1 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7 2018-03-23 Richard Biener <rguenther@suse.de>
10 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11 we are going to emit early debug for LTO.
13 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16 * function.c (match_asm_constraints_1): Don't optimize if input
17 doesn't satisfy general_operand predicate for output's mode.
20 * alias.c (write_dependence_p): Don't require for x_canonicalized
21 non-VOIDmode if x has VOIDmode.
24 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
25 just don't try to optimize it rather than assert it never happens.
27 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
29 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
30 macro expansions for definition of ST_INTERNAL_<mode> and
31 LD_INTERNAL_<mode> builtins.
32 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
34 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
36 (altivec_expand_st_builtin): Likewise.
37 (altivec_expand_builtin): Remove calls to deleted functions.
38 (rs6000_address_for_altivec): Delete this function.
39 * config/rs6000/vector.md: Remove expands for
40 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
42 2018-03-22 Sudakshina Das <sudi.das@arm.com>
45 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
46 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
47 re-computing once computed.
48 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
49 (arm_init_machine_status): Initialize
50 machine->static_chain_stack_bytes.
52 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
55 * doc/extend.texi: Add four new prototypes for vec_ld.
56 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
57 definitions for more logical presentation.
58 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
59 entries for V1TI variants of __builtin_altivec_ld builtin.
60 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
61 handling of V1TI variant of LVX icode pattern.
62 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
63 (rs6000_gimple_fold_builtin): Likewise.
64 (altivec_init_builtins): Add code to define
65 __builtin_altivec_lvx_v1ti function.
67 2018-03-22 Jakub Jelinek <jakub@redhat.com>
70 * function.c (match_asm_constraints_1): Don't do the optimization
71 if input isn't a REG, SUBREG, MEM or constant.
73 2018-03-22 Tom de Vries <tom@codesourcery.com>
75 PR tree-optimization/84956
76 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
79 2018-03-22 Jakub Jelinek <jakub@redhat.com>
82 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
83 DECL_INITIAL (decl) to decl at the end.
84 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
87 2018-03-21 Joseph Myers <joseph@codesourcery.com>
89 * doc/extend.texi (__builtin_tgmath): Document when complex
90 integer types are treated as _Complex _Float64.
92 2018-03-21 Tom de Vries <tom@codesourcery.com>
94 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
96 2018-03-21 Jakub Jelinek <jakub@redhat.com>
98 PR tree-optimization/84960
99 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
100 if it is ENTRY block, move them into single succ of ENTRY in that case.
102 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
104 PR tree-optimization/84811
105 * poly-int.h (poly_span_traits): Remove the T3 parameter and
106 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
107 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
108 (known_subrange_p): Update accordingly. Cast each value involved
109 in the size comparison, rather than casting the result of the
112 2018-03-21 Jakub Jelinek <jakub@redhat.com>
114 PR tree-optimization/84982
115 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
116 by flipping the least significant bit rather than all bits from
117 bitpos to bitpos + bitsize - 1.
119 2018-03-21 Nathan Sidwell <nathan@acm.org>
121 * doc/extend.texi (Deprecated Features): Remove mention of
122 long-deleted deprecations.
124 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
127 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
128 * configure: Regenerate.
130 2018-03-21 Tom de Vries <tom@codesourcery.com>
132 PR tree-optimization/83126
133 * tree-parloops.c (num_phis): New function.
134 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
136 2018-03-21 Nathan Sidwell <nathan@acm.org>
138 * doc/extend.texi (Deprecated Features): Update deprecated flags,
139 mention anon-struct/union members and trailing attributes.
141 2018-03-21 Bin Cheng <bin.cheng@arm.com>
143 PR tree-optimization/84969
144 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
145 builtin memset partitions if they set different rhs values.
147 2018-03-21 Jakub Jelinek <jakub@redhat.com>
149 PR rtl-optimization/84989
150 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
151 VEC_DUPLICATE with scalar result mode.
153 2018-03-21 Martin Liska <mliska@suse.cz>
156 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
157 not intended return statement.
159 2018-03-21 Martin Liska <mliska@suse.cz>
162 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
163 (chkp_find_bound_slots_1): Limit number of iterations.
165 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
168 * Minor grammar fixes for x86 options.
170 2018-03-20 Jakub Jelinek <jakub@redhat.com>
173 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
174 holding REG_CFA_RESTORE notes, instead turn them into a USE.
176 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
179 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
180 (altivec_lvx_<mode>_1op): Likewise.
181 (altivec_stvx_<mode>_2op): Likewise.
182 (altivec_stvx_<mode>_1op): Likewise.
183 (altivec_lvx_<VM2:mode>): New define_expand.
184 (altivec_stvx_<VM2:mode>): Likewise.
185 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
186 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
187 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
188 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
189 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
190 (rs6000_gen_lvx): Likewise.
191 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
192 (altivec_expand_stv_builtin): Likewise.
193 (altivec_expand_builtin): Likewise.
194 * config/rs6000/vector.md: Likewise.
196 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
199 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
202 2018-03-20 Richard Biener <rguenther@suse.de>
205 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
206 sign-conversions as zero, fall back to standard scalar_stmt
209 2018-03-20 Martin Liska <mliska@suse.cz>
212 * predict.c (rebuild_frequencies): Handle case when we have
213 PROFILE_ABSENT, but flag_guess_branch_prob is false.
215 2018-03-20 Jakub Jelinek <jakub@redhat.com>
218 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
219 flag_section_anchors.
220 * varasm.c (use_blocks_for_decl_p): Remove hack for
224 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
226 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
227 be created, use lowpart_subreg of operands[0] rather than operands[0]
229 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
230 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
231 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
232 and n constraint instead of aarch64_shift_imm_di and Usd.
233 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
234 (*aarch64_<optab>_reg_minus<mode>3): ... this.
236 2018-03-20 Sudakshina Das <sudi.das@arm.com>
239 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
240 to favor GPR over NEON registers.
241 (<shift>di3_neon): Likewise.
243 2018-03-20 Tom de Vries <tom@codesourcery.com>
246 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
247 (nvptx_process_pars): Emit bar.sync asap and alap.
249 2018-03-20 Tom de Vries <tom@codesourcery.com>
252 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
253 seen_label if seen_label is already set.
255 2018-03-20 Jakub Jelinek <jakub@redhat.com>
258 * config/i386/i386.c (fold_builtin_cpu): For features above 31
259 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
260 Use 1U instead of 1. Formatting fixes.
263 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
264 instead of TREE_TYPE (s1) for the return value.
266 2018-03-19 Jakub Jelinek <jakub@redhat.com>
268 PR tree-optimization/84946
269 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
270 bitsize + bitsize in poly_uint64 rather than poly_int64.
273 * dwarf2asm.c: Include fold-const.c.
274 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
275 of decl rather than decl itself.
277 PR rtl-optimization/84643
278 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
280 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
283 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
284 calling assemble_variable.
286 2018-03-19 Sudakshina Das <sudi.das@arm.com>
289 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
290 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
292 2018-03-19 Jim Wilson <jimw@sifive.com>
295 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
296 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
297 (riscv_first_stack_step): Likewise.
298 (riscv_option_override): Use STACK_BOUNDARY instead of
300 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
302 (BIGGEST_ALIGNMENT): Set to 128.
303 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
304 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
307 2018-03-19 Richard Biener <rguenther@suse.de>
309 PR tree-optimization/84933
310 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
311 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
313 2018-03-19 Richard Biener <rguenther@suse.de>
315 PR tree-optimization/84859
316 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
317 (cond_if_else_store_replacement): Perform sinking operation on
318 single-store BBs regardless of MAX_STORES_TO_SINK setting.
319 Generalize what a BB with a single eligible store is.
321 2018-03-19 Richard Biener <rguenther@suse.de>
323 PR tree-optimization/84929
324 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
325 chrec_is_positive against non-chrec arg.
327 2018-03-19 Tamar Christina <tamar.christina@arm.com>
330 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
332 2018-03-18 Martin Liska <mliska@suse.cz>
334 PR rtl-optimization/84635
335 * regrename.c (build_def_use): Use matches_mode only when
338 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
340 PR tree-optimization/84913
341 * tree-vect-loop.c (vectorizable_reduction): Don't try to
342 vectorize chains of COND_EXPRs.
344 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
346 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
348 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
350 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
352 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
354 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
356 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
357 Kito Cheng <kito.cheng@gmail.com>
359 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
360 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
361 (nds32_adjust_reg_alloc_order): New function.
362 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
364 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
366 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
367 nds32_print_operand, nds32_print_operand_address): Use
368 HOST_WIDE_INT_PRINT_DEC instead.
370 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
372 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
374 2018-03-17 Jakub Jelinek <jakub@redhat.com>
377 * config/i386/i386.c (initial_ix86_tune_features,
378 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
380 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
381 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
382 rather than 1u << ix86_tune. Formatting fix.
383 (ix86_option_override_internal): Change ix86_arch_mask from
384 unsigned int to unsigned HOST_WIDE_INT, initialize to
385 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
386 (ix86_function_specific_restore): Likewise.
388 2018-03-16 Jakub Jelinek <jakub@redhat.com>
391 * postreload.c (reload_combine_recognize_pattern): Perform
392 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
393 truncate_int_for_mode the result for the destination's mode.
396 * hsa-gen.c (mem_type_for_type): Fix comment typo.
397 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
399 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
402 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
405 * lra-assigns.c (lra_split_hard_reg_for): Don't use
406 regno_allocno_class_array and sorted_pseudos.
407 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
408 insns where regno is used.
410 2018-03-16 Martin Liska <mliska@suse.cz>
413 * multiple_target.c (create_dispatcher_calls): Redirect
414 reference in the symbol table.
416 2018-03-16 Martin Liska <mliska@suse.cz>
419 * multiple_target.c (create_dispatcher_calls): Redirect also
422 2018-03-16 Jakub Jelinek <jakub@redhat.com>
426 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
427 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
428 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
430 2018-03-16 Julia Koval <julia.koval@intel.com>
432 * doc/invoke.texi (Skylake Server): Add CLWB.
433 Cannonlake): Remove CLWB.
435 2018-03-16 Jakub Jelinek <jakub@redhat.com>
437 PR tree-optimization/84841
438 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
440 (FLOAT_ONE_CONST_TYPE): Define.
441 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
442 (sort_by_operand_rank): Put entries with higher constant_type last
443 rather than first to match comments.
445 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
447 * config/nios2/nios2.md (movsi_internal): Fix thinko in
450 2018-03-15 Jakub Jelinek <jakub@redhat.com>
453 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
454 check and use address of target always.
456 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
459 * config/i386/i386.c (indirect_thunk_needed): Update comments.
460 (indirect_thunk_bnd_needed): Likewise.
461 (indirect_thunks_used): Likewise.
462 (indirect_thunks_bnd_used): Likewise.
463 (indirect_return_needed): New.
464 (indirect_return_bnd_needed): Likewise.
465 (output_indirect_thunk_function): Add a bool argument for
467 (output_indirect_thunk_function): Don't generate alias for
468 function return thunk.
469 (ix86_code_end): Call output_indirect_thunk_function to generate
470 function return thunks.
471 (ix86_output_function_return): Set indirect_return_bnd_needed
472 and indirect_return_needed instead of indirect_thunk_bnd_needed
473 and indirect_thunk_needed.
475 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
477 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
478 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
479 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
481 2018-03-15 David Malcolm <dmalcolm@redhat.com>
482 Paul Hua <paul.hua.gm@gmail.com>
485 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
487 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
489 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
490 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
493 2018-03-15 Tamar Christina <tamar.christina@arm.com>
496 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
497 instead of GET_MODE_SIZE when comparing Units.
499 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
502 * varasm.c (hash_section): Return an unchangeble hash value
503 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
504 Return !aarch64_can_use_per_function_literal_pools_p ().
506 2018-03-15 Jakub Jelinek <jakub@redhat.com>
509 * optabs.c (emit_conditional_move): Pass address of cmode's copy
510 rather than address of cmode as last argument to prepare_cmp_insn.
512 2018-03-15 Julia Koval <julia.koval@intel.com>
514 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
515 F_AVX512VNNI, F_AVX512BITALG): New.
517 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
520 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
521 insn for floating-point loads and stores.
523 2018-03-14 Carl Love <cel@us.ibm.com>
525 * config/rs6000/rs6000-c.c: Add macro definitions for
526 ALTIVEC_BUILTIN_VEC_PERMXOR.
527 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
528 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
529 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
530 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
532 * config/doc/extend.texi: Add prototypes for vec_permxor.
534 2018-03-14 David Malcolm <dmalcolm@redhat.com>
537 * diagnostic-show-locus.c (class layout_point): Convert m_line
538 from int to linenum_type.
539 (line_span::comparator): Use linenum "compare" function when
540 comparing line numbers.
541 (test_line_span): New function.
542 (layout_range::contains_point): Convert param "row" from int to
544 (layout_range::intersects_line_p): Likewise.
545 (layout::will_show_line_p): Likewise.
546 (layout::print_source_line): Likewise.
547 (layout::should_print_annotation_line_p): Likewise.
548 (layout::print_annotation_line): Likewise.
549 (layout::print_leading_fixits): Likewise.
550 (layout::annotation_line_showed_range_p): Likewise.
551 (struct line_corrections): Likewise for field m_row.
552 (line_corrections::line_corrections): Likewise for param "row".
553 (layout::print_trailing_fixits): Likewise.
554 (layout::get_state_at_point): Likewise.
555 (layout::get_x_bound_for_row): Likewise.
556 (layout::print_line): Likewise.
557 (diagnostic_show_locus): Likewise for locals "last_line" and
559 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
560 * input.c (selftest::test_linenum_comparisons): New function.
561 (selftest::input_c_tests): Call it.
562 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
563 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
564 * selftest.h (ASSERT_GT): New macro.
565 (ASSERT_GT_AT): New macro.
566 (ASSERT_LT): New macro.
567 (ASSERT_LT_AT): New macro.
569 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
571 PR rtl-optimization/84780
572 * combine.c (distribute_links): Don't make a link based on pc_rtx.
574 2018-03-14 Martin Liska <mliska@suse.cz>
576 * tree.c (record_node_allocation_statistics): Use
578 (get_stats_node_kind): New function extracted from
579 record_node_allocation_statistics.
580 (free_node): Use get_stats_node_kind.
582 2018-03-14 Richard Biener <rguenther@suse.de>
584 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
585 that the value-set of ANTIC_IN doesn't grow.
588 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
590 (BB_VISITED_WITH_VISITED_SUCCS): New define.
591 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
593 2018-03-14 Julia Koval <julia.koval@intel.com>
595 * config.gcc (icelake-client, icelake-server): New.
597 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
598 (initial_ix86_arch_features): Ditto.
599 (PTA_SKYLAKE): Add SGX.
600 (PTA_ICELAKE): Remove.
601 (PTA_ICELAKE_CLIENT): New.
602 (PTA_ICELAKE_SERVER): New.
603 (ix86_option_override_internal): Split up icelake on icelake client and
605 (get_builtin_code_for_version): Ditto.
606 (fold_builtin_cpu): Ditto.
607 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
608 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
609 * config/i386/i386.h (processor_type): Ditto.
610 * doc/invoke.texi: Ditto.
612 2018-03-14 Jakub Jelinek <jakub@redhat.com>
615 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
616 INTEGER_CST offset, add it together with bitpos / 8 and
617 sign extend based on POINTER_SIZE.
621 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
624 * config/i386/constraints.md (Yc): New register constraint.
625 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
626 Use Yc constraint for alternative 2 of operand 0. Remove
627 preferred_for_speed attribute.
629 2018-03-14 Richard Biener <rguenther@suse.de>
631 PR tree-optimization/84830
632 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
633 with the old one to avoid oscillations.
635 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
638 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
640 (assign_by_spills): Return a flag of reload assignment failure.
641 Do not process the reload assignment failures. Do not spill other
642 reload pseudos if they has the same reg class. Update n if
644 (lra_assign): Add a return arg. Set up from the result of
645 assign_by_spills call.
646 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
647 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
648 usage_insns if it is not NULL.
649 (spill_hard_reg_in_range): New function.
650 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
651 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
653 (lra_assign): Change prototype.
654 * lra.c (lra): Add code to deal with fails by splitting hard reg
657 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
659 * config/riscv/riscv.opt (mrelax): New option.
660 * config/riscv/riscv.c (riscv_file_start): Emit ".option
661 "norelax" when riscv_mrelax is disabled.
662 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
664 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
667 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
668 reassociation for int modes.
670 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
672 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
673 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
675 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
676 * config/aarch64/aarch64-sve.md
677 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
678 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
679 (*extend<mode><Vwide>2): Rename to...
