773 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2776 * dwarf2out.c (is_trivial_indirect_ref): New function.
2777 (dwarf2out_late_global_decl): Do not generate a location
2778 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2779 and that are not defined in the current unit.
2781 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2783 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2784 instead of a libcall for UNORDERED.
2786 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2789 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2790 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2792 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2795 * config/s390/s390.c (s390_set_current_function): Invoke
2796 s390_indirect_branch_settings also if fndecl didn't change.
2798 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2800 * config/rs6000/rs6000.md (blockage): Set length to zero.
2802 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2804 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2806 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2809 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2810 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2814 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2815 PARALLEL incoming that failed vt_get_decl_and_offset check.
2818 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2819 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2821 (get_variable_section): For decls in named .bss* sections pass true as
2822 second argument to bss_initializer_p.
2824 2018-02-09 Marek Polacek <polacek@redhat.com>
2825 Jakub Jelinek <jakub@redhat.com>
2828 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2829 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2830 Sync some changes from cxx_fold_indirect_ref.
2832 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2834 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2836 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2837 (BLOCK_INLINE_ENTRY_LABEL): New.
2838 (dwarf2out_var_location): Disregard inline entry markers.
2839 (inline_entry_data): New struct.
2840 (inline_entry_data_hasher): New hashtable type.
2841 (inline_entry_data_hasher::hash): New.
2842 (inline_entry_data_hasher::equal): New.
2843 (inline_entry_data_table): New variable.
2844 (add_high_low_attributes): Add DW_AT_entry_pc and
2845 DW_AT_GNU_entry_view attributes if a pending entry is found
2846 in inline_entry_data_table. Add old entry_pc attribute only
2847 if debug nonbinding markers are disabled.
2848 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2849 markers are enabled.
2850 (block_within_block_p, dwarf2out_inline_entry): New.
2851 (dwarf2out_finish): Check that no entries remained in
2852 inline_entry_data_table.
2853 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2854 (final_scan_insn, notice_source_line): Likewise.
2855 (rest_of_clean_state): Skip inline entry markers.
2856 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2858 * gimple.c (gimple_build_debug_inline_entry): New.
2859 * gimple.h (enum gimple_debug_subcode): Add
2860 GIMPLE_DEBUG_INLINE_ENTRY.
2861 (gimple_build_debug_inline_entry): Declare.
2862 (gimple_debug_inline_entry_p): New.
2863 (gimple_debug_nonbind_marker_p): Adjust.
2864 * insn-notes.def (INLINE_ENTRY): New.
2865 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2866 inline entry marker notes.
2867 (print_insn): Likewise.
2868 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2869 (INSN_DEBUG_MARKER_KIND): Likewise.
2870 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2871 * tree-inline.c (expand_call_inline): Build and insert
2872 debug_inline_entry stmt.
2873 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2874 inline entry blocks early, if nonbind markers are enabled.
2875 (dump_scope_block): Dump fragment info.
2876 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2877 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2878 (gimple_build_debug_inline_entry): New.
2879 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2880 Enable/disable inline entry points too.
2881 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2882 (DEBUG_INSN): Describe inline entry markers.
2884 * common.opt (gvariable-location-views): New.
2885 (gvariable-location-views=incompat5): New.
2886 * config.in: Rebuilt.
2887 * configure: Rebuilt.
2888 * configure.ac: Test assembler for view support.
2889 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2890 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2891 * dwarf2out.c (var_loc_view): New typedef.
2892 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2893 (dwarf2out_locviews_in_attribute): New.
2894 (dwarf2out_locviews_in_loclist): New.
2895 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2896 (enum dw_line_info_opcode): Add LI_adv_address.
2897 (struct dw_line_info_table): Add view.
2898 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2899 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2900 (zero_view_p): New variable.
2901 (ZERO_VIEW_P): New macro.
2902 (output_asm_line_debug_info): New.
2903 (struct var_loc_node): Add view.
2904 (add_AT_view_list, AT_loc_list): New.
2905 (add_var_loc_to_decl): Add view param. Test it against last.
2906 (new_loc_list): Add view params. Record them.
2907 (AT_loc_list_ptr): Handle loc and view lists.
2908 (view_list_to_loc_list_val_node): New.
2909 (print_dw_val): Handle dw_val_class_view_list.
2910 (size_of_die): Likewise.
2911 (value_format): Likewise.
2912 (loc_list_has_views): New.
2913 (gen_llsym): Set vl_symbol too.
2914 (maybe_gen_llsym, skip_loc_list_entry): New.
2915 (dwarf2out_maybe_output_loclist_view_pair): New.
2916 (output_loc_list): Output view list or entries too.
2917 (output_view_list_offset): New.
2918 (output_die): Handle dw_val_class_view_list.
2919 (output_dwarf_version): New.
2920 (output_compilation_unit_header): Use it.
2921 (output_skeleton_debug_sections): Likewise.
2922 (output_rnglists, output_line_info): Likewise.
2923 (output_pubnames, output_aranges): Update version comments.
2924 (output_one_line_info_table): Output view numbers in asm comments.
2925 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2926 Call maybe_gen_llsym.
2927 (loc_list_from_tree_1): Adjust.
2928 (add_AT_location_description): Create view list attribute if
2929 needed, check it's absent otherwise.
2930 (convert_cfa_to_fb_loc_list): Adjust.
2931 (maybe_emit_file): Call output_asm_line_debug_info for test.
2932 (dwarf2out_var_location): Reset views as needed. Precompute
2933 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2934 attribute. Set view.
2935 (new_line_info_table): Reset next view.
2936 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2937 (dwarf2out_source_line): Likewise. Output view resets and labels to
2938 the assembler, or select appropriate line info opcodes.
2939 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2940 (optimize_string_length): Catch it. Adjust.
2941 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2942 dw_val_class_view_list, and remove it if no longer needed.
2943 (hash_loc_list): Hash view numbers.
2944 (loc_list_hasher::equal): Compare them.
2945 (optimize_location_lists): Check whether a view list symbol is
2946 needed, and whether the locview attribute is present, and
2947 whether they match. Remove the locview attribute if no longer
2949 (index_location_lists): Call skip_loc_list_entry for test.
2950 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2951 Use output_dwarf_version.
2952 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2953 (struct dw_val_node): Add val_view_list.
2954 * final.c (SEEN_NEXT_VIEW): New.
2955 (set_next_view_needed): New.
2956 (clear_next_view_needed): New.
2957 (maybe_output_next_view): New.
2958 (final_start_function): Rename to...
2959 (final_start_function_1): ... this. Take pointer to FIRST,
2960 add SEEN parameter. Emit param bindings in the initial view.
2961 (final_start_function): Reintroduce SEEN-less interface.
2962 (final): Rename to...
2963 (final_1): ... this. Take SEEN parameter. Output final pending
2964 next view at the end.
2965 (final): Reintroduce seen-less interface.
2966 (final_scan_insn): Output pending next view before switching
2967 sections or ending a block. Mark the next view as needed when
2968 outputting variable locations. Notify debug backend of section
2969 changes, and of location view changes.
2970 (rest_of_handle_final): Adjust.
2971 * toplev.c (process_options): Autodetect value for debug variable
2972 location views option. Warn on incompat5 without -gdwarf-5.
2973 * doc/invoke.texi (gvariable-location-views): New.
2974 (gvariable-location-views=incompat5): New.
2975 (gno-variable-location-views): New.
2977 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2979 PR tree-optimization/84136
2980 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2981 that the result of find_edge is non-NULL.
2983 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2986 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2987 storing integer register in SImode. Fix cost of 256 and 512
2988 byte aligned SSE register store.
2990 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2992 * config/i386/i386.c (ix86_multiplication_cost): Fix
2993 multiplication cost for TARGET_AVX512DQ.
2995 2018-02-08 Marek Polacek <polacek@redhat.com>
2997 PR tree-optimization/84238
2998 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3001 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3003 PR tree-optimization/84265
3004 * tree-vect-stmts.c (vectorizable_store): Don't treat
3005 VMAT_CONTIGUOUS accesses as grouped.
3006 (vectorizable_load): Likewise.
3008 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3010 PR tree-optimization/81635
3011 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3012 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3013 (test_round_for_mask): New functions.
3014 (wide_int_cc_tests): Call test_round_for_mask.
3015 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3016 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3017 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3018 range returned by get_range_info.
3020 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3023 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3024 * symtab.c: Include builtins.h
3025 (symtab_node::output_to_lto_symbol_table_p): Move here
3026 from lto-streamer-out.c:output_symbol_p.
3027 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3028 (output_symbol_p): Move all logic to symtab.c
3029 (produce_symtab): Update.
3031 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3033 * config/s390/s390-opts.h (enum indirect_branch): Define.
3034 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3035 (s390_indirect_branch_via_thunk)
3036 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3037 (enum s390_indirect_branch_type): Define.
3038 * config/s390/s390.c (struct s390_frame_layout, struct
3039 machine_function): Remove.
3040 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3041 (indirect_branch_table_label_no, indirect_branch_table_name):
3043 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3044 (enum s390_indirect_branch_option): Define.
3045 (s390_return_addr_from_memory): New function.
3046 (s390_handle_string_attribute): New function.
3047 (s390_attribute_table): Add new attribute handler.
3048 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3049 (s390_indirect_branch_via_thunk): New function.
3050 (s390_indirect_branch_via_inline_thunk): New function.
3051 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3052 sibling call optimization for non z10 compiles.
3053 (s390_emit_call): Force indirect branch target to be a single
3054 register. Add r1 clobber for non-z10 compiles.
3055 (s390_emit_epilogue): Emit return jump via return_use expander.
3056 (s390_reorg): Handle JUMP_INSNs as execute targets.
3057 (s390_option_override_internal): Perform validity checks for the
3058 new command line options.
3059 (s390_indirect_branch_attrvalue): New function.
3060 (s390_indirect_branch_settings): New function.
3061 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3062 (s390_output_indirect_thunk_function): New function.
3063 (s390_code_end): Implement target hook.
3064 (s390_case_values_threshold): Implement target hook.
3065 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3067 * config/s390/s390.h (struct s390_frame_layout)
3068 (struct machine_function): Move here from s390.c.
3069 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3070 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3071 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3072 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3073 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3074 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3075 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3076 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3077 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3078 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3079 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3080 (mnemonic attribute): Add values which aren't recognized
3082 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3083 pattern for branch conversion. Fix mnemonic attribute.
3084 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3085 indirect branch via thunk if requested.
3086 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3087 ("*indirect_jump"): Disable for branch conversion using out of
3089 ("indirect_jump_via_thunk<mode>_z10")
3090 ("indirect_jump_via_thunk<mode>")
3091 ("indirect_jump_via_inlinethunk<mode>_z10")
3092 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3093 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3094 ("casesi_jump_via_inlinethunk<mode>_z10")
3095 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3096 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3097 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3098 ("*indirect2_jump"): Disable for branch conversion.
3099 ("casesi_jump"): Turn into expander and expand patterns for branch
3101 ("return_use"): New expander.
3102 ("*return"): Emit return via thunk and rename it to ...
3103 ("*return<mode>"): ... this one.
3104 * config/s390/s390.opt: Add new options and and enum for the
3107 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3109 * lra-constraints.c (match_reload): Unconditionally use
3110 gen_lowpart_SUBREG, rather than selecting between that
3111 and equivalent gen_rtx_SUBREG code.
3113 2018-02-08 Richard Biener <rguenther@suse.de>
3115 PR tree-optimization/84233
3116 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3117 changed flag instead of boguously re-using phi_inserted.
3119 2018-02-08 Martin Jambor <mjambor@suse.cz>
3121 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3122 static local variables.
3124 2018-02-08 Richard Biener <rguenther@suse.de>
3126 PR tree-optimization/84278
3127 * tree-vect-stmts.c (vectorizable_store): When looking for
3128 smaller vector types to perform grouped strided loads/stores
3129 make sure the mode is supported by the target.
3130 (vectorizable_load): Likewise.
3132 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3134 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3135 Increase LDP/STP opportunities by adding adjacent callee-saves.
3137 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3139 PR rtl-optimization/84068
3140 PR rtl-optimization/83459
3141 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3143 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3145 PR tree-optimization/84224
3146 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3147 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3150 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3153 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3154 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3155 count, remove one USE.
3157 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3159 * doc/install.texi (Configuration): Document the
3160 --with-long-double-format={ibm,ieee} PowerPC configuration
3164 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3165 Convert from define_expand to be define_insn_and_split. Rework
3166 float/double/_Float128 conversions to QI/HI/SImode to work with
3167 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3168 conversions to QI/HImode types did a store and then a load to
3169 truncate the value. For conversions to VSX registers, don't split
3170 the insn, instead emit the code directly. Use the code iterator
3171 any_fix to combine signed and unsigned conversions.
3172 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3173 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3174 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3175 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3176 (fix_<mode>di2_hw): Likewise.
3177 (fixuns_<mode>di2_hw): Likewise.
3178 (fix_<mode>si2_hw): Likewise.
3179 (fixuns_<mode>si2_hw): Likewise.
3180 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3181 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3182 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3183 fix<uns>_trunc<SFDF:mode>si2_p8.
3184 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3186 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3187 (fix<uns>_<mode>_mem): Likewise.
3188 (fctiw<u>z_<mode>_mem): Likewise.
3189 (fix<uns>_<mode>_mem): Likewise.
3190 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3191 the register allocator from doing a direct move to the GPRs to do
3192 a store, and instead use the ISA 3.0 store byte/half-word from
3193 vector register instruction. For IEEE 128-bit floating point,
3194 also optimize stores of 32-bit ints.
3195 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3197 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3199 * genextract.c (push_pathstr_operand): New function to support
3201 (walk_rtx): Call push_pathstr_operand.
3202 (print_path): Support [a-zA-Z].
3204 2018-02-07 Richard Biener <rguenther@suse.de>
3206 PR tree-optimization/84037
3207 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3208 (cse_and_gimplify_to_preheader): Declare.
3209 (vect_get_place_in_interleaving_chain): Likewise.
3210 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3212 (_loop_vec_info::~_loop_vec_info): Delete it.
3213 (cse_and_gimplify_to_preheader): New function.
3214 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3215 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3216 (vectorizable_load): Likewise. For grouped stores always base
3217 the IV on the first element.
3218 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3219 condition before gimplifying.
3221 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3223 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3224 *DIV_EXPR and *MOD_EXPR.
3226 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3229 * config/i386/i386.c (ix86_option_override_internal): Mask out
3230 the CF_SET bit when checking -fcf-protection.
3232 2018-02-07 Tom de Vries <tom@codesourcery.com>
3235 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3238 2018-02-07 Richard Biener <rguenther@suse.de>
3240 PR tree-optimization/84204
3241 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3244 PR tree-optimization/84205
3245 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3246 special-case isl_ast_op_zdiv_r.
3248 PR tree-optimization/84223
3249 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3250 Only add conditions from within the region.
3251 (gather_bbs::after_dom_children): Adjust.
3253 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3256 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3257 * config/avr/avr.md: Only post-reload split REG-REG moves if
3258 either register is GENERAL_REG_P.
3260 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3262 PR tree-optimization/84235
3263 * tree-ssa-scopedtables.c
3264 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3265 if the subtraction is performed in floating point type where NaNs are
3266 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3267 build 1. Formatting fix.
3269 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3272 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3273 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3274 and skip it regardless of bb boundaries. Use CALL_P macro,
3275 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3276 unnecessarily, formatting fix.
3278 2018-02-06 Michael Collison <michael.collison@arm.com>
3280 * config/arm/thumb2.md:
3281 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3282 (*thumb_mov_notscc): Ditto.
3284 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3287 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3288 unsigned_fix, not "s".
3290 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3292 * configure.ac (gcc_fn_eh_frame_ro): New function.
3293 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3294 correct .eh_frame permissions.
3295 * configure: Regenerate.
3297 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3299 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3302 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3304 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3305 Display warning message for -mno-speculate-indirect-jumps.
3307 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3309 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3311 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3313 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3315 PR tree-optimization/84225
3316 * tree-eh.c (find_trapping_overflow): Only call
3317 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3319 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3322 * config/i386/i386.c: Reimplement the check of possible options
3323 -mibt/-mshstk conbination. Change error messages.
3324 * doc/invoke.texi: Fix a typo: remove extra '='.
3326 2018-02-06 Marek Polacek <polacek@redhat.com>
3328 PR tree-optimization/84228
3329 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3331 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3334 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3335 emitted arch directives.
3336 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3337 __ARM_FEATURE_COPROC before changing architectures.
3339 2018-02-06 Richard Biener <rguenther@suse.de>
3341 * config/i386/i386.c (print_reg): Fix typo.
3342 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3344 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3346 * configure: Regenerate.
3348 2018-02-05 Martin Sebor <msebor@redhat.com>
3350 PR tree-optimization/83369
3351 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3354 2018-02-05 Martin Liska <mliska@suse.cz>
3356 * doc/invoke.texi: Cherry-pick upstream r323995.
3358 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3360 * ira.c (ira_init_register_move_cost): Adjust comment.
3362 2018-02-05 Martin Liska <mliska@suse.cz>
3364 PR gcov-profile/84137
3365 * doc/gcov.texi: Fix typo in documentation.
3367 2018-02-05 Martin Liska <mliska@suse.cz>
3369 PR gcov-profile/83879
3370 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3371 using dlopen functionality.
3373 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3375 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3376 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3377 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3378 _mm_maskz_range_round_ss): New intrinsics.
3379 (__builtin_ia32_rangesd128_round)
3380 (__builtin_ia32_rangess128_round): Remove.
3381 (__builtin_ia32_rangesd128_mask_round,
3382 __builtin_ia32_rangess128_mask_round): New builtins.
3383 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3384 __builtin_ia32_rangess128_round): Remove.
3385 (__builtin_ia32_rangesd128_mask_round,
3386 __builtin_ia32_rangess128_mask_round): New builtins.
3387 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3388 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3389 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3390 "<round_saeonly_constraint>")): Changed to ...
3391 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3392 "<round_saeonly_scalar_constraint>")): ... this.
3393 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3394 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3395 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3396 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3397 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3399 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3401 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3403 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3404 Remove all values except native, 8540 and 8548.
3406 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3408 * config/i386/i386.c (ix86_output_function_return): Pass
3409 INVALID_REGNUM, instead of -1, as invalid register number to
3410 indirect_thunk_name and output_indirect_thunk.
3412 2018-02-02 Julia Koval <julia.koval@intel.com>
3414 * config.gcc: Add -march=icelake.
3415 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3416 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3417 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3418 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3419 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3420 (processor_target_table): Add icelake.
3421 (ix86_option_override_internal): Handle new PTAs.
3422 (get_builtin_code_for_version): Handle icelake.
3423 (M_INTEL_COREI7_ICELAKE): New.
3424 (fold_builtin_cpu): Handle icelake.
3425 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3426 * doc/invoke.texi: Add -march=icelake.
3428 2018-02-02 Julia Koval <julia.koval@intel.com>
3430 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3431 to wide_int_bitmask.
3432 * wide-int-bitmask.h: New.
3434 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3437 * config/i386/i386.md: Replace Pmode with word_mode in
3438 builtin_setjmp_setup and builtin_longjmp to support x32.
3440 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3444 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3446 (rs6000_supported_cpu_names): New static variable.
3447 (linux_cpu_translation_table): Likewise.
3448 (elf_platform) <cpu>: Define new static variable and use it.
3449 Translate kernel AT_PLATFORM name to canonical name if needed.
3450 Error if platform name is unknown.
3452 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3455 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3457 2018-02-01 Jeff Law <law@redhat.com>
3460 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3461 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3462 the scratch if RELEASE_VIA_POP is false.
3463 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3464 If we have to save a temporary register, decrement SIZE appropriately.
3465 Pass new arguments to release_scratch_register_on_entry.
3466 (ix86_adjust_stack_and_probe): Likewise.
3467 (ix86_emit_probe_stack_range): Pass new arguments to
3468 release_scratch_register_on_entry.
3470 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3472 PR rtl-optimization/84157
3473 * combine.c (change_zero_ext): Use REG_P predicate in
3474 front of HARD_REGISTER_P predicate.
3476 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3478 * config/avr/avr.c (avr_option_override): Move disabling of
3479 -fdelete-null-pointer-checks to...
3480 * common/config/avr/avr-common.c (avr_option_optimization_table):
3483 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3485 PR tree-optimization/81635
3486 * tree-data-ref.c (split_constant_offset_1): For types that
3487 wrap on overflow, try to use range info to prove that wrapping
3490 2018-02-01 Renlin Li <renlin.li@arm.com>
3493 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3495 (aarch64_register_move_cost): Likewise.
3496 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3498 (REG_CLASS_NAMES): Likewise.
3499 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3500 TAILCALL_ADDR_REGS. Remove IP registers.
3501 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3503 2018-02-01 Richard Biener <rguenther@suse.de>
3505 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3506 for specifying RPO order and allow NULL for that.
3507 * domwalk.c (dom_walker::dom_walker): Likewise.
3508 (dom_walker::walk): Handle NULL RPO order.
3509 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3511 (rewrite_update_dom_walker): Likewise.
3512 (mark_def_dom_walker): Likewise.
3514 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3516 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3517 (aarch64_maybe_expand_sve_subreg_move): Declare.
3518 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3519 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3521 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3522 that are semantically a reverse operation.
3523 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3524 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3525 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3527 (aarch64_can_change_mode_class): For big-endian, forbid changes
3528 between two SVE modes if they have different element sizes.
3530 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3532 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3533 the TImode handling for big-endian targets.
3535 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3537 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3538 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3540 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3541 Remove BSWAP handing for big-endian targets and use the form of
3542 LD1RQ appropariate for the mode.
3544 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3546 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3547 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3550 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3553 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3554 check for operands that need to go through aarch64_sve_reload_be.
3556 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3558 PR tree-optimization/81661
3559 PR tree-optimization/84117
3560 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3561 * tree-eh.c: Include gimplify.h.
3562 (find_trapping_overflow, replace_trapping_overflow,
3563 rewrite_to_non_trapping_overflow): New functions.
3564 * tree-vect-loop.c: Include tree-eh.h.
3565 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3566 * tree-data-ref.c: Include tree-eh.h.
3567 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3569 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3571 PR rtl-optimization/84123
3572 * combine.c (change_zero_ext): Check if hard register satisfies
3573 can_change_dest_mode before calling gen_lowpart_SUBREG.
3575 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3578 * ira.c (ira_init_register_move_cost): Remove assert.
3580 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3582 PR rtl-optimization/84071
3583 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3584 * doc/tm.texi: Regenerate.
3586 2018-01-31 Richard Biener <rguenther@suse.de>
3588 PR tree-optimization/84132
3589 * tree-data-ref.c (analyze_miv_subscript): Properly
3590 check whether evolution_function_is_affine_multivariate_p
3591 before calling gcd_of_steps_may_divide_p.
3593 2018-01-31 Julia Koval <julia.koval@intel.com>
3596 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3597 * config/i386/i386.md (rdpid_rex64) New.
3598 (rdpid): Make 32bit only.
3600 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3603 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3604 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3606 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3609 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3611 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3613 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3615 PR rtl-optimization/84071
3616 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3617 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3619 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3621 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3622 (arc_attribute_table): Add 'aux' attribute.
3623 (arc_in_small_data_p): Consider aux like variables.
3624 (arc_is_aux_reg_p): New function.
3625 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3626 (arc_get_aux_arg): New function.
3627 (prepare_move_operands): Handle aux-register access.
3628 (arc_handle_aux_attribute): New function.
3629 * doc/extend.texi (ARC Variable attributes): Add subsection.
3631 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3633 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3634 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3635 (arc_attribute_table): Add 'uncached' attribute.
3636 (arc_print_operand): Print '.di' flag for uncached memory
3638 (arc_in_small_data_p): Do not consider for small data the uncached
3640 (arc_is_uncached_mem_p): New function.
3641 * config/arc/predicates.md (compact_store_memory_operand): Check
3642 for uncached memory accesses.
3643 (nonvol_nonimm_operand): Likewise.
3644 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3646 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3649 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3650 falign-loops=): Add Optimization flag.
3652 2018-01-30 Jeff Law <law@redhat.com>
3655 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3656 INT_REGISTERS_SAVED. Check it prior to calling
3657 get_scratch_register_on_entry.
3658 (ix86_adjust_stack_and_probe): Similarly.
3659 (ix86_emit_probe_stack_range): Similarly.
3660 (ix86_expand_prologue): Corresponding changes.
3662 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3665 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3666 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3668 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3671 * lra-constraints.c (curr_insn_transform): Process AND in the
3674 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3676 PR rtl-optimization/83986
3677 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3678 dependence against last_pending_memory_flush in addition to
3681 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3683 PR tree-optimization/81611
3684 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3687 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3690 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3693 2018-01-30 Richard Biener <rguenther@suse.de>
3694 Jakub Jelinek <jakub@redhat.com>
3696 PR tree-optimization/84111
3697 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3698 inner loops added during recursion, as they don't have up-to-date
3701 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3704 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3705 (can_inline_edge_by_limits_p): ... here.
3706 (can_early_inline_edge_p, check_callers,
3707 update_caller_keys, update_callee_keys, recursive_inlining,
3708 add_new_edges_to_heap, speculation_useful_p,
3709 inline_small_functions,
3710 inline_small_functions, flatten_function,
3711 inline_to_all_callers_1): Update.
3713 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3715 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3718 2018-01-30 Richard Biener <rguenther@suse.de>
3720 PR tree-optimization/83008
3721 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3722 invariant and constant vector uses in stmts when they need
3725 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3728 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3729 * configure: Regenerate.
3731 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3733 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3735 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3736 Use gen_rtx_REG rather than gen_lowpart.
3738 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3740 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3741 rather than 0 when creating partial subregs.
3743 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3745 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3748 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3751 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3752 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3753 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3754 flags. This restores the settings used before the 2017-07-24.
3755 Turning off pre increment/decrement/modify allows IVOPTS to
3756 optimize DF/SF loops where the index is an int.
3758 2018-01-29 Richard Biener <rguenther@suse.de>
3759 Kelvin Nilsen <kelvin@gcc.gnu.org>
3762 * tree-vect-stmts.c (vectorizable_call): Don't call
3763 targetm.vectorize_builtin_md_vectorized_function if callee is
3766 2018-01-22 Carl Love <cel@us.ibm.com>
3768 * doc/extend.tex: Fix typo in second arg in
3769 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3771 2018-01-29 Richard Biener <rguenther@suse.de>
3773 PR tree-optimization/84086
3774 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3775 (flush_ssaname_freelist): When SSA names were released reset
3776 the SCEV hash table.
3778 2018-01-29 Richard Biener <rguenther@suse.de>
3780 PR tree-optimization/84057
3781 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3782 removed paths when removing edges.
3784 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3786 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3787 -mfunction-return=@var{choice}.
3789 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3792 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3794 (layout::print_source_line): Likewise.
3795 (test_get_line_width_without_trailing_whitespace): Add test cases.
3797 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3800 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3803 2018-01-26 Jim Wilson <jimw@sifive.com>
3805 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3807 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3810 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3812 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3813 and CMP + SUB-immediate -> SUBS.
3815 2018-01-26 Martin Sebor <msebor@redhat.com>
3817 PR tree-optimization/83896
3818 * tree-ssa-strlen.c (get_string_len): Rename...
3819 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3820 Avoid assuming length is constant.
3821 (handle_char_store): Use HOST_WIDE_INT for string length.
3823 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3826 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3827 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3829 2018-01-26 Richard Biener <rguenther@suse.de>
3831 PR rtl-optimization/84003
3832 * dse.c (record_store): Only record redundant stores when
3833 the earlier store aliases at least all accesses the later one does.
3835 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3837 PR rtl-optimization/83985
3838 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3839 REG_CFA_RESTORE insns.
3840 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3841 REG_CFA_RESTORE insns here.
3844 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3845 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3847 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3849 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3850 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3851 (arc_init): Likewise.
3852 (arc_override_options): Likewise.
3853 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3855 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3857 * config/arc/arc.h (TARGET_DBNZ): Define.
3858 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3859 properly set the tune attribute.
3860 (dbnz): Use TARGET_DBNZ guard.
3861 * config/arc/arc.opt (mtune): Add core3 option.
3863 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3865 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3866 recognize new pic like addresses.
3867 (arc_delegitimize_address): Clean up.
3869 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3871 * config/arc/arc-arches.def: Option mrf16 valid for all
3873 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3874 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3875 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3876 * config/arc/arc-tables.opt: Regenerate.
3877 * config/arc/arc.c (arc_conditional_register_usage): Handle
3878 reduced register file case.
3879 (arc_file_start): Set must have build attributes.
3880 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3882 * config/arc/arc.opt (mrf16): Add new option.
3883 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3884 * config/arc/genmultilib.awk: Handle new mrf16 option.
3885 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3886 * config/arc/t-multilib: Regenerate.
3887 * doc/invoke.texi (ARC Options): Document mrf16 option.
3889 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3891 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3892 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3893 (arc_attribute_table): Add 'secure_call' attribute.
3894 (arc_print_operand): Print secure call operand.
3895 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3897 (arc_is_secure_call_p): New function. * config/arc/arc.md
3898 (call_i): Add support for sjli instruction.
3899 (call_value_i): Likewise.
3900 * config/arc/constraints.md (Csc): New constraint.
3902 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3903 John Eric Martin <John.Martin@emmicro-us.com>
3905 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3906 * config/arc/arc.c (_arc_jli_section): New struct.
3907 (arc_jli_section): New type.
3908 (rc_jli_sections): New static variable.
3909 (arc_handle_jli_attribute): New function.
3910 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3911 (arc_file_end): New function.
3912 (TARGET_ASM_FILE_END): Define.
3913 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3914 (arc_add_jli_section): New function.
3915 (jli_call_scan): Likewise.
3916 (arc_reorg): Call jli_call_scan.
3917 (arc_output_addsi): Remove 'S' from printing asm operand.
3918 (arc_is_jli_call_p): New function.
3919 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3921 (movhi_insn): Likewise.
3922 (movsi_insn): Likewise.
3923 (movsi_set_cc_insn): Likewise.
3924 (loadqi_update): Likewise.
3925 (load_zeroextendqisi_update): Likewise.
3926 (load_signextendqisi_update): Likewise.
3927 (loadhi_update): Likewise.
3928 (load_zeroextendhisi_update): Likewise.
3929 (load_signextendhisi_update): Likewise.
3930 (loadsi_update): Likewise.
3931 (loadsf_update): Likewise.
3932 (movsicc_insn): Likewise.
3933 (bset_insn): Likewise.
3934 (bxor_insn): Likewise.
3935 (bclr_insn): Likewise.
3936 (bmsk_insn): Likewise.
3937 (bicsi3_insn): Likewise.
3938 (cmpsi_cc_c_insn): Likewise.
3939 (movsi_ne): Likewise.
3940 (movsi_cond_exec): Likewise.
3941 (clrsbsi2): Likewise.
3952 (call_i): Remove 'S' asm letter, add jli instruction.
3953 (call_value_i): Likewise.
3954 * config/arc/arc.op (mjli-always): New option.
3955 * config/arc/constraints.md (Cji): New constraint.
3956 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3958 (subsf3_fpx): Likewise.
3959 (mulsf3_fpx): Likewise.
3960 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3962 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3964 * doc/invoke.texi (ARC): Document mjli-always option.
3966 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3968 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3969 avoid addition with 0 and use incw and decw where possible.
3971 2018-01-26 Richard Biener <rguenther@suse.de>
3973 PR tree-optimization/81082
3974 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3975 association if it requires casting to unsigned.
3976 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3977 from fold_plusminus_mult_expr to catch important cases late when
3978 range info is available.
3980 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3982 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3983 * configure.ac (hidden_linkonce): New test.
3984 * configure: Regenerate.
3985 * config.in: Regenerate.
3987 2018-01-26 Julia Koval <julia.koval@intel.com>
3989 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3990 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3991 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3992 _mm_mask_bitshuffle_epi64_mask): Fix type.
3993 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3994 USI_FTYPE_V4DI_V4DI_USI): Remove.
3995 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3996 __builtin_ia32_vpshufbitqmb256_mask,
3997 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3998 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3999 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4001 2018-01-26 Alan Modra <amodra@gmail.com>
4004 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4005 UNSPEC_VBPERMQ. Sort other unspecs.
4007 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4009 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4011 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4014 * predict.c (drop_profile): Do not push/pop cfun; update also
4016 (handle_missing_profiles): Fix logic looking for zero profiles.
4018 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4021 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4022 on functions with #pragma omp declare simd or functions with simd
4024 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4025 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4026 Remove trailing \n from warning_at calls.
4028 2018-01-25 Tom de Vries <tom@codesourcery.com>
4031 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4032 for neutered workers.
4034 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4037 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4038 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4040 2018-01-24 Jeff Law <law@redhat.com>
4043 * i386.c (get_probe_interval): Move to earlier point.
4044 (ix86_compute_frame_layout): If -fstack-clash-protection and
4045 the frame is larger than the probe interval, then use pushes
4046 to save registers rather than reg->mem moves.
4047 (ix86_expand_prologue): Remove conditional for int_registers_saved
4050 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4053 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4054 min/max for never referenced object.
4056 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4059 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4061 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4062 attributes from DECL_ATTRIBUTES (decl) without affecting
4063 DECL_ATTRIBUTES (current_function_decl).
4064 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4065 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4067 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4069 PR tree-optimization/83979
4070 * fold-const.c (fold_comparison): Use constant_boolean_node
4071 instead of boolean_{true,false}_node.
4073 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4075 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4078 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4080 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4081 Simplify the clause that sets the length attribute.
4082 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4083 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4084 clause that sets the length attribute.
4085 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4087 2018-01-24 Tom de Vries <tom@codesourcery.com>
4090 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4091 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4092 Add strict parameter.
4093 (prevent_branch_around_nothing): Insert dummy insn between branch to
4094 label and label with no ptx insn inbetween.
4095 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4097 2018-01-24 Tom de Vries <tom@codesourcery.com>
4100 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4101 for neutered threads in warp.
4102 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4104 2018-01-24 Richard Biener <rguenther@suse.de>
4106 PR tree-optimization/83176
4107 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4110 2018-01-24 Richard Biener <rguenther@suse.de>
4112 PR tree-optimization/82819
4113 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4114 code generating pluses that are no-ops in the target precision.
4116 2018-01-24 Richard Biener <rguenther@suse.de>
4119 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4121 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4123 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4124 to merge probabilities.
4125 * predict.c (probably_never_executed): Also mark as cold functions
4126 with global 0 profile and guessed local profile.
4127 * profile-count.c (profile_probability::combine_with_count): New
4129 * profile-count.h (profile_probability::operator*,
4130 profile_probability::operator*=, profile_probability::operator/,
4131 profile_probability::operator/=): Reduce precision to adjusted
4132 and set value to guessed on contradictory divisions.
4133 (profile_probability::combine_with_freq): Remove.
4134 (profile_probability::combine_wiht_count): Declare.
4135 (profile_count::force_nonzero):: Set to adjusted.
4136 (profile_count::probability_in):: Set quality to adjusted.
4137 * tree-ssa-tail-merge.c (replace_block_by): Use
4140 2018-01-23 Andrew Waterman <andrew@sifive.com>
4141 Jim Wilson <jimw@sifive.com>
4143 * config/riscv/riscv.c (riscv_stack_boundary): New.
4144 (riscv_option_override): Set riscv_stack_boundary. Handle
4145 riscv_preferred_stack_boundary_arg.
4146 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4147 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4148 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4149 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4150 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4151 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4153 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4156 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4157 of struct ix86_frame.
4158 (ix86_expand_epilogue): Likewise. Add a local variable for
4159 the reg_save_offset field in struct ix86_frame.
4161 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4163 PR tree-optimization/82604
4164 * tree-loop-distribution.c (enum partition_kind): New enum item
4165 PKIND_PARTIAL_MEMSET.
4166 (partition_builtin_p): Support above new enum item.
4167 (generate_code_for_partition): Ditto.
4168 (compute_access_range): Differentiate cases that equality can be
4169 proven at all loops, the innermost loops or no loops.
4170 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4171 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4172 (finalize_partitions, distribute_loop): Don't fuse partition of
4173 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4174 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4177 2018-01-23 Martin Liska <mliska@suse.cz>
4179 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4180 order to ignore the predictor.
4181 (PRED_POLYMORPHIC_CALL): Likewise.
4182 (PRED_RECURSIVE_CALL): Likewise.
4184 2018-01-23 Martin Liska <mliska@suse.cz>
4186 * tree-profile.c (tree_profiling): Print function header to
4187 aware reader which function we are working on.
4188 * value-prof.c (gimple_find_values_to_profile): Do not print
4189 not interesting value histograms.
4191 2018-01-23 Martin Liska <mliska@suse.cz>
4193 * profile-count.h (enum profile_quality): Add
4194 profile_uninitialized as the first value. Do not number values
4195 as they are zero based.
4196 (profile_count::verify): Update sanity check.
4197 (profile_probability::verify): Likewise.
4199 2018-01-23 Nathan Sidwell <nathan@acm.org>
4201 * doc/invoke.texi (ffor-scope): Deprecate.
4203 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4205 PR tree-optimization/83510
4206 * domwalk.c (set_all_edges_as_executable): New function.
4207 (dom_walker::dom_walker): Convert bool param
4208 "skip_unreachable_blocks" to enum reachability. Move setup of
4209 edge flags to set_all_edges_as_executable and only do it when
4210 reachability is REACHABLE_BLOCKS.
4211 * domwalk.h (enum dom_walker::reachability): New enum.
4212 (dom_walker::dom_walker): Convert bool param
4213 "skip_unreachable_blocks" to enum reachability.
4214 (set_all_edges_as_executable): New decl.
4215 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4216 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4218 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4219 but converting true to REACHABLE_BLOCKS.
4220 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4222 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4223 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4224 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4226 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4227 if check_all_array_refs will be called.
4229 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4231 * tree.c (selftest::test_location_wrappers): Add more test
4234 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4236 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4237 (selftest::test_bit_in_range): Likewise.
4239 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4242 * doc/sourcebuild.texi (vect_float): Say that the selector
4243 only describes the situation when -funsafe-math-optimizations is on.
4244 (vect_float_strict): Document.
4246 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4248 PR tree-optimization/83965
4249 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4250 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4251 instead of checking only for a reduction.
4252 (vect_recog_widen_sum_pattern): Likewise.
4254 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4256 * predict.c (probably_never_executed): Only use precise profile info.
4257 (compute_function_frequency): Skip after inlining hack since we now
4258 have quality checking.
4260 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4262 * profile-count.h (profile_probability::very_unlikely,
4263 profile_probability::unlikely, profile_probability::even): Set
4264 precision to guessed.
4266 2018-01-23 Richard Biener <rguenther@suse.de>
4268 PR tree-optimization/83963
4269 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4270 Properly terminate dominator walk when crossing the exit edge not
4271 when visiting its source block.
4273 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4276 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4277 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4279 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4281 PR tree-optimization/83957
4282 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4283 semicolon after for body surrounded by braces.
4285 PR tree-optimization/83081
4286 * profile-count.h (profile_probability::split): New method.
4287 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4288 Use profile_probability::split.
4289 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4290 when splitting a single conditional jump into 2.
4292 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4294 PR tree-optimization/69452
4295 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4298 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4300 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4301 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4302 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4304 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4306 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4307 * config/rl78/rl78.md (movdi): New define_expand.
4308 * config/rl78/rl78.c (rl78_split_movdi): New function.
4310 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4313 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4315 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4316 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4317 128-bit to produce an UNSPEC move to get the double word with the
4318 signbit and then a shift directly to do signbit.
4319 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4320 implementation with a new version that just does either a direct
4321 move or a regular move. Move memory interface to separate insns.
4322 Move insns so they are next to the expander.
4323 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4324 with signbit move. Split big and little endian case.
4325 (signbit<mode>2_dm_mem_le): Likewise.
4326 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4327 (signbit<mode>2_dm2): Likewise.
4329 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4331 * config/rl78/rl78.md (anddi3): New define_expand.
4333 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4335 * config/rl78/rl78.md (umindi3): New define_expand.
4337 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4339 * config/rl78/rl78.md (smindi3): New define_expand.
4341 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4343 * config/rl78/rl78.md (smaxdi3): New define_expand.
4345 2018-01-22 Carl Love <cel@us.ibm.com>
4347 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4348 LVX_V1TI): Add macro expansion.
4349 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4350 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4351 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4352 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4353 Change check to determine if the instruction is a byte reversing
4354 entry. Fix typo in comment.
4355 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4356 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4357 Add def_builtin calls for new builtins.
4358 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4359 Add define_insn expansion.
4361 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4363 * config/rl78/rl78.md (umaxdi3): New define_expand.
4365 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4367 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4368 for non-QImode registers.
4370 2018-01-22 Richard Biener <rguenther@suse.de>
4372 PR tree-optimization/83963
4373 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4374 including the loop exit block.
4375 (scop_detection::merge_sese): Likewise.
4376 (scop_detection::add_scop): Do it here instead.
4378 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4380 * doc/sourcebuild.texi (arm_softfloat): Document.
4382 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4385 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4386 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4387 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4389 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4390 David Edelsohn <dje.gcc@gmail.com>
4393 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4394 Change "crset eq" to "crset 2".
4395 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4396 (*call_indirect_aix<mode>_nospec): Likewise.
4397 (*call_value_indirect_aix<mode>_nospec): Likewise.
4398 (*call_indirect_elfv2<mode>_nospec): Likewise.
4399 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4400 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4401 change assembly output from . to $.
4402 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4403 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4404 (*tablejump<mode>_internal1_nospec): Likewise.
4406 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4409 * config/sh/sh_optimize_sett_clrt.cc:
4410 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4412 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4414 PR tree-optimization/83940
4415 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4416 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4417 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4418 use it to pass back the definition type.
4419 (vect_check_store_rhs): Likewise rhs_dt_out.
4420 (vect_build_gather_load_calls): Add a mask_dt argument and use
4421 it instead of a call to vect_is_simple_use.
4422 (vectorizable_store): Update calls to vect_check_load_store_mask
4423 and vect_check_store_rhs. Use the dt returned by the latter instead
4424 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4425 instead of calls to vect_is_simple_use. Pass the scalar rather
4426 than the vector operand to vect_is_simple_use when handling
4427 second and subsequent copies of an rhs value.
4428 (vectorizable_load): Update calls to vect_check_load_store_mask
4429 and vect_build_gather_load_calls. Use the cached mask_dt and
4430 gs_info.offset_dt instead of calls to vect_is_simple_use.
4432 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4435 * tree-emutls.c: Include gimplify.h.
4436 (lower_emutls_2): New function.
4437 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4438 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4439 it before further processing.
4442 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4443 UINTVAL (trueop1) instead of INTVAL (op1).
4445 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4449 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4450 INCOMING_FRAME_SP_OFFSET if not defined.
4451 (scan_trace): Add ENTRY argument. If true and
4452 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4453 emit a note to adjust the CFA offset.
4454 (create_cfi_notes): Adjust scan_trace callers.
4455 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4456 INCOMING_FRAME_SP_OFFSET in the CIE.
4457 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4458 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4460 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4461 * doc/tm.texi: Regenerated.
4463 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4465 PR rtl-optimization/83147
4466 * lra-constraints.c (remove_inheritance_pseudos): Use
4467 lra_substitute_pseudo_within_insn.
4469 2018-01-19 Tom de Vries <tom@codesourcery.com>
4470 Cesar Philippidis <cesar@codesourcery.com>
4473 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4475 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4478 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4479 spaces for function labels.
4481 2018-01-19 Martin Liska <mliska@suse.cz>
4483 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4484 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4485 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4486 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4487 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4488 (PRED_CONST_RETURN): Change from 69 to 65.
4489 (PRED_NULL_RETURN): Change from 91 to 71.
4490 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4491 (PRED_LOOP_GUARD): Change from 66 to 73.
4493 2018-01-19 Martin Liska <mliska@suse.cz>
4495 * predict.c (predict_insn_def): Add new assert.
4496 (struct branch_predictor): Change type to signed integer.
4497 (test_prediction_value_range): Amend test to cover
4499 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4500 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4501 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4502 (PRED_LOOP_IV_COMPARE): Likewise.
4503 * predict.h (PROB_UNINITIALIZED): Define new constant.
4505 2018-01-19 Martin Liska <mliska@suse.cz>
4507 * predict.c (dump_prediction): Add new format for
4508 analyze_brprob.py script which is enabled with -details
4510 * profile-count.h (precise_p): New function.
4512 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4514 PR tree-optimization/83922
4515 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4516 there are no statements that need masking.
4517 (vect_active_double_reduction_p): New function.
4518 (vect_analyze_loop_operations): Use it when handling phis that
4519 are not in the loop header.
4521 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4523 PR tree-optimization/83914
4524 * tree-vect-loop.c (vectorizable_induction): Don't convert
4525 init_expr or apply the peeling adjustment for inductions
4526 that are nested within the vectorized loop.
4528 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4530 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4533 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4537 * function.h (gimplify_parameters): Add gimple_seq * argument.
4538 * function.c: Include gimple.h and options.h.
4539 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4540 for the added local temporaries if needed.
4541 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4542 if there are any parameter cleanups, wrap whole body into a
4543 try/finally with the cleanups.
4545 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4548 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4549 Use GET_MODE_CLASS for scalar floating point.
4551 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4555 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4556 Fix call of call_cgraph_insertion_hooks.
4558 2018-01-18 Martin Sebor <msebor@redhat.com>
4560 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4562 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4565 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4568 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4571 * common.opt: (-ffile-prefix-map): New option.
4572 * opts.c (common_handle_option): Defer it.
4573 * opts-global.c (handle_common_deferred_options): Handle it.
4574 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4575 * file-prefix-map.h: New file.
4576 (remap_debug_filename, add_debug_prefix_map): ...here.
4577 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4578 * final.c (debug_prefix_map, add_debug_prefix_map
4579 remap_debug_filename): Move to...
4580 * file-prefix-map.c: New file.
4581 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4582 generalize, get rid of alloca(), use strrchr() instead of strchr().
4583 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4584 Implement in terms of add_prefix_map().
4585 (remap_macro_filename, remap_debug_filename): Implement in term of
4587 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4588 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4589 * dbxout.c: Include file-prefix-map.h.
4590 * varasm.c: Likewise.
4591 * vmsdbgout.c: Likewise.
4592 * xcoffout.c: Likewise.
4593 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4594 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4595 * doc/invoke.texi (-ffile-prefix-map): Document.
4596 (-fdebug-prefix-map): Update description.