680 (aarch64_sve_extend<mode><Vwide>2): ...this.
681 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
682 renaming the old pattern to...
683 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
685 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
686 define_expand, renaming the old pattern to...
687 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
688 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
689 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
690 account when deciding which SVE instruction the optab should use.
691 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
693 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
695 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
696 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
697 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
698 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
699 (tlsdesc_small_<mode>): Turn a define_expand and use
700 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
701 (tlsdesc_small_advsimd_<mode>): ...this.
702 (tlsdesc_small_sve_<mode>): New pattern.
704 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
706 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
707 (UNSPEC_UMUL_HIGHPART): New constants.
708 (MUL_HIGHPART): New int iteraor.
709 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
710 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
712 (*<su>mul<mode>3_highpart): New define_insn.
714 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
717 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
720 2018-03-13 Martin Liska <mliska@suse.cz>
723 * (sem_item_optimizer::sem_item_optimizer): Initialize new
725 (sem_item_optimizer::~sem_item_optimizer): Release it.
726 (sem_item_optimizer::merge_classes): Register variable aliases.
727 (sem_item_optimizer::fixup_pt_set): New function.
728 (sem_item_optimizer::fixup_points_to_sets): Likewise.
729 * ipa-icf.h: Declare new variables and functions.
731 2018-03-13 Jakub Jelinek <jakub@redhat.com>
734 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
735 integer_pow2p@2 and test integer_pow2p in condition.
736 (A < 0 ? C : 0): Similarly for @1.
739 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
740 characters starting at p contain '\0' character, don't look beyond
744 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
745 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
748 * reg-stack.c (change_stack): Change update_end var from int to
749 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
750 also call set_block_for_insn on the newly added insns and rescan.
753 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
757 * tree.c (stabilize_reference_1): Return save_expr (e) for
758 STATEMENT_LIST even if it doesn't have side-effects.
760 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
762 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
764 2018-03-12 Renlin Li <renlin.li@arm.com>
766 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
767 aarch64_output_scalar_simd_mov_immediate.
769 2018-03-12 Martin Sebor <msebor@redhat.com>
771 PR tree-optimization/83456
772 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
773 for perfectly overlapping calls to memcpy.
774 (gimple_fold_builtin_memory_chk): Same.
775 (gimple_fold_builtin_strcpy): Handle no-warning.
776 (gimple_fold_builtin_stxcpy_chk): Same.
777 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
779 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
781 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
782 parameter. Use it for SFmode.
783 (rs6000_function_arg_advance_1): Adjust.
784 (rs6000_function_arg): Adjust.
785 (rs6000_gimplify_va_arg): Pass false for that new parameter.
787 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
789 PR rtl-optimization/84169
790 PR rtl-optimization/84780
791 * combine.c (can_combine_p): Check for a 2-insn combination whether
792 the destination register is used between the two insns, too.
794 2018-03-12 Richard Biener <rguenther@suse.de>
796 PR tree-optimization/84803
797 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
798 for refs DR analysis didn't process.
800 2018-03-12 Richard Biener <rguenther@suse.de>
802 PR tree-optimization/84777
803 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
804 force-vectorize loops ignore whether we are optimizing for size.
806 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
808 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
809 (TARGET_MD_ASM_ADJUST): Define.
811 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
812 Kito Cheng <kito.cheng@gmail.com>
813 Chung-Ju Wu <jasonwucj@gmail.com>
815 * config/nds32/nds32.c (nds32_compute_stack_frame,
816 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
817 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
818 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
819 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
820 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
821 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
822 * config/nds32/nds32.md (prologue, epilogue): Use macro
823 NDS32_V3PUSH_AVAILABLE_P to do checking.
825 2018-03-11 Jakub Jelinek <jakub@redhat.com>
828 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
829 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
830 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
831 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
832 addition of most attributes on !orig_type_die or the attribute not
833 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
835 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
836 Chung-Ju Wu <jasonwucj@gmail.com>
838 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
840 * config/nds32/nds32.opt (mvh): New option.
842 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
843 Chung-Ju Wu <jasonwucj@gmail.com>
845 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
847 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
848 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
851 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
852 Chung-Ju Wu <jasonwucj@gmail.com>
854 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
856 * config/nds32/nds32-multiple.md (strlensi): New pattern.
857 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
859 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
860 Kito Cheng <kito.cheng@gmail.com>
861 Chung-Ju Wu <jasonwucj@gmail.com>
863 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
864 UNSPEC_FFMISM and UNSPEC_FLMISM.
865 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
866 for ffb, ffmism and flmism.
867 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
868 (unspec_ffmism): Ditto.
869 (unspec_flmism): Ditto.
870 (nds32_expand_builtin_impl): Check if string extension is available.
871 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
872 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
874 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
877 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
880 * lra-assigns.c (assign_by_spills): Return a flag of reload
881 assignment failure. Do not process the reload assignment
882 failures. Do not spill other reload pseudos if they has the same
884 (lra_assign): Add a return arg. Set up from the result of
885 assign_by_spills call.
886 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
887 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
888 usage_insns if it is not NULL.
889 (spill_hard_reg_in_range): New function.
890 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
891 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
893 (lra_assign): Change prototype.
894 * lra.c (lra): Add code to deal with fails by splitting hard reg
897 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
900 * config/i386/i386.opt: Replace Enforcment with Enforcement.
902 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
905 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
906 (dw_val_node): Add val_symbolic_view.
907 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
908 (symview_upper_bound): New.
909 (new_line_info_table): Initialize symviews_since_reset.
910 (dwarf2out_source_line): Count symviews_since_reset and set
912 (dw_val_equal_p): Handle symview.
913 (add_AT_symview): New.
914 (print_dw_val): Handle symview.
915 (attr_checksum, attr_checksum_ordered): Likewise.
916 (same_dw_val_p, size_of_die): Likewise.
917 (value_format, output_die): Likewise.
918 (add_high_low_attributes): Use add_AT_symview for entry_view.
919 (dwarf2out_finish): Reset symview_upper_bound, clear
922 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
925 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
926 Add strict argument and use it.
927 (rs6000_split_multireg_move): Update for new strict argument.
928 (mem_operand_gpr): Disallow all non-offsettable addresses.
929 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
931 2018-03-09 Jakub Jelinek <jakub@redhat.com>
934 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
935 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
936 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
939 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
940 decl, use remap_type if we want to use the type.
942 2018-03-09 Martin Sebor <msebor@redhat.com>
944 PR tree-optimization/84526
945 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
947 (builtin_access::generic_overlap): Be prepared to handle non-array
950 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
952 PR rtl-optimization/84682
953 * lra-constraints.c (process_address_1): Check is_address flag
954 for address constraints.
955 (process_alt_operands): Likewise.
956 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
957 preprocess_constraints.
958 * recog.h (preprocess_constraints): Add oploc parameter.
960 * recog.c (preprocess_constraints): Test address_operand for
961 CT_ADDRESS constraints.
963 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
966 * lra-assigns.c (assign_by_spills): Return a flag of reload
967 assignment failure. Do not process the reload assignment
968 failures. Do not spill other reload pseudos if they has the same
970 (lra_assign): Add a return arg. Set up from the result of
971 assign_by_spills call.
972 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
973 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
974 usage_insns if it is not NULL.
975 (spill_hard_reg_in_range): New function.
976 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
977 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
979 (lra_assign): Change prototype.
980 * lra.c (lra): Add code to deal with fails by splitting hard reg
983 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
986 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
987 Accept complain bool parameter. Only emit errors if it is true.
988 (arm_parse_cpu_option_name): Likewise.
989 (arm_target_thumb_only): Adjust callers of the above.
990 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
991 prototype to take a default true bool parameter.
992 (arm_parse_arch_option_name): Likewise.
994 2018-03-09 David Malcolm <dmalcolm@redhat.com>
995 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
999 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1000 * configure: Regenerate.
1001 * configure.ac ("linker --version-script option"): New.
1002 ("linker soname option"): New.
1004 2018-03-09 Richard Biener <rguenther@suse.de>
1006 PR tree-optimization/84775
1007 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1008 immediate uses of predicate stmts and mark them modified.
1011 PR tree-optimization/84178
1012 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1014 (version_loop_for_if_conversion): Delay update_ssa call.
1015 (tree_if_conversion): Delay update_ssa until after predicate
1018 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1021 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1022 when the function accesses prior frames.
1024 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1027 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1028 gen_llsym, otherwise call maybe_gen_llsym.
1031 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1032 has ',' character inside of it.
1034 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1037 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1038 as clobbering CC_REGNUM.
1040 2018-03-08 Richard Biener <rguenther@suse.de>
1043 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1044 (follow_copies_to_constant): Do not follow SSA names registered
1047 2018-03-08 Richard Biener <rguenther@suse.de>
1049 PR tree-optimization/84178
1050 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1052 (version_loop_for_if_conversion): Delay update_ssa call.
1053 (tree_if_conversion): Delay update_ssa until after predicate
1056 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1058 PR tree-optimization/84178
1059 * tree-if-conv.c (release_bb_predicate): Remove the
1060 the assertion that the stmts have NULL use_ops.
1061 Discard the statements, asserting that they haven't
1062 yet been added to a BB.
1064 2018-03-08 Richard Biener <rguenther@suse.de>
1066 PR tree-optimization/84746
1067 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1068 (phi_translate): Pass in destination ANTIC_OUT set.
1069 (phi_translate_1): Likewise. For a simplified result lookup
1070 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1071 (phi_translate_set): Adjust.
1072 (do_pre_regular_insertion): Likewise.
1073 (do_pre_partial_partial_insertion): Likewise.
1075 2018-03-08 Martin Liska <mliska@suse.cz>
1077 PR gcov-profile/84735
1078 * doc/gcov.texi: Document usage of profile files.
1079 * gcov-io.h: Document changes in the format.
1081 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1085 * dwarf2out.c (struct dw_line_info_table): Update comments for
1087 (FORCE_RESET_NEXT_VIEW): New.
1088 (FORCE_RESETTING_VIEW_P): New.
1089 (RESETTING_VIEW_P): Check for -1 too.
1090 (ZERO_VIEW_P): Likewise.
1091 (new_line_info_table): Force-reset next view.
1092 (dwarf2out_begin_function): Likewise.
1093 (dwarf2out_source_line): Simplify zero_view_p initialization.
1094 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1095 view directly. Omit view when omitting .loc at line 0.
1097 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1099 PR tree-optimization/84740
1100 * tree-switch-conversion.c (process_switch): Call build_constructors
1101 only if info.phi_count is non-zero.
1103 PR tree-optimization/84739
1104 * tree-tailcall.c (find_tail_calls): Check call arguments against
1105 DECL_ARGUMENTS (current_function_decl) rather than
1106 DECL_ARGUMENTS (func) when checking for tail recursion.
1108 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1110 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1111 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1112 Volker Reichelt's entry and add entries for people that perform
1113 GCC fuzzy testing and report numerous bugs.
1115 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1118 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1119 readonly data in sdata, if that is disabled.
1120 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1121 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1122 -mreadonly-in-sdata option.
1124 2018-03-07 Martin Sebor <msebor@redhat.com>
1126 PR tree-optimization/84468
1127 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1128 basic block when looking for nul assignment.
1130 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1133 * except.h (output_function_exception_table): Adjust prototype.
1134 * except.c (output_function_exception_table): Remove FNNAME parameter
1135 and add SECTION parameter. Ouput one part of the table at a time.
1136 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1137 the first part of the exception table and emit unwind directives.
1138 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1139 (i386_pe_seh_cold_init): Likewise.
1140 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1141 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1142 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1143 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1144 * config/i386/winnt.c: Include except.h.
1145 (struct seh_frame_state): Add reg_offset, after_prologue and
1146 in_cold_section fields.
1147 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1148 (i386_pe_seh_cold_init): New function.
1149 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1150 to seh->in_cold_section.
1151 (seh_emit_push): Record the offset of the push.
1152 (seh_emit_save): Record the offet of the save.
1153 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1154 Test seh->after_prologue to disregard the epilogue.
1155 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1156 (i386_pe_end_cold_function): New function.
1158 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1161 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1162 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1165 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1167 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1168 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1171 * multiple_target.c: Include tree-inline.h and intl.h.
1172 (expand_target_clones): Diagnose and fail if node->definition and
1173 !tree_versionable_function_p (node->decl).
1175 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1177 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1179 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1180 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1181 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1183 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1186 * combine.c (try_combine): Use reg_or_subregno instead of handling
1187 just paradoxical SUBREGs and REGs.
1189 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1191 * config/arc/arc.c (arc_finalize_pic): Remove function.
1192 (arc_must_save_register): We use single base PIC register, remove
1193 checks to save/restore the PIC register.
1194 (arc_expand_prologue): Likewise.
1195 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1197 (arc_verify_short): Likewise.
1198 (arc_attr_type): Likewise.
1199 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1200 (walk_stores): Likewise.
1201 (arc_address_cost): Make it static.
1202 (arc_verify_short): Likewise.
1203 (branch_dest): Likewise.
1204 (arc_attr_type): Likewise.
1205 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1206 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1207 (arc_final_prescan_insn): Remove inserting the nops due to
1208 hardware hazards. It is done in reorg step.
1209 (insn_length_variant_t): Remove.
1210 (insn_length_parameters_t): Likewise.
1211 (arc_insn_length_parameters): Likewise.
1212 (arc_get_insn_variants): Likewise.
1213 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1215 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1218 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1221 PR tree-optimization/84687
1222 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1224 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1226 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1228 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1229 Rename to ppc_speculation_barrier.
1230 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1231 __builtin_ppc_speculation_barrier.
1233 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1236 * combine.c (combine_simplify_rtx): Don't try to simplify if
1237 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1240 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1242 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1243 to 32 bytes when compiling for POWER9.
1245 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1248 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1249 regparm >= 3 with no arg reg available also for calls with
1250 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1253 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1255 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1257 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1260 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1262 2018-03-05 Richard Biener <rguenther@suse.de>
1264 PR tree-optimization/84486
1265 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1266 When inserting a __builtin_assume_aligned call set the LHS
1267 SSA name alignment info accordingly.
1269 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1271 PR tree-optimization/84114
1272 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1273 Avoid reassociation of FLOAT_MODE addition.
1275 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1277 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1278 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1279 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1280 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1281 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1282 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1283 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1285 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1286 __builtin_ia32_wbinvd): New builtins.
1287 (SPECIAL_ARGS2): New.
1288 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1289 (SPECIAL_ARGS2): New.
1290 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1291 (ix86_valid_target_attribute_inner_p): Ditto.
1292 (ix86_init_mmx_sse_builtins): Add special_args2.
1293 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1294 TARGET_WBNOINVD_P): New.
1295 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1296 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1297 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1298 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1299 * config/i386/pconfigintrin.h: New file.
1300 * config/i386/wbnoinvdintrin.h: Ditto.
1301 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1302 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1304 2018-03-05 Richard Biener <rguenther@suse.de>
1306 PR tree-optimization/84670
1307 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1309 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1310 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1311 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1312 doesn't grow if all successors (recursively) were visited at least
1315 2018-03-05 Richard Biener <rguenther@suse.de>
1317 PR tree-optimization/84650
1318 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1319 if executed in the loop pipeline.
1321 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1323 * doc/configfiles.texi (Configuration Files): Move info about
1324 conditionalizing $target-protos.h to...
1325 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1326 differs from $target-protos.h.
1328 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1329 Chung-Ju Wu <jasonwucj@gmail.com>
1331 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1332 * config/nds32/nds32-multiple.md (setmemsi): Define.
1333 * config/nds32/nds32-memory-manipulation.c
1334 (nds32_gen_dup_4_byte_to_word_value): New.
1335 (emit_setmem_word_loop): New.
1336 (emit_setmem_byte_loop): New.
1337 (nds32_expand_setmem_loop): New.
1338 (nds32_expand_setmem_loop_v3m): New.
1339 (nds32_expand_setmem_unroll): New.
1340 (nds32_expand_setmem): New.
1342 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1343 Chung-Ju Wu <jasonwucj@gmail.com>
1345 * config/nds32/nds32-memory-manipulation.c
1346 (nds32_emit_load_store): New.
1347 (nds32_emit_post_inc_load_store): New.
1348 (nds32_emit_mem_move): New.
1349 (nds32_emit_mem_move_block): New.
1350 (nds32_expand_movmemsi_loop_unknown_size): New.
1351 (nds32_expand_movmemsi_loop_known_size): New.
1352 (nds32_expand_movmemsi_loop): New.