4598 2018-01-18 Martin Liska <mliska@suse.cz>
4600 * config/i386/i386.c (indirect_thunk_name): Document that also
4602 (output_indirect_thunk): Document why both instructions
4603 (pause and lfence) are generated.
4605 2018-01-18 Richard Biener <rguenther@suse.de>
4607 PR tree-optimization/83887
4608 * graphite-scop-detection.c
4609 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4610 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4611 (scop_detection::merge_sese): Re-implement with a flood-fill
4612 algorithm that properly finds a SESE region if it exists.
4614 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4617 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4618 pointer_diff optimizations use view_convert instead of convert.
4620 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4622 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4623 Generate different code for -mno-speculate-indirect-jumps.
4624 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4625 (*call_indirect_aix<mode>): Disable for
4626 -mno-speculate-indirect-jumps.
4627 (*call_indirect_aix<mode>_nospec): New define_insn.
4628 (*call_value_indirect_aix<mode>): Disable for
4629 -mno-speculate-indirect-jumps.
4630 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4631 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4632 -mno-speculate-indirect-jumps.
4633 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4635 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4637 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4638 long double type, set the flags for noting the default long double
4639 type, even if we don't pass or return a long double type.
4641 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4644 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4647 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4649 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4650 support for merge[hl].
4651 (fold_mergehl_helper): New helper function.
4652 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4653 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4654 (altivec_vmrglw_direct): Add xxmrglw insn.
4656 2018-01-17 Andrew Waterman <andrew@sifive.com>
4658 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4659 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4661 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4664 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4665 call the lto_location_cache before reading the
4666 DECL_SOURCE_LOCATION of the types.
4668 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4669 Richard Sandiford <richard.sandiford@linaro.org>
4671 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4672 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4673 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4674 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4675 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4677 * config/aarch64/constraints.md (aarch64_movti_operand):
4679 * config/aarch64/predicates.md (Uti): Add new constraint.
4681 2018-01-17 Carl Love <cel@us.ibm.com>
4683 * config/rs6000/vsx.md (define_expand xl_len_r,
4684 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4685 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4687 (define_expand, define_insn): Move the shift left from the
4688 define_insn to the define_expand for lxvl and stxvl instructions.
4689 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4690 and XL_LEN_R definitions to PURE.
4692 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4694 * config/i386/i386.c (indirect_thunk_name): Declare regno
4695 as unsigned int. Compare regno with INVALID_REGNUM.
4696 (output_indirect_thunk): Ditto.
4697 (output_indirect_thunk_function): Ditto.
4698 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4699 in the call to output_indirect_thunk_function.
4701 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4704 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4705 rather than the size of inner_type to determine the stack slot size
4706 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4708 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4711 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4714 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4716 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4717 endian Linux systems to optionally enable multilibs for selecting
4718 the long double type if the user configured an explicit type.
4719 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4720 have no long double multilibs if not defined.
4721 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4722 warn if the user used -mabi={ieee,ibm}longdouble and we built
4723 multilibs for long double.
4724 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4725 appropriate multilib option.
4726 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4728 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4729 for building long double multilibs.
4730 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4732 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4734 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4737 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4739 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4742 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4745 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4748 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4750 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4751 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4753 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4755 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4756 different rtl trees depending on TARGET_64BIT.
4757 (rs6000_gen_lvx): Likewise.
4759 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4761 * config/visium/visium.md (nop): Tweak comment.
4762 (hazard_nop): Likewise.
4764 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4766 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4767 -mspeculate-indirect-jumps.
4768 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4769 for -mno-speculate-indirect-jumps.
4770 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4771 (*call_value_indirect_elfv2<mode>): Disable for
4772 -mno-speculate-indirect-jumps.
4773 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4774 (indirect_jump): Emit different RTL for
4775 -mno-speculate-indirect-jumps.
4776 (*indirect_jump<mode>): Disable for
4777 -mno-speculate-indirect-jumps.
4778 (*indirect_jump<mode>_nospec): New define_insn.
4779 (tablejump): Emit different RTL for
4780 -mno-speculate-indirect-jumps.
4781 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4782 (tablejumpsi_nospec): New define_expand.
4783 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4784 (tablejumpdi_nospec): New define_expand.
4785 (*tablejump<mode>_internal1): Disable for
4786 -mno-speculate-indirect-jumps.
4787 (*tablejump<mode>_internal1_nospec): New define_insn.
4788 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4791 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4793 * caller-save.c (insert_save): Drop unnecessary parameter. All
4796 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4797 Richard Biener <rguenth@suse.de>
4800 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4801 return early, inline manually is_gimple_sizepos. Make sure if we
4802 call gimplify_expr we don't end up with a gimple constant.
4803 * tree.c (variably_modified_type_p): Don't return true for
4804 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4805 * gimplify.h (is_gimple_sizepos): Remove.
4807 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4809 PR tree-optimization/83857
4810 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4811 vectorizable_live_operation for pure SLP statements.
4812 (vectorizable_live_operation): Handle PHIs.
4814 2018-01-16 Richard Biener <rguenther@suse.de>
4816 PR tree-optimization/83867
4817 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4818 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4820 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4823 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4824 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4825 If off is not INTEGER_CST, issue a may not be aligned warning
4826 rather than isn't aligned. Use isn%'t rather than isn't.
4827 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4829 <case MULT_EXPR>: Improve the case when bottom and one of the
4830 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4831 operand, in that case check if the other operand is multiple of
4832 bottom divided by the INTEGER_CST operand.
4834 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4837 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4838 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4839 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4840 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4841 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4842 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4843 (pa_function_arg_size): New function.
4845 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4847 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4848 in a separate statement.
4850 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4852 PR tree-optimization/83847
4853 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4854 group gathers and scatters.
4856 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4858 PR rtl-optimization/86620
4859 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4861 PR rtl-optimization/83213
4862 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4863 to last if both are JUMP_INSNs.
4865 PR tree-optimization/83843
4866 * gimple-ssa-store-merging.c
4867 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4868 store_immediate_info for bswap/nop orig_stores.
4870 2018-01-15 Andrew Waterman <andrew@sifive.com>
4872 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4874 <UDIV>: Increase cost if !TARGET_DIV.
4876 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4878 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4879 (define_attr "cr_logical_3op"): New.
4880 (cceq_ior_compare): Adjust.
4881 (cceq_ior_compare_complement): Adjust.
4882 (*cceq_rev_compare): Adjust.
4883 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4884 (is_cracked_insn): Adjust.
4885 (insn_must_be_first_in_group): Adjust.
4886 * config/rs6000/40x.md: Adjust.
4887 * config/rs6000/440.md: Adjust.
4888 * config/rs6000/476.md: Adjust.
4889 * config/rs6000/601.md: Adjust.
4890 * config/rs6000/603.md: Adjust.
4891 * config/rs6000/6xx.md: Adjust.
4892 * config/rs6000/7450.md: Adjust.
4893 * config/rs6000/7xx.md: Adjust.
4894 * config/rs6000/8540.md: Adjust.
4895 * config/rs6000/cell.md: Adjust.
4896 * config/rs6000/e300c2c3.md: Adjust.
4897 * config/rs6000/e500mc.md: Adjust.
4898 * config/rs6000/e500mc64.md: Adjust.
4899 * config/rs6000/e5500.md: Adjust.
4900 * config/rs6000/e6500.md: Adjust.
4901 * config/rs6000/mpc.md: Adjust.
4902 * config/rs6000/power4.md: Adjust.
4903 * config/rs6000/power5.md: Adjust.
4904 * config/rs6000/power6.md: Adjust.
4905 * config/rs6000/power7.md: Adjust.
4906 * config/rs6000/power8.md: Adjust.
4907 * config/rs6000/power9.md: Adjust.
4908 * config/rs6000/rs64.md: Adjust.
4909 * config/rs6000/titan.md: Adjust.
4911 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4913 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4914 ix86_indirect_branch_register logic.
4916 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4918 * config/i386/constraints.md (Bs): Update
4919 ix86_indirect_branch_register check. Don't check
4920 ix86_indirect_branch_register with GOT_memory_operand.
4922 * config/i386/predicates.md (GOT_memory_operand): Don't check
4923 ix86_indirect_branch_register here.
4924 (GOT32_symbol_operand): Likewise.
4926 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4928 * config/i386/predicates.md (constant_call_address_operand):
4929 Rewrite ix86_indirect_branch_register logic.
4930 (sibcall_insn_operand): Likewise.
4932 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4934 * config/i386/constraints.md (Bs): Replace
4935 ix86_indirect_branch_thunk_register with
4936 ix86_indirect_branch_register.
4938 * config/i386/i386.md (indirect_jump): Likewise.
4939 (tablejump): Likewise.
4940 (*sibcall_memory): Likewise.
4941 (*sibcall_value_memory): Likewise.
4942 Peepholes of indirect call and jump via memory: Likewise.
4943 * config/i386/i386.opt: Likewise.
4944 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4945 (GOT_memory_operand): Likewise.
4946 (call_insn_operand): Likewise.
4947 (sibcall_insn_operand): Likewise.
4948 (GOT32_symbol_operand): Likewise.
4950 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4953 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4954 type rather than type addr's type points to.
4955 (expand_omp_atomic_mutex): Likewise.
4956 (expand_omp_atomic): Likewise.
4958 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4961 * config/i386/i386.c (output_indirect_thunk_function): Use
4962 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4963 for __x86_return_thunk.
4965 2018-01-15 Richard Biener <rguenther@suse.de>
4968 * expmed.c (extract_bit_field_1): Fix typo.
4970 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4973 * config/arm/iterators.md (VF): New mode iterator.
4974 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4975 Remove integer-related logic from pattern.
4976 (neon_vabd<mode>_3): Likewise.
4978 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4981 * common.opt (fstrict-overflow): No longer an alias.
4982 (fwrapv-pointer): New option.
4983 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4984 also for pointer types based on flag_wrapv_pointer.
4985 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4986 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4987 opts->x_flag_wrapv got set.
4988 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4989 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4990 POINTER_TYPE_OVERFLOW_UNDEFINED.
4991 * match.pd: Likewise in address comparison pattern.
4992 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4994 2018-01-15 Richard Biener <rguenther@suse.de>
4997 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4998 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4999 Reset type names to their identifier if their TYPE_DECL doesn't
5000 have linkage (and thus is used for ODR and devirt).
5001 (save_debug_info_for_decl): Remove.
5002 (save_debug_info_for_type): Likewise.
5003 (add_tree_to_fld_list): Adjust.
5004 * tree-pretty-print.c (dump_generic_node): Make dumping of
5005 type names more robust.
5007 2018-01-15 Richard Biener <rguenther@suse.de>
5009 * BASE-VER: Bump to 8.0.1.
5011 2018-01-14 Martin Sebor <msebor@redhat.com>
5014 * builtins.c (check_access): Avoid warning when the no-warning bit
5017 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5019 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5020 * ira-color (allocno_hard_regs_compare): Likewise.
5022 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5025 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5026 Use .pushsection/.popsection.
5028 2018-01-14 Martin Sebor <msebor@redhat.com>
5031 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5033 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5035 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5036 entry from extra_headers.
5037 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5038 extra_headers, make the list bitwise identical to the i?86-*-* one.
5040 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5042 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5043 -mcmodel=large with -mindirect-branch=thunk,
5044 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5045 -mfunction-return=thunk-extern.
5046 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5047 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5048 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5050 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5052 * config/i386/i386.c (print_reg): Print the name of the full
5053 integer register without '%'.
5054 (ix86_print_operand): Handle 'V'.
5055 * doc/extend.texi: Document 'V' modifier.
5057 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5059 * config/i386/constraints.md (Bs): Disallow memory operand for
5060 -mindirect-branch-register.
5062 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5063 (GOT_memory_operand): Likewise.
5064 (call_insn_operand): Likewise.
5065 (sibcall_insn_operand): Likewise.
5066 (GOT32_symbol_operand): Likewise.
5067 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5068 for -mindirect-branch-register.
5069 (tablejump): Likewise.
5070 (*sibcall_memory): Likewise.
5071 (*sibcall_value_memory): Likewise.
5072 Disallow peepholes of indirect call and jump via memory for
5073 -mindirect-branch-register.
5074 (*call_pop): Replace m with Bw.
5075 (*call_value_pop): Likewise.
5076 (*sibcall_pop_memory): Replace m with Bs.
5077 * config/i386/i386.opt (mindirect-branch-register): New option.
5078 * doc/invoke.texi: Document -mindirect-branch-register option.
5080 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5082 * config/i386/i386-protos.h (ix86_output_function_return): New.
5083 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5084 set function_return_type.
5085 (indirect_thunk_name): Add ret_p to indicate thunk for function
5087 (output_indirect_thunk_function): Pass false to
5088 indirect_thunk_name.
5089 (ix86_output_indirect_branch_via_reg): Likewise.
5090 (ix86_output_indirect_branch_via_push): Likewise.
5091 (output_indirect_thunk_function): Create alias for function
5092 return thunk if regno < 0.
5093 (ix86_output_function_return): New function.
5094 (ix86_handle_fndecl_attribute): Handle function_return.
5095 (ix86_attribute_table): Add function_return.
5096 * config/i386/i386.h (machine_function): Add
5097 function_return_type.
5098 * config/i386/i386.md (simple_return_internal): Use
5099 ix86_output_function_return.
5100 (simple_return_internal_long): Likewise.
5101 * config/i386/i386.opt (mfunction-return=): New option.
5102 (indirect_branch): Mention -mfunction-return=.
5103 * doc/extend.texi: Document function_return function attribute.
5104 * doc/invoke.texi: Document -mfunction-return= option.
5106 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5108 * config/i386/i386-opts.h (indirect_branch): New.
5109 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5110 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5111 with local indirect jump when converting indirect call and jump.
5112 (ix86_set_indirect_branch_type): New.
5113 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5114 (indirectlabelno): New.
5115 (indirect_thunk_needed): Likewise.
5116 (indirect_thunk_bnd_needed): Likewise.
5117 (indirect_thunks_used): Likewise.
5118 (indirect_thunks_bnd_used): Likewise.
5119 (INDIRECT_LABEL): Likewise.
5120 (indirect_thunk_name): Likewise.
5121 (output_indirect_thunk): Likewise.
5122 (output_indirect_thunk_function): Likewise.
5123 (ix86_output_indirect_branch_via_reg): Likewise.
5124 (ix86_output_indirect_branch_via_push): Likewise.
5125 (ix86_output_indirect_branch): Likewise.
5126 (ix86_output_indirect_jmp): Likewise.
5127 (ix86_code_end): Call output_indirect_thunk_function if needed.
5128 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5130 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5131 (ix86_attribute_table): Add indirect_branch.
5132 * config/i386/i386.h (machine_function): Add indirect_branch_type
5133 and has_local_indirect_jump.
5134 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5136 (tablejump): Likewise.
5137 (*indirect_jump): Use ix86_output_indirect_jmp.
5138 (*tablejump_1): Likewise.
5139 (simple_return_indirect_internal): Likewise.
5140 * config/i386/i386.opt (mindirect-branch=): New option.
5141 (indirect_branch): New.
5144 (thunk-inline): Likewise.
5145 (thunk-extern): Likewise.
5146 * doc/extend.texi: Document indirect_branch function attribute.
5147 * doc/invoke.texi: Document -mindirect-branch= option.
5149 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5152 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5154 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5156 * ipa-inline.c (want_inline_small_function_p): Return false if
5157 inlining has already failed with CIF_FINAL_ERROR.
5158 (update_caller_keys): Call want_inline_small_function_p before
5160 (update_callee_keys): Likewise.
5162 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5164 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5166 (rs6000_quadword_masked_address_p): Likewise.
5167 (quad_aligned_load_p): Likewise.
5168 (quad_aligned_store_p): Likewise.
5169 (const_load_sequence_p): Add comment to describe the outer-most loop.
5170 (mimic_memory_attributes_and_flags): New function.
5171 (rs6000_gen_stvx): Likewise.
5172 (replace_swapped_aligned_store): Likewise.
5173 (rs6000_gen_lvx): Likewise.
5174 (replace_swapped_aligned_load): Likewise.
5175 (replace_swapped_load_constant): Capitalize argument name in
5176 comment describing this function.
5177 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5178 and stores that access quad-word aligned addresses and replace
5179 with stvx or lvx instructions when appropriate.
5180 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5181 New function prototype.
5182 (rs6000_quadword_masked_address_p): Likewise.
5183 (rs6000_gen_lvx): Likewise.
5184 (rs6000_gen_stvx): Likewise.
5185 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5186 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5187 when memory address is aligned.
5188 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5189 this split to select lvx instruction when memory address is aligned.
5190 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5191 instruction when memory address is aligned.
5192 (*vsx_le_perm_load_v16qi): Likewise.
5193 (four unnamed splitters): Modify to select the stvx instruction
5194 when memory is aligned.
5196 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5198 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5199 which appears in the queue multiple times.
5201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5202 Alan Hayward <alan.hayward@arm.com>
5203 David Sherwood <david.sherwood@arm.com>
5205 * tree-vectorizer.h (vec_lower_bound): New structure.
5206 (_loop_vec_info): Add check_nonzero and lower_bounds.
5207 (LOOP_VINFO_CHECK_NONZERO): New macro.
5208 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5209 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5210 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5211 fields. Make seg_len the distance travelled, not including the
5213 (dr_direction_indicator): Declare.
5214 (dr_zero_step_indicator): Likewise.
5215 (dr_known_forward_stride_p): Likewise.
5216 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5218 (runtime_alias_check_p): Allow runtime alias checks with
5220 (operator ==): Compare access_size and align.
5221 (prune_runtime_alias_test_list): Rework for new distinction between
5222 the access_size and seg_len.
5223 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5225 (get_segment_min_max): New function.
5226 (create_intersect_range_checks): Use it.
5227 (dr_step_indicator): New function.
5228 (dr_direction_indicator): Likewise.
5229 (dr_zero_step_indicator): Likewise.
5230 (dr_known_forward_stride_p): Likewise.
5231 * tree-loop-distribution.c (data_ref_segment_size): Return
5232 DR_STEP * (niters - 1).
5233 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5235 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5236 (vect_preserves_scalar_order_p): New function, split out from...
5237 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5238 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5239 (vect_vfa_access_size): New function.
5240 (vect_vfa_align): Likewise.
5241 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5242 (dump_lower_bound): New function.
5243 (vect_check_lower_bound): Likewise.
5244 (vect_small_gap_p): Likewise.
5245 (vectorizable_with_step_bound_p): Likewise.
5246 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5247 depencies if the vectorization factor is 1. Convert the checks
5248 for nonzero steps into checks on the bounds of DR_STEP. Try using
5249 a bunds check for variable steps if the minimum required step is
5250 relatively small. Update calls to the dr_with_seg_len
5251 constructor and to vect_compile_time_alias.
5252 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5254 (vect_loop_versioning): Call it.
5255 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5257 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5260 Alan Hayward <alan.hayward@arm.com>
5261 David Sherwood <david.sherwood@arm.com>
5263 * doc/sourcebuild.texi (vect_scatter_store): Document.
5264 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5266 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5268 * genopinit.c (main): Add supports_vec_scatter_store and
5269 supports_vec_scatter_store_cached to target_optabs.
5270 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5271 IFN_MASK_SCATTER_STORE.
5272 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5274 * internal-fn.h (internal_store_fn_p): Declare.
5275 (internal_fn_stored_value_index): Likewise.
5276 * internal-fn.c (scatter_store_direct): New macro.
5277 (expand_scatter_store_optab_fn): New function.
5278 (direct_scatter_store_optab_supported_p): New macro.
5279 (internal_store_fn_p): New function.
5280 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5281 IFN_MASK_SCATTER_STORE.
5282 (internal_fn_mask_index): Likewise.
5283 (internal_fn_stored_value_index): New function.
5284 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5286 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5287 * optabs-query.c (supports_vec_scatter_store_p): New function.
5288 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5289 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5290 true for scatter stores.
5291 (vect_gather_scatter_fn_p): Handle scatter stores too.
5292 (vect_check_gather_scatter): Consider using scatter stores if
5293 supports_vec_scatter_store_p.
5294 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5296 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5297 internal_fn_stored_value_index.
5298 (check_load_store_masking): Handle scatter stores too.
5299 (vect_get_store_rhs): Make public.
5300 (vectorizable_call): Use internal_store_fn_p.
5301 (vectorizable_store): Handle scatter store internal functions.
5302 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5303 when deciding whether the end of the group has been reached.
5304 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5305 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5306 (mask_scatter_store<mode>): New insns.
5308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5309 Alan Hayward <alan.hayward@arm.com>
5310 David Sherwood <david.sherwood@arm.com>
5312 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5313 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5314 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5316 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5317 Use vect_truncate_gather_scatter_offset if we can't treat the
5318 operation as a normal gather load or scatter store.
5319 (get_group_load_store_type): Take the gather_scatter_info
5320 as argument. Try using a gather load or scatter store for
5321 single-element groups.
5322 (get_load_store_type): Update calls to get_group_load_store_type
5323 and vect_use_strided_gather_scatters_p.
5325 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5326 Alan Hayward <alan.hayward@arm.com>
5327 David Sherwood <david.sherwood@arm.com>
5329 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5330 optional tree argument.
5331 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5333 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5334 but continue to use the current value as a fallback.
5335 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5336 to compare the updates.
5337 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5338 (get_load_store_type): Use it when handling a strided access.
5339 (vect_get_strided_load_store_ops): New function.
5340 (vect_get_data_ptr_increment): Likewise.
5341 (vectorizable_load): Handle strided gather loads. Always pass
5342 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5345 Alan Hayward <alan.hayward@arm.com>
5346 David Sherwood <david.sherwood@arm.com>
5348 * doc/md.texi (gather_load@var{m}): Document.
5349 (mask_gather_load@var{m}): Likewise.
5350 * genopinit.c (main): Add supports_vec_gather_load and
5351 supports_vec_gather_load_cached to target_optabs.
5352 * optabs-tree.c (init_tree_optimization_optabs): Use
5353 ggc_cleared_alloc to allocate target_optabs.
5354 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5355 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5357 * internal-fn.h (internal_load_fn_p): Declare.
5358 (internal_gather_scatter_fn_p): Likewise.
5359 (internal_fn_mask_index): Likewise.
5360 (internal_gather_scatter_fn_supported_p): Likewise.
5361 * internal-fn.c (gather_load_direct): New macro.
5362 (expand_gather_load_optab_fn): New function.
5363 (direct_gather_load_optab_supported_p): New macro.
5364 (direct_internal_fn_optab): New function.
5365 (internal_load_fn_p): Likewise.
5366 (internal_gather_scatter_fn_p): Likewise.
5367 (internal_fn_mask_index): Likewise.
5368 (internal_gather_scatter_fn_supported_p): Likewise.
5369 * optabs-query.c (supports_at_least_one_mode_p): New function.
5370 (supports_vec_gather_load_p): Likewise.
5371 * optabs-query.h (supports_vec_gather_load_p): Declare.
5372 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5373 and memory_type field.
5374 (NUM_PATTERNS): Bump to 15.
5375 * tree-vect-data-refs.c: Include internal-fn.h.
5376 (vect_gather_scatter_fn_p): New function.
5377 (vect_describe_gather_scatter_call): Likewise.
5378 (vect_check_gather_scatter): Try using internal functions for
5379 gather loads. Recognize existing calls to a gather load function.
5380 (vect_analyze_data_refs): Consider using gather loads if
5381 supports_vec_gather_load_p.
5382 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5383 (vect_get_gather_scatter_offset_type): Likewise.
5384 (vect_convert_mask_for_vectype): Likewise.
5385 (vect_add_conversion_to_patterm): Likewise.
5386 (vect_try_gather_scatter_pattern): Likewise.
5387 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5388 (vect_vect_recog_func_ptrs): Add it.
5389 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5390 internal_fn_mask_index and internal_gather_scatter_fn_p.
5391 (check_load_store_masking): Take the gather_scatter_info as an
5392 argument and handle gather loads.
5393 (vect_get_gather_scatter_ops): New function.
5394 (vectorizable_call): Check internal_load_fn_p.
5395 (vectorizable_load): Likewise. Handle gather load internal
5397 (vectorizable_store): Update call to check_load_store_masking.
5398 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5399 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5400 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5401 (aarch64_gather_scale_operand_d): New predicates.
5402 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5403 (mask_gather_load<mode>): New insns.
5405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5406 Alan Hayward <alan.hayward@arm.com>
5407 David Sherwood <david.sherwood@arm.com>
5409 * optabs.def (fold_left_plus_optab): New optab.
5410 * doc/md.texi (fold_left_plus_@var{m}): Document.
5411 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5412 * internal-fn.c (fold_left_direct): Define.
5413 (expand_fold_left_optab_fn): Likewise.
5414 (direct_fold_left_optab_supported_p): Likewise.
5415 * fold-const-call.c (fold_const_fold_left): New function.
5416 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5417 * tree-parloops.c (valid_reduction_p): New function.
5418 (gather_scalar_reductions): Use it.
5419 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5420 (vect_finish_replace_stmt): Declare.
5421 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5422 (needs_fold_left_reduction_p): New function, split out from...
5423 (vect_is_simple_reduction): ...here. Accept reductions that
5424 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5425 (vect_force_simple_reduction): Also store the reduction type in
5426 the assignment's STMT_VINFO_REDUC_TYPE.
5427 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5428 (merge_with_identity): New function.
5429 (vect_expand_fold_left): Likewise.
5430 (vectorize_fold_left_reduction): Likewise.
5431 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5432 scalar phi in place for it. Check for target support and reject
5433 cases that would reassociate the operation. Defer the transform
5434 phase to vectorize_fold_left_reduction.
5435 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5436 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5437 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5441 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5442 call to ifc_temp_var.
5444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5445 Alan Hayward <alan.hayward@arm.com>
5446 David Sherwood <david.sherwood@arm.com>
5448 * target.def (legitimize_address_displacement): Take the original
5449 offset as a poly_int.
5450 * targhooks.h (default_legitimize_address_displacement): Update
5452 * targhooks.c (default_legitimize_address_displacement): Likewise.
5453 * doc/tm.texi: Regenerate.
5454 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5455 as an argument, moving assert of ad->disp == ad->disp_term to...
5456 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5457 Try calling targetm.legitimize_address_displacement before expanding
5458 the address rather than afterwards, and adjust for the new interface.
5459 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5460 Match the new hook interface. Handle SVE addresses.
5461 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5466 * Makefile.in (OBJS): Add early-remat.o.
5467 * target.def (select_early_remat_modes): New hook.
5468 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5469 * doc/tm.texi: Regenerate.
5470 * targhooks.h (default_select_early_remat_modes): Declare.
5471 * targhooks.c (default_select_early_remat_modes): New function.
5472 * timevar.def (TV_EARLY_REMAT): New timevar.
5473 * passes.def (pass_early_remat): New pass.
5474 * tree-pass.h (make_pass_early_remat): Declare.
5475 * early-remat.c: New file.
5476 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5478 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5481 Alan Hayward <alan.hayward@arm.com>
5482 David Sherwood <david.sherwood@arm.com>
5484 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5485 vfm1 with a bound_epilog parameter.
5486 (vect_do_peeling): Update calls accordingly, and move the prologue
5487 call earlier in the function. Treat the base bound_epilog as 0 for
5488 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5489 this base when peeling for gaps.
5490 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5491 with fully-masked loops.
5492 (vect_estimate_min_profitable_iters): Handle the single peeled
5493 iteration in that case.
5495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5496 Alan Hayward <alan.hayward@arm.com>
5497 David Sherwood <david.sherwood@arm.com>
5499 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5500 single-element interleaving even if the size is not a power of 2.
5501 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5502 accesses for single-element interleaving if the group size is
5505 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5506 Alan Hayward <alan.hayward@arm.com>
5507 David Sherwood <david.sherwood@arm.com>
5509 * doc/md.texi (fold_extract_last_@var{m}): Document.
5510 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5511 * optabs.def (fold_extract_last_optab): New optab.
5512 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5513 * internal-fn.c (fold_extract_direct): New macro.
5514 (expand_fold_extract_optab_fn): Likewise.
5515 (direct_fold_extract_optab_supported_p): Likewise.
5516 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5517 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5518 EXTRACT_LAST_REDUCTION.
5519 (get_initial_def_for_reduction): Do not create an initial vector
5520 for EXTRACT_LAST_REDUCTION reductions.
5521 (vectorizable_reduction): Leave the scalar phi in place for
5522 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5523 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5524 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5525 transform phase to vectorizable_condition.
5526 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5528 (vect_finish_stmt_generation): ...here.
5529 (vect_finish_replace_stmt): New function.
5530 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5531 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5533 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5536 Alan Hayward <alan.hayward@arm.com>
5537 David Sherwood <david.sherwood@arm.com>
5539 * doc/md.texi (extract_last_@var{m}): Document.
5540 * optabs.def (extract_last_optab): New optab.
5541 * internal-fn.def (EXTRACT_LAST): New internal function.
5542 * internal-fn.c (cond_unary_direct): New macro.
5543 (expand_cond_unary_optab_fn): Likewise.
5544 (direct_cond_unary_optab_supported_p): Likewise.
5545 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5546 loops using EXTRACT_LAST.
5547 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5548 (extract_last_<mode>): ...this optab.
5549 (vec_extract<mode><Vel>): Update accordingly.
5551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5552 Alan Hayward <alan.hayward@arm.com>
5553 David Sherwood <david.sherwood@arm.com>
5555 * target.def (empty_mask_is_expensive): New hook.
5556 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5557 * doc/tm.texi: Regenerate.
5558 * targhooks.h (default_empty_mask_is_expensive): Declare.
5559 * targhooks.c (default_empty_mask_is_expensive): New function.
5560 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5561 if the target says that empty masks are expensive.
5562 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5564 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5567 Alan Hayward <alan.hayward@arm.com>
5568 David Sherwood <david.sherwood@arm.com>
5570 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5571 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5572 (vect_use_loop_mask_for_alignment_p): New function.
5573 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5574 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5575 niters_skip argument. Make sure that the first niters_skip elements
5576 of the first iteration are inactive.
5577 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5578 Update call to vect_set_loop_masks_directly.
5579 (get_misalign_in_elems): New function, split out from...
5580 (vect_gen_prolog_loop_niters): ...here.
5581 (vect_update_init_of_dr): Take a code argument that specifies whether
5582 the adjustment should be added or subtracted.
5583 (vect_update_init_of_drs): Likewise.
5584 (vect_prepare_for_masked_peels): New function.
5585 (vect_do_peeling): Skip prologue peeling if we're using a mask
5586 instead. Update call to vect_update_inits_of_drs.
5587 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5589 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5590 alignment. Do not include the number of peeled iterations in
5591 the minimum threshold in that case.
5592 (vectorizable_induction): Adjust the start value down by
5593 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5594 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5595 Take the number of skipped iterations into account when calculating
5597 * tree-vect-stmts.c (vect_gen_while_not): New function.
5599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5600 Alan Hayward <alan.hayward@arm.com>
5601 David Sherwood <david.sherwood@arm.com>
5603 * doc/sourcebuild.texi (vect_fully_masked): Document.
5604 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5606 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5608 (vect_analyze_loop_2): ...here. Don't check the vectorization
5609 factor against the number of loop iterations if the loop is
5612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5613 Alan Hayward <alan.hayward@arm.com>
5614 David Sherwood <david.sherwood@arm.com>
5616 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5617 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5618 (dump_groups): Update accordingly.
5619 (iv_use::mem_type): New member variable.
5620 (address_p): New function.
5621 (record_use): Add a mem_type argument and initialize the new
5623 (record_group_use): Add a mem_type argument. Use address_p.
5624 Remove obsolete null checks of base_object. Update call to record_use.
5625 (find_interesting_uses_op): Update call to record_group_use.
5626 (find_interesting_uses_cond): Likewise.
5627 (find_interesting_uses_address): Likewise.
5628 (get_mem_type_for_internal_fn): New function.
5629 (find_address_like_use): Likewise.
5630 (find_interesting_uses_stmt): Try find_address_like_use before
5631 calling find_interesting_uses_op.
5632 (addr_offset_valid_p): Use the iv mem_type field as the type
5633 of the addressed memory.
5634 (add_autoinc_candidates): Likewise.
5635 (get_address_cost): Likewise.
5636 (split_small_address_groups_p): Use address_p.
5637 (split_address_groups): Likewise.
5638 (add_iv_candidate_for_use): Likewise.
5639 (autoinc_possible_for_pair): Likewise.
5640 (rewrite_groups): Likewise.
5641 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5642 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5643 (get_alias_ptr_type_for_ptr_address): New function.
5644 (rewrite_use_address): Rewrite address uses in calls that were
5645 identified by find_address_like_use.
5647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5648 Alan Hayward <alan.hayward@arm.com>
5649 David Sherwood <david.sherwood@arm.com>
5651 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5653 * gimple-expr.h (is_gimple_addressable: Likewise.
5654 * gimple-expr.c (is_gimple_address): Likewise.
5655 * internal-fn.c (expand_call_mem_ref): New function.
5656 (expand_mask_load_optab_fn): Use it.
5657 (expand_mask_store_optab_fn): Likewise.
5659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5660 Alan Hayward <alan.hayward@arm.com>
5661 David Sherwood <david.sherwood@arm.com>
5663 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5664 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5665 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5666 (cond_umax@var{mode}): Document.
5667 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5668 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5669 (cond_umin_optab, cond_umax_optab): New optabs.
5670 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5671 (COND_IOR, COND_XOR): New internal functions.
5672 * internal-fn.h (get_conditional_internal_fn): Declare.
5673 * internal-fn.c (cond_binary_direct): New macro.
5674 (expand_cond_binary_optab_fn): Likewise.
5675 (direct_cond_binary_optab_supported_p): Likewise.
5676 (get_conditional_internal_fn): New function.
5677 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5678 Cope with reduction statements that are vectorized as calls rather
5680 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5681 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5682 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5683 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5684 (UNSPEC_COND_EOR): New unspecs.
5685 (optab): Add mappings for them.
5686 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5687 (sve_int_op, sve_fp_op): New int attributes.
5689 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5690 Alan Hayward <alan.hayward@arm.com>
5691 David Sherwood <david.sherwood@arm.com>
5693 * optabs.def (while_ult_optab): New optab.
5694 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5695 * internal-fn.def (WHILE_ULT): New internal function.
5696 * internal-fn.h (direct_internal_fn_supported_p): New override
5697 that takes two types as argument.
5698 * internal-fn.c (while_direct): New macro.
5699 (expand_while_optab_fn): New function.
5700 (convert_optab_supported_p): Likewise.
5701 (direct_while_optab_supported_p): New macro.
5702 * wide-int.h (wi::udiv_ceil): New function.
5703 * tree-vectorizer.h (rgroup_masks): New structure.
5704 (vec_loop_masks): New typedef.
5705 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5707 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5708 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5709 (vect_max_vf): New function.
5710 (slpeel_make_loop_iterate_ntimes): Delete.
5711 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5712 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5713 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5714 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5715 internal-fn.h, stor-layout.h and optabs-query.h.
5716 (vect_set_loop_mask): New function.
5717 (add_preheader_seq): Likewise.
5718 (add_header_seq): Likewise.
5719 (interleave_supported_p): Likewise.
5720 (vect_maybe_permute_loop_masks): Likewise.
5721 (vect_set_loop_masks_directly): Likewise.
5722 (vect_set_loop_condition_masked): Likewise.
5723 (vect_set_loop_condition_unmasked): New function, split out from
5724 slpeel_make_loop_iterate_ntimes.
5725 (slpeel_make_loop_iterate_ntimes): Rename to..
5726 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5727 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5728 (vect_do_peeling): Update call accordingly.
5729 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5731 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5732 mask_compare_type, can_fully_mask_p and fully_masked_p.
5733 (release_vec_loop_masks): New function.
5734 (_loop_vec_info): Use it to free the loop masks.
5735 (can_produce_all_loop_masks_p): New function.
5736 (vect_get_max_nscalars_per_iter): Likewise.
5737 (vect_verify_full_masking): Likewise.
5738 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5739 retries, and free the mask rgroups before retrying. Check loop-wide
5740 reasons for disallowing fully-masked loops. Make the final decision
5741 about whether use a fully-masked loop or not.
5742 (vect_estimate_min_profitable_iters): Do not assume that peeling
5743 for the number of iterations will be needed for fully-masked loops.
5744 (vectorizable_reduction): Disable fully-masked loops.
5745 (vectorizable_live_operation): Likewise.
5746 (vect_halve_mask_nunits): New function.
5747 (vect_double_mask_nunits): Likewise.
5748 (vect_record_loop_mask): Likewise.
5749 (vect_get_loop_mask): Likewise.
5750 (vect_transform_loop): Handle the case in which the final loop
5751 iteration might handle a partial vector. Call vect_set_loop_condition
5752 instead of slpeel_make_loop_iterate_ntimes.
5753 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5754 (check_load_store_masking): New function.
5755 (prepare_load_store_mask): Likewise.
5756 (vectorizable_store): Handle fully-masked loops.
5757 (vectorizable_load): Likewise.
5758 (supportable_widening_operation): Use vect_halve_mask_nunits for
5760 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5761 (vect_gen_while): New function.
5762 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5763 (aarch64_uqdec<mode>): New insn.
5765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5766 Alan Hayward <alan.hayward@arm.com>
5767 David Sherwood <david.sherwood@arm.com>
5769 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5770 (reduc_xor_scal_optab): New optabs.
5771 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5772 (reduc_xor_scal_@var{m}): Document.
5773 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5774 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5776 * fold-const-call.c (fold_const_call): Handle them.
5777 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5778 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5779 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5780 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5781 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5782 (UNSPEC_XORV): New unspecs.
5783 (optab): Add entries for them.
5784 (BITWISEV): New int iterator.
5785 (bit_reduc_op): New int attributes.
5787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5788 Alan Hayward <alan.hayward@arm.com>
5789 David Sherwood <david.sherwood@arm.com>
5791 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5792 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5793 * optabs.def (vec_shl_insert_optab): New optab.
5794 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5795 (duplicate_and_interleave): Likewise.
5796 * tree-vect-loop.c: Include internal-fn.h.
5797 (neutral_op_for_slp_reduction): New function, split out from
5798 get_initial_defs_for_reduction.
5799 (get_initial_def_for_reduction): Handle option 2 for variable-length
5800 vectors by loading the neutral value into a vector and then shifting
5801 the initial value into element 0.
5802 (get_initial_defs_for_reduction): Replace the code argument with
5803 the neutral value calculated by neutral_op_for_slp_reduction.
5804 Use gimple_build_vector for constant-length vectors.
5805 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5806 but the first group_size elements have a neutral value.
5807 Use duplicate_and_interleave otherwise.
5808 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5809 Update call to get_initial_defs_for_reduction. Handle SLP
5810 reductions for variable-length vectors by creating one vector
5811 result for each scalar result, with the elements associated
5812 with other scalar results stubbed out with the neutral value.
5813 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5814 Require IFN_VEC_SHL_INSERT for double reductions on
5815 variable-length vectors, or SLP reductions that have
5816 a neutral value. Require can_duplicate_and_interleave_p
5817 support for variable-length unchained SLP reductions if there
5818 is no neutral value, such as for MIN/MAX reductions. Also require
5819 the number of vector elements to be a multiple of the number of
5820 SLP statements when doing variable-length unchained SLP reductions.
5821 Update call to vect_create_epilog_for_reduction.
5822 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5823 and remove initial values.
5824 (duplicate_and_interleave): Make public.
5825 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5826 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5829 Alan Hayward <alan.hayward@arm.com>
5830 David Sherwood <david.sherwood@arm.com>
5832 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5833 (can_duplicate_and_interleave_p): New function.
5834 (vect_get_and_check_slp_defs): Take the vector of statements
5835 rather than just the current one. Remove excess parentheses.
5836 Restriction rejectinon of vect_constant_def and vect_external_def
5837 for variable-length vectors to boolean types, or types for which
5838 can_duplicate_and_interleave_p is false.
5839 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5840 (duplicate_and_interleave): New function.
5841 (vect_get_constant_vectors): Use gimple_build_vector for
5842 constant-length vectors and suitable variable-length constant
5843 vectors. Use duplicate_and_interleave for other variable-length
5844 vectors. Don't defer the update when inserting new statements.
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-loop.c (vect_estimate_min_profitable_iters): Make sure
5851 min_profitable_iters doesn't go negative.
5853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5854 Alan Hayward <alan.hayward@arm.com>
5855 David Sherwood <david.sherwood@arm.com>
5857 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5858 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5859 * optabs.def (vec_mask_load_lanes_optab): New optab.
5860 (vec_mask_store_lanes_optab): Likewise.
5861 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5862 (MASK_STORE_LANES): Likewise.
5863 * internal-fn.c (mask_load_lanes_direct): New macro.
5864 (mask_store_lanes_direct): Likewise.
5865 (expand_mask_load_optab_fn): Handle masked operations.
5866 (expand_mask_load_lanes_optab_fn): New macro.
5867 (expand_mask_store_optab_fn): Handle masked operations.
5868 (expand_mask_store_lanes_optab_fn): New macro.
5869 (direct_mask_load_lanes_optab_supported_p): Likewise.
5870 (direct_mask_store_lanes_optab_supported_p): Likewise.
5871 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5873 (vect_load_lanes_supported): Likewise.
5874 * tree-vect-data-refs.c (strip_conversion): New function.
5875 (can_group_stmts_p): Likewise.
5876 (vect_analyze_data_ref_accesses): Use it instead of checking
5877 for a pair of assignments.
5878 (vect_store_lanes_supported): Take a masked_p parameter.
5879 (vect_load_lanes_supported): Likewise.
5880 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5881 vect_store_lanes_supported and vect_load_lanes_supported.
5882 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5883 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5884 parameter. Don't allow gaps for masked accesses.
5885 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5886 and vect_load_lanes_supported.
5887 (get_load_store_type): Take a masked_p parameter and update
5888 call to get_group_load_store_type.
5889 (vectorizable_store): Update call to get_load_store_type.
5890 Handle IFN_MASK_STORE_LANES.
5891 (vectorizable_load): Update call to get_load_store_type.
5892 Handle IFN_MASK_LOAD_LANES.
5894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5895 Alan Hayward <alan.hayward@arm.com>
5896 David Sherwood <david.sherwood@arm.com>
5898 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5900 * config/aarch64/aarch64-protos.h
5901 (aarch64_sve_struct_memory_operand_p): Declare.
5902 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5903 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5904 (VPRED, vpred): Handle SVE structure modes.
5905 * config/aarch64/constraints.md (Utx): New constraint.
5906 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5907 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5908 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5909 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5910 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5911 structure modes. Split into pieces after RA.
5912 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5913 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5915 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5916 SVE structure modes.
5917 (aarch64_classify_address): Likewise.
5918 (sizetochar): Move earlier in file.
5919 (aarch64_print_operand): Handle SVE register lists.
5920 (aarch64_array_mode): New function.
5921 (aarch64_sve_struct_memory_operand_p): Likewise.
5922 (TARGET_ARRAY_MODE): Redefine.
5924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5925 Alan Hayward <alan.hayward@arm.com>
5926 David Sherwood <david.sherwood@arm.com>
5928 * target.def (array_mode): New target hook.
5929 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5930 * doc/tm.texi: Regenerate.
5931 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5932 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5933 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5935 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5939 Alan Hayward <alan.hayward@arm.com>
5940 David Sherwood <david.sherwood@arm.com>
5942 * fold-const.c (fold_binary_loc): Check the argument types
5943 rather than the result type when testing for a vector operation.
5945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5947 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5948 * doc/tm.texi: Regenerate.
5950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5951 Alan Hayward <alan.hayward@arm.com>
5952 David Sherwood <david.sherwood@arm.com>
5954 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5955 (sve): Document new AArch64 extension.
5956 * doc/md.texi (w): Extend the description of the AArch64
5957 constraint to include SVE vectors.
5958 (Upl, Upa): Document new AArch64 predicate constraints.
5959 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5961 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5962 (msve-vector-bits=): New option.
5963 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5964 SVE when these are disabled.
5965 (sve): New extension.
5966 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5967 modes. Adjust their number of units based on aarch64_sve_vg.
5968 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5969 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5970 aarch64_addr_query_type.
5971 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5972 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5973 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5974 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5975 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5976 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5977 (aarch64_simd_imm_zero_p): Delete.
5978 (aarch64_check_zero_based_sve_index_immediate): Declare.
5979 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5980 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5981 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5982 (aarch64_sve_float_mul_immediate_p): Likewise.
5983 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5985 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5986 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5987 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5988 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5989 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5990 (aarch64_regmode_natural_size): Likewise.
5991 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5992 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5994 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5995 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5996 for VG and the SVE predicate registers.
5997 (V_ALIASES): Add a "z"-prefixed alias.
5998 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5999 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6000 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6001 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6002 (REG_CLASS_NAMES): Add entries for them.
6003 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6004 and the predicate registers.
6005 (aarch64_sve_vg): Declare.
6006 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6007 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6008 (REGMODE_NATURAL_SIZE): Define.
6009 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6011 * config/aarch64/aarch64.c: Include cfgrtl.h.
6012 (simd_immediate_info): Add a constructor for series vectors,
6013 and an associated step field.
6014 (aarch64_sve_vg): New variable.
6015 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6016 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6017 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6018 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6019 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6020 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6021 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6022 (aarch64_get_mask_mode): New functions.
6023 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6024 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6025 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6026 predicate modes and predicate registers. Explicitly restrict
6027 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6028 to store a vector mode if it is recognized by
6029 aarch64_classify_vector_mode.
6030 (aarch64_regmode_natural_size): New function.
6031 (aarch64_hard_regno_caller_save_mode): Return the original mode
6033 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6034 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6035 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6036 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6038 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6039 does not overlap dest if the function is frame-related. Handle
6041 (aarch64_split_add_offset): New function.
6042 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6043 them aarch64_add_offset.
6044 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6045 and update call to aarch64_sub_sp.
6046 (aarch64_add_cfa_expression): New function.
6047 (aarch64_expand_prologue): Pass extra temporary registers to the
6048 functions above. Handle the case in which we need to emit new
6049 DW_CFA_expressions for registers that were originally saved
6050 relative to the stack pointer, but now have to be expressed
6051 relative to the frame pointer.
6052 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6054 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6055 IP0 and IP1 values for SVE frames.
6056 (aarch64_expand_vec_series): New function.
6057 (aarch64_expand_sve_widened_duplicate): Likewise.
6058 (aarch64_expand_sve_const_vector): Likewise.
6059 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6060 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6061 into the register, rather than emitting a SET directly.