1353 (nds32_expand_movmemsi_unroll): New.
1354 (nds32_expand_movmemqi): Rename ...
1355 (nds32_expand_movmemsi): ... to this.
1356 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1357 (movmemsi): ... to this.
1358 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1359 (nds32_expand_movmemsi): ... to this.
1361 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1362 Monk Chiang <sh.chiang04@gmail.com>
1363 Chung-Ju Wu <jasonwucj@gmail.com>
1365 * config/nds32/nds32-protos.h
1366 (nds32_expand_load_multiple): New arguments.
1367 (nds32_expand_store_multiple): Ditto.
1368 (nds32_valid_multiple_load_store): Rename ...
1369 (nds32_valid_multiple_load_store_p): ... to this.
1370 * config/nds32/nds32-memory-manipulation.c
1371 (nds32_expand_load_multiple): Refine implementation.
1372 (nds32_expand_store_multiple): Ditto.
1373 * config/nds32/nds32-multiple.md
1374 (load_multiple): Update nds32_expand_load_multiple interface.
1375 (store_multiple): Update nds32_expand_store_multiple interface.
1376 * config/nds32/nds32-predicates.c
1377 (nds32_valid_multiple_load_store): Rename ...
1378 (nds32_valid_multiple_load_store_p): ... to this and refine
1380 * config/nds32/predicates.md
1381 (nds32_load_multiple_and_update_address_operation): New predicate.
1382 (nds32_store_multiple_and_update_address_operation): New predicate.
1384 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1385 Chung-Ju Wu <jasonwucj@gmail.com>
1387 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1388 (combo): New attribute.
1389 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1391 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1393 * config/nds32/nds32.opt: Change -mcmodel= default value.
1395 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1396 Monk Chiang <sh.chiang04@gmail.com>
1397 Chung-Ju Wu <jasonwucj@gmail.com>
1399 * config/nds32/constants.md (unspec_element): New enum.
1400 * config/nds32/constraints.md (Umw): New constraint.
1401 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1402 * config/nds32/nds32-intrinsic.md: Likewise.
1403 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1404 (nds32_valid_smw_lwm_base_p): New.
1405 (nds32_output_smw_single_word): New.
1406 (nds32_output_lmw_single_word): New.
1407 (nds32_expand_unaligned_load): New.
1408 (nds32_expand_unaligned_store): New.
1409 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1410 (nds32_output_smw_single_word): Declare.
1411 (nds32_output_lmw_single_word): Declare.
1412 (nds32_expand_unaligned_load): Declare.
1413 (nds32_expand_unaligned_store): Declare.
1414 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1415 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1416 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1417 NDS32_BUILTIN_UASTORE_DW.
1418 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1421 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1422 Kito Cheng <kito.cheng@gmail.com>
1423 Chung-Ju Wu <jasonwucj@gmail.com>
1425 * config/nds32/nds32-intrinsic.c
1426 (nds32_expand_builtin_null_ftype_reg): Delete.
1427 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1428 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1429 (nds32_read_argument): New.
1430 (nds32_legitimize_target): Ditto.
1431 (nds32_legitimize_argument): Ditto.
1432 (nds32_check_constant_argument): Ditto.
1433 (nds32_expand_unop_builtin): Ditto.
1434 (nds32_expand_unopimm_builtin): Ditto.
1435 (nds32_expand_binop_builtin): Ditto.
1436 (nds32_builtin_decl_impl): Ditto.
1437 (builtin_description): Ditto.
1438 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1439 (nds32_init_builtins_impl): Ditto.
1440 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1441 (nds32_builtin_decl): New.
1442 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1443 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1445 2018-03-02 Jeff Law <law@redhat.com>
1447 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1448 (redundant_insn, fill_simple_delay_slots): Likewise.
1449 (fill_slots_from_thread): Likewise.
1450 * resource.c (mark_referenced_resources): Likewise.
1451 (mark_set_resources, find_dead_or_set_registers): Likewise.
1453 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1455 * substring-locations.h (format_warning_va): Formatting fix for
1457 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1459 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1460 * substring-locations.c: Include intl.h.
1461 (format_warning_va): Turned into small wrapper around
1462 format_warning_n_va, renamed to ...
1463 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1464 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1466 (format_warning_at_substring_n): New function.
1467 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1468 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1469 format_warning_at_substring with just a shorter name instead of
1470 const function pointer.
1471 (fmtwarn_n): New function.
1472 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1473 appropriate, get rid of all the fmtstr temporaries, move conditionals
1474 with G_() wrapped string literals directly into fmtwarn arguments,
1475 cast dir.len to (int), formatting fixes.
1477 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1479 * doc/invoke.texi: Remove "Cilk Plus" references.
1481 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1482 Richard Biener <rguenther@suse.de>
1485 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1486 for error or warning attributes if CALL_FROM_THUNK_P is set.
1489 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1492 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1493 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1496 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1497 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1499 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1500 instead of -1U in last predictors element's probability member.
1502 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1505 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1506 arguments if they are comparable.
1508 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1510 PR tree-optimization/84634
1511 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1512 masks and masked_loop_p with a single loop_masks, making sure it's
1513 null for bb vectorization.
1515 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1517 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1518 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1519 loop->force_vectorize to check whether there is no alias.
1521 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1524 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1526 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1528 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1529 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1530 instead of a loop around prev_real_insn.
1531 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1535 * config/i386/i386.c (ix86_print_operand): Use conditional
1536 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1539 2018-03-02 Richard Biener <rguenther@suse.de>
1541 PR tree-optimization/84427
1542 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1543 (bitmap_set_subtract_values): Rewrite to handle multiple
1546 (prune_clobbered_mems): Likewise.
1547 (phi_translate): Take edge instead of pred/phiblock.
1548 (phi_translate_1): Likewise.
1549 (phi_translate_set): Likewise. Insert all translated
1550 exprs for a value into the set, keeping possibly multiple
1551 expressions per value.
1552 (compute_antic_aux): Adjust for phi_translate changes.
1553 When intersecting union the expressions and prune those
1554 not in the final value set, keeping possibly multiple
1555 expressions per value. Do not use value-insertion
1556 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1557 all expressions. Add verification that the value-sets
1558 only shrink during iteration.
1559 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1560 (do_pre_regular_insertion): Likewise.
1561 (do_pre_partial_partial_insertion): Likewise.
1563 2018-03-02 Richard Biener <rguenther@suse.de>
1566 * config/darwin.c (saved_debug_info_level): New static global.
1567 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1568 (darwin_asm_lto_end): Restore debug info generation settings.
1570 2018-03-01 Martin Liska <mliska@suse.cz>
1573 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1576 2018-03-01 Richard Biener <rguenther@suse.de>
1579 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1580 type annotation with LTO when debug was disabled at compile-time.
1582 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1584 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1586 (mips_final_postscan_insn): Likewise.
1588 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1590 PR rtl-optimization/84528
1591 * alias.c (init_alias_target): Add commentary.
1592 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1593 a unique base value if the frame pointer is not eliminated
1594 to the stack pointer.
1596 2018-03-01 Tom de Vries <tom@codesourcery.com>
1598 PR rtl-optimization/83327
1599 * lra-int.h (hard_regs_spilled_into): Declare.
1600 * lra.c (hard_regs_spilled_into): Define.
1601 (init_reg_info): Init hard_regs_spilled_into.
1602 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1603 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1604 (process_bb_lives): Handle hard_regs_spilled_into.
1605 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1606 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1608 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1610 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1611 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1612 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1613 * config/rs6000/aix72.h: New file.
1615 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1617 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1618 instead of warning_at with conditional singular and plural messages
1622 * stor-layout.c (update_alignment_for_field): For
1623 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1624 && !DECL_PACKED (field), do the alignment update, just use
1625 only desired_align instead of MAX (type_align, desired_align)
1627 (place_field): Don't do known_align < desired_align handling
1628 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1629 is non-NULL, instead do it after rli->prev_field handling and
1630 only if not within a bitfield word. For DECL_PACKED (field)
1631 use type_align of BITS_PER_UNIT.
1633 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1635 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1636 superfluous parentheses and trailing spaces.
1638 2018-02-28 Richard Biener <rguenther@suse.de>
1640 PR tree-optimization/84584
1641 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1642 SCoPs with fake exit edge.
1644 2018-02-28 Martin Liska <mliska@suse.cz>
1647 * timevar.c (timer::print): Fix format to properly print 100%
1650 2018-02-28 Richard Biener <rguenther@suse.de>
1653 * genmatch.c (capture_info::walk_match): Do not mark
1654 captured expressions without operands as expr_p given
1655 they act more like predicates and should be subject to
1656 "lost tail" side-effect preserving.
1658 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1660 PR rtl-optimization/81611
1661 * auto-inc-dec.c (attempt_change): Move dead note from
1662 mem_insn if it's the next use of regno
1663 (find_address): Take address use of reg holding
1664 non-incremented value. Add parm to limit search to the named
1666 (merge_in_block): Attempt to use a mem insn that is the next
1667 use of the original regno.
1669 2018-02-27 Martin Sebor <msebor@redhat.com>
1672 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1673 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1675 2018-02-27 Martin Sebor <msebor@redhat.com>
1677 PR translation/84207
1678 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1679 n argument to unsigned HOST_WIDE_INT.
1680 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1681 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1682 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1683 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1685 2018-02-27 Richard Biener <rguenther@suse.de>
1687 PR tree-optimization/84512
1688 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1689 Do not use the estimate returned from record_stmt_cost for
1690 the scalar iteration cost but sum properly using add_stmt_cost.
1692 2018-02-27 Richard Biener <rguenther@suse.de>
1694 PR tree-optimization/84466
1695 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1696 Adjust last change to less strictly validate use operands.
1698 2018-02-27 Martin Liska <mliska@suse.cz>
1700 PR gcov-profile/84548
1701 * gcov.c (process_file): Allow partial overlap and consider it
1702 also as group functions.
1703 (output_lines): Properly calculate range of lines for a group.
1705 2018-02-27 Martin Liska <mliska@suse.cz>
1707 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1708 'ggc' suffixes. Change first column width.
1709 (timer::print): Fix formatting of the column.
1711 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1713 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1714 preserve inline entry blocks for the sake of debug inline
1715 entry point markers alone.
1716 (remove_unused_locals): Suggest in comments a better place to
1717 force the preservation of inline entry blocks that are
1718 otherwise unused, but do not preserve them.
1720 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1722 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1724 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1727 * config/i386/constraints.md (Bs): Replace
1728 ix86_indirect_branch_register with
1729 TARGET_INDIRECT_BRANCH_REGISTER.
1731 * config/i386/i386.md (indirect_jump): Likewise.
1732 (tablejump): Likewise.
1733 (*sibcall_memory): Likewise.
1734 (*sibcall_value_memory): Likewise.
1735 Peepholes of indirect call and jump via memory: Likewise.
1736 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1737 (*sibcall_value_GOT_32): Likewise.
1738 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1739 (GOT_memory_operand): Likewise.
1740 (call_insn_operand): Likewise.
1741 (sibcall_insn_operand): Likewise.
1742 (GOT32_symbol_operand): Likewise.
1743 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1745 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1747 PR rtl-optimization/83496
1748 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1749 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1750 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1751 redundant insn, if any.
1752 (relax_delay_slots): Likewise.
1753 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1755 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1757 PR tree-optimization/83965
1758 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1759 that grouped statements are part of a reduction chain. Return
1760 true if the statement is not marked as a reduction itself but
1762 (vect_recog_dot_prod_pattern): Don't check whether the statement
1763 is part of a group here.
1764 (vect_recog_sad_pattern): Likewise.
1765 (vect_recog_widen_sum_pattern): Likewise.
1767 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1770 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1772 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1775 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1777 (ix86_output_indirect_function_return): New prototype.
1778 (ix86_split_simple_return_pop_internal): Likewise.
1779 * config/i386/i386.c (indirect_return_via_cx): New.
1780 (indirect_return_via_cx_bnd): Likewise.
1781 (indirect_thunk_name): Handle return va CX_REG.
1782 (output_indirect_thunk_function): Create alias for
1783 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1784 (ix86_output_indirect_jmp): Remove the bool argument.
1785 (ix86_output_indirect_function_return): New function.
1786 (ix86_split_simple_return_pop_internal): Likewise.
1787 * config/i386/i386.md (*indirect_jump): Don't pass false
1788 to ix86_output_indirect_jmp.
1789 (*tablejump_1): Likewise.
1790 (simple_return_pop_internal): Change it to define_insn_and_split.
1791 Call ix86_split_simple_return_pop_internal to split it for
1793 (simple_return_indirect_internal): Call
1794 ix86_output_indirect_function_return instead of
1795 ix86_output_indirect_jmp.
1797 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1800 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1801 memset and value initialization afterwards.
1803 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1805 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1807 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1810 * common/config/aarch64/aarch64-common.c
1811 (aarch_option_optimization_table[]): Switch
1812 off fomit-frame-pointer
1814 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1815 Chung-Ju Wu <jasonwucj@gmail.com>
1817 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1819 (store_multiple): Ditto.
1821 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1823 * config.gcc: Add --with-cpu support for nds32 target.
1824 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1825 * config/nds32/nds32.opt: Add -mcpu= option.
1827 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1829 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1830 isel=yes): Warn for these deprecated options.
1832 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1834 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1835 ISA_2_5_MASKS_EMBEDDED.
1837 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1839 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1840 p->max as pointers rather than using iterative_hash_expr.
1842 2018-02-23 Carl Love <cel@us.ibm.com>
1844 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1845 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1847 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1848 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1849 P8V_BUILTIN_VEC_VUNSIGNED2.
1851 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1854 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1855 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1857 * lra-constraints.c (curr_insn_transform): Set up
1858 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1860 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1861 * lra-eliminations.c (spill_pseudos): Ditto.
1862 (process_insn_for_elimination): Ditto.
1863 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1864 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1865 LRA_NON_CLOBBERED_ALT.
1867 2018-02-22 Martin Sebor <msebor@redhat.com>
1869 PR tree-optimization/84480
1870 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1871 to maybe_diag_stxncpy_trunc. Call it.
1872 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1873 from gimple_fold_builtin_strcpy. Print inlining stack.
1874 (handle_builtin_stxncpy): Print inlining stack.
1875 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1877 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1880 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1881 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1882 and -fcheck-pointer-bounds are used together.
1883 (indirect_thunk_prefix): New enum.
1884 (indirect_thunk_need_prefix): New function.
1885 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1886 "_nt" instead of "_bnd" for NOTRACK prefix.
1887 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1888 (output_indirect_thunk_function): Likewise.
1890 (ix86_code_end): Update output_indirect_thunk_function calls.
1891 (ix86_output_indirect_branch_via_reg): Replace
1892 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1893 (ix86_output_indirect_branch_via_push): Likewise.
1894 (ix86_output_function_return): Likewise.
1895 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1896 incompatible with -fcf-protection=branch and
1897 -fcheck-pointer-bounds.
1899 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1902 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1903 Change gcc_assert call to output_operand_lossage.
1905 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1907 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1909 2018-02-22 DJ Delorie <dj@redhat.com>
1910 Sebastian Perta <sebastian.perta@renesas.com>
1911 Oleg Endo <olegendo@gcc.gnu.org>
1913 * config/rx/rx.c (rx_rtx_costs): New function.
1914 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1916 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1918 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1920 2018-02-22 Martin Liska <mliska@suse.cz>
1923 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1924 Add "native" as a possible value.
1926 2018-02-22 Martin Liska <mliska@suse.cz>
1929 * config/i386/i386.c (ix86_option_override_internal):
1930 Add "native" as a possible value for -march and -mtune.
1932 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1935 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1936 to all type variants.
1938 PR tree-optimization/84503
1939 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1940 width as info->bitpos + info->bitsize - start.
1941 (merged_store_group::merge_overlapping): Simplify width computation.
1942 (check_no_overlap): New function.
1943 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1944 start + width and last_order of the group, fail if check_no_overlap
1946 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1947 to group if check_no_overlap fails.
1949 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1951 * config/rs6000/altivec.md: Delete contraint arguments to
1952 define_expand, define_split, and define_peephole2, and in
1953 define_insn_and_split if always unused.
1954 * config/rs6000/darwin.md: Ditto.
1955 * config/rs6000/dfp.md: Ditto.
1956 * config/rs6000/rs6000.md: Ditto.
1957 * config/rs6000/sync.md: Ditto.
1958 * config/rs6000/vector.md: Ditto.
1959 * config/rs6000/vsx.md: Ditto.
1961 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1963 * config/rs6000/altivec.md: Write output control strings as braced
1964 blocks instead of double-quoted strings.
1965 * config/rs6000/darwin.md: Ditto.
1966 * config/rs6000/rs6000.md: Ditto.
1967 * config/rs6000/vector.md: Ditto.
1968 * config/rs6000/vsx.md: Ditto.
1970 2018-02-21 Jason Merrill <jason@redhat.com>
1972 PR c++/84314 - ICE with templates and fastcall attribute.
1973 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1975 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1977 * ipa-cp.c (determine_versionability): Fix comment typos.
1979 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1982 * ipa-cp.c (determine_versionability): Do not version functions caling
1985 2018-02-21 Martin Liska <mliska@suse.cz>
1988 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1989 Add "native" as a possible value.
1990 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1991 the macro when native cpu detection is available.
1993 2018-02-21 Martin Liska <mliska@suse.cz>
1996 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1997 Add "native" as a possible value.
1998 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1999 when native cpu detection is available.
2001 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2002 Martin Sebor <msebor@redhat.com>
2004 PR tree-optimization/84478
2005 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2007 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2008 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2009 support which is conservatively correct, for 2 only stay conservative
2010 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2011 argument to the 2 argument get_range_strlen, adjust 6 arg
2012 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2014 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2015 (gimple_fold_builtin_strlen): Pass true as last argument to
2018 2018-02-20 Martin Sebor <msebor@redhat.com>
2021 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2022 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2023 (builtin_memref::builtin_memref): Factor out parts into
2024 set_base_and_offset and call it.
2026 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2029 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2030 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2031 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2032 search at the associated MODE_INT.
2034 2018-02-20 Jeff Law <law@redhat.com>
2037 PR tree-optimization/81592
2039 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2040 for range data rather than using global data.
2041 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2042 range data rather than using global data.
2043 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2044 pass it to children as needed.
2045 (struct directive::fmtresult): Similarly.
2046 (struct directive::set_width): Similarly.
2047 (struct directive::set_precision): Similarly.
2048 (format_integer, format_directive, parse_directive): Similarly.
2049 (format_none): Accept unnamed vr_values parameter.
2050 (format_percent, format_floating, format_character): Similarly.
2051 (format_string, format_plain): Similarly.
2052 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2053 the EVRP range analyzer for range data rather than using global data.
2054 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2055 gimple-ssa-evrp-analyze.h
2056 (class sprintf_dom_walker): Add after_dom_children member function.
2057 Add evrp_range_analyzer member.
2058 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2059 range analyzer as needed.
2060 (sprintf_dom_walker::after_dom_children): New member function.
2061 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2063 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2064 (evrp_range_analyzer::pop_to_marker): Likewise.
2066 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2068 PR tree-optimization/84419
2069 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2070 with the required type if its current type is compatible but
2073 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2076 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2077 after vectorization.
2079 2018-02-20 Martin Liska <mliska@suse.cz>
2082 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2083 possible values if we don't have a hint.
2085 2018-02-20 Martin Liska <mliska@suse.cz>
2089 * final.c (shorten_branches): Build align_tab array with one
2091 * opts.c (finish_options): Add alignment option limit check.
2092 (MAX_CODE_ALIGN): Likewise.
2093 (MAX_CODE_ALIGN_VALUE): Likewise.
2094 * doc/invoke.texi: Document maximum allowed option value for
2095 all -falign-* options.
2097 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2100 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2101 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2102 * var-tracking.c (emit_note_insn_var_location): Remove all references
2103 to NOTE_INSN_CALL_ARG_LOCATION.
2104 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2105 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2106 Use copy_rtx_if_shared.
2107 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2108 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2109 (dwarf2out_var_location): Remove handling of
2110 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2112 * final.c (final_scan_insn): Remove all references to
2113 NOTE_INSN_CALL_ARG_LOCATION.
2114 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2115 before dumping final insns.
2116 * except.c (emit_note_eh_region_end): Remove all references to
2117 NOTE_INSN_CALL_ARG_LOCATION.
2118 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2119 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2120 * config/arc/arc.c (hwloop_optimize): Likewise.
2121 * config/arm/arm.c (create_fix_barrier): Likewise.
2122 * config/s390/s390.c (s390_chunkify_start): Likewise.
2123 * config/sh/sh.c (find_barrier): Likewise.
2124 * config/i386/i386.c (rest_of_insert_endbranch,
2125 ix86_seh_fixup_eh_fallthru): Likewise.
2126 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2127 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2128 * config/frv/frv.c (frv_function_prologue): Likewise.
2129 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2131 (note_outside_basic_block_p): Remove all references to
2132 NOTE_INSN_CALL_ARG_LOCATION.
2133 * gengtype.c (adjust_field_rtx_def): Likewise.
2134 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2136 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2137 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2140 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2143 PR tree-optimization/84452
2144 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2145 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2148 2018-02-19 Martin Liska <mliska@suse.cz>
2151 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2153 2018-02-19 Martin Liska <mliska@suse.cz>
2154 Richard Sandiford <richard.sandiford@linaro.org>
2156 PR tree-optimization/82491
2157 * gimple-fold.c (get_base_constructor): Make earlier bail out
2160 2018-02-19 Carl Love <cel@us.ibm.com>
2162 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2163 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2165 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2166 P8V_BUILTIN_VEC_NEG.
2168 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2170 * config/rl78/rl78.md (movdf): New define expand.
2172 2018-02-19 Martin Liska <mliska@suse.cz>
2175 * doc/invoke.texi: Fix typo.
2176 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2178 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2180 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2181 handle rs6000_single_float and rs6000_double_float specially for
2184 2018-02-16 Jeff Law <law@redhat.com>
2186 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2189 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2192 * ipa-inline.c (inline_small_functions): Fix a typo.
2194 2018-02-16 Nathan Sidwell <nathan@acm.org>
2196 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2198 2018-02-16 Carl Love <cel@us.ibm.com>
2200 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2201 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2202 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2203 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2204 expansion to P8V_BUILTIN_VEC_FLOAT2.
2206 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2208 PR rtl-optimization/70023
2209 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2210 src_regno into account.
2212 2018-02-16 Carl Love <cel@us.ibm.com>
2214 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2215 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2216 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2217 * config/rs6000/rs6000.c: Remove case statements for
2218 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2219 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2220 and P9V_BUILTIN_VEC_VINSERT4B.
2221 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2222 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2223 * config/rs6000/vsx.md:
2224 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2227 2018-02-16 Carl Love <cel@us.ibm.com>
2229 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2231 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2233 * config/rs6000/rs6000-c.c: Add the definitions for
2234 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2235 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2236 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2237 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2238 definition for insert4b and define insn *insert3b_internal.
2239 * doc/extend.texi: Add documentation for vec_extract4b.
2241 2018-02-16 Nathan Sidwell <nathan@acm.org>
2243 * doc/extend.texi (Backwards Compatibility): Mention friend
2244 injection. Note for-scope is deprecated.
2245 * doc/invoke.texi (-ffriend-injection): Deprecate.
2247 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2249 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2250 that moved to I2, also allow destinations that are a paradoxical
2251 subreg (instead of a normal reg).
2253 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2256 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2259 2018-02-16 Richard Biener <rguenther@suse.de>
2261 PR tree-optimization/84037
2262 PR tree-optimization/84016
2264 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2265 Adjust vec_construct for the fact we need additional higher latency
2266 128bit inserts for AVX256 and AVX512 vector builds.
2267 (ix86_add_stmt_cost): Scale vector construction cost for
2270 2018-02-16 Richard Biener <rguenther@suse.de>
2272 PR tree-optimization/84417
2273 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2274 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2275 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2277 2018-02-16 Martin Liska <mliska@suse.cz>
2280 * internal-fn.def (ASAN_CHECK): Set proper flags.
2281 (ASAN_MARK): Likewise.
2283 2018-02-16 Julia Koval <julia.koval@intel.com>
2285 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2286 from PTA_CANNONLAKE.
2288 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2291 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2292 Use ++iter rather than iter++ for std::list iterators.
2293 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2294 defer deleting them until all nodes in the forest are processed. Do
2295 free even leaf nodes. Change to_process into auto_vec.
2298 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2299 * vec.h (vec_default_construct): Use memset instead of placement new
2300 if BROKEN_VALUE_INITIALIZATION is defined.
2301 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2302 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2305 PR rtl-optimization/83723
2306 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2307 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2308 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2310 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2312 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2314 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2316 PR rtl-optimization/81443
2317 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2318 from inner REGs to paradoxical SUBREGs.
2320 2018-02-16 Richard Biener <rguenther@suse.de>
2322 PR tree-optimization/84399
2323 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2324 For operands we can analyze at their definition make sure we can
2325 analyze them at each use as well.
2327 2018-02-16 Richard Biener <rguenther@suse.de>
2329 PR tree-optimization/84190
2330 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2331 volatile accesses if the decl isn't volatile.
2333 2018-02-15 Jason Merrill <jason@redhat.com>
2335 PR c++/84314 - ICE with templates and fastcall attribute.
2336 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2337 TYPE_CANONICAL on an existing type.
2339 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2341 PR tree-optimization/84383
2342 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2343 dstoff nor call operand_equal_p if dstbase is NULL.
2345 PR tree-optimization/84334
2346 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2347 also a CONSTANT_CLASS_P, punt.
2349 2018-02-14 Jim Wilson <jimw@sifive.com>
2351 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2352 first SMALL_OPERAND check. New local min_second_step. Move assert
2353 to where locals are set. Add TARGET_RVC support.
2354 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2356 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2358 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2360 2018-02-14 Martin Sebor <msebor@redhat.com>
2362 PR tree-optimization/83698
2363 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2364 arrays constrain the offset range to their bounds.
2365 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2366 (builtin_access::overlap): Avoid setting the size of overlap if it's
2368 (maybe_diag_overlap): Also consider arrays when deciding what values
2369 of offsets to include in diagnostics.
2371 2018-02-14 Martin Sebor <msebor@redhat.com>
2374 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2375 that correspond to the kind of a declaration.
2377 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2380 * config/pa/pa.md: Load address of PIC label using the linkage table
2381 if the label is nonlocal.
2383 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2385 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2386 warning message if user requests -maltivec=be.
2387 * doc/invoke.texi: Document deprecation of -maltivec=be.
2389 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2392 * config/rs6000/rs6000-c.c: Update definitions for
2393 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2394 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2396 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2399 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2400 add _get_ssp intrinsics. Remove argument from
2401 __builtin_ia32_rdssp[d|q].
2402 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2403 * config/i386/i386-builtin.def: Remove argument from
2404 __builtin_ia32_rdssp[d|q].
2405 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2406 ix86_expand_special_args_builtin for _rdssp[d|q].
2407 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2408 Clear register before usage.
2409 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2410 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2412 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2414 PR tree-optimization/84357
2415 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2416 operand 1 of an ARRAY_REF too.
2418 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2421 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2422 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2424 (set_of_reg): New struct.
2425 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2426 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2427 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2429 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2430 Split into bitclr, bitset, bitinvert patterns if appropriate.
2431 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2432 use rx_fuse_in_memory_bitop.
2433 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2434 to named insn, correct maximum insn length.
2436 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2439 * machmode.def: Define a complex mode for PARTIAL_INT.
2440 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2442 * doc/rtl.texi: Document CSPImode.
2443 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2445 (msp430_hard_regno_nregs_with_padding): Likewise.
2447 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2450 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2452 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2454 PR rtl-optimization/84169
2455 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2456 we generated a parallel as new i3 and we split that to new i2 and i3
2457 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2458 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2459 those to i2, not i1. Partially rewrite this scan code.
2461 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2464 * stor-layout.c (place_field): For variable length fields, adjust
2465 offset_align afterwards not just based on the field's alignment,
2466 but also on the size.
2469 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2470 of exps and logs in the use_exp2 case.
2472 2018-02-13 Jeff Law <law@redhat.com>
2474 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2477 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2480 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2484 * common.opt (gas-loc-support, gas-locview-support): New.
2485 (ginline-points, ginternal-reset-location-views): New.
2486 * doc/invoke.texi: Document them. Use @itemx where intended.
2487 (gvariable-location-views): Adjust.
2488 * target.def (reset_location_view): New.
2489 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2490 (TARGET_RESET_LOCATION_VIEW): New.
2491 * doc/tm.texi: Rebuilt.
2492 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2493 (dwarf2out_default_as_locview_support): New.
2494 (output_asm_line_debug_info): Use option variables.
2495 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2496 (output_loc_list): Likewise.
2497 (add_high_low_attributes): Check option variables.
2498 Don't output entry view attribute in strict mode.
2499 (gen_inlined_subroutine_die): Check option variables.
2500 (dwarf2out_inline_entry): Likewise.
2501 (init_sections_and_labels): Likewise.
2502 (dwarf2out_early_finish): Likewise.
2503 (maybe_reset_location_view): New, from...
2504 (dwarf2out_var_location): ... here. Call it.
2505 * debug.h (dwarf2out_default_as_loc_support): Declare.
2506 (dwarf2out_default_as_locview_support): Declare.
2507 * hooks.c (hook_int_rtx_insn_0): New.
2508 * hooks.h (hook_int_rtx_insn_0): Declare.
2509 * toplev.c (process_options): Take -gas-loc-support and
2510 -gas-locview-support from dwarf2out. Enable
2511 -gvariable-location-views by default only with locview
2512 assembler support. Enable -ginternal-reset-location-views by
2513 default only if the target defines the corresponding hook.
2514 Enable -ginline-points by default if location views are
2515 enabled; force it disabled if statement frontiers are
2517 * tree-inline.c (expand_call_inline): Check option variables.
2518 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2520 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2522 PR tree-optimization/84321
2523 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2524 handling. Also check whether the anti-range contains any values
2525 that satisfy the mask; switch to a VR_RANGE if not.
2527 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2530 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2532 2018-02-13 Martin Jambor <mjambor@suse.cz>
2535 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2536 of call statements, also set location of a load to a temporary.
2538 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2540 * config/rl78/rl78.c (add_vector_labels): New function.
2541 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2542 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2543 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2544 which checks that no arguments are passed.
2545 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2546 * doc/extend.texi: Documentation for the new attribute.
2548 2018-02-13 Andreas Schwab <schwab@suse.de>
2550 * config/riscv/linux.h (CPP_SPEC): Define.
2552 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2555 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2556 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2557 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2558 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2559 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2560 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2561 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2562 temporarily for AES and PCLMUL builtins.
2564 PR tree-optimization/84339
2565 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2566 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2570 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2571 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2572 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2574 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2576 * omp-simd-clone.h: New file.
2577 * omp-simd-clone.c: Include omp-simd-clone.h.
2578 (expand_simd_clones): No longer static.
2579 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2580 cgraph.h and omp-simd-clone.h.
2581 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2582 (vect_recog_widen_shift_pattern): Formatting fix.
2583 (vect_pattern_recog_1): Don't check optab for calls.
2586 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2587 operands[2] into a REG before using gen_lowpart on it.
2589 2018-02-12 Jeff Law <law@redhat.com>
2592 * config/sh/sh.c (find_barrier): Consider a sibling call
2595 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2596 successfully back substituting a reg.
2598 2018-02-12 Richard Biener <rguenther@suse.de>
2600 PR tree-optimization/84037
2601 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2602 parameter, move visited init to caller.
2603 (vect_slp_analyze_operations): Separate cost from validity
2604 check, initialize visited once for all instances.
2605 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2607 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2609 (vect_model_promotion_demotion_cost): Likewise.
2610 (vectorizable_bswap): Guard cost modeling with !slp_node
2611 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2613 (vectorizable_call): Likewise.
2614 (vectorizable_conversion): Likewise.
2615 (vectorizable_assignment): Likewise.
2616 (vectorizable_shift): Likewise.
2617 (vectorizable_operation): Likewise.
2618 (vectorizable_store): Likewise.
2619 (vectorizable_load): Likewise.
2620 (vectorizable_condition): Likewise.
2621 (vectorizable_comparison): Likewise.
2623 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2626 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2627 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2628 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2630 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2633 * params.def (inline-min-speedup): Increase from 8 to 15.
2634 (max-inline-insns-auto): Decrease from 40 to 30.
2635 * ipa-split.c (consider_split): Add some buffer for function to
2636 be considered inlining candidate.
2637 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2640 2018-02-12 Richard Biener <rguenther@suse.de>
2642 PR tree-optimization/84037
2643 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2644 matched stmts if we cannot swap the non-matched ones.
2646 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2648 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2649 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2650 _mm_maskz_scalef_round_ss): New intrinsics.
2651 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2652 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2653 __builtin_ia32_scalefss_round): Remove.