6062 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6063 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6064 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6065 (offset_9bit_signed_scaled_p): New functions.
6066 (aarch64_replicate_bitmask_imm): New function.
6067 (aarch64_bitmask_imm): Use it.
6068 (aarch64_cannot_force_const_mem): Reject expressions involving
6069 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6070 (aarch64_classify_index): Handle SVE indices, by requiring
6071 a plain register index with a scale that matches the element size.
6072 (aarch64_classify_address): Handle SVE addresses. Assert that
6073 the mode of the address is VOIDmode or an integer mode.
6074 Update call to aarch64_classify_symbol.
6075 (aarch64_classify_symbolic_expression): Update call to
6076 aarch64_classify_symbol.
6077 (aarch64_const_vec_all_in_range_p): New function.
6078 (aarch64_print_vector_float_operand): Likewise.
6079 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6080 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6081 and the FP immediates 1.0 and 0.5.
6082 (aarch64_print_address_internal): Handle SVE addresses.
6083 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6084 (aarch64_regno_regclass): Handle predicate registers.
6085 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6087 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6088 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6089 (aarch64_convert_sve_vector_bits): New function.
6090 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6091 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6093 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6094 Handle SVE vector and predicate modes. Accept VL-based constants
6095 that need only one temporary register, and VL offsets that require
6096 no temporary registers.
6097 (aarch64_conditional_register_usage): Mark the predicate registers
6098 as fixed if SVE isn't available.
6099 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6100 Return true for SVE vector and predicate modes.
6101 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6102 rather than an unsigned int. Handle SVE modes.
6103 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6105 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6107 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6108 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6109 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6110 (aarch64_sve_float_mul_immediate_p): New functions.
6111 (aarch64_sve_valid_immediate): New function.
6112 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6113 Explicitly reject structure modes. Check for INDEX constants.
6114 Handle PTRUE and PFALSE constants.
6115 (aarch64_check_zero_based_sve_index_immediate): New function.
6116 (aarch64_simd_imm_zero_p): Delete.
6117 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6118 vector modes. Accept constants in the range of CNT[BHWD].
6119 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6120 ask for an Advanced SIMD mode.
6121 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6122 (aarch64_simd_vector_alignment): Handle SVE predicates.
6123 (aarch64_vectorize_preferred_vector_alignment): New function.
6124 (aarch64_simd_vector_alignment_reachable): Use it instead of
6126 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6127 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6129 (MAX_VECT_LEN): Delete.
6130 (expand_vec_perm_d): Add a vec_flags field.
6131 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6132 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6133 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6135 (aarch64_evpc_rev): Rename to...
6136 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6137 (aarch64_evpc_rev_global): New function.
6138 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6139 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6141 (aarch64_evpc_sve_tbl): New function.
6142 (aarch64_expand_vec_perm_const_1): Update after rename of
6143 aarch64_evpc_rev. Handle SVE permutes too, trying
6144 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6145 than aarch64_evpc_tbl.
6146 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6147 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6148 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6149 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6150 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6151 (aarch64_expand_sve_vcond): New functions.
6152 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6153 of aarch64_vector_mode_p.
6154 (aarch64_dwarf_poly_indeterminate_value): New function.
6155 (aarch64_compute_pressure_classes): Likewise.
6156 (aarch64_can_change_mode_class): Likewise.
6157 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6158 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6159 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6160 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6161 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6162 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6163 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6164 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6166 (Dn, Dl, Dr): Accept const as well as const_vector.
6167 (Dz): Likewise. Compare against CONST0_RTX.
6168 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6169 of "vector" where appropriate.
6170 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6171 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6172 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6173 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6174 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6175 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6176 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6177 (v_int_equiv): Extend to SVE modes.
6178 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6180 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6181 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6182 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6183 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6184 (SVE_COND_FP_CMP): New int iterators.
6185 (perm_hilo): Handle the new unpack unspecs.
6186 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6188 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6189 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6190 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6191 (aarch64_equality_operator, aarch64_constant_vector_operand)
6192 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6193 (aarch64_sve_nonimmediate_operand): Likewise.
6194 (aarch64_sve_general_operand): Likewise.
6195 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6196 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6197 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6198 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6199 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6200 (aarch64_sve_float_arith_immediate): Likewise.
6201 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6202 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6203 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6204 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6205 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6206 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6207 (aarch64_sve_float_arith_operand): Likewise.
6208 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6209 (aarch64_sve_float_mul_operand): Likewise.
6210 (aarch64_sve_vec_perm_operand): Likewise.
6211 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6212 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6213 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6214 as well as const_vector.
6215 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6216 in file. Use CONST0_RTX and CONSTM1_RTX.
6217 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6218 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6219 Use aarch64_simd_imm_zero.
6220 * config/aarch64/aarch64-sve.md: New file.
6221 * config/aarch64/aarch64.md: Include it.
6222 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6223 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6224 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6225 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6226 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6227 (sve): New attribute.
6228 (enabled): Disable instructions with the sve attribute unless
6230 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6231 aarch64_expand_mov_immediate.
6232 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6233 CNT[BHSD] immediates.
6234 (movti): Split CONST_POLY_INT moves into two halves.
6235 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6236 Split additions that need a temporary here if the destination
6237 is the stack pointer.
6238 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6239 (*add<mode>3_poly_1): New instruction.
6240 (set_clobber_cc): New expander.
6242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6244 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6245 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6246 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6247 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6248 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6249 Change innermode from fixed_mode_size to machine_mode.
6250 (simplify_subreg): Update call accordingly. Handle a constant-sized
6251 subreg of a variable-length CONST_VECTOR.
6253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6254 Alan Hayward <alan.hayward@arm.com>
6255 David Sherwood <david.sherwood@arm.com>
6257 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6258 (add_offset_to_base): New function, split out from...
6259 (create_mem_ref): ...here. When handling a scale other than 1,
6260 check first whether the address is valid without the offset.
6261 Add it into the base if so, leaving the index and scale as-is.
6263 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6266 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6267 fold_for_warn before checking if arg2 is INTEGER_CST.
6269 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6271 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6272 (store_multiple_operation): Delete.
6273 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6274 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6275 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6276 guarded by TARGET_STRING.
6277 (rs6000_output_load_multiple): Delete.
6278 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6279 OPTION_MASK_STRING / TARGET_STRING handling.
6280 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6281 (const rs6000_opt_masks) <"string">: Change mask to 0.
6282 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6283 (MASK_STRING): Delete.
6284 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6286 (load_multiple): Delete.
6293 (store_multiple): Delete.
6300 (movmemsi_8reg): Delete.
6301 (corresponding unnamed define_insn): Delete.
6302 (movmemsi_6reg): Delete.
6303 (corresponding unnamed define_insn): Delete.
6304 (movmemsi_4reg): Delete.
6305 (corresponding unnamed define_insn): Delete.
6306 (movmemsi_2reg): Delete.
6307 (corresponding unnamed define_insn): Delete.
6308 (movmemsi_1reg): Delete.
6309 (corresponding unnamed define_insn): Delete.
6310 * config/rs6000/rs6000.opt (mno-string): New.
6311 (mstring): Replace by deprecation warning stub.
6312 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6314 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6316 * regrename.c (regrename_do_replace): If replacing the same
6317 reg multiple times, try to reuse last created gen_raw_REG.
6320 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6321 main to workaround a bug in GDB.
6323 2018-01-12 Tom de Vries <tom@codesourcery.com>
6326 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6328 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6330 PR rtl-optimization/80481
6331 * ira-color.c (get_cap_member): New function.
6332 (allocnos_conflict_by_live_ranges_p): Use it.
6333 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6334 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6336 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6339 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6340 (*saddl_se_1): Ditto.
6342 (*ssubl_se_1): Ditto.
6344 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6346 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6347 rather than wi::to_widest for DR_INITs.
6348 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6349 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6350 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6352 (vect_analyze_group_access_1): Note that here.
6354 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6356 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6357 polynomial type sizes.
6359 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6361 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6362 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6363 (gimple_add_tmp_var): Likewise.
6365 2018-01-12 Martin Liska <mliska@suse.cz>
6367 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6368 (gimple_alloc_sizes): Likewise.
6369 (dump_gimple_statistics): Use PRIu64 in printf format.
6370 * gimple.h: Change uint64_t to int.
6372 2018-01-12 Martin Liska <mliska@suse.cz>
6374 * tree-core.h: Use uint64_t instead of int.
6375 * tree.c (tree_node_counts): Likewise.
6376 (tree_node_sizes): Likewise.
6377 (dump_tree_statistics): Use PRIu64 in printf format.
6379 2018-01-12 Martin Liska <mliska@suse.cz>
6381 * Makefile.in: As qsort_chk is implemented in vec.c, add
6382 vec.o to linkage of gencfn-macros.
6383 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6384 passing the info to record_node_allocation_statistics.
6385 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6387 * ggc-common.c (struct ggc_usage): Add operator== and use
6388 it in operator< and compare function.
6389 * mem-stats.h (struct mem_usage): Likewise.
6390 * vec.c (struct vec_usage): Remove operator< and compare
6391 function. Can be simply inherited.
6393 2018-01-12 Martin Jambor <mjambor@suse.cz>
6396 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6397 * tree-ssa-math-opts.c: Include domwalk.h.
6398 (convert_mult_to_fma_1): New function.
6399 (fma_transformation_info): New type.
6400 (fma_deferring_state): Likewise.
6401 (cancel_fma_deferring): New function.
6402 (result_of_phi): Likewise.
6403 (last_fma_candidate_feeds_initial_phi): Likewise.
6404 (convert_mult_to_fma): Added deferring logic, split actual
6405 transformation to convert_mult_to_fma_1.
6406 (math_opts_dom_walker): New type.
6407 (math_opts_dom_walker::after_dom_children): New method, body moved
6408 here from pass_optimize_widening_mul::execute, added deferring logic
6410 (pass_optimize_widening_mul::execute): Moved most of code to
6411 math_opts_dom_walker::after_dom_children.
6412 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6413 * config/i386/i386.c (ix86_option_override_internal): Added
6414 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6416 2018-01-12 Richard Biener <rguenther@suse.de>
6419 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6420 inline instance vars.
6422 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6425 * config/rx/rx.c (rx_is_restricted_memory_address):
6428 2018-01-12 Richard Biener <rguenther@suse.de>
6430 PR tree-optimization/80846
6431 * target.def (split_reduction): New target hook.
6432 * targhooks.c (default_split_reduction): New function.
6433 * targhooks.h (default_split_reduction): Declare.
6434 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6435 target requests first reduce vectors by combining low and high
6437 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6438 (get_vectype_for_scalar_type_and_size): Export.
6439 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6440 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6441 * doc/tm.texi: Regenerate.
6442 * config/i386/i386.c (ix86_split_reduction): Implement
6443 TARGET_VECTORIZE_SPLIT_REDUCTION.
6445 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6448 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6449 in PIC mode except for TARGET_VXWORKS_RTP.
6450 * config/sparc/sparc.c: Include cfgrtl.h.
6451 (TARGET_INIT_PIC_REG): Define.
6452 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6453 (sparc_pic_register_p): New predicate.
6454 (sparc_legitimate_address_p): Use it.
6455 (sparc_legitimize_pic_address): Likewise.
6456 (sparc_delegitimize_address): Likewise.
6457 (sparc_mode_dependent_address_p): Likewise.
6458 (gen_load_pcrel_sym): Remove 4th parameter.
6459 (load_got_register): Adjust call to above. Remove obsolete stuff.
6460 (sparc_expand_prologue): Do not call load_got_register here.
6461 (sparc_flat_expand_prologue): Likewise.
6462 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6463 (sparc_use_pseudo_pic_reg): New function.
6464 (sparc_init_pic_reg): Likewise.
6465 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6466 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6468 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6470 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6471 Add item for branch_cost.
6473 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6475 PR rtl-optimization/83565
6476 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6477 not extend the result to a larger mode for rotate operations.
6478 (num_sign_bit_copies1): Likewise.
6480 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6483 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6485 Use values-Xc.o for -pedantic.
6486 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6488 2018-01-12 Martin Liska <mliska@suse.cz>
6491 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6493 (possible_polymorphic_call_targets): Use it.
6494 (ipa_devirt): Likewise.
6496 2018-01-12 Martin Liska <mliska@suse.cz>
6498 * profile-count.h (enum profile_quality): Use 0 as invalid
6499 enum value of profile_quality.
6501 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6503 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6504 -mext-string options.
6506 2018-01-12 Richard Biener <rguenther@suse.de>
6508 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6509 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6510 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6512 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6514 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6516 * configure.ac (--with-long-double-format): Add support for the
6517 configuration option to change the default long double format on
6519 * config.gcc (powerpc*-linux*-*): Likewise.
6520 * configure: Regenerate.
6521 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6522 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6523 used without modification.
6525 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6527 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6528 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6529 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6530 MISC_BUILTIN_SPEC_BARRIER.
6531 (rs6000_init_builtins): Likewise.
6532 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6534 (speculation_barrier): New define_insn.
6535 * doc/extend.texi: Document __builtin_speculation_barrier.
6537 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6540 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6541 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6542 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6544 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6545 integral modes instead of "ss" and "sd".
6546 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6547 vectors with 32-bit and 64-bit elements.
6548 (vecdupssescalarmodesuffix): New mode attribute.
6549 (vec_dup<mode>): Use it.
6551 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6554 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6555 frame if argument is passed on stack.
6557 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6560 * ree.c (combine_reaching_defs): Optimize also
6561 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6562 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6564 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6567 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6569 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6572 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6573 after they are computed.
6575 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6577 PR tree-optimization/83695
6578 * gimple-loop-linterchange.cc
6579 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6580 reset cached scev information after interchange.
6581 (pass_linterchange::execute): Remove call to scev_reset_htab.
6583 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6585 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6586 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6587 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6588 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6589 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6590 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6591 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6592 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6593 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6594 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6595 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6596 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6597 (V_lane_reg): Likewise.
6598 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6600 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6601 (vfmal_lane_low<mode>_intrinsic,
6602 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6603 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6604 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6605 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6606 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6607 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6609 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6611 * config/arm/arm-cpus.in (fp16fml): New feature.
6612 (ALL_SIMD): Add fp16fml.
6613 (armv8.2-a): Add fp16fml as an option.
6614 (armv8.3-a): Likewise.
6615 (armv8.4-a): Add fp16fml as part of fp16.
6616 * config/arm/arm.h (TARGET_FP16FML): Define.
6617 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6619 * config/arm/arm-modes.def (V2HF): Define.
6620 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6621 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6622 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6623 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6624 vfmsl_low, vfmsl_high): New set of builtins.
6625 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6626 (vfml_op): New code attribute.
6627 (VFMLHALVES): New int iterator.
6628 (VFML, VFMLSEL): New mode attributes.
6629 (V_reg): Define mapping for V2HF.
6630 (V_hi, V_lo): New mode attributes.
6631 (VF_constraint): Likewise.
6632 (vfml_half, vfml_half_selector): New int attributes.
6633 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6635 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6636 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6638 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6639 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6640 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6641 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6643 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6644 Document new effective target and option set.
6646 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6648 * config/arm/arm-cpus.in (armv8_4): New feature.
6649 (ARMv8_4a): New fgroup.
6650 (armv8.4-a): New arch.
6651 * config/arm/arm-tables.opt: Regenerate.
6652 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6653 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6654 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6655 Add matching rules for -march=armv8.4-a and extensions.
6656 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6658 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6661 * config/rx/rx.md (BW): New mode attribute.
6662 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6664 2018-01-11 Richard Biener <rguenther@suse.de>
6666 PR tree-optimization/83435
6667 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6668 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6669 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6671 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6672 Alan Hayward <alan.hayward@arm.com>
6673 David Sherwood <david.sherwood@arm.com>
6675 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6677 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6678 (aarch64_print_address_internal): Use it to check for a zero offset.
6680 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6681 Alan Hayward <alan.hayward@arm.com>
6682 David Sherwood <david.sherwood@arm.com>
6684 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6685 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6686 Return a poly_int64 rather than a HOST_WIDE_INT.
6687 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6688 rather than a HOST_WIDE_INT.
6689 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6690 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6691 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6692 final_offset from HOST_WIDE_INT to poly_int64.
6693 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6694 to_constant when getting the number of units in an Advanced SIMD
6696 (aarch64_builtin_vectorized_function): Check for a constant number
6698 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6700 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6701 attribute instead of GET_MODE_NUNITS.
6702 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6703 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6704 GET_MODE_SIZE for fixed-size registers.
6705 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6706 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6707 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6708 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6709 (aarch64_print_operand, aarch64_print_address_internal)
6710 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6711 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6712 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6713 Handle polynomial GET_MODE_SIZE.
6714 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6715 wider than SImode without modification.
6716 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6717 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6718 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6719 passing and returning SVE modes.
6720 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6721 rather than GEN_INT.
6722 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6723 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6724 (aarch64_allocate_and_probe_stack_space): Likewise.
6725 (aarch64_layout_frame): Cope with polynomial offsets.
6726 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6727 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6729 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6730 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6731 poly_int64 rather than a HOST_WIDE_INT.
6732 (aarch64_get_separate_components, aarch64_process_components)
6733 (aarch64_expand_prologue, aarch64_expand_epilogue)
6734 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6735 (aarch64_anchor_offset): New function, split out from...
6736 (aarch64_legitimize_address): ...here.
6737 (aarch64_builtin_vectorization_cost): Handle polynomial
6738 TYPE_VECTOR_SUBPARTS.
6739 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6741 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6742 number of elements from the PARALLEL rather than the mode.
6743 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6744 rather than GET_MODE_BITSIZE.
6745 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6746 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6747 (aarch64_expand_vec_perm_const_1): Handle polynomial
6748 d->perm.length () and d->perm elements.
6749 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6750 Apply to_constant to d->perm elements.
6751 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6752 polynomial CONST_VECTOR_NUNITS.
6753 (aarch64_move_pointer): Take amount as a poly_int64 rather
6755 (aarch64_progress_pointer): Avoid temporary variable.
6756 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6757 the mode attribute instead of GET_MODE.
6759 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6760 Alan Hayward <alan.hayward@arm.com>
6761 David Sherwood <david.sherwood@arm.com>
6763 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6764 x exists before using it.
6765 (aarch64_add_constant_internal): Rename to...
6766 (aarch64_add_offset_1): ...this. Replace regnum with separate
6767 src and dest rtxes. Handle the case in which they're different,
6768 including when the offset is zero. Replace scratchreg with an rtx.
6769 Use 2 additions if there is no spare register into which we can
6770 move a 16-bit constant.
6771 (aarch64_add_constant): Delete.
6772 (aarch64_add_offset): Replace reg with separate src and dest
6773 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6774 Use aarch64_add_offset_1.
6775 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6776 an rtx rather than an int. Take the delta as a poly_int64
6777 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6778 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6779 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6780 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6781 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6783 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6784 aarch64_add_constant.
6786 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6788 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6789 Use scalar_float_mode.
6791 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6793 * config/aarch64/aarch64-simd.md
6794 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6795 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6796 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6797 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6798 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6799 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6800 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6801 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6802 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6803 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6805 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6808 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6809 targ_options->x_arm_arch_string is non NULL.
6811 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6813 * config/aarch64/aarch64.h
6814 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6816 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6819 * expmed.c (emit_store_flag_force): Swap if const op0
6820 and change VOIDmode to mode of op0.
6822 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6824 PR rtl-optimization/83761
6825 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6826 than bytes to mode_for_size.
6828 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6831 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6832 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6835 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6838 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6839 when in layout mode.
6840 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6841 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6844 2018-01-10 Michael Collison <michael.collison@arm.com>
6846 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6847 * config/aarch64/aarch64-option-extension.def: Add
6848 AARCH64_OPT_EXTENSION of 'fp16fml'.
6849 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6850 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6851 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6852 * config/aarch64/constraints.md (Ui7): New constraint.
6853 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6854 (VFMLA_SEL_W): Ditto.
6857 (VFMLA16_LOW): New int iterator.
6858 (VFMLA16_HIGH): Ditto.
6859 (UNSPEC_FMLAL): New unspec.
6860 (UNSPEC_FMLSL): Ditto.
6861 (UNSPEC_FMLAL2): Ditto.
6862 (UNSPEC_FMLSL2): Ditto.
6863 (f16mac): New code attribute.
6864 * config/aarch64/aarch64-simd-builtins.def
6865 (aarch64_fmlal_lowv2sf): Ditto.
6866 (aarch64_fmlsl_lowv2sf): Ditto.
6867 (aarch64_fmlalq_lowv4sf): Ditto.
6868 (aarch64_fmlslq_lowv4sf): Ditto.
6869 (aarch64_fmlal_highv2sf): Ditto.
6870 (aarch64_fmlsl_highv2sf): Ditto.
6871 (aarch64_fmlalq_highv4sf): Ditto.
6872 (aarch64_fmlslq_highv4sf): Ditto.
6873 (aarch64_fmlal_lane_lowv2sf): Ditto.
6874 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6875 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6876 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6877 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6878 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6879 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6880 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6881 (aarch64_fmlal_lane_highv2sf): Ditto.
6882 (aarch64_fmlsl_lane_highv2sf): Ditto.
6883 (aarch64_fmlal_laneq_highv2sf): Ditto.
6884 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6885 (aarch64_fmlalq_lane_highv4sf): Ditto.
6886 (aarch64_fmlsl_lane_highv4sf): Ditto.
6887 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6888 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6889 * config/aarch64/aarch64-simd.md:
6890 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6891 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6892 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6893 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6894 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6895 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6896 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6897 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6898 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6899 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6900 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6901 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6902 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6903 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6904 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6905 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6906 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6907 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6908 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6909 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6910 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6911 (vfmlsl_low_u32): Ditto.
6912 (vfmlalq_low_u32): Ditto.
6913 (vfmlslq_low_u32): Ditto.
6914 (vfmlal_high_u32): Ditto.
6915 (vfmlsl_high_u32): Ditto.
6916 (vfmlalq_high_u32): Ditto.
6917 (vfmlslq_high_u32): Ditto.
6918 (vfmlal_lane_low_u32): Ditto.
6919 (vfmlsl_lane_low_u32): Ditto.
6920 (vfmlal_laneq_low_u32): Ditto.
6921 (vfmlsl_laneq_low_u32): Ditto.
6922 (vfmlalq_lane_low_u32): Ditto.
6923 (vfmlslq_lane_low_u32): Ditto.
6924 (vfmlalq_laneq_low_u32): Ditto.
6925 (vfmlslq_laneq_low_u32): Ditto.
6926 (vfmlal_lane_high_u32): Ditto.
6927 (vfmlsl_lane_high_u32): Ditto.
6928 (vfmlal_laneq_high_u32): Ditto.
6929 (vfmlsl_laneq_high_u32): Ditto.
6930 (vfmlalq_lane_high_u32): Ditto.
6931 (vfmlslq_lane_high_u32): Ditto.
6932 (vfmlalq_laneq_high_u32): Ditto.
6933 (vfmlslq_laneq_high_u32): Ditto.
6934 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6935 (AARCH64_FL_FOR_ARCH8_4): New.
6936 (AARCH64_ISA_F16FML): New ISA flag.
6937 (TARGET_F16FML): New feature flag for fp16fml.
6938 (doc/invoke.texi): Document new fp16fml option.