2654 (__builtin_ia32_scalefsd_mask_round,
2655 __builtin_ia32_scalefss_mask_round): New intrinsics.
2656 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2657 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2658 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2659 "<round_constraint>")): Changed to ...
2660 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2661 "<round_scalar_constraint>")): ... this.
2662 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2663 %0, %1, %2<round_op3>}"): Changed to ...
2664 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2665 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2666 %2<round_scalar_mask_op3>}"): ... this.
2667 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2669 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2671 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2672 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2673 (_mm_maskz_sqrt_round_ss): New intrinsics.
2674 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2675 (__builtin_ia32_sqrtsd_mask_round)
2676 (__builtin_ia32_sqrtss_mask_round): New builtins.
2677 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2678 (__builtin_ia32_sqrtss_round): Remove.
2679 (__builtin_ia32_sqrtsd_mask_round)
2680 (__builtin_ia32_sqrtss_mask_round): New builtins.
2681 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2682 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2683 ((match_operand:VF_128 1 "vector_operand"
2684 "xBm,<round_constraint>")): Changed to ...
2685 ((match_operand:VF_128 1 "vector_operand"
2686 "xBm,<round_scalar_constraint>")): ... this.
2687 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2688 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2689 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2690 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2691 %<iptr>1<round_scalar_mask_op3>}): ... this.
2692 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2693 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2695 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2698 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2699 Cast vec_cmpeq result to correct type.
2700 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2701 Cast vec_cmpgt result to correct type.
2703 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2705 * final.c (final_scan_insn_1): Renamed from...
2706 (final_scan_insn): ... this. New wrapper, to recover
2707 seen from the outermost call in recursive ones.
2708 * config/sparc/sparc.c (output_return): Drop seen from call.
2709 (output_sibcall): Likewise.
2710 * config/visium/visium.c (output_branch): Likewise.
2712 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2714 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2717 2018-02-10 Alan Modra <amodra@gmail.com>
2720 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2721 Specify LR as an input.
2723 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2726 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2727 remove_member_access_dummy_vars): New functions.
2728 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2729 lower_omp_1, execute_lower_omp): Use them.
2731 PR rtl-optimization/84308
2732 * shrink-wrap.c (spread_components): Release todo vector.
2734 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2736 PR rtl-optimization/57193
2737 * ira-color.c (struct allocno_color_data): Add member
2738 conflict_allocno_hard_prefs.
2739 (update_conflict_allocno_hard_prefs): New.
2740 (bucket_allocno_compare_func): Add a preference based on
2741 conflict_allocno_hard_prefs.
2742 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2743 (color_allocnos): Remove a dead code. Initiate
2744 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2746 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2749 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2750 constraint from =wa to wa. Avoid a subreg on the output operand,
2751 instead use a pseudo and subreg it in a move.
2752 (p9_xxbrd_<mode>): Changed to ...
2753 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2754 (p9_xxbrd_v2df): New expander.
2755 (p9_xxbrw_<mode>): Changed to ...
2756 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2757 (p9_xxbrw_v4sf): New expander.
2759 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2761 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2763 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2766 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2767 multiply in 32-bit mode.
2768 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2769 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2772 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2774 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2775 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2776 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2777 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2779 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2782 * dwarf2out.c (is_trivial_indirect_ref): New function.
2783 (dwarf2out_late_global_decl): Do not generate a location
2784 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2785 and that are not defined in the current unit.
2787 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2789 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2790 instead of a libcall for UNORDERED.
2792 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2795 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2796 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2798 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2801 * config/s390/s390.c (s390_set_current_function): Invoke
2802 s390_indirect_branch_settings also if fndecl didn't change.
2804 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2806 * config/rs6000/rs6000.md (blockage): Set length to zero.
2808 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2810 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2812 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2815 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2816 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2820 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2821 PARALLEL incoming that failed vt_get_decl_and_offset check.
2824 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2825 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2827 (get_variable_section): For decls in named .bss* sections pass true as
2828 second argument to bss_initializer_p.
2830 2018-02-09 Marek Polacek <polacek@redhat.com>
2831 Jakub Jelinek <jakub@redhat.com>
2834 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2835 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2836 Sync some changes from cxx_fold_indirect_ref.
2838 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2840 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2842 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2843 (BLOCK_INLINE_ENTRY_LABEL): New.
2844 (dwarf2out_var_location): Disregard inline entry markers.
2845 (inline_entry_data): New struct.
2846 (inline_entry_data_hasher): New hashtable type.
2847 (inline_entry_data_hasher::hash): New.
2848 (inline_entry_data_hasher::equal): New.
2849 (inline_entry_data_table): New variable.
2850 (add_high_low_attributes): Add DW_AT_entry_pc and
2851 DW_AT_GNU_entry_view attributes if a pending entry is found
2852 in inline_entry_data_table. Add old entry_pc attribute only
2853 if debug nonbinding markers are disabled.
2854 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2855 markers are enabled.
2856 (block_within_block_p, dwarf2out_inline_entry): New.
2857 (dwarf2out_finish): Check that no entries remained in
2858 inline_entry_data_table.
2859 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2860 (final_scan_insn, notice_source_line): Likewise.
2861 (rest_of_clean_state): Skip inline entry markers.
2862 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2864 * gimple.c (gimple_build_debug_inline_entry): New.
2865 * gimple.h (enum gimple_debug_subcode): Add
2866 GIMPLE_DEBUG_INLINE_ENTRY.
2867 (gimple_build_debug_inline_entry): Declare.
2868 (gimple_debug_inline_entry_p): New.
2869 (gimple_debug_nonbind_marker_p): Adjust.
2870 * insn-notes.def (INLINE_ENTRY): New.
2871 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2872 inline entry marker notes.
2873 (print_insn): Likewise.
2874 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2875 (INSN_DEBUG_MARKER_KIND): Likewise.
2876 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2877 * tree-inline.c (expand_call_inline): Build and insert
2878 debug_inline_entry stmt.
2879 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2880 inline entry blocks early, if nonbind markers are enabled.
2881 (dump_scope_block): Dump fragment info.
2882 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2883 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2884 (gimple_build_debug_inline_entry): New.
2885 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2886 Enable/disable inline entry points too.
2887 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2888 (DEBUG_INSN): Describe inline entry markers.
2890 * common.opt (gvariable-location-views): New.
2891 (gvariable-location-views=incompat5): New.
2892 * config.in: Rebuilt.
2893 * configure: Rebuilt.
2894 * configure.ac: Test assembler for view support.
2895 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2896 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2897 * dwarf2out.c (var_loc_view): New typedef.
2898 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2899 (dwarf2out_locviews_in_attribute): New.
2900 (dwarf2out_locviews_in_loclist): New.
2901 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2902 (enum dw_line_info_opcode): Add LI_adv_address.
2903 (struct dw_line_info_table): Add view.
2904 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2905 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2906 (zero_view_p): New variable.
2907 (ZERO_VIEW_P): New macro.
2908 (output_asm_line_debug_info): New.
2909 (struct var_loc_node): Add view.
2910 (add_AT_view_list, AT_loc_list): New.
2911 (add_var_loc_to_decl): Add view param. Test it against last.
2912 (new_loc_list): Add view params. Record them.
2913 (AT_loc_list_ptr): Handle loc and view lists.
2914 (view_list_to_loc_list_val_node): New.
2915 (print_dw_val): Handle dw_val_class_view_list.
2916 (size_of_die): Likewise.
2917 (value_format): Likewise.
2918 (loc_list_has_views): New.
2919 (gen_llsym): Set vl_symbol too.
2920 (maybe_gen_llsym, skip_loc_list_entry): New.
2921 (dwarf2out_maybe_output_loclist_view_pair): New.
2922 (output_loc_list): Output view list or entries too.
2923 (output_view_list_offset): New.
2924 (output_die): Handle dw_val_class_view_list.
2925 (output_dwarf_version): New.
2926 (output_compilation_unit_header): Use it.
2927 (output_skeleton_debug_sections): Likewise.
2928 (output_rnglists, output_line_info): Likewise.
2929 (output_pubnames, output_aranges): Update version comments.
2930 (output_one_line_info_table): Output view numbers in asm comments.
2931 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2932 Call maybe_gen_llsym.
2933 (loc_list_from_tree_1): Adjust.
2934 (add_AT_location_description): Create view list attribute if
2935 needed, check it's absent otherwise.
2936 (convert_cfa_to_fb_loc_list): Adjust.
2937 (maybe_emit_file): Call output_asm_line_debug_info for test.
2938 (dwarf2out_var_location): Reset views as needed. Precompute
2939 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2940 attribute. Set view.
2941 (new_line_info_table): Reset next view.
2942 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2943 (dwarf2out_source_line): Likewise. Output view resets and labels to
2944 the assembler, or select appropriate line info opcodes.
2945 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2946 (optimize_string_length): Catch it. Adjust.
2947 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2948 dw_val_class_view_list, and remove it if no longer needed.
2949 (hash_loc_list): Hash view numbers.
2950 (loc_list_hasher::equal): Compare them.
2951 (optimize_location_lists): Check whether a view list symbol is
2952 needed, and whether the locview attribute is present, and
2953 whether they match. Remove the locview attribute if no longer
2955 (index_location_lists): Call skip_loc_list_entry for test.
2956 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2957 Use output_dwarf_version.
2958 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2959 (struct dw_val_node): Add val_view_list.
2960 * final.c (SEEN_NEXT_VIEW): New.
2961 (set_next_view_needed): New.
2962 (clear_next_view_needed): New.
2963 (maybe_output_next_view): New.
2964 (final_start_function): Rename to...
2965 (final_start_function_1): ... this. Take pointer to FIRST,
2966 add SEEN parameter. Emit param bindings in the initial view.
2967 (final_start_function): Reintroduce SEEN-less interface.
2968 (final): Rename to...
2969 (final_1): ... this. Take SEEN parameter. Output final pending
2970 next view at the end.
2971 (final): Reintroduce seen-less interface.
2972 (final_scan_insn): Output pending next view before switching
2973 sections or ending a block. Mark the next view as needed when
2974 outputting variable locations. Notify debug backend of section
2975 changes, and of location view changes.
2976 (rest_of_handle_final): Adjust.
2977 * toplev.c (process_options): Autodetect value for debug variable
2978 location views option. Warn on incompat5 without -gdwarf-5.
2979 * doc/invoke.texi (gvariable-location-views): New.
2980 (gvariable-location-views=incompat5): New.
2981 (gno-variable-location-views): New.
2983 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2985 PR tree-optimization/84136
2986 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2987 that the result of find_edge is non-NULL.
2989 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2992 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2993 storing integer register in SImode. Fix cost of 256 and 512
2994 byte aligned SSE register store.
2996 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2998 * config/i386/i386.c (ix86_multiplication_cost): Fix
2999 multiplication cost for TARGET_AVX512DQ.
3001 2018-02-08 Marek Polacek <polacek@redhat.com>
3003 PR tree-optimization/84238
3004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3007 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3009 PR tree-optimization/84265
3010 * tree-vect-stmts.c (vectorizable_store): Don't treat
3011 VMAT_CONTIGUOUS accesses as grouped.
3012 (vectorizable_load): Likewise.
3014 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3016 PR tree-optimization/81635
3017 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3018 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3019 (test_round_for_mask): New functions.
3020 (wide_int_cc_tests): Call test_round_for_mask.
3021 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3022 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3023 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3024 range returned by get_range_info.
3026 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3029 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3030 * symtab.c: Include builtins.h
3031 (symtab_node::output_to_lto_symbol_table_p): Move here
3032 from lto-streamer-out.c:output_symbol_p.
3033 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3034 (output_symbol_p): Move all logic to symtab.c
3035 (produce_symtab): Update.
3037 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3039 * config/s390/s390-opts.h (enum indirect_branch): Define.
3040 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3041 (s390_indirect_branch_via_thunk)
3042 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3043 (enum s390_indirect_branch_type): Define.
3044 * config/s390/s390.c (struct s390_frame_layout, struct
3045 machine_function): Remove.
3046 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3047 (indirect_branch_table_label_no, indirect_branch_table_name):
3049 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3050 (enum s390_indirect_branch_option): Define.
3051 (s390_return_addr_from_memory): New function.
3052 (s390_handle_string_attribute): New function.
3053 (s390_attribute_table): Add new attribute handler.
3054 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3055 (s390_indirect_branch_via_thunk): New function.
3056 (s390_indirect_branch_via_inline_thunk): New function.
3057 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3058 sibling call optimization for non z10 compiles.
3059 (s390_emit_call): Force indirect branch target to be a single
3060 register. Add r1 clobber for non-z10 compiles.
3061 (s390_emit_epilogue): Emit return jump via return_use expander.
3062 (s390_reorg): Handle JUMP_INSNs as execute targets.
3063 (s390_option_override_internal): Perform validity checks for the
3064 new command line options.
3065 (s390_indirect_branch_attrvalue): New function.
3066 (s390_indirect_branch_settings): New function.
3067 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3068 (s390_output_indirect_thunk_function): New function.
3069 (s390_code_end): Implement target hook.
3070 (s390_case_values_threshold): Implement target hook.
3071 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3073 * config/s390/s390.h (struct s390_frame_layout)
3074 (struct machine_function): Move here from s390.c.
3075 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3076 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3077 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3078 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3079 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3080 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3081 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3082 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3083 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3084 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3085 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3086 (mnemonic attribute): Add values which aren't recognized
3088 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3089 pattern for branch conversion. Fix mnemonic attribute.
3090 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3091 indirect branch via thunk if requested.
3092 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3093 ("*indirect_jump"): Disable for branch conversion using out of
3095 ("indirect_jump_via_thunk<mode>_z10")
3096 ("indirect_jump_via_thunk<mode>")
3097 ("indirect_jump_via_inlinethunk<mode>_z10")
3098 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3099 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3100 ("casesi_jump_via_inlinethunk<mode>_z10")
3101 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3102 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3103 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3104 ("*indirect2_jump"): Disable for branch conversion.
3105 ("casesi_jump"): Turn into expander and expand patterns for branch
3107 ("return_use"): New expander.
3108 ("*return"): Emit return via thunk and rename it to ...
3109 ("*return<mode>"): ... this one.
3110 * config/s390/s390.opt: Add new options and and enum for the
3113 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3115 * lra-constraints.c (match_reload): Unconditionally use
3116 gen_lowpart_SUBREG, rather than selecting between that
3117 and equivalent gen_rtx_SUBREG code.
3119 2018-02-08 Richard Biener <rguenther@suse.de>
3121 PR tree-optimization/84233
3122 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3123 changed flag instead of boguously re-using phi_inserted.
3125 2018-02-08 Martin Jambor <mjambor@suse.cz>
3127 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3128 static local variables.
3130 2018-02-08 Richard Biener <rguenther@suse.de>
3132 PR tree-optimization/84278
3133 * tree-vect-stmts.c (vectorizable_store): When looking for
3134 smaller vector types to perform grouped strided loads/stores
3135 make sure the mode is supported by the target.
3136 (vectorizable_load): Likewise.
3138 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3140 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3141 Increase LDP/STP opportunities by adding adjacent callee-saves.
3143 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3145 PR rtl-optimization/84068
3146 PR rtl-optimization/83459
3147 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3149 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3151 PR tree-optimization/84224
3152 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3153 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3156 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3159 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3160 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3161 count, remove one USE.
3163 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3165 * doc/install.texi (Configuration): Document the
3166 --with-long-double-format={ibm,ieee} PowerPC configuration
3170 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3171 Convert from define_expand to be define_insn_and_split. Rework
3172 float/double/_Float128 conversions to QI/HI/SImode to work with
3173 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3174 conversions to QI/HImode types did a store and then a load to
3175 truncate the value. For conversions to VSX registers, don't split
3176 the insn, instead emit the code directly. Use the code iterator
3177 any_fix to combine signed and unsigned conversions.
3178 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3179 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3180 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3181 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3182 (fix_<mode>di2_hw): Likewise.
3183 (fixuns_<mode>di2_hw): Likewise.
3184 (fix_<mode>si2_hw): Likewise.
3185 (fixuns_<mode>si2_hw): Likewise.
3186 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3187 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3188 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3189 fix<uns>_trunc<SFDF:mode>si2_p8.
3190 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3192 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3193 (fix<uns>_<mode>_mem): Likewise.
3194 (fctiw<u>z_<mode>_mem): Likewise.
3195 (fix<uns>_<mode>_mem): Likewise.
3196 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3197 the register allocator from doing a direct move to the GPRs to do
3198 a store, and instead use the ISA 3.0 store byte/half-word from
3199 vector register instruction. For IEEE 128-bit floating point,
3200 also optimize stores of 32-bit ints.
3201 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3203 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3205 * genextract.c (push_pathstr_operand): New function to support
3207 (walk_rtx): Call push_pathstr_operand.
3208 (print_path): Support [a-zA-Z].
3210 2018-02-07 Richard Biener <rguenther@suse.de>
3212 PR tree-optimization/84037
3213 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3214 (cse_and_gimplify_to_preheader): Declare.
3215 (vect_get_place_in_interleaving_chain): Likewise.
3216 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3218 (_loop_vec_info::~_loop_vec_info): Delete it.
3219 (cse_and_gimplify_to_preheader): New function.
3220 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3221 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3222 (vectorizable_load): Likewise. For grouped stores always base
3223 the IV on the first element.
3224 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3225 condition before gimplifying.
3227 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3229 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3230 *DIV_EXPR and *MOD_EXPR.
3232 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3235 * config/i386/i386.c (ix86_option_override_internal): Mask out
3236 the CF_SET bit when checking -fcf-protection.
3238 2018-02-07 Tom de Vries <tom@codesourcery.com>
3241 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3244 2018-02-07 Richard Biener <rguenther@suse.de>
3246 PR tree-optimization/84204
3247 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3250 PR tree-optimization/84205
3251 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3252 special-case isl_ast_op_zdiv_r.
3254 PR tree-optimization/84223
3255 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3256 Only add conditions from within the region.
3257 (gather_bbs::after_dom_children): Adjust.
3259 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3262 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3263 * config/avr/avr.md: Only post-reload split REG-REG moves if
3264 either register is GENERAL_REG_P.
3266 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3268 PR tree-optimization/84235
3269 * tree-ssa-scopedtables.c
3270 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3271 if the subtraction is performed in floating point type where NaNs are
3272 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3273 build 1. Formatting fix.
3275 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3278 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3279 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3280 and skip it regardless of bb boundaries. Use CALL_P macro,
3281 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3282 unnecessarily, formatting fix.
3284 2018-02-06 Michael Collison <michael.collison@arm.com>
3286 * config/arm/thumb2.md:
3287 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3288 (*thumb_mov_notscc): Ditto.
3290 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3293 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3294 unsigned_fix, not "s".
3296 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3298 * configure.ac (gcc_fn_eh_frame_ro): New function.
3299 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3300 correct .eh_frame permissions.
3301 * configure: Regenerate.
3303 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3305 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3308 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3310 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3311 Display warning message for -mno-speculate-indirect-jumps.
3313 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3315 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3317 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3319 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3321 PR tree-optimization/84225
3322 * tree-eh.c (find_trapping_overflow): Only call
3323 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3325 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3328 * config/i386/i386.c: Reimplement the check of possible options
3329 -mibt/-mshstk conbination. Change error messages.
3330 * doc/invoke.texi: Fix a typo: remove extra '='.
3332 2018-02-06 Marek Polacek <polacek@redhat.com>
3334 PR tree-optimization/84228
3335 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3337 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3340 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3341 emitted arch directives.
3342 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3343 __ARM_FEATURE_COPROC before changing architectures.
3345 2018-02-06 Richard Biener <rguenther@suse.de>
3347 * config/i386/i386.c (print_reg): Fix typo.
3348 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3350 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3352 * configure: Regenerate.
3354 2018-02-05 Martin Sebor <msebor@redhat.com>
3356 PR tree-optimization/83369
3357 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3360 2018-02-05 Martin Liska <mliska@suse.cz>
3362 * doc/invoke.texi: Cherry-pick upstream r323995.
3364 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3366 * ira.c (ira_init_register_move_cost): Adjust comment.
3368 2018-02-05 Martin Liska <mliska@suse.cz>
3370 PR gcov-profile/84137
3371 * doc/gcov.texi: Fix typo in documentation.
3373 2018-02-05 Martin Liska <mliska@suse.cz>
3375 PR gcov-profile/83879
3376 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3377 using dlopen functionality.
3379 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3381 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3382 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3383 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3384 _mm_maskz_range_round_ss): New intrinsics.
3385 (__builtin_ia32_rangesd128_round)
3386 (__builtin_ia32_rangess128_round): Remove.
3387 (__builtin_ia32_rangesd128_mask_round,
3388 __builtin_ia32_rangess128_mask_round): New builtins.
3389 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3390 __builtin_ia32_rangess128_round): Remove.
3391 (__builtin_ia32_rangesd128_mask_round,
3392 __builtin_ia32_rangess128_mask_round): New builtins.
3393 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3394 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3395 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3396 "<round_saeonly_constraint>")): Changed to ...
3397 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3398 "<round_saeonly_scalar_constraint>")): ... this.
3399 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3400 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3401 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3402 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3403 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3405 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3407 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3409 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3410 Remove all values except native, 8540 and 8548.
3412 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3414 * config/i386/i386.c (ix86_output_function_return): Pass
3415 INVALID_REGNUM, instead of -1, as invalid register number to
3416 indirect_thunk_name and output_indirect_thunk.
3418 2018-02-02 Julia Koval <julia.koval@intel.com>
3420 * config.gcc: Add -march=icelake.
3421 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3422 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3423 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3424 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3425 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3426 (processor_target_table): Add icelake.
3427 (ix86_option_override_internal): Handle new PTAs.
3428 (get_builtin_code_for_version): Handle icelake.
3429 (M_INTEL_COREI7_ICELAKE): New.
3430 (fold_builtin_cpu): Handle icelake.
3431 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3432 * doc/invoke.texi: Add -march=icelake.
3434 2018-02-02 Julia Koval <julia.koval@intel.com>
3436 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3437 to wide_int_bitmask.
3438 * wide-int-bitmask.h: New.
3440 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3443 * config/i386/i386.md: Replace Pmode with word_mode in
3444 builtin_setjmp_setup and builtin_longjmp to support x32.
3446 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3450 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3452 (rs6000_supported_cpu_names): New static variable.
3453 (linux_cpu_translation_table): Likewise.
3454 (elf_platform) <cpu>: Define new static variable and use it.
3455 Translate kernel AT_PLATFORM name to canonical name if needed.
3456 Error if platform name is unknown.
3458 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3461 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3463 2018-02-01 Jeff Law <law@redhat.com>
3466 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3467 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3468 the scratch if RELEASE_VIA_POP is false.
3469 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3470 If we have to save a temporary register, decrement SIZE appropriately.
3471 Pass new arguments to release_scratch_register_on_entry.
3472 (ix86_adjust_stack_and_probe): Likewise.
3473 (ix86_emit_probe_stack_range): Pass new arguments to
3474 release_scratch_register_on_entry.
3476 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3478 PR rtl-optimization/84157
3479 * combine.c (change_zero_ext): Use REG_P predicate in
3480 front of HARD_REGISTER_P predicate.
3482 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3484 * config/avr/avr.c (avr_option_override): Move disabling of
3485 -fdelete-null-pointer-checks to...
3486 * common/config/avr/avr-common.c (avr_option_optimization_table):
3489 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3491 PR tree-optimization/81635
3492 * tree-data-ref.c (split_constant_offset_1): For types that
3493 wrap on overflow, try to use range info to prove that wrapping
3496 2018-02-01 Renlin Li <renlin.li@arm.com>
3499 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3501 (aarch64_register_move_cost): Likewise.
3502 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3504 (REG_CLASS_NAMES): Likewise.
3505 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3506 TAILCALL_ADDR_REGS. Remove IP registers.
3507 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3509 2018-02-01 Richard Biener <rguenther@suse.de>
3511 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3512 for specifying RPO order and allow NULL for that.
3513 * domwalk.c (dom_walker::dom_walker): Likewise.
3514 (dom_walker::walk): Handle NULL RPO order.
3515 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3517 (rewrite_update_dom_walker): Likewise.
3518 (mark_def_dom_walker): Likewise.
3520 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3522 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3523 (aarch64_maybe_expand_sve_subreg_move): Declare.
3524 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3525 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3527 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3528 that are semantically a reverse operation.
3529 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3530 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3531 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3533 (aarch64_can_change_mode_class): For big-endian, forbid changes
3534 between two SVE modes if they have different element sizes.
3536 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3538 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3539 the TImode handling for big-endian targets.
3541 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3543 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3544 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3546 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3547 Remove BSWAP handing for big-endian targets and use the form of
3548 LD1RQ appropariate for the mode.
3550 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3552 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3553 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3556 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3559 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3560 check for operands that need to go through aarch64_sve_reload_be.
3562 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3564 PR tree-optimization/81661
3565 PR tree-optimization/84117
3566 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3567 * tree-eh.c: Include gimplify.h.
3568 (find_trapping_overflow, replace_trapping_overflow,
3569 rewrite_to_non_trapping_overflow): New functions.
3570 * tree-vect-loop.c: Include tree-eh.h.
3571 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3572 * tree-data-ref.c: Include tree-eh.h.
3573 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3575 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3577 PR rtl-optimization/84123
3578 * combine.c (change_zero_ext): Check if hard register satisfies
3579 can_change_dest_mode before calling gen_lowpart_SUBREG.
3581 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3584 * ira.c (ira_init_register_move_cost): Remove assert.
3586 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3588 PR rtl-optimization/84071
3589 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3590 * doc/tm.texi: Regenerate.
3592 2018-01-31 Richard Biener <rguenther@suse.de>
3594 PR tree-optimization/84132
3595 * tree-data-ref.c (analyze_miv_subscript): Properly
3596 check whether evolution_function_is_affine_multivariate_p
3597 before calling gcd_of_steps_may_divide_p.
3599 2018-01-31 Julia Koval <julia.koval@intel.com>
3602 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3603 * config/i386/i386.md (rdpid_rex64) New.
3604 (rdpid): Make 32bit only.
3606 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3609 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3610 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3612 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3615 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3617 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3619 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3621 PR rtl-optimization/84071
3622 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3623 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3625 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3627 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3628 (arc_attribute_table): Add 'aux' attribute.
3629 (arc_in_small_data_p): Consider aux like variables.
3630 (arc_is_aux_reg_p): New function.
3631 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3632 (arc_get_aux_arg): New function.
3633 (prepare_move_operands): Handle aux-register access.
3634 (arc_handle_aux_attribute): New function.
3635 * doc/extend.texi (ARC Variable attributes): Add subsection.
3637 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3639 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3640 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3641 (arc_attribute_table): Add 'uncached' attribute.
3642 (arc_print_operand): Print '.di' flag for uncached memory
3644 (arc_in_small_data_p): Do not consider for small data the uncached
3646 (arc_is_uncached_mem_p): New function.
3647 * config/arc/predicates.md (compact_store_memory_operand): Check
3648 for uncached memory accesses.
3649 (nonvol_nonimm_operand): Likewise.
3650 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3652 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3655 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3656 falign-loops=): Add Optimization flag.
3658 2018-01-30 Jeff Law <law@redhat.com>
3661 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3662 INT_REGISTERS_SAVED. Check it prior to calling
3663 get_scratch_register_on_entry.
3664 (ix86_adjust_stack_and_probe): Similarly.
3665 (ix86_emit_probe_stack_range): Similarly.
3666 (ix86_expand_prologue): Corresponding changes.
3668 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3671 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3672 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3674 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3677 * lra-constraints.c (curr_insn_transform): Process AND in the
3680 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3682 PR rtl-optimization/83986
3683 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3684 dependence against last_pending_memory_flush in addition to
3687 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3689 PR tree-optimization/81611
3690 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3693 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3696 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3699 2018-01-30 Richard Biener <rguenther@suse.de>
3700 Jakub Jelinek <jakub@redhat.com>
3702 PR tree-optimization/84111
3703 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3704 inner loops added during recursion, as they don't have up-to-date
3707 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3710 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3711 (can_inline_edge_by_limits_p): ... here.
3712 (can_early_inline_edge_p, check_callers,
3713 update_caller_keys, update_callee_keys, recursive_inlining,
3714 add_new_edges_to_heap, speculation_useful_p,
3715 inline_small_functions,
3716 inline_small_functions, flatten_function,
3717 inline_to_all_callers_1): Update.
3719 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3721 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3724 2018-01-30 Richard Biener <rguenther@suse.de>
3726 PR tree-optimization/83008
3727 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3728 invariant and constant vector uses in stmts when they need
3731 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3734 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3735 * configure: Regenerate.
3737 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3739 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3741 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3742 Use gen_rtx_REG rather than gen_lowpart.
3744 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3746 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3747 rather than 0 when creating partial subregs.
3749 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3751 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3754 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3757 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3758 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3759 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3760 flags. This restores the settings used before the 2017-07-24.
3761 Turning off pre increment/decrement/modify allows IVOPTS to
3762 optimize DF/SF loops where the index is an int.
3764 2018-01-29 Richard Biener <rguenther@suse.de>
3765 Kelvin Nilsen <kelvin@gcc.gnu.org>
3768 * tree-vect-stmts.c (vectorizable_call): Don't call
3769 targetm.vectorize_builtin_md_vectorized_function if callee is
3772 2018-01-22 Carl Love <cel@us.ibm.com>
3774 * doc/extend.tex: Fix typo in second arg in
3775 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3777 2018-01-29 Richard Biener <rguenther@suse.de>
3779 PR tree-optimization/84086
3780 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3781 (flush_ssaname_freelist): When SSA names were released reset
3782 the SCEV hash table.
3784 2018-01-29 Richard Biener <rguenther@suse.de>
3786 PR tree-optimization/84057
3787 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3788 removed paths when removing edges.
3790 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3792 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3793 -mfunction-return=@var{choice}.
3795 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3798 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3800 (layout::print_source_line): Likewise.
3801 (test_get_line_width_without_trailing_whitespace): Add test cases.
3803 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3806 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3809 2018-01-26 Jim Wilson <jimw@sifive.com>
3811 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3813 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3816 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3818 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3819 and CMP + SUB-immediate -> SUBS.
3821 2018-01-26 Martin Sebor <msebor@redhat.com>
3823 PR tree-optimization/83896
3824 * tree-ssa-strlen.c (get_string_len): Rename...
3825 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3826 Avoid assuming length is constant.
3827 (handle_char_store): Use HOST_WIDE_INT for string length.
3829 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3832 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3833 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3835 2018-01-26 Richard Biener <rguenther@suse.de>
3837 PR rtl-optimization/84003
3838 * dse.c (record_store): Only record redundant stores when
3839 the earlier store aliases at least all accesses the later one does.
3841 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3843 PR rtl-optimization/83985
3844 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3845 REG_CFA_RESTORE insns.
3846 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3847 REG_CFA_RESTORE insns here.
3850 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3851 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3853 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3855 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3856 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3857 (arc_init): Likewise.
3858 (arc_override_options): Likewise.
3859 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3861 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3863 * config/arc/arc.h (TARGET_DBNZ): Define.
3864 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3865 properly set the tune attribute.
3866 (dbnz): Use TARGET_DBNZ guard.
3867 * config/arc/arc.opt (mtune): Add core3 option.
3869 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3871 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3872 recognize new pic like addresses.
3873 (arc_delegitimize_address): Clean up.
3875 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3877 * config/arc/arc-arches.def: Option mrf16 valid for all
3879 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3880 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3881 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3882 * config/arc/arc-tables.opt: Regenerate.
3883 * config/arc/arc.c (arc_conditional_register_usage): Handle
3884 reduced register file case.
3885 (arc_file_start): Set must have build attributes.
3886 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3888 * config/arc/arc.opt (mrf16): Add new option.
3889 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3890 * config/arc/genmultilib.awk: Handle new mrf16 option.
3891 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3892 * config/arc/t-multilib: Regenerate.
3893 * doc/invoke.texi (ARC Options): Document mrf16 option.
3895 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3897 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3898 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3899 (arc_attribute_table): Add 'secure_call' attribute.
3900 (arc_print_operand): Print secure call operand.
3901 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3903 (arc_is_secure_call_p): New function. * config/arc/arc.md
3904 (call_i): Add support for sjli instruction.
3905 (call_value_i): Likewise.
3906 * config/arc/constraints.md (Csc): New constraint.
3908 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3909 John Eric Martin <John.Martin@emmicro-us.com>
3911 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3912 * config/arc/arc.c (_arc_jli_section): New struct.
3913 (arc_jli_section): New type.
3914 (rc_jli_sections): New static variable.
3915 (arc_handle_jli_attribute): New function.
3916 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3917 (arc_file_end): New function.
3918 (TARGET_ASM_FILE_END): Define.
3919 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3920 (arc_add_jli_section): New function.
3921 (jli_call_scan): Likewise.
3922 (arc_reorg): Call jli_call_scan.