6940 2018-01-10 Michael Collison <michael.collison@arm.com>
6942 * config/aarch64/aarch64-builtins.c:
6943 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6944 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6945 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6946 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6947 (AARCH64_ISA_SHA3): New ISA flag.
6948 (TARGET_SHA3): New feature flag for sha3.
6949 * config/aarch64/iterators.md (sha512_op): New int attribute.
6950 (CRYPTO_SHA512): New int iterator.
6951 (UNSPEC_SHA512H): New unspec.
6952 (UNSPEC_SHA512H2): Ditto.
6953 (UNSPEC_SHA512SU0): Ditto.
6954 (UNSPEC_SHA512SU1): Ditto.
6955 * config/aarch64/aarch64-simd-builtins.def
6956 (aarch64_crypto_sha512hqv2di): New builtin.
6957 (aarch64_crypto_sha512h2qv2di): Ditto.
6958 (aarch64_crypto_sha512su0qv2di): Ditto.
6959 (aarch64_crypto_sha512su1qv2di): Ditto.
6960 (aarch64_eor3qv8hi): Ditto.
6961 (aarch64_rax1qv2di): Ditto.
6962 (aarch64_xarqv2di): Ditto.
6963 (aarch64_bcaxqv8hi): Ditto.
6964 * config/aarch64/aarch64-simd.md:
6965 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6966 (aarch64_crypto_sha512su0qv2di): Ditto.
6967 (aarch64_crypto_sha512su1qv2di): Ditto.
6968 (aarch64_eor3qv8hi): Ditto.
6969 (aarch64_rax1qv2di): Ditto.
6970 (aarch64_xarqv2di): Ditto.
6971 (aarch64_bcaxqv8hi): Ditto.
6972 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6973 (vsha512h2q_u64): Ditto.
6974 (vsha512su0q_u64): Ditto.
6975 (vsha512su1q_u64): Ditto.
6976 (veor3q_u16): Ditto.
6977 (vrax1q_u64): Ditto.
6979 (vbcaxq_u16): Ditto.
6980 * config/arm/types.md (crypto_sha512): New type attribute.
6981 (crypto_sha3): Ditto.
6982 (doc/invoke.texi): Document new sha3 option.
6984 2018-01-10 Michael Collison <michael.collison@arm.com>
6986 * config/aarch64/aarch64-builtins.c:
6987 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6988 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6989 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6990 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6991 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6992 (AARCH64_ISA_SM4): New ISA flag.
6993 (TARGET_SM4): New feature flag for sm4.
6994 * config/aarch64/aarch64-simd-builtins.def
6995 (aarch64_sm3ss1qv4si): Ditto.
6996 (aarch64_sm3tt1aq4si): Ditto.
6997 (aarch64_sm3tt1bq4si): Ditto.
6998 (aarch64_sm3tt2aq4si): Ditto.
6999 (aarch64_sm3tt2bq4si): Ditto.
7000 (aarch64_sm3partw1qv4si): Ditto.
7001 (aarch64_sm3partw2qv4si): Ditto.
7002 (aarch64_sm4eqv4si): Ditto.
7003 (aarch64_sm4ekeyqv4si): Ditto.
7004 * config/aarch64/aarch64-simd.md:
7005 (aarch64_sm3ss1qv4si): Ditto.
7006 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7007 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7008 (aarch64_sm4eqv4si): Ditto.
7009 (aarch64_sm4ekeyqv4si): Ditto.
7010 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7011 (sm3part_op): Ditto.
7012 (CRYPTO_SM3TT): Ditto.
7013 (CRYPTO_SM3PART): Ditto.
7014 (UNSPEC_SM3SS1): New unspec.
7015 (UNSPEC_SM3TT1A): Ditto.
7016 (UNSPEC_SM3TT1B): Ditto.
7017 (UNSPEC_SM3TT2A): Ditto.
7018 (UNSPEC_SM3TT2B): Ditto.
7019 (UNSPEC_SM3PARTW1): Ditto.
7020 (UNSPEC_SM3PARTW2): Ditto.
7021 (UNSPEC_SM4E): Ditto.
7022 (UNSPEC_SM4EKEY): Ditto.
7023 * config/aarch64/constraints.md (Ui2): New constraint.
7024 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7025 * config/arm/types.md (crypto_sm3): New type attribute.
7026 (crypto_sm4): Ditto.
7027 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7028 (vsm3tt1aq_u32): Ditto.
7029 (vsm3tt1bq_u32): Ditto.
7030 (vsm3tt2aq_u32): Ditto.
7031 (vsm3tt2bq_u32): Ditto.
7032 (vsm3partw1q_u32): Ditto.
7033 (vsm3partw2q_u32): Ditto.
7034 (vsm4eq_u32): Ditto.
7035 (vsm4ekeyq_u32): Ditto.
7036 (doc/invoke.texi): Document new sm4 option.
7038 2018-01-10 Michael Collison <michael.collison@arm.com>
7040 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7041 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7042 (AARCH64_FL_FOR_ARCH8_4): New.
7043 (AARCH64_FL_V8_4): New flag.
7044 (doc/invoke.texi): Document new armv8.4-a option.
7046 2018-01-10 Michael Collison <michael.collison@arm.com>
7048 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7049 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7050 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7051 * config/aarch64/aarch64-option-extension.def: Add
7052 AARCH64_OPT_EXTENSION of 'sha2'.
7053 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7054 (crypto): Disable sha2 and aes if crypto disabled.
7055 (crypto): Enable aes and sha2 if enabled.
7056 (simd): Disable sha2 and aes if simd disabled.
7057 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7059 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7060 (TARGET_SHA2): New feature flag for sha2.
7061 (TARGET_AES): New feature flag for aes.
7062 * config/aarch64/aarch64-simd.md:
7063 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7064 conditional on TARGET_AES.
7065 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7066 (aarch64_crypto_sha1hsi): Make pattern conditional
7068 (aarch64_crypto_sha1hv4si): Ditto.
7069 (aarch64_be_crypto_sha1hv4si): Ditto.
7070 (aarch64_crypto_sha1su1v4si): Ditto.
7071 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7072 (aarch64_crypto_sha1su0v4si): Ditto.
7073 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7074 (aarch64_crypto_sha256su0v4si): Ditto.
7075 (aarch64_crypto_sha256su1v4si): Ditto.
7076 (doc/invoke.texi): Document new aes and sha2 options.
7078 2018-01-10 Martin Sebor <msebor@redhat.com>
7080 PR tree-optimization/83781
7081 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7084 2018-01-11 Martin Sebor <msebor@gmail.com>
7085 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7087 PR tree-optimization/83501
7088 PR tree-optimization/81703
7090 * tree-ssa-strlen.c (get_string_cst): Rename...
7091 (get_string_len): ...to this. Handle global constants.
7092 (handle_char_store): Adjust.
7094 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7095 Jim Wilson <jimw@sifive.com>
7097 * config/riscv/riscv-protos.h (riscv_output_return): New.
7098 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7099 (riscv_attribute_table, riscv_output_return),
7100 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7101 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7102 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7104 (riscv_expand_prologue): Add early return for naked function.
7105 (riscv_expand_epilogue): Likewise.
7106 (riscv_function_ok_for_sibcall): Return false for naked function.
7107 (riscv_set_current_function): New.
7108 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7109 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7110 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7111 * doc/extend.texi (RISC-V Function Attributes): New.
7113 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7115 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7116 check for 128-bit long double before checking TCmode.
7117 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7118 128-bit long doubles before checking TFmode or TCmode.
7119 (FLOAT128_IBM_P): Likewise.
7121 2018-01-10 Martin Sebor <msebor@redhat.com>
7123 PR tree-optimization/83671
7124 * builtins.c (c_strlen): Unconditionally return zero for the empty
7126 Use -Warray-bounds for warnings.
7127 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7128 for non-constant array indices with COMPONENT_REF, arrays of
7129 arrays, and pointers to arrays.
7130 (gimple_fold_builtin_strlen): Determine and set length range for
7131 non-constant character arrays.
7133 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7136 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7139 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7141 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7143 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7146 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7147 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7148 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7149 indexed_or_indirect_operand predicate.
7150 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7151 (*vsx_le_perm_load_v8hi): Likewise.
7152 (*vsx_le_perm_load_v16qi): Likewise.
7153 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7154 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7155 (*vsx_le_perm_store_v8hi): Likewise.
7156 (*vsx_le_perm_store_v16qi): Likewise.
7157 (eight unnamed splitters): Likewise.
7159 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7161 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7162 * config/rs6000/emmintrin.h: Likewise.
7163 * config/rs6000/mmintrin.h: Likewise.
7164 * config/rs6000/xmmintrin.h: Likewise.
7166 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7169 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7171 * tree.c (tree_nop_conversion): Return true for location wrapper
7173 (maybe_wrap_with_location): New function.
7174 (selftest::check_strip_nops): New function.
7175 (selftest::test_location_wrappers): New function.
7176 (selftest::tree_c_tests): Call it.
7177 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7178 (maybe_wrap_with_location): New decl.
7179 (EXPR_LOCATION_WRAPPER_P): New macro.
7180 (location_wrapper_p): New inline function.
7181 (tree_strip_any_location_wrapper): New inline function.
7183 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7186 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7187 stack_realign_offset for the largest alignment of stack slot
7189 (ix86_find_max_used_stack_alignment): New function.
7190 (ix86_finalize_stack_frame_flags): Use it. Set
7191 max_used_stack_alignment if we don't realign stack.
7192 * config/i386/i386.h (machine_function): Add
7193 max_used_stack_alignment.
7195 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7197 * config/arm/arm.opt (-mbranch-cost): New option.
7198 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7201 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7204 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7205 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7207 2018-01-10 Richard Biener <rguenther@suse.de>
7210 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7211 early out so it also covers the case where we have a non-NULL
7214 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7216 PR tree-optimization/83753
7217 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7218 for non-strided grouped accesses if the number of elements is 1.
7220 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7223 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7224 * i386.h (TARGET_USE_GATHER): Define.
7225 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7227 2018-01-10 Martin Liska <mliska@suse.cz>
7230 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7231 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7233 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7234 CLEANUP_NO_PARTITIONING is not set.
7236 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7238 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7239 for vectors, as a partial revert of r254296.
7240 * rtl.h (const_vec_p): Delete.
7241 (const_vec_duplicate_p): Don't test for vector CONSTs.
7242 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7243 * expmed.c (make_tree): Likewise.
7246 * common.md (E, F): Use CONSTANT_P instead of checking for
7248 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7249 checking for CONST_VECTOR.
7251 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7254 * predict.c (force_edge_cold): Handle in more sane way edges
7257 2018-01-09 Carl Love <cel@us.ibm.com>
7259 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7261 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7262 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7263 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7264 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7265 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7266 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7267 * config/rs6000/rs6000-protos.h: Add extern defition for
7268 rs6000_generate_float2_double_code.
7269 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7271 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7272 (float2_v2df): Add define_expand.
7274 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7277 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7278 op_mode in the force_to_mode call.
7280 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7282 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7283 instead of checking each element individually.
7284 (aarch64_evpc_uzp): Likewise.
7285 (aarch64_evpc_zip): Likewise.
7286 (aarch64_evpc_ext): Likewise.
7287 (aarch64_evpc_rev): Likewise.
7288 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7289 instead of checking each element individually. Return true without
7291 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7292 whether all selected elements come from the same input, instead of
7293 checking each element individually. Remove calls to gen_rtx_REG,
7294 start_sequence and end_sequence and instead assert that no rtl is
7297 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7299 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7300 order of HIGH and CONST checks.
7302 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7304 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7305 if the destination isn't an SSA_NAME.
7307 2018-01-09 Richard Biener <rguenther@suse.de>
7309 PR tree-optimization/83668
7310 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7312 (canonicalize_loop_form): ... here, renamed from ...
7313 (canonicalize_loop_closed_ssa_form): ... this and amended to
7314 swap successor edges for loop exit blocks to make us use
7315 the RPO order we need for initial schedule generation.
7317 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7319 PR tree-optimization/64811
7320 * match.pd: When optimizing comparisons with Inf, avoid
7321 introducing or losing exceptions from comparisons with NaN.
7323 2018-01-09 Martin Liska <mliska@suse.cz>
7326 * asan.c (shadow_mem_size): Add gcc_assert.
7328 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7330 Don't save registers in main().
7333 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7334 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7335 * config/avr/avr.c (avr_set_current_function): Don't error if
7336 naked, OS_task or OS_main are specified at the same time.
7337 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7339 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7341 * common/config/avr/avr-common.c (avr_option_optimization_table):
7342 Switch on -mmain-is-OS_task for optimizing compilations.
7344 2018-01-09 Richard Biener <rguenther@suse.de>
7346 PR tree-optimization/83572
7347 * graphite.c: Include cfganal.h.
7348 (graphite_transform_loops): Connect infinite loops to exit
7349 and remove fake edges at the end.
7351 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7353 * ipa-inline.c (edge_badness): Revert accidental checkin.
7355 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7358 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7359 symbols; not inline clones.
7361 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7364 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7365 hard registers. Formatting fixes.
7367 PR preprocessor/83722
7368 * gcc.c (try_generate_repro): Pass
7369 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7370 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7373 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7374 Kito Cheng <kito.cheng@gmail.com>
7376 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7377 (riscv_leaf_function_p): Delete.
7378 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7380 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7382 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7384 (do_ifelse): New function.
7385 (do_isel): New function.
7386 (do_sub3): New function.
7387 (do_add3): New function.
7388 (do_load_mask_compare): New function.
7389 (do_overlap_load_compare): New function.
7390 (expand_compare_loop): New function.
7391 (expand_block_compare): Call expand_compare_loop() when appropriate.
7392 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7394 (-mblock-compare-inline-loop-limit): New option.
7396 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7399 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7400 Reverse order of second and third operands in first alternative.
7401 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7402 of first and second elements in UNSPEC_VPERMR vector.
7403 (altivec_expand_vec_perm_le): Likewise.
7405 2018-01-08 Jeff Law <law@redhat.com>
7407 PR rtl-optimizatin/81308
7408 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7409 (process_switch): If group_case_labels makes a change, then set
7411 (pass_convert_switch::execute): If a switch is converted, then
7412 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7414 PR rtl-optimization/81308
7415 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7418 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7420 PR target/83663 - Revert r255946
7421 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7422 generation for cases where splatting a value is not useful.
7423 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7424 across a vec_duplicate and a paradoxical subreg forming a vector
7425 mode to a vec_concat.
7427 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7429 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7430 -march=armv8.3-a variants.
7431 * config/arm/t-multilib: Likewise.
7432 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7434 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7436 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7438 (cceq_ior_compare_complement): Give it a name so I can use it, and
7439 change boolean_or_operator predicate to boolean_operator so it can
7440 be used to generate a crand.
7441 (eqne): New code iterator.
7442 (bd/bd_neg): New code_attrs.
7443 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7444 a single define_insn.
7445 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7446 decrement (bdnzt/bdnzf/bdzt/bdzf).
7447 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7448 with the new names of the branch decrement patterns, and added the
7449 names of the branch decrement conditional patterns.
7451 2018-01-08 Richard Biener <rguenther@suse.de>
7453 PR tree-optimization/83563
7454 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7457 2018-01-08 Richard Biener <rguenther@suse.de>
7460 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7462 2018-01-08 Richard Biener <rguenther@suse.de>
7464 PR tree-optimization/83685
7465 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7466 references to abnormals.
7468 2018-01-08 Richard Biener <rguenther@suse.de>
7471 * dwarf2out.c (output_indirect_strings): Handle empty
7472 skeleton_debug_str_hash.
7473 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7475 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7477 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7478 (emit_store_direct): Likewise.
7479 (arc_trampoline_adjust_address): Likewise.
7480 (arc_asm_trampoline_template): New function.
7481 (arc_initialize_trampoline): Use asm_trampoline_template.
7482 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7483 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7484 * config/arc/arc.md (flush_icache): Delete pattern.
7486 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7488 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7489 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7492 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7495 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7496 by not USED_FOR_TARGET.
7497 (make_pass_resolve_sw_modes): Likewise.
7499 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7501 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7504 2018-01-08 Richard Biener <rguenther@suse.de>
7507 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7509 2018-01-08 Richard Biener <rguenther@suse.de>
7512 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7514 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7517 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7518 basic blocks with a small number of successors.
7519 (convert_control_dep_chain_into_preds): Improve handling of
7521 (dump_predicates): Split apart into...
7522 (dump_pred_chain): ...here...
7523 (dump_pred_info): ...and here.
7524 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7525 (can_chain_union_be_invalidated_p): Improve check for invalidation
7527 (uninit_uses_cannot_happen): Avoid unnecessary if
7528 convert_control_dep_chain_into_preds yielded nothing.
7530 2018-01-06 Martin Sebor <msebor@redhat.com>
7532 PR tree-optimization/83640
7533 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7534 subtracting negative offset from size.
7535 (builtin_access::overlap): Adjust offset bounds of the access to fall
7536 within the size of the object if possible.
7538 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7540 PR rtl-optimization/83699
7541 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7542 extract_bit_field_as_subreg to cases in which the extracted
7543 value is also a vector.
7545 * lra-constraints.c (process_alt_operands): Test for the equivalence
7546 substitutions when detecting a possible reload cycle.
7548 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7551 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7552 by default if flag_selective_schedling{,2}. Formatting fixes.
7554 PR rtl-optimization/83682
7555 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7556 if it has non-VECTOR_MODE element mode.
7557 (vec_duplicate_p): Likewise.
7560 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7561 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7563 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7566 * config/i386/i386-builtin.def
7567 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7568 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7569 Require also OPTION_MASK_ISA_AVX512F in addition to
7570 OPTION_MASK_ISA_GFNI.
7571 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7572 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7573 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7574 to OPTION_MASK_ISA_GFNI.
7575 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7576 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7577 OPTION_MASK_ISA_AVX512BW.
7578 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7579 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7580 addition to OPTION_MASK_ISA_GFNI.
7581 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7582 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7583 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7584 to OPTION_MASK_ISA_GFNI.
7585 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7586 a requirement for all ISAs rather than any of them with a few
7588 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7590 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7591 bitmasks to be enabled with 3 exceptions, instead of requiring any
7592 enabled ISA with lots of exceptions.
7593 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7594 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7595 Change avx512bw in isa attribute to avx512f.
7596 * config/i386/sgxintrin.h: Add license boilerplate.
7597 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7598 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7599 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7600 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7602 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7603 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7604 temporarily sse2 rather than sse if not enabled already.
7607 * config/i386/sse.md (VI248_VLBW): Rename to ...
7608 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7609 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7610 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7611 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7612 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7613 mode iterator instead of VI248_VLBW.
7615 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7617 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7618 (record_modified): Skip clobbers; add debug output.
7619 (param_change_prob): Use sreal frequencies.
7621 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7623 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7624 punt for user-aligned variables.
7626 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7628 * tree-chrec.c (chrec_contains_symbols): Return true for
7631 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7634 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7635 of (x|y) == x for BICS pattern.
7637 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7639 PR tree-optimization/83605
7640 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7641 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7644 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7646 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7647 * config/epiphany/rtems.h: New file.
7649 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7650 Uros Bizjak <ubizjak@gmail.com>
7653 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7654 QIreg_operand instead of register_operand predicate.
7655 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7656 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7657 comments instead of -fmitigate[-_]rop.
7659 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7662 * cgraphunit.c (symbol_table::compile): Switch to text_section
7663 before calling assembly_start debug hook.
7664 * run-rtl-passes.c (run_rtl_passes): Likewise.
7667 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7669 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7670 range_int_cst_p rather than !symbolic_range_p before calling
7671 extract_range_from_multiplicative_op_1.
7673 2018-01-04 Jeff Law <law@redhat.com>
7675 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7676 redundant test in assertion.
7678 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7680 * doc/rtl.texi: Document machine_mode wrapper classes.
7682 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7684 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7687 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7689 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7690 the VEC_PERM_EXPR fold to fail.
7692 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7695 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7696 to switched_sections.
7698 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7701 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7704 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7707 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7708 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7710 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7713 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7714 is BLKmode and bitpos not zero or mode change is needed.
7716 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7719 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7722 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7725 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7726 instead of MULT rtx. Update all corresponding splitters.
7728 (*ssub<modesuffix>): Ditto.
7730 (*cmp_sadd_di): Update split patterns.
7731 (*cmp_sadd_si): Ditto.
7732 (*cmp_sadd_sidi): Ditto.
7733 (*cmp_ssub_di): Ditto.
7734 (*cmp_ssub_si): Ditto.
7735 (*cmp_ssub_sidi): Ditto.
7736 * config/alpha/predicates.md (const23_operand): New predicate.
7737 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7738 Look for ASHIFT, not MULT inner operand.
7739 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7741 2018-01-04 Martin Liska <mliska@suse.cz>
7743 PR gcov-profile/83669
7744 * gcov.c (output_intermediate_file): Add version to intermediate
7746 * doc/gcov.texi: Document new field 'version' in intermediate
7747 file format. Fix location of '-k' option of gcov command.
7749 2018-01-04 Martin Liska <mliska@suse.cz>
7752 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7754 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7756 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7758 2018-01-03 Martin Sebor <msebor@redhat.com>
7760 PR tree-optimization/83655
7761 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7762 checking calls with invalid arguments.
7764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7766 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7767 (vectorizable_mask_load_store): Delete.
7768 (vectorizable_call): Return false for masked loads and stores.
7769 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7770 instead of gimple_assign_rhs1.
7771 (vectorizable_load): Handle IFN_MASK_LOAD.
7772 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7776 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7778 (vectorizable_mask_load_store): ...here.
7779 (vectorizable_load): ...and here.
7781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7783 * tree-vect-stmts.c (vect_build_all_ones_mask)
7784 (vect_build_zero_merge_argument): New functions, split out from...
7785 (vectorizable_load): ...here.
7787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7789 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7791 (vectorizable_mask_load_store): ...here.
7792 (vectorizable_store): ...and here.
7794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7796 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7798 (vectorizable_mask_load_store): ...here.
7800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7802 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7803 (vect_model_store_cost): Take a vec_load_store_type instead of a
7805 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7806 (vect_model_store_cost): Take a vec_load_store_type instead of a
7808 (vectorizable_mask_load_store): Update accordingly.
7809 (vectorizable_store): Likewise.
7810 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7814 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7815 IFN_MASK_LOAD calls here rather than...
7816 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7819 Alan Hayward <alan.hayward@arm.com>
7820 David Sherwood <david.sherwood@arm.com>
7822 * expmed.c (extract_bit_field_1): For vector extracts,
7823 fall back to extract_bit_field_as_subreg if vec_extract
7826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7827 Alan Hayward <alan.hayward@arm.com>
7828 David Sherwood <david.sherwood@arm.com>
7830 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7831 they are variable or constant sized.
7832 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7833 slots for constant-sized data.
7835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7836 Alan Hayward <alan.hayward@arm.com>
7837 David Sherwood <david.sherwood@arm.com>
7839 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7840 handling COND_EXPRs with boolean comparisons, try to find a better
7841 basis for the mask type than the boolean itself.
7843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7845 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7846 is calculated and how it can be overridden.
7847 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7848 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7850 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
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 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7858 Remove the mode argument.
7859 (aarch64_simd_valid_immediate): Remove the mode and inverse
7861 * config/aarch64/iterators.md (bitsize): New iterator.
7862 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7863 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7864 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7865 aarch64_simd_valid_immediate.
7866 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7867 (aarch64_reg_or_bic_imm): Likewise.