3923 (arc_output_addsi): Remove 'S' from printing asm operand.
3924 (arc_is_jli_call_p): New function.
3925 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3927 (movhi_insn): Likewise.
3928 (movsi_insn): Likewise.
3929 (movsi_set_cc_insn): Likewise.
3930 (loadqi_update): Likewise.
3931 (load_zeroextendqisi_update): Likewise.
3932 (load_signextendqisi_update): Likewise.
3933 (loadhi_update): Likewise.
3934 (load_zeroextendhisi_update): Likewise.
3935 (load_signextendhisi_update): Likewise.
3936 (loadsi_update): Likewise.
3937 (loadsf_update): Likewise.
3938 (movsicc_insn): Likewise.
3939 (bset_insn): Likewise.
3940 (bxor_insn): Likewise.
3941 (bclr_insn): Likewise.
3942 (bmsk_insn): Likewise.
3943 (bicsi3_insn): Likewise.
3944 (cmpsi_cc_c_insn): Likewise.
3945 (movsi_ne): Likewise.
3946 (movsi_cond_exec): Likewise.
3947 (clrsbsi2): Likewise.
3958 (call_i): Remove 'S' asm letter, add jli instruction.
3959 (call_value_i): Likewise.
3960 * config/arc/arc.op (mjli-always): New option.
3961 * config/arc/constraints.md (Cji): New constraint.
3962 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3964 (subsf3_fpx): Likewise.
3965 (mulsf3_fpx): Likewise.
3966 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3968 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3970 * doc/invoke.texi (ARC): Document mjli-always option.
3972 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3974 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3975 avoid addition with 0 and use incw and decw where possible.
3977 2018-01-26 Richard Biener <rguenther@suse.de>
3979 PR tree-optimization/81082
3980 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3981 association if it requires casting to unsigned.
3982 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3983 from fold_plusminus_mult_expr to catch important cases late when
3984 range info is available.
3986 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3988 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3989 * configure.ac (hidden_linkonce): New test.
3990 * configure: Regenerate.
3991 * config.in: Regenerate.
3993 2018-01-26 Julia Koval <julia.koval@intel.com>
3995 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3996 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3997 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3998 _mm_mask_bitshuffle_epi64_mask): Fix type.
3999 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4000 USI_FTYPE_V4DI_V4DI_USI): Remove.
4001 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4002 __builtin_ia32_vpshufbitqmb256_mask,
4003 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4004 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4005 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4007 2018-01-26 Alan Modra <amodra@gmail.com>
4010 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4011 UNSPEC_VBPERMQ. Sort other unspecs.
4013 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4015 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4017 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4020 * predict.c (drop_profile): Do not push/pop cfun; update also
4022 (handle_missing_profiles): Fix logic looking for zero profiles.
4024 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4027 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4028 on functions with #pragma omp declare simd or functions with simd
4030 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4031 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4032 Remove trailing \n from warning_at calls.
4034 2018-01-25 Tom de Vries <tom@codesourcery.com>
4037 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4038 for neutered workers.
4040 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4043 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4044 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4046 2018-01-24 Jeff Law <law@redhat.com>
4049 * i386.c (get_probe_interval): Move to earlier point.
4050 (ix86_compute_frame_layout): If -fstack-clash-protection and
4051 the frame is larger than the probe interval, then use pushes
4052 to save registers rather than reg->mem moves.
4053 (ix86_expand_prologue): Remove conditional for int_registers_saved
4056 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4059 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4060 min/max for never referenced object.
4062 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4065 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4067 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4068 attributes from DECL_ATTRIBUTES (decl) without affecting
4069 DECL_ATTRIBUTES (current_function_decl).
4070 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4071 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4073 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4075 PR tree-optimization/83979
4076 * fold-const.c (fold_comparison): Use constant_boolean_node
4077 instead of boolean_{true,false}_node.
4079 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4081 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4084 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4086 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4087 Simplify the clause that sets the length attribute.
4088 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4089 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4090 clause that sets the length attribute.
4091 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4093 2018-01-24 Tom de Vries <tom@codesourcery.com>
4096 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4097 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4098 Add strict parameter.
4099 (prevent_branch_around_nothing): Insert dummy insn between branch to
4100 label and label with no ptx insn inbetween.
4101 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4103 2018-01-24 Tom de Vries <tom@codesourcery.com>
4106 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4107 for neutered threads in warp.
4108 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4110 2018-01-24 Richard Biener <rguenther@suse.de>
4112 PR tree-optimization/83176
4113 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4116 2018-01-24 Richard Biener <rguenther@suse.de>
4118 PR tree-optimization/82819
4119 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4120 code generating pluses that are no-ops in the target precision.
4122 2018-01-24 Richard Biener <rguenther@suse.de>
4125 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4127 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4129 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4130 to merge probabilities.
4131 * predict.c (probably_never_executed): Also mark as cold functions
4132 with global 0 profile and guessed local profile.
4133 * profile-count.c (profile_probability::combine_with_count): New
4135 * profile-count.h (profile_probability::operator*,
4136 profile_probability::operator*=, profile_probability::operator/,
4137 profile_probability::operator/=): Reduce precision to adjusted
4138 and set value to guessed on contradictory divisions.
4139 (profile_probability::combine_with_freq): Remove.
4140 (profile_probability::combine_wiht_count): Declare.
4141 (profile_count::force_nonzero):: Set to adjusted.
4142 (profile_count::probability_in):: Set quality to adjusted.
4143 * tree-ssa-tail-merge.c (replace_block_by): Use
4146 2018-01-23 Andrew Waterman <andrew@sifive.com>
4147 Jim Wilson <jimw@sifive.com>
4149 * config/riscv/riscv.c (riscv_stack_boundary): New.
4150 (riscv_option_override): Set riscv_stack_boundary. Handle
4151 riscv_preferred_stack_boundary_arg.
4152 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4153 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4154 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4155 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4156 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4157 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4159 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4162 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4163 of struct ix86_frame.
4164 (ix86_expand_epilogue): Likewise. Add a local variable for
4165 the reg_save_offset field in struct ix86_frame.
4167 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4169 PR tree-optimization/82604
4170 * tree-loop-distribution.c (enum partition_kind): New enum item
4171 PKIND_PARTIAL_MEMSET.
4172 (partition_builtin_p): Support above new enum item.
4173 (generate_code_for_partition): Ditto.
4174 (compute_access_range): Differentiate cases that equality can be
4175 proven at all loops, the innermost loops or no loops.
4176 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4177 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4178 (finalize_partitions, distribute_loop): Don't fuse partition of
4179 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4180 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4183 2018-01-23 Martin Liska <mliska@suse.cz>
4185 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4186 order to ignore the predictor.
4187 (PRED_POLYMORPHIC_CALL): Likewise.
4188 (PRED_RECURSIVE_CALL): Likewise.
4190 2018-01-23 Martin Liska <mliska@suse.cz>
4192 * tree-profile.c (tree_profiling): Print function header to
4193 aware reader which function we are working on.
4194 * value-prof.c (gimple_find_values_to_profile): Do not print
4195 not interesting value histograms.
4197 2018-01-23 Martin Liska <mliska@suse.cz>
4199 * profile-count.h (enum profile_quality): Add
4200 profile_uninitialized as the first value. Do not number values
4201 as they are zero based.
4202 (profile_count::verify): Update sanity check.
4203 (profile_probability::verify): Likewise.
4205 2018-01-23 Nathan Sidwell <nathan@acm.org>
4207 * doc/invoke.texi (ffor-scope): Deprecate.
4209 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4211 PR tree-optimization/83510
4212 * domwalk.c (set_all_edges_as_executable): New function.
4213 (dom_walker::dom_walker): Convert bool param
4214 "skip_unreachable_blocks" to enum reachability. Move setup of
4215 edge flags to set_all_edges_as_executable and only do it when
4216 reachability is REACHABLE_BLOCKS.
4217 * domwalk.h (enum dom_walker::reachability): New enum.
4218 (dom_walker::dom_walker): Convert bool param
4219 "skip_unreachable_blocks" to enum reachability.
4220 (set_all_edges_as_executable): New decl.
4221 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4222 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4224 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4225 but converting true to REACHABLE_BLOCKS.
4226 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4228 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4229 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4230 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4232 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4233 if check_all_array_refs will be called.
4235 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4237 * tree.c (selftest::test_location_wrappers): Add more test
4240 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4242 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4243 (selftest::test_bit_in_range): Likewise.
4245 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4248 * doc/sourcebuild.texi (vect_float): Say that the selector
4249 only describes the situation when -funsafe-math-optimizations is on.
4250 (vect_float_strict): Document.
4252 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4254 PR tree-optimization/83965
4255 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4256 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4257 instead of checking only for a reduction.
4258 (vect_recog_widen_sum_pattern): Likewise.
4260 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4262 * predict.c (probably_never_executed): Only use precise profile info.
4263 (compute_function_frequency): Skip after inlining hack since we now
4264 have quality checking.
4266 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4268 * profile-count.h (profile_probability::very_unlikely,
4269 profile_probability::unlikely, profile_probability::even): Set
4270 precision to guessed.
4272 2018-01-23 Richard Biener <rguenther@suse.de>
4274 PR tree-optimization/83963
4275 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4276 Properly terminate dominator walk when crossing the exit edge not
4277 when visiting its source block.
4279 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4282 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4283 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4285 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4287 PR tree-optimization/83957
4288 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4289 semicolon after for body surrounded by braces.
4291 PR tree-optimization/83081
4292 * profile-count.h (profile_probability::split): New method.
4293 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4294 Use profile_probability::split.
4295 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4296 when splitting a single conditional jump into 2.
4298 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4300 PR tree-optimization/69452
4301 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4304 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4306 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4307 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4308 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4310 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4312 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4313 * config/rl78/rl78.md (movdi): New define_expand.
4314 * config/rl78/rl78.c (rl78_split_movdi): New function.
4316 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4319 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4321 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4322 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4323 128-bit to produce an UNSPEC move to get the double word with the
4324 signbit and then a shift directly to do signbit.
4325 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4326 implementation with a new version that just does either a direct
4327 move or a regular move. Move memory interface to separate insns.
4328 Move insns so they are next to the expander.
4329 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4330 with signbit move. Split big and little endian case.
4331 (signbit<mode>2_dm_mem_le): Likewise.
4332 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4333 (signbit<mode>2_dm2): Likewise.
4335 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4337 * config/rl78/rl78.md (anddi3): New define_expand.
4339 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4341 * config/rl78/rl78.md (umindi3): New define_expand.
4343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4345 * config/rl78/rl78.md (smindi3): New define_expand.
4347 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4349 * config/rl78/rl78.md (smaxdi3): New define_expand.
4351 2018-01-22 Carl Love <cel@us.ibm.com>
4353 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4354 LVX_V1TI): Add macro expansion.
4355 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4356 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4357 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4358 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4359 Change check to determine if the instruction is a byte reversing
4360 entry. Fix typo in comment.
4361 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4362 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4363 Add def_builtin calls for new builtins.
4364 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4365 Add define_insn expansion.
4367 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4369 * config/rl78/rl78.md (umaxdi3): New define_expand.
4371 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4373 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4374 for non-QImode registers.
4376 2018-01-22 Richard Biener <rguenther@suse.de>
4378 PR tree-optimization/83963
4379 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4380 including the loop exit block.
4381 (scop_detection::merge_sese): Likewise.
4382 (scop_detection::add_scop): Do it here instead.
4384 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4386 * doc/sourcebuild.texi (arm_softfloat): Document.
4388 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4391 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4392 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4393 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4395 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4396 David Edelsohn <dje.gcc@gmail.com>
4399 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4400 Change "crset eq" to "crset 2".
4401 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4402 (*call_indirect_aix<mode>_nospec): Likewise.
4403 (*call_value_indirect_aix<mode>_nospec): Likewise.
4404 (*call_indirect_elfv2<mode>_nospec): Likewise.
4405 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4406 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4407 change assembly output from . to $.
4408 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4409 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4410 (*tablejump<mode>_internal1_nospec): Likewise.
4412 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4415 * config/sh/sh_optimize_sett_clrt.cc:
4416 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4418 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4420 PR tree-optimization/83940
4421 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4422 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4423 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4424 use it to pass back the definition type.
4425 (vect_check_store_rhs): Likewise rhs_dt_out.
4426 (vect_build_gather_load_calls): Add a mask_dt argument and use
4427 it instead of a call to vect_is_simple_use.
4428 (vectorizable_store): Update calls to vect_check_load_store_mask
4429 and vect_check_store_rhs. Use the dt returned by the latter instead
4430 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4431 instead of calls to vect_is_simple_use. Pass the scalar rather
4432 than the vector operand to vect_is_simple_use when handling
4433 second and subsequent copies of an rhs value.
4434 (vectorizable_load): Update calls to vect_check_load_store_mask
4435 and vect_build_gather_load_calls. Use the cached mask_dt and
4436 gs_info.offset_dt instead of calls to vect_is_simple_use.
4438 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4441 * tree-emutls.c: Include gimplify.h.
4442 (lower_emutls_2): New function.
4443 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4444 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4445 it before further processing.
4448 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4449 UINTVAL (trueop1) instead of INTVAL (op1).
4451 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4455 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4456 INCOMING_FRAME_SP_OFFSET if not defined.
4457 (scan_trace): Add ENTRY argument. If true and
4458 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4459 emit a note to adjust the CFA offset.
4460 (create_cfi_notes): Adjust scan_trace callers.
4461 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4462 INCOMING_FRAME_SP_OFFSET in the CIE.
4463 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4464 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4466 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4467 * doc/tm.texi: Regenerated.
4469 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4471 PR rtl-optimization/83147
4472 * lra-constraints.c (remove_inheritance_pseudos): Use
4473 lra_substitute_pseudo_within_insn.
4475 2018-01-19 Tom de Vries <tom@codesourcery.com>
4476 Cesar Philippidis <cesar@codesourcery.com>
4479 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4481 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4484 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4485 spaces for function labels.
4487 2018-01-19 Martin Liska <mliska@suse.cz>
4489 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4490 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4491 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4492 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4493 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4494 (PRED_CONST_RETURN): Change from 69 to 65.
4495 (PRED_NULL_RETURN): Change from 91 to 71.
4496 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4497 (PRED_LOOP_GUARD): Change from 66 to 73.
4499 2018-01-19 Martin Liska <mliska@suse.cz>
4501 * predict.c (predict_insn_def): Add new assert.
4502 (struct branch_predictor): Change type to signed integer.
4503 (test_prediction_value_range): Amend test to cover
4505 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4506 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4507 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4508 (PRED_LOOP_IV_COMPARE): Likewise.
4509 * predict.h (PROB_UNINITIALIZED): Define new constant.
4511 2018-01-19 Martin Liska <mliska@suse.cz>
4513 * predict.c (dump_prediction): Add new format for
4514 analyze_brprob.py script which is enabled with -details
4516 * profile-count.h (precise_p): New function.
4518 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4520 PR tree-optimization/83922
4521 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4522 there are no statements that need masking.
4523 (vect_active_double_reduction_p): New function.
4524 (vect_analyze_loop_operations): Use it when handling phis that
4525 are not in the loop header.
4527 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4529 PR tree-optimization/83914
4530 * tree-vect-loop.c (vectorizable_induction): Don't convert
4531 init_expr or apply the peeling adjustment for inductions
4532 that are nested within the vectorized loop.
4534 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4536 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4539 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4543 * function.h (gimplify_parameters): Add gimple_seq * argument.
4544 * function.c: Include gimple.h and options.h.
4545 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4546 for the added local temporaries if needed.
4547 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4548 if there are any parameter cleanups, wrap whole body into a
4549 try/finally with the cleanups.
4551 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4554 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4555 Use GET_MODE_CLASS for scalar floating point.
4557 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4561 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4562 Fix call of call_cgraph_insertion_hooks.
4564 2018-01-18 Martin Sebor <msebor@redhat.com>
4566 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4568 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4571 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4574 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4577 * common.opt: (-ffile-prefix-map): New option.
4578 * opts.c (common_handle_option): Defer it.
4579 * opts-global.c (handle_common_deferred_options): Handle it.
4580 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4581 * file-prefix-map.h: New file.
4582 (remap_debug_filename, add_debug_prefix_map): ...here.
4583 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4584 * final.c (debug_prefix_map, add_debug_prefix_map
4585 remap_debug_filename): Move to...
4586 * file-prefix-map.c: New file.
4587 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4588 generalize, get rid of alloca(), use strrchr() instead of strchr().
4589 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4590 Implement in terms of add_prefix_map().