7868 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7869 with an insn_type enum and msl with a modifier_type enum.
7870 Replace element_width with a scalar_mode. Change the shift
7871 to unsigned int. Add constructors for scalar_float_mode and
7872 scalar_int_mode elements.
7873 (aarch64_vect_float_const_representable_p): Delete.
7874 (aarch64_can_const_movi_rtx_p)
7875 (aarch64_simd_scalar_immediate_valid_for_move)
7876 (aarch64_simd_make_constant): Update call to
7877 aarch64_simd_valid_immediate.
7878 (aarch64_advsimd_valid_immediate_hs): New function.
7879 (aarch64_advsimd_valid_immediate): Likewise.
7880 (aarch64_simd_valid_immediate): Remove mode and inverse
7881 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7882 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7883 and aarch64_float_const_representable_p on the result.
7884 (aarch64_output_simd_mov_immediate): Remove mode argument.
7885 Update call to aarch64_simd_valid_immediate and use of
7886 simd_immediate_info.
7887 (aarch64_output_scalar_simd_mov_immediate): Update call
7890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7891 Alan Hayward <alan.hayward@arm.com>
7892 David Sherwood <david.sherwood@arm.com>
7894 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7895 (mode_nunits): Likewise CONST_MODE_NUNITS.
7896 * machmode.def (ADJUST_NUNITS): Document.
7897 * genmodes.c (mode_data::need_nunits_adj): New field.
7898 (blank_mode): Update accordingly.
7899 (adj_nunits): New variable.
7900 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7902 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7903 listed in adj_nunits.
7904 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7905 listed in adj_nunits. Don't emit case statements for such modes.
7906 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7907 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7908 nothing if adj_nunits is nonnull.
7909 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7910 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7911 (emit_mode_fbit): Update use of print_maybe_const_decl.
7912 (emit_move_size): Likewise. Treat the array as non-const
7914 (emit_mode_adjustments): Handle adj_nunits.
7916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7918 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7919 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7920 (VECTOR_MODES): Use it.
7921 (make_vector_modes): Take the prefix as an argument.
7923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7924 Alan Hayward <alan.hayward@arm.com>
7925 David Sherwood <david.sherwood@arm.com>
7927 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7928 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7929 for MODE_VECTOR_BOOL.
7930 * machmode.def (VECTOR_BOOL_MODE): Document.
7931 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7932 (make_vector_bool_mode): New function.
7933 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7935 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7936 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7938 * stor-layout.c (int_mode_for_mode): Likewise.
7939 * tree.c (build_vector_type_for_mode): Likewise.
7940 * varasm.c (output_constant_pool_2): Likewise.
7941 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7942 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7943 for MODE_VECTOR_BOOL.
7944 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7945 of mode class checks.
7946 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7947 instead of a list of mode class checks.
7948 (expand_vector_scalar_condition): Likewise.
7949 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7952 Alan Hayward <alan.hayward@arm.com>
7953 David Sherwood <david.sherwood@arm.com>
7955 * machmode.h (mode_size): Change from unsigned short to
7957 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7958 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7959 or if measurement_type is not polynomial.
7960 (fixed_size_mode::includes_p): Check for constant-sized modes.
7961 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7962 return a poly_uint16 rather than an unsigned short.
7963 (emit_mode_size): Change the type of mode_size from unsigned short
7964 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7965 (emit_mode_adjustments): Cope with polynomial vector sizes.
7966 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7968 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7970 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7971 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7972 * caller-save.c (setup_save_areas): Likewise.
7973 (replace_reg_with_saved_mem): Likewise.
7974 * calls.c (emit_library_call_value_1): Likewise.
7975 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7976 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7977 (gen_lowpart_for_combine): Likewise.
7978 * convert.c (convert_to_integer_1): Likewise.
7979 * cse.c (equiv_constant, cse_insn): Likewise.
7980 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7981 (cselib_subst_to_values): Likewise.
7982 * dce.c (word_dce_process_block): Likewise.
7983 * df-problems.c (df_word_lr_mark_ref): Likewise.
7984 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7985 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7986 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7987 (rtl_for_decl_location): Likewise.
7988 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7989 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7990 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7991 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7992 (expand_expr_real_1): Likewise.
7993 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7994 (pad_below): Likewise.
7995 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7996 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7997 * ira.c (get_subreg_tracking_sizes): Likewise.
7998 * ira-build.c (ira_create_allocno_objects): Likewise.
7999 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8000 (ira_sort_regnos_for_alter_reg): Likewise.
8001 * ira-costs.c (record_operand_costs): Likewise.
8002 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8003 (resolve_simple_move): Likewise.
8004 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8005 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8006 (lra_constraints): Likewise.
8007 (CONST_POOL_OK_P): Reject variable-sized modes.
8008 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8009 (add_pseudo_to_slot, lra_spill): Likewise.
8010 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8011 * optabs-query.c (get_best_extraction_insn): Likewise.
8012 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8013 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8014 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8015 * recog.c (offsettable_address_addr_space_p): Likewise.
8016 * regcprop.c (maybe_mode_change): Likewise.
8017 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8018 * regrename.c (build_def_use): Likewise.
8019 * regstat.c (dump_reg_info): Likewise.
8020 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8021 (find_reloads, find_reloads_subreg_address): Likewise.
8022 * reload1.c (eliminate_regs_1): Likewise.
8023 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8024 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8025 (simplify_binary_operation_1, simplify_subreg): Likewise.
8026 * targhooks.c (default_function_arg_padding): Likewise.
8027 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8028 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8029 (verify_gimple_assign_ternary): Likewise.
8030 * tree-inline.c (estimate_move_cost): Likewise.
8031 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8032 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8033 (get_address_cost_ainc): Likewise.
8034 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8035 (vect_supportable_dr_alignment): Likewise.
8036 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8037 (vectorizable_reduction): Likewise.
8038 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8039 (vectorizable_operation, vectorizable_load): Likewise.
8040 * tree.c (build_same_sized_truth_vector_type): Likewise.
8041 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8042 * var-tracking.c (emit_note_insn_var_location): Likewise.
8043 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8044 (ADDR_VEC_ALIGN): Likewise.
8046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8047 Alan Hayward <alan.hayward@arm.com>
8048 David Sherwood <david.sherwood@arm.com>
8050 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8052 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8053 or if measurement_type is polynomial.
8054 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8055 * combine.c (make_extraction): Likewise.
8056 * dse.c (find_shift_sequence): Likewise.
8057 * dwarf2out.c (mem_loc_descriptor): Likewise.
8058 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8059 (extract_bit_field, extract_low_bits): Likewise.
8060 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8061 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8062 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8063 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8064 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8065 * reload.c (find_reloads): Likewise.
8066 * reload1.c (alter_reg): Likewise.
8067 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8068 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8069 * tree-if-conv.c (predicate_mem_writes): Likewise.
8070 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8071 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8072 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8073 * valtrack.c (dead_debug_insert_temp): Likewise.
8074 * varasm.c (mergeable_constant_section): Likewise.
8075 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8078 Alan Hayward <alan.hayward@arm.com>
8079 David Sherwood <david.sherwood@arm.com>
8081 * expr.c (expand_assignment): Cope with polynomial mode sizes
8082 when assigning to a CONCAT.
8084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8085 Alan Hayward <alan.hayward@arm.com>
8086 David Sherwood <david.sherwood@arm.com>
8088 * machmode.h (mode_precision): Change from unsigned short to
8090 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8092 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8093 or if measurement_type is not polynomial.
8094 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8095 in which the mode is already known to be a scalar_int_mode.
8096 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8097 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8099 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8100 for GET_MODE_PRECISION.
8101 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8102 for GET_MODE_PRECISION.
8103 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8105 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8106 (expand_field_assignment, make_extraction): Likewise.
8107 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8108 (get_last_value): Likewise.
8109 * convert.c (convert_to_integer_1): Likewise.
8110 * cse.c (cse_insn): Likewise.
8111 * expr.c (expand_expr_real_1): Likewise.
8112 * lra-constraints.c (simplify_operand_subreg): Likewise.
8113 * optabs-query.c (can_atomic_load_p): Likewise.
8114 * optabs.c (expand_atomic_load): Likewise.
8115 (expand_atomic_store): Likewise.
8116 * ree.c (combine_reaching_defs): Likewise.
8117 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8118 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8119 * tree.h (type_has_mode_precision_p): Likewise.
8120 * ubsan.c (instrument_si_overflow): Likewise.
8122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8123 Alan Hayward <alan.hayward@arm.com>
8124 David Sherwood <david.sherwood@arm.com>
8126 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8127 polynomial numbers of units.
8128 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8129 (valid_vector_subparts_p): New function.
8130 (build_vector_type): Remove temporary shim and take the number
8131 of units as a poly_uint64 rather than an int.
8132 (build_opaque_vector_type): Take the number of units as a
8133 poly_uint64 rather than an int.
8134 * tree.c (build_vector_from_ctor): Handle polynomial
8135 TYPE_VECTOR_SUBPARTS.
8136 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8137 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8138 (build_vector_from_val): If the number of units is variable,
8139 use build_vec_duplicate_cst for constant operands and
8140 VEC_DUPLICATE_EXPR otherwise.
8141 (make_vector_type): Remove temporary is_constant ().
8142 (build_vector_type, build_opaque_vector_type): Take the number of
8143 units as a poly_uint64 rather than an int.
8144 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8146 * cfgexpand.c (expand_debug_expr): Likewise.
8147 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8148 (store_constructor, expand_expr_real_1): Likewise.
8149 (const_scalar_mask_from_tree): Likewise.
8150 * fold-const-call.c (fold_const_reduction): Likewise.
8151 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8152 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8153 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8154 (fold_relational_const): Likewise.
8155 (native_interpret_vector): Likewise. Change the size from an
8156 int to an unsigned int.
8157 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8158 TYPE_VECTOR_SUBPARTS.
8159 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8160 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8161 duplicating a non-constant operand into a variable-length vector.
8162 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8163 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8164 * ipa-icf.c (sem_variable::equals): Likewise.
8165 * match.pd: Likewise.
8166 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8167 * print-tree.c (print_node): Likewise.
8168 * stor-layout.c (layout_type): Likewise.
8169 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8170 * tree-cfg.c (verify_gimple_comparison): Likewise.
8171 (verify_gimple_assign_binary): Likewise.
8172 (verify_gimple_assign_ternary): Likewise.
8173 (verify_gimple_assign_single): Likewise.
8174 * tree-pretty-print.c (dump_generic_node): Likewise.
8175 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8176 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8177 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8178 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8179 (vect_shift_permute_load_chain): Likewise.
8180 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8181 (expand_vector_condition, optimize_vector_constructor): Likewise.
8182 (lower_vec_perm, get_compute_type): Likewise.
8183 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8184 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8185 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8186 (vect_recog_mask_conversion_pattern): Likewise.
8187 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8188 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8189 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8190 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8191 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8192 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8193 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8194 (supportable_widening_operation): Likewise.
8195 (supportable_narrowing_operation): Likewise.
8196 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8198 * varasm.c (output_constant): Likewise.
8200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8201 Alan Hayward <alan.hayward@arm.com>
8202 David Sherwood <david.sherwood@arm.com>
8204 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8205 so that both the length == 3 and length != 3 cases set up their
8206 own permute vectors. Add comments explaining why we know the
8207 number of elements is constant.
8208 (vect_permute_load_chain): Likewise.
8210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8211 Alan Hayward <alan.hayward@arm.com>
8212 David Sherwood <david.sherwood@arm.com>
8214 * machmode.h (mode_nunits): Change from unsigned char to
8216 (ONLY_FIXED_SIZE_MODES): New macro.
8217 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8218 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8219 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8221 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8222 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8223 or if measurement_type is not polynomial.
8224 * genmodes.c (ZERO_COEFFS): New macro.
8225 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8227 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8228 Use ZERO_COEFFS when emitting initializers.
8229 * data-streamer.h (bp_pack_poly_value): New function.
8230 (bp_unpack_poly_value): Likewise.
8231 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8232 for GET_MODE_NUNITS.
8233 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8234 for GET_MODE_NUNITS.
8235 * tree.c (make_vector_type): Remove temporary shim and make
8236 the real function take the number of units as a poly_uint64
8238 (build_vector_type_for_mode): Handle polynomial nunits.
8239 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8240 * emit-rtl.c (const_vec_series_p_1): Likewise.
8241 (gen_rtx_CONST_VECTOR): Likewise.
8242 * fold-const.c (test_vec_duplicate_folding): Likewise.
8243 * genrecog.c (validate_pattern): Likewise.
8244 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8245 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8246 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8247 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8248 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8249 * rtlanal.c (subreg_get_info): Likewise.
8250 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8251 (vect_grouped_load_supported): Likewise.
8252 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8253 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8254 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8255 (simplify_const_unary_operation, simplify_binary_operation_1)
8256 (simplify_const_binary_operation, simplify_ternary_operation)
8257 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8258 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8259 instead of CONST_VECTOR_NUNITS.
8260 * varasm.c (output_constant_pool_2): Likewise.
8261 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8262 explicit-encoded elements in the XVEC for variable-length vectors.
8264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8266 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8269 Alan Hayward <alan.hayward@arm.com>
8270 David Sherwood <david.sherwood@arm.com>
8272 * coretypes.h (fixed_size_mode): Declare.
8273 (fixed_size_mode_pod): New typedef.
8274 * builtins.h (target_builtins::x_apply_args_mode)
8275 (target_builtins::x_apply_result_mode): Change type to
8276 fixed_size_mode_pod.
8277 * builtins.c (apply_args_size, apply_result_size, result_vector)
8278 (expand_builtin_apply_args_1, expand_builtin_apply)
8279 (expand_builtin_return): Update accordingly.
8281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8283 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8284 * cselib.c (cselib_hash_rtx): Likewise.
8285 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8286 CONST_VECTOR encoding.
8288 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8289 Jeff Law <law@redhat.com>
8292 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8293 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8294 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8295 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8298 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8299 explicitly probe *sp in a noreturn function if there were any callee
8300 register saves or frame pointer is needed.
8302 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8305 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8306 BLKmode for ternary, binary or unary expressions.
8309 * var-tracking.c (delete_vta_debug_insn): New inline function.
8310 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8311 insns from get_insns () to NULL instead of each bb separately.
8312 Use delete_vta_debug_insn. No longer static.
8313 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8314 delete_vta_debug_insns callers.
8315 * rtl.h (delete_vta_debug_insns): Declare.
8316 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8317 instead of variable_tracking_main.
8319 2018-01-03 Martin Sebor <msebor@redhat.com>
8321 PR tree-optimization/83603
8322 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8323 arguments past the endof the argument list in functions declared
8324 without a prototype.
8325 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8326 Avoid checking when arguments are null.
8328 2018-01-03 Martin Sebor <msebor@redhat.com>
8331 * doc/extend.texi (attribute const): Fix a typo.
8332 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8333 issuing -Wsuggest-attribute for void functions.
8335 2018-01-03 Martin Sebor <msebor@redhat.com>
8337 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8338 offset_int::from instead of wide_int::to_shwi.
8339 (maybe_diag_overlap): Remove assertion.
8340 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8341 * gimple-ssa-sprintf.c (format_directive): Same.
8342 (parse_directive): Same.
8343 (sprintf_dom_walker::compute_format_length): Same.
8344 (try_substitute_return_value): Same.
8346 2018-01-03 Jeff Law <law@redhat.com>
8349 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8350 non-constant residual for zero at runtime and avoid probing in
8351 that case. Reorganize code for trailing problem to mirror handling
8354 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8356 PR tree-optimization/83501
8357 * tree-ssa-strlen.c (get_string_cst): New.
8358 (handle_char_store): Call get_string_cst.
8360 2018-01-03 Martin Liska <mliska@suse.cz>
8362 PR tree-optimization/83593
8363 * tree-ssa-strlen.c: Include tree-cfg.h.
8364 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8365 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8366 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8368 (strlen_dom_walker::before_dom_children): Call
8369 gimple_purge_dead_eh_edges. Dump tranformation with details
8371 (strlen_dom_walker::before_dom_children): Update call by adding
8372 new argument cleanup_eh.
8373 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8375 2018-01-03 Martin Liska <mliska@suse.cz>
8378 * cif-code.def (VARIADIC_THUNK): New enum value.
8379 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8382 2018-01-03 Jan Beulich <jbeulich@suse.com>
8384 * sse.md (mov<mode>_internal): Tighten condition for when to use
8385 vmovdqu<ssescalarsize> for TI and OI modes.
8387 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8389 Update copyright years.
8391 2018-01-03 Martin Liska <mliska@suse.cz>
8394 * ipa-visibility.c (function_and_variable_visibility): Skip
8395 functions with noipa attribure.
8397 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8399 * gcc.c (process_command): Update copyright notice dates.
8400 * gcov-dump.c (print_version): Ditto.
8401 * gcov.c (print_version): Ditto.
8402 * gcov-tool.c (print_version): Ditto.
8403 * gengtype.c (create_file): Ditto.
8404 * doc/cpp.texi: Bump @copying's copyright year.
8405 * doc/cppinternals.texi: Ditto.
8406 * doc/gcc.texi: Ditto.
8407 * doc/gccint.texi: Ditto.
8408 * doc/gcov.texi: Ditto.
8409 * doc/install.texi: Ditto.
8410 * doc/invoke.texi: Ditto.
8412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8414 * vector-builder.h (vector_builder::m_full_nelts): Change from
8415 unsigned int to poly_uint64.
8416 (vector_builder::full_nelts): Update prototype accordingly.
8417 (vector_builder::new_vector): Likewise.
8418 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8419 (vector_builder::operator ==): Likewise.
8420 (vector_builder::finalize): Likewise.
8421 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8422 Take the number of elements as a poly_uint64 rather than an
8424 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8425 from unsigned int to poly_uint64.
8426 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8427 (vec_perm_indices::new_vector): Likewise.
8428 (vec_perm_indices::length): Likewise.
8429 (vec_perm_indices::nelts_per_input): Likewise.
8430 (vec_perm_indices::input_nelts): Likewise.
8431 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8432 number of elements per input as a poly_uint64 rather than an
8433 unsigned int. Use the original encoding for variable-length
8434 vectors, rather than clamping each individual element.
8435 For the second and subsequent elements in each pattern,
8436 clamp the step and base before clamping their sum.
8437 (vec_perm_indices::series_p): Handle polynomial element counts.
8438 (vec_perm_indices::all_in_range_p): Likewise.
8439 (vec_perm_indices_to_tree): Likewise.
8440 (vec_perm_indices_to_rtx): Likewise.
8441 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8442 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8443 (tree_vector_builder::new_binary_operation): Handle polynomial
8444 element counts. Return false if we need to know the number
8445 of elements at compile time.
8446 * fold-const.c (fold_vec_perm): Punt if the number of elements
8447 isn't known at compile time.
8449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8451 * vec-perm-indices.h (vec_perm_builder): Change element type
8452 from HOST_WIDE_INT to poly_int64.
8453 (vec_perm_indices::element_type): Update accordingly.
8454 (vec_perm_indices::clamp): Handle polynomial element_types.
8455 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8456 (vec_perm_indices::all_in_range_p): Likewise.
8457 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8459 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8460 polynomial vec_perm_indices element types.
8461 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8462 * fold-const.c (fold_vec_perm): Likewise.
8463 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8464 * tree-vect-generic.c (lower_vec_perm): Likewise.
8465 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8466 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8467 element type to HOST_WIDE_INT.
8469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8470 Alan Hayward <alan.hayward@arm.com>
8471 David Sherwood <david.sherwood@arm.com>
8473 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8474 rather than an int. Use plus_constant.
8475 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8476 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8479 Alan Hayward <alan.hayward@arm.com>
8480 David Sherwood <david.sherwood@arm.com>
8482 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8483 a HOST_WIDE_INT to a poly_int64.
8485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8486 Alan Hayward <alan.hayward@arm.com>
8487 David Sherwood <david.sherwood@arm.com>
8489 * calls.c (load_register_parameters): Cope with polynomial
8490 mode sizes. Require a constant size for BLKmode parameters
8491 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8492 forces a parameter to be padded at the lsb end in order to
8493 fill a complete number of words, require the parameter size
8494 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
8501 from unsigned int to poly_uint64_pod.
8502 (alter_reg): Treat mode sizes as polynomial.
8504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8505 Alan Hayward <alan.hayward@arm.com>
8506 David Sherwood <david.sherwood@arm.com>
8508 * reload.c (complex_word_subreg_p): New function.
8509 (reload_inner_reg_of_subreg, push_reload): Use it.
8511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8512 Alan Hayward <alan.hayward@arm.com>
8513 David Sherwood <david.sherwood@arm.com>
8515 * lra-constraints.c (process_alt_operands): Reject matched
8516 operands whose sizes aren't ordered.
8517 (match_reload): Refer to this check here.
8519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8520 Alan Hayward <alan.hayward@arm.com>
8521 David Sherwood <david.sherwood@arm.com>
8523 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8524 that the mode size is in the set {1, 2, 4, 8, 16}.
8526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8527 Alan Hayward <alan.hayward@arm.com>
8528 David Sherwood <david.sherwood@arm.com>
8530 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8531 Use plus_constant instead of gen_rtx_PLUS.
8533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8534 Alan Hayward <alan.hayward@arm.com>
8535 David Sherwood <david.sherwood@arm.com>
8537 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8538 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8539 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8540 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8541 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8542 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8543 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8544 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8545 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8546 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8548 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8549 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8550 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8551 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8552 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8553 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8554 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8555 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8556 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8557 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8559 * expr.c (emit_move_resolve_push): Treat the input and result
8560 of PUSH_ROUNDING as a poly_int64.
8561 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8562 (emit_push_insn): Likewise.
8563 * lra-eliminations.c (mark_not_eliminable): Likewise.
8564 * recog.c (push_operand): Likewise.
8565 * reload1.c (elimination_effects): Likewise.
8566 * rtlanal.c (nonzero_bits1): Likewise.
8567 * calls.c (store_one_arg): Likewise. Require the padding to be
8568 known at compile time.
8570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8571 Alan Hayward <alan.hayward@arm.com>
8572 David Sherwood <david.sherwood@arm.com>
8574 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8575 Use plus_constant instead of gen_rtx_PLUS.
8577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8578 Alan Hayward <alan.hayward@arm.com>
8579 David Sherwood <david.sherwood@arm.com>
8581 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8585 Alan Hayward <alan.hayward@arm.com>
8586 David Sherwood <david.sherwood@arm.com>
8588 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8589 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8590 via stack temporaries. Treat the mode size as polynomial too.
8592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8593 Alan Hayward <alan.hayward@arm.com>
8594 David Sherwood <david.sherwood@arm.com>
8596 * expr.c (expand_expr_real_2): When handling conversions involving
8597 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8598 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8599 as a poly_uint64 too.
8601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8602 Alan Hayward <alan.hayward@arm.com>
8603 David Sherwood <david.sherwood@arm.com>
8605 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8608 Alan Hayward <alan.hayward@arm.com>
8609 David Sherwood <david.sherwood@arm.com>
8611 * combine.c (can_change_dest_mode): Handle polynomial
8612 REGMODE_NATURAL_SIZE.
8613 * expmed.c (store_bit_field_1): Likewise.
8614 * expr.c (store_constructor): Likewise.
8615 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8616 and polynomial REGMODE_NATURAL_SIZE.
8617 (gen_lowpart_common): Likewise.
8618 * reginfo.c (record_subregs_of_mode): Likewise.
8619 * rtlanal.c (read_modify_subreg_p): Likewise.
8621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8622 Alan Hayward <alan.hayward@arm.com>
8623 David Sherwood <david.sherwood@arm.com>
8625 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8626 numbers of elements.
8628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8629 Alan Hayward <alan.hayward@arm.com>
8630 David Sherwood <david.sherwood@arm.com>
8632 * match.pd: Cope with polynomial numbers of vector elements.
8634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8635 Alan Hayward <alan.hayward@arm.com>
8636 David Sherwood <david.sherwood@arm.com>
8638 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8639 in a POINTER_PLUS_EXPR.
8641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8642 Alan Hayward <alan.hayward@arm.com>
8643 David Sherwood <david.sherwood@arm.com>
8645 * omp-simd-clone.c (simd_clone_subparts): New function.
8646 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8647 (ipa_simd_modify_function_body): Likewise.
8649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8650 Alan Hayward <alan.hayward@arm.com>
8651 David Sherwood <david.sherwood@arm.com>
8653 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8654 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8655 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8656 (expand_vector_condition, vector_element): Likewise.
8657 (subparts_gt): New function.
8658 (get_compute_type): Use subparts_gt.
8659 (count_type_subparts): Delete.
8660 (expand_vector_operations_1): Use subparts_gt instead of
8661 count_type_subparts.
8663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8664 Alan Hayward <alan.hayward@arm.com>
8665 David Sherwood <david.sherwood@arm.com>
8667 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8668 (vect_compile_time_alias): ...this new function. Do the calculation
8669 on poly_ints rather than trees.
8670 (vect_prune_runtime_alias_test_list): Update call accordingly.
8672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8673 Alan Hayward <alan.hayward@arm.com>
8674 David Sherwood <david.sherwood@arm.com>
8676 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8678 (vect_schedule_slp_instance): Likewise.
8680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8681 Alan Hayward <alan.hayward@arm.com>
8682 David Sherwood <david.sherwood@arm.com>
8684 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8685 constant and extern definitions for variable-length vectors.
8686 (vect_get_constant_vectors): Note that the number of units
8687 is known to be constant.
8689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8690 Alan Hayward <alan.hayward@arm.com>
8691 David Sherwood <david.sherwood@arm.com>
8693 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8694 of units as polynomial. Choose between WIDE and NARROW based
8697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8698 Alan Hayward <alan.hayward@arm.com>
8699 David Sherwood <david.sherwood@arm.com>
8701 * tree-vect-stmts.c (simd_clone_subparts): New function.
8702 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8705 Alan Hayward <alan.hayward@arm.com>
8706 David Sherwood <david.sherwood@arm.com>
8708 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8709 vectors as polynomial. Use build_index_vector for
8712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8713 Alan Hayward <alan.hayward@arm.com>
8714 David Sherwood <david.sherwood@arm.com>
8716 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8717 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8718 for variable-length vectors.
8719 (vectorizable_mask_load_store): Treat the number of units as
8720 polynomial, asserting that it is constant if the condition has
8721 already been enforced.
8722 (vectorizable_store, vectorizable_load): Likewise.
8724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8725 Alan Hayward <alan.hayward@arm.com>
8726 David Sherwood <david.sherwood@arm.com>
8728 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8729 of units as polynomial. Punt if we can't tell at compile time
8730 which vector contains the final result.
8732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8733 Alan Hayward <alan.hayward@arm.com>
8734 David Sherwood <david.sherwood@arm.com>
8736 * tree-vect-loop.c (vectorizable_induction): Treat the number
8737 of units as polynomial. Punt on SLP inductions. Use an integer
8738 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8739 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
8747 * tree.c (build_index_vector): New function.
8748 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8749 of units as polynomial, forcibly converting it to a constant if
8750 vectorizable_reduction has already enforced the condition.
8751 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8752 to create a {1,2,3,...} vector.
8753 (vectorizable_reduction): Treat the number of units as polynomial.
8754 Choose vectype_in based on the largest scalar element size rather
8755 than the smallest number of units. Enforce the restrictions
8758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8759 Alan Hayward <alan.hayward@arm.com>
8760 David Sherwood <david.sherwood@arm.com>
8762 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8763 number of units as polynomial.
8765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8766 Alan Hayward <alan.hayward@arm.com>
8767 David Sherwood <david.sherwood@arm.com>
8769 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8770 * target.def (autovectorize_vector_sizes): Return the vector sizes
8771 by pointer, using vector_sizes rather than a bitmask.
8772 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8773 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8774 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8776 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8777 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8778 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8779 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8780 * omp-general.c (omp_max_vf): Likewise.
8781 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8782 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8783 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8784 * tree-vect-slp.c (vect_slp_bb): Likewise.
8785 * doc/tm.texi: Regenerate.
8786 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8788 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8789 the vector size as a poly_uint64 rather than an unsigned int.
8790 (current_vector_size): Change from an unsigned int to a poly_uint64.
8791 (get_vectype_for_scalar_type): Update accordingly.
8792 * tree.h (build_truth_vector_type): Take the size and number of
8793 units as a poly_uint64 rather than an unsigned int.
8794 (build_vector_type): Add a temporary overload that takes
8795 the number of units as a poly_uint64 rather than an unsigned int.
8796 * tree.c (make_vector_type): Likewise.
8797 (build_truth_vector_type): Take the number of units as a poly_uint64
8798 rather than an unsigned int.
8800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8801 Alan Hayward <alan.hayward@arm.com>
8802 David Sherwood <david.sherwood@arm.com>
8804 * target.def (get_mask_mode): Take the number of units and length
8805 as poly_uint64s rather than unsigned ints.
8806 * targhooks.h (default_get_mask_mode): Update accordingly.
8807 * targhooks.c (default_get_mask_mode): Likewise.
8808 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8809 * doc/tm.texi: Regenerate.
8811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8812 Alan Hayward <alan.hayward@arm.com>
8813 David Sherwood <david.sherwood@arm.com>
8815 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8816 * omp-general.c (omp_max_vf): Likewise.
8817 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8818 (expand_omp_simd): Handle polynomial safelen.
8819 * omp-low.c (omplow_simd_context): Add a default constructor.
8820 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8821 (lower_rec_simd_input_clauses): Update accordingly.
8822 (lower_rec_input_clauses): Likewise.
8824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8825 Alan Hayward <alan.hayward@arm.com>
8826 David Sherwood <david.sherwood@arm.com>
8828 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8829 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8830 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8831 (vect_analyze_slp_cost): Likewise.
8832 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8833 (vect_model_load_cost): Likewise.
8835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8836 Alan Hayward <alan.hayward@arm.com>
8837 David Sherwood <david.sherwood@arm.com>
8839 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8840 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8841 from an unsigned int * to a poly_uint64_pod *.
8842 (calculate_unrolling_factor): New function.
8843 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8846 Alan Hayward <alan.hayward@arm.com>
8847 David Sherwood <david.sherwood@arm.com>
8849 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8850 from an unsigned int to a poly_uint64.
8851 (_loop_vec_info::slp_unrolling_factor): Likewise.
8852 (_loop_vec_info::vectorization_factor): Change from an int
8854 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8855 (vect_get_num_vectors): New function.
8856 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8857 (vect_get_num_copies): Use vect_get_num_vectors.
8858 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8859 to an unsigned int *.
8860 (vect_analyze_data_refs): Change min_vf from an int * to a
8862 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8863 than an unsigned HOST_WIDE_INT.
8864 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8865 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8866 to an unsigned int *.
8867 (vect_analyze_data_ref_dependences): Likewise.
8868 (vect_compute_data_ref_alignment): Handle polynomial vf.
8869 (vect_enhance_data_refs_alignment): Likewise.
8870 (vect_prune_runtime_alias_test_list): Likewise.
8871 (vect_shift_permute_load_chain): Likewise.
8872 (vect_supportable_dr_alignment): Likewise.
8873 (dependence_distance_ge_vf): Take the vectorization factor as a
8874 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8875 (vect_analyze_data_refs): Change min_vf from an int * to a
8877 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8878 vfm1 as a poly_uint64 rather than an int. Make the same change
8879 for the returned bound_scalar.
8880 (vect_gen_vector_loop_niters): Handle polynomial vf.
8881 (vect_do_peeling): Likewise. Update call to
8882 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8883 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8885 * tree-vect-loop.c (vect_determine_vectorization_factor)
8886 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8887 (vect_get_known_peeling_cost): Likewise.
8888 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8889 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8890 (vect_transform_loop): Likewise. Use the lowest possible VF when
8891 updating the upper bounds of the loop.
8892 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8894 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8895 polynomial unroll factors.
8896 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8897 (vect_make_slp_decision): Likewise.
8898 (vect_supported_load_permutation_p): Likewise, and polynomial
8900 (vect_analyze_slp_cost): Handle polynomial vf.
8901 (vect_slp_analyze_node_operations): Likewise.
8902 (vect_slp_analyze_bb_1): Likewise.
8903 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8904 than an unsigned HOST_WIDE_INT.
8905 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8906 (vectorizable_load): Handle polynomial vf.
8907 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8909 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8912 Alan Hayward <alan.hayward@arm.com>
8913 David Sherwood <david.sherwood@arm.com>
8915 * match.pd: Handle bit operations involving three constants
8916 and try to fold one pair.
8918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8920 * tree-vect-loop-manip.c: Include gimple-fold.h.
8921 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8922 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8923 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8924 Add a path that uses a step of VF instead of 1, but disable it
8926 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8927 and niters_no_overflow parameters. Update calls to
8928 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8929 Create a new SSA name if the latter choses to use a ste other
8930 than zero, and return it via niters_vector_mult_vf_var.
8931 * tree-vect-loop.c (vect_transform_loop): Update calls to
8932 vect_do_peeling, vect_gen_vector_loop_niters and
8933 slpeel_make_loop_iterate_ntimes.
8934 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8935 (vect_gen_vector_loop_niters): Update declarations after above changes.
8937 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8939 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8940 128-bit round to integer instructions.
8941 (ceil<mode>2): Likewise.
8942 (btrunc<mode>2): Likewise.
8943 (round<mode>2): Likewise.
8945 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8947 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8948 unaligned VSX load/store on P8/P9.
8949 (expand_block_clear): Allow the use of unaligned VSX
8950 load/store on P8/P9.
8952 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8954 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8956 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8957 swap associated with both a load and a store.
8959 2018-01-02 Andrew Waterman <andrew@sifive.com>
8961 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8962 * config/riscv/riscv.md (clear_cache): Use it.
8964 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8966 * web.c: Remove out-of-date comment.
8968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8970 * expr.c (fixup_args_size_notes): Check that any existing
8971 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8972 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8973 (emit_single_push_insn): ...here.
8975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8977 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8978 (const_vector_encoded_nelts): New function.
8979 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8980 (const_vector_int_elt, const_vector_elt): Declare.
8981 * emit-rtl.c (const_vector_int_elt_1): New function.
8982 (const_vector_elt): Likewise.
8983 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8984 of CONST_VECTOR_ELT.
8986 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8988 * expr.c: Include rtx-vector-builder.h.
8989 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8990 directly on the tree encoding.
8991 (const_vector_from_tree): Likewise.
8992 * optabs.c: Include rtx-vector-builder.h.
8993 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8994 sequence of "u" values.
8995 * vec-perm-indices.c: Include rtx-vector-builder.h.
8996 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8997 directly on the vec_perm_indices encoding.
8999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9001 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9002 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9003 * rtx-vector-builder.h: New file.
9004 * rtx-vector-builder.c: Likewise.
9005 * rtl.h (rtx_def::u2): Add a const_vector field.
9006 (CONST_VECTOR_NPATTERNS): New macro.
9007 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9008 (CONST_VECTOR_DUPLICATE_P): Likewise.
9009 (CONST_VECTOR_STEPPED_P): Likewise.
9010 (CONST_VECTOR_ENCODED_ELT): Likewise.
9011 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9012 (unwrap_const_vec_duplicate): Likewise.
9013 (const_vec_series_p): Check for a non-duplicated vector encoding.
9014 Say that the function only returns true for integer vectors.
9015 * emit-rtl.c: Include rtx-vector-builder.h.
9016 (gen_const_vec_duplicate_1): Delete.
9017 (gen_const_vector): Call gen_const_vec_duplicate instead of
9018 gen_const_vec_duplicate_1.
9019 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9020 (gen_const_vec_duplicate): Use rtx_vector_builder.
9021 (gen_const_vec_series): Likewise.
9022 (gen_rtx_CONST_VECTOR): Likewise.
9023 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9024 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9025 Build a new vector rather than modifying a CONST_VECTOR in-place.
9026 (handle_special_swappables): Update call accordingly.
9027 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9028 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9029 Build a new vector rather than modifying a CONST_VECTOR in-place.
9030 (handle_special_swappables): Update call accordingly.
9032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9034 * simplify-rtx.c (simplify_const_binary_operation): Use
9035 CONST_VECTOR_ELT instead of XVECEXP.
9037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9039 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9040 the selector elements to be different from the data elements
9041 if the selector is a VECTOR_CST.
9042 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9043 ssizetype for the selector.
9045 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9047 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9048 before testing each element individually.
9049 * tree-vect-generic.c (lower_vec_perm): Likewise.
9051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9053 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9054 * selftest-run-tests.c (selftest::run_tests): Call it.
9055 * vector-builder.h (vector_builder::operator ==): New function.
9056 (vector_builder::operator !=): Likewise.
9057 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9058 (vec_perm_indices::all_from_input_p): New function.
9059 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9060 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9061 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9062 instead of reading the VECTOR_CST directly. Detect whether both
9063 vector inputs are the same before constructing the vec_perm_indices,
9064 and update the number of inputs argument accordingly. Use the
9065 utility functions added above. Only construct sel2 if we need to.
9067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9069 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9070 the broadcast of the low byte.
9071 (expand_mult_highpart): Use an explicit encoding for the permutes.
9072 * optabs-query.c (can_mult_highpart_p): Likewise.
9073 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9074 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9075 (vectorizable_bswap): Likewise.
9076 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9077 explicit encoding for the power-of-2 permutes.
9078 (vect_permute_store_chain): Likewise.
9079 (vect_grouped_load_supported): Likewise.
9080 (vect_permute_load_chain): Likewise.
9082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9084 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9085 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9086 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9087 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9088 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9089 (vect_gen_perm_mask_any): Likewise.
9091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9093 * int-vector-builder.h: New file.
9094 * vec-perm-indices.h: Include int-vector-builder.h.
9095 (vec_perm_indices): Redefine as an int_vector_builder.
9096 (auto_vec_perm_indices): Delete.
9097 (vec_perm_builder): Redefine as a stand-alone class.
9098 (vec_perm_indices::vec_perm_indices): New function.
9099 (vec_perm_indices::clamp): Likewise.
9100 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9101 (vec_perm_indices::new_vector): New function.
9102 (vec_perm_indices::new_expanded_vector): Update for new
9103 vec_perm_indices class.
9104 (vec_perm_indices::rotate_inputs): New function.
9105 (vec_perm_indices::all_in_range_p): Operate directly on the
9106 encoded form, without computing elided elements.
9107 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9108 encoding. Update for new vec_perm_indices class.
9109 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9110 the given vec_perm_builder.
9111 (expand_vec_perm_var): Update vec_perm_builder constructor.
9112 (expand_mult_highpart): Use vec_perm_builder instead of
9113 auto_vec_perm_indices.
9114 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9115 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9116 or double series encoding as appropriate.
9117 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9118 vec_perm_indices instead of auto_vec_perm_indices.
9119 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9120 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9121 (vect_permute_store_chain): Likewise.
9122 (vect_grouped_load_supported): Likewise.
9123 (vect_permute_load_chain): Likewise.
9124 (vect_shift_permute_load_chain): Likewise.
9125 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9126 (vect_transform_slp_perm_load): Likewise.
9127 (vect_schedule_slp_instance): Likewise.
9128 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9129 (vectorizable_mask_load_store): Likewise.
9130 (vectorizable_bswap): Likewise.
9131 (vectorizable_store): Likewise.
9132 (vectorizable_load): Likewise.
9133 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9134 vec_perm_indices instead of auto_vec_perm_indices. Use
9135 tree_to_vec_perm_builder to read the vector from a tree.
9136 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9137 vec_perm_builder instead of a vec_perm_indices.
9138 (have_whole_vector_shift): Use vec_perm_builder and
9139 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9140 truncation to calc_vec_perm_mask_for_shift.
9141 (vect_create_epilog_for_reduction): Likewise.
9142 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9143 from auto_vec_perm_indices to vec_perm_indices.
9144 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9145 instead of changing individual elements.
9146 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9147 the vector in d.perm.
9148 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9149 from auto_vec_perm_indices to vec_perm_indices.
9150 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9151 instead of changing individual elements.
9152 (arm_vectorize_vec_perm_const): Use new_vector to install
9153 the vector in d.perm.
9154 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9155 Update vec_perm_builder constructor.
9156 (rs6000_expand_interleave): Likewise.
9157 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9158 (rs6000_expand_interleave): Likewise.
9160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9162 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9163 to qimode could truncate the indices.
9164 * optabs.c (expand_vec_perm_var): Likewise.
9166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9168 * Makefile.in (OBJS): Add vec-perm-indices.o.
9169 * vec-perm-indices.h: New file.
9170 * vec-perm-indices.c: Likewise.
9171 * target.h (vec_perm_indices): Replace with a forward class
9173 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9174 * optabs.h: Include vec-perm-indices.h.
9175 (expand_vec_perm): Delete.
9176 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9177 (expand_vec_perm_const): Declare.
9178 * target.def (vec_perm_const_ok): Replace with...
9179 (vec_perm_const): ...this new hook.
9180 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9181 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9182 * doc/tm.texi: Regenerate.
9183 * optabs.def (vec_perm_const): Delete.
9184 * doc/md.texi (vec_perm_const): Likewise.
9185 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9186 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9187 expand_vec_perm for constant permutation vectors. Assert that
9188 the mode of variable permutation vectors is the integer equivalent
9189 of the mode that is being permuted.
9190 * optabs-query.h (selector_fits_mode_p): Declare.
9191 * optabs-query.c: Include vec-perm-indices.h.
9192 (selector_fits_mode_p): New function.
9193 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9194 is defined, instead of checking whether the vec_perm_const_optab
9195 exists. Use targetm.vectorize.vec_perm_const instead of
9196 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9197 fit in the vector mode before using a variable permute.
9198 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9199 vec_perm_indices instead of an rtx.
9200 (expand_vec_perm): Replace with...
9201 (expand_vec_perm_const): ...this new function. Take the selector
9202 as a vec_perm_indices rather than an rtx. Also take the mode of
9203 the selector. Update call to shift_amt_for_vec_perm_mask.
9204 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9205 Use vec_perm_indices::new_expanded_vector to expand the original
9206 selector into bytes. Check whether the indices fit in the vector
9207 mode before using a variable permute.
9208 (expand_vec_perm_var): Make global.
9209 (expand_mult_highpart): Use expand_vec_perm_const.
9210 * fold-const.c: Includes vec-perm-indices.h.
9211 * tree-ssa-forwprop.c: Likewise.
9212 * tree-vect-data-refs.c: Likewise.
9213 * tree-vect-generic.c: Likewise.
9214 * tree-vect-loop.c: Likewise.
9215 * tree-vect-slp.c: Likewise.
9216 * tree-vect-stmts.c: Likewise.
9217 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9219 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9220 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9221 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9222 (aarch64_vectorize_vec_perm_const): ...this new function.
9223 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9224 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9225 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9226 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9227 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9228 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9229 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9231 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9232 check for NEON modes.
9233 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9234 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9235 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9236 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9238 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9239 the old VEC_PERM_CONST conditions.
9240 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9241 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9242 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9243 (ia64_vectorize_vec_perm_const_ok): Merge into...
9244 (ia64_vectorize_vec_perm_const): ...this new function.
9245 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9246 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9247 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9248 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9249 * config/mips/mips.c (mips_expand_vec_perm_const)
9250 (mips_vectorize_vec_perm_const_ok): Merge into...
9251 (mips_vectorize_vec_perm_const): ...this new function.
9252 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9253 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9254 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9255 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9256 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9257 (rs6000_expand_vec_perm_const): Delete.
9258 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9260 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9261 (altivec_expand_vec_perm_const_le): Take each operand individually.
9262 Operate on constant selectors rather than rtxes.
9263 (altivec_expand_vec_perm_const): Likewise. Update call to
9264 altivec_expand_vec_perm_const_le.
9265 (rs6000_expand_vec_perm_const): Delete.
9266 (rs6000_vectorize_vec_perm_const_ok): Delete.
9267 (rs6000_vectorize_vec_perm_const): New function.
9268 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9269 an element count and rtx array.
9270 (rs6000_expand_extract_even): Update call accordingly.
9271 (rs6000_expand_interleave): Likewise.
9272 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9273 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9274 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9275 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9276 (rs6000_expand_vec_perm_const): Delete.
9277 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
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. Remove stray
9286 reference to the SPE evmerge intructions.
9287 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9288 an element count and rtx array.
9289 (rs6000_expand_extract_even): Update call accordingly.
9290 (rs6000_expand_interleave): Likewise.
9291 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9292 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9294 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9298 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9299 vector mode and that that mode matches the mode of the data
9301 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9302 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9303 directly using expand_vec_perm_1 when forcing selectors into
9305 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9307 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9309 * optabs-query.h (can_vec_perm_p): Delete.
9310 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9311 * optabs-query.c (can_vec_perm_p): Split into...
9312 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9313 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9314 particular selector is valid.
9315 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9316 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9317 (vect_grouped_load_supported): Likewise.
9318 (vect_shift_permute_load_chain): Likewise.
9319 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9320 (vect_transform_slp_perm_load): Likewise.
9321 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9322 (vectorizable_bswap): Likewise.
9323 (vect_gen_perm_mask_checked): Likewise.
9324 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9325 implementations of variable permutation vectors into account
9326 when deciding which selector to use.
9327 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9328 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9329 with a false third argument.
9330 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9331 to test whether the constant selector is valid and can_vec_perm_var_p
9332 to test whether a variable selector is valid.
9334 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9336 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9337 * optabs-query.c (can_vec_perm_p): Likewise.
9338 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9339 instead of vec_perm_indices.
9340 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9341 (vect_gen_perm_mask_checked): Likewise,
9342 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9343 (vect_gen_perm_mask_checked): Likewise,
9345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9347 * optabs-query.h (qimode_for_vec_perm): Declare.
9348 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9349 (qimode_for_vec_perm): ...this new function.
9350 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9352 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9354 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9355 does not have a conditional at the top.
9357 2018-01-02 Richard Biener <rguenther@suse.de>
9359 * ipa-inline.c (big_speedup_p): Fix expression.
9361 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9364 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9367 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9371 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9372 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9373 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9374 cond_taken_branch_cost 3->4.
9376 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9378 PR tree-optimization/83581
9379 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9380 TODO_cleanup_cfg if any changes have been made.
9383 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9384 convert_modes if target mode has the right side, but different mode
9388 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9389 last argument when extracting from CONCAT. If either from_real or
9390 from_imag is NULL, use expansion through memory. If result is not
9391 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9392 the parts directly to inner mode, if even that fails, use expansion
9396 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9397 check for bswap in mode rather than HImode and use that in expand_unop
9400 Copyright (C) 2018 Free Software Foundation, Inc.
9402 Copying and distribu