4591 (remap_macro_filename, remap_debug_filename): Implement in term of
4593 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4594 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4595 * dbxout.c: Include file-prefix-map.h.
4596 * varasm.c: Likewise.
4597 * vmsdbgout.c: Likewise.
4598 * xcoffout.c: Likewise.
4599 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4600 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4601 * doc/invoke.texi (-ffile-prefix-map): Document.
4602 (-fdebug-prefix-map): Update description.
4604 2018-01-18 Martin Liska <mliska@suse.cz>
4606 * config/i386/i386.c (indirect_thunk_name): Document that also
4608 (output_indirect_thunk): Document why both instructions
4609 (pause and lfence) are generated.
4611 2018-01-18 Richard Biener <rguenther@suse.de>
4613 PR tree-optimization/83887
4614 * graphite-scop-detection.c
4615 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4616 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4617 (scop_detection::merge_sese): Re-implement with a flood-fill
4618 algorithm that properly finds a SESE region if it exists.
4620 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4623 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4624 pointer_diff optimizations use view_convert instead of convert.
4626 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4628 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4629 Generate different code for -mno-speculate-indirect-jumps.
4630 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4631 (*call_indirect_aix<mode>): Disable for
4632 -mno-speculate-indirect-jumps.
4633 (*call_indirect_aix<mode>_nospec): New define_insn.
4634 (*call_value_indirect_aix<mode>): Disable for
4635 -mno-speculate-indirect-jumps.
4636 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4637 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4638 -mno-speculate-indirect-jumps.
4639 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4641 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4643 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4644 long double type, set the flags for noting the default long double
4645 type, even if we don't pass or return a long double type.
4647 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4650 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4653 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4655 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4656 support for merge[hl].
4657 (fold_mergehl_helper): New helper function.
4658 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4659 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4660 (altivec_vmrglw_direct): Add xxmrglw insn.
4662 2018-01-17 Andrew Waterman <andrew@sifive.com>
4664 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4665 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4667 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4670 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4671 call the lto_location_cache before reading the
4672 DECL_SOURCE_LOCATION of the types.
4674 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4675 Richard Sandiford <richard.sandiford@linaro.org>
4677 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4678 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4679 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4680 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4681 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4683 * config/aarch64/constraints.md (aarch64_movti_operand):
4685 * config/aarch64/predicates.md (Uti): Add new constraint.
4687 2018-01-17 Carl Love <cel@us.ibm.com>
4689 * config/rs6000/vsx.md (define_expand xl_len_r,
4690 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4691 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4693 (define_expand, define_insn): Move the shift left from the
4694 define_insn to the define_expand for lxvl and stxvl instructions.
4695 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4696 and XL_LEN_R definitions to PURE.
4698 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4700 * config/i386/i386.c (indirect_thunk_name): Declare regno
4701 as unsigned int. Compare regno with INVALID_REGNUM.
4702 (output_indirect_thunk): Ditto.
4703 (output_indirect_thunk_function): Ditto.
4704 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4705 in the call to output_indirect_thunk_function.
4707 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4710 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4711 rather than the size of inner_type to determine the stack slot size
4712 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4714 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4717 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4720 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4722 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4723 endian Linux systems to optionally enable multilibs for selecting
4724 the long double type if the user configured an explicit type.
4725 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4726 have no long double multilibs if not defined.
4727 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4728 warn if the user used -mabi={ieee,ibm}longdouble and we built
4729 multilibs for long double.
4730 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4731 appropriate multilib option.
4732 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4734 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4735 for building long double multilibs.
4736 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4738 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4740 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4743 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4745 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4748 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4751 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4754 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4756 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4757 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4759 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4761 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4762 different rtl trees depending on TARGET_64BIT.
4763 (rs6000_gen_lvx): Likewise.
4765 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4767 * config/visium/visium.md (nop): Tweak comment.
4768 (hazard_nop): Likewise.
4770 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4772 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4773 -mspeculate-indirect-jumps.
4774 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4775 for -mno-speculate-indirect-jumps.
4776 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4777 (*call_value_indirect_elfv2<mode>): Disable for
4778 -mno-speculate-indirect-jumps.
4779 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4780 (indirect_jump): Emit different RTL for
4781 -mno-speculate-indirect-jumps.
4782 (*indirect_jump<mode>): Disable for
4783 -mno-speculate-indirect-jumps.
4784 (*indirect_jump<mode>_nospec): New define_insn.
4785 (tablejump): Emit different RTL for
4786 -mno-speculate-indirect-jumps.
4787 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4788 (tablejumpsi_nospec): New define_expand.
4789 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4790 (tablejumpdi_nospec): New define_expand.
4791 (*tablejump<mode>_internal1): Disable for
4792 -mno-speculate-indirect-jumps.
4793 (*tablejump<mode>_internal1_nospec): New define_insn.
4794 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4797 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4799 * caller-save.c (insert_save): Drop unnecessary parameter. All
4802 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4803 Richard Biener <rguenth@suse.de>
4806 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4807 return early, inline manually is_gimple_sizepos. Make sure if we
4808 call gimplify_expr we don't end up with a gimple constant.
4809 * tree.c (variably_modified_type_p): Don't return true for
4810 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4811 * gimplify.h (is_gimple_sizepos): Remove.
4813 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4815 PR tree-optimization/83857
4816 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4817 vectorizable_live_operation for pure SLP statements.
4818 (vectorizable_live_operation): Handle PHIs.
4820 2018-01-16 Richard Biener <rguenther@suse.de>
4822 PR tree-optimization/83867
4823 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4824 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4826 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4829 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4830 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4831 If off is not INTEGER_CST, issue a may not be aligned warning
4832 rather than isn't aligned. Use isn%'t rather than isn't.
4833 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4835 <case MULT_EXPR>: Improve the case when bottom and one of the
4836 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4837 operand, in that case check if the other operand is multiple of
4838 bottom divided by the INTEGER_CST operand.
4840 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4843 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4844 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4845 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4846 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4847 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4848 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4849 (pa_function_arg_size): New function.
4851 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4853 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4854 in a separate statement.
4856 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4858 PR tree-optimization/83847
4859 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4860 group gathers and scatters.
4862 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4864 PR rtl-optimization/86620
4865 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4867 PR rtl-optimization/83213
4868 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4869 to last if both are JUMP_INSNs.
4871 PR tree-optimization/83843
4872 * gimple-ssa-store-merging.c
4873 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4874 store_immediate_info for bswap/nop orig_stores.
4876 2018-01-15 Andrew Waterman <andrew@sifive.com>
4878 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4880 <UDIV>: Increase cost if !TARGET_DIV.
4882 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4884 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4885 (define_attr "cr_logical_3op"): New.
4886 (cceq_ior_compare): Adjust.
4887 (cceq_ior_compare_complement): Adjust.
4888 (*cceq_rev_compare): Adjust.
4889 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4890 (is_cracked_insn): Adjust.
4891 (insn_must_be_first_in_group): Adjust.
4892 * config/rs6000/40x.md: Adjust.
4893 * config/rs6000/440.md: Adjust.
4894 * config/rs6000/476.md: Adjust.
4895 * config/rs6000/601.md: Adjust.
4896 * config/rs6000/603.md: Adjust.
4897 * config/rs6000/6xx.md: Adjust.
4898 * config/rs6000/7450.md: Adjust.
4899 * config/rs6000/7xx.md: Adjust.
4900 * config/rs6000/8540.md: Adjust.
4901 * config/rs6000/cell.md: Adjust.
4902 * config/rs6000/e300c2c3.md: Adjust.
4903 * config/rs6000/e500mc.md: Adjust.
4904 * config/rs6000/e500mc64.md: Adjust.
4905 * config/rs6000/e5500.md: Adjust.
4906 * config/rs6000/e6500.md: Adjust.
4907 * config/rs6000/mpc.md: Adjust.
4908 * config/rs6000/power4.md: Adjust.
4909 * config/rs6000/power5.md: Adjust.
4910 * config/rs6000/power6.md: Adjust.
4911 * config/rs6000/power7.md: Adjust.
4912 * config/rs6000/power8.md: Adjust.
4913 * config/rs6000/power9.md: Adjust.
4914 * config/rs6000/rs64.md: Adjust.
4915 * config/rs6000/titan.md: Adjust.
4917 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4919 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4920 ix86_indirect_branch_register logic.
4922 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4924 * config/i386/constraints.md (Bs): Update
4925 ix86_indirect_branch_register check. Don't check
4926 ix86_indirect_branch_register with GOT_memory_operand.
4928 * config/i386/predicates.md (GOT_memory_operand): Don't check
4929 ix86_indirect_branch_register here.
4930 (GOT32_symbol_operand): Likewise.
4932 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4934 * config/i386/predicates.md (constant_call_address_operand):
4935 Rewrite ix86_indirect_branch_register logic.
4936 (sibcall_insn_operand): Likewise.
4938 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4940 * config/i386/constraints.md (Bs): Replace
4941 ix86_indirect_branch_thunk_register with
4942 ix86_indirect_branch_register.
4944 * config/i386/i386.md (indirect_jump): Likewise.
4945 (tablejump): Likewise.
4946 (*sibcall_memory): Likewise.
4947 (*sibcall_value_memory): Likewise.
4948 Peepholes of indirect call and jump via memory: Likewise.
4949 * config/i386/i386.opt: Likewise.
4950 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4951 (GOT_memory_operand): Likewise.
4952 (call_insn_operand): Likewise.
4953 (sibcall_insn_operand): Likewise.
4954 (GOT32_symbol_operand): Likewise.
4956 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4959 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4960 type rather than type addr's type points to.
4961 (expand_omp_atomic_mutex): Likewise.
4962 (expand_omp_atomic): Likewise.
4964 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4967 * config/i386/i386.c (output_indirect_thunk_function): Use
4968 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4969 for __x86_return_thunk.
4971 2018-01-15 Richard Biener <rguenther@suse.de>
4974 * expmed.c (extract_bit_field_1): Fix typo.
4976 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4979 * config/arm/iterators.md (VF): New mode iterator.
4980 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4981 Remove integer-related logic from pattern.
4982 (neon_vabd<mode>_3): Likewise.
4984 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4987 * common.opt (fstrict-overflow): No longer an alias.
4988 (fwrapv-pointer): New option.
4989 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4990 also for pointer types based on flag_wrapv_pointer.
4991 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4992 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4993 opts->x_flag_wrapv got set.
4994 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4995 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4996 POINTER_TYPE_OVERFLOW_UNDEFINED.
4997 * match.pd: Likewise in address comparison pattern.
4998 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5000 2018-01-15 Richard Biener <rguenther@suse.de>
5003 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5004 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5005 Reset type names to their identifier if their TYPE_DECL doesn't
5006 have linkage (and thus is used for ODR and devirt).
5007 (save_debug_info_for_decl): Remove.
5008 (save_debug_info_for_type): Likewise.
5009 (add_tree_to_fld_list): Adjust.
5010 * tree-pretty-print.c (dump_generic_node): Make dumping of
5011 type names more robust.
5013 2018-01-15 Richard Biener <rguenther@suse.de>
5015 * BASE-VER: Bump to 8.0.1.
5017 2018-01-14 Martin Sebor <msebor@redhat.com>
5020 * builtins.c (check_access): Avoid warning when the no-warning bit
5023 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5025 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5026 * ira-color (allocno_hard_regs_compare): Likewise.
5028 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5031 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5032 Use .pushsection/.popsection.
5034 2018-01-14 Martin Sebor <msebor@redhat.com>
5037 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5039 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5041 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5042 entry from extra_headers.
5043 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5044 extra_headers, make the list bitwise identical to the i?86-*-* one.
5046 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5048 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5049 -mcmodel=large with -mindirect-branch=thunk,
5050 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5051 -mfunction-return=thunk-extern.
5052 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5053 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5054 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5056 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5058 * config/i386/i386.c (print_reg): Print the name of the full
5059 integer register without '%'.
5060 (ix86_print_operand): Handle 'V'.
5061 * doc/extend.texi: Document 'V' modifier.
5063 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5065 * config/i386/constraints.md (Bs): Disallow memory operand for
5066 -mindirect-branch-register.
5068 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5069 (GOT_memory_operand): Likewise.
5070 (call_insn_operand): Likewise.
5071 (sibcall_insn_operand): Likewise.
5072 (GOT32_symbol_operand): Likewise.
5073 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5074 for -mindirect-branch-register.
5075 (tablejump): Likewise.
5076 (*sibcall_memory): Likewise.
5077 (*sibcall_value_memory): Likewise.
5078 Disallow peepholes of indirect call and jump via memory for
5079 -mindirect-branch-register.
5080 (*call_pop): Replace m with Bw.
5081 (*call_value_pop): Likewise.
5082 (*sibcall_pop_memory): Replace m with Bs.
5083 * config/i386/i386.opt (mindirect-branch-register): New option.
5084 * doc/invoke.texi: Document -mindirect-branch-register option.
5086 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5088 * config/i386/i386-protos.h (ix86_output_function_return): New.
5089 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5090 set function_return_type.
5091 (indirect_thunk_name): Add ret_p to indicate thunk for function
5093 (output_indirect_thunk_function): Pass false to
5094 indirect_thunk_name.
5095 (ix86_output_indirect_branch_via_reg): Likewise.
5096 (ix86_output_indirect_branch_via_push): Likewise.
5097 (output_indirect_thunk_function): Create alias for function
5098 return thunk if regno < 0.
5099 (ix86_output_function_return): New function.
5100 (ix86_handle_fndecl_attribute): Handle function_return.
5101 (ix86_attribute_table): Add function_return.
5102 * config/i386/i386.h (machine_function): Add
5103 function_return_type.
5104 * config/i386/i386.md (simple_return_internal): Use
5105 ix86_output_function_return.
5106 (simple_return_internal_long): Likewise.
5107 * config/i386/i386.opt (mfunction-return=): New option.
5108 (indirect_branch): Mention -mfunction-return=.
5109 * doc/extend.texi: Document function_return function attribute.
5110 * doc/invoke.texi: Document -mfunction-return= option.
5112 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5114 * config/i386/i386-opts.h (indirect_branch): New.
5115 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5116 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5117 with local indirect jump when converting indirect call and jump.
5118 (ix86_set_indirect_branch_type): New.
5119 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5120 (indirectlabelno): New.
5121 (indirect_thunk_needed): Likewise.
5122 (indirect_thunk_bnd_needed): Likewise.
5123 (indirect_thunks_used): Likewise.
5124 (indirect_thunks_bnd_used): Likewise.
5125 (INDIRECT_LABEL): Likewise.
5126 (indirect_thunk_name): Likewise.
5127 (output_indirect_thunk): Likewise.
5128 (output_indirect_thunk_function): Likewise.
5129 (ix86_output_indirect_branch_via_reg): Likewise.
5130 (ix86_output_indirect_branch_via_push): Likewise.
5131 (ix86_output_indirect_branch): Likewise.
5132 (ix86_output_indirect_jmp): Likewise.
5133 (ix86_code_end): Call output_indirect_thunk_function if needed.
5134 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5136 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5137 (ix86_attribute_table): Add indirect_branch.
5138 * config/i386/i386.h (machine_function): Add indirect_branch_type
5139 and has_local_indirect_jump.
5140 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5142 (tablejump): Likewise.
5143 (*indirect_jump): Use ix86_output_indirect_jmp.
5144 (*tablejump_1): Likewise.
5145 (simple_return_indirect_internal): Likewise.
5146 * config/i386/i386.opt (mindirect-branch=): New option.
5147 (indirect_branch): New.
5150 (thunk-inline): Likewise.
5151 (thunk-extern): Likewise.
5152 * doc/extend.texi: Document indirect_branch function attribute.
5153 * doc/invoke.texi: Document -mindirect-branch= option.
5155 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5158 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5160 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5162 * ipa-inline.c (want_inline_small_function_p): Return false if
5163 inlining has already failed with CIF_FINAL_ERROR.
5164 (update_caller_keys): Call want_inline_small_function_p before
5166 (update_callee_keys): Likewise.
5168 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5170 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5172 (rs6000_quadword_masked_address_p): Likewise.
5173 (quad_aligned_load_p): Likewise.
5174 (quad_aligned_store_p): Likewise.
5175 (const_load_sequence_p): Add comment to describe the outer-most loop.
5176 (mimic_memory_attributes_and_flags): New function.
5177 (rs6000_gen_stvx): Likewise.
5178 (replace_swapped_aligned_store): Likewise.
5179 (rs6000_gen_lvx): Likewise.
5180 (replace_swapped_aligned_load): Likewise.
5181 (replace_swapped_load_constant): Capitalize argument name in
5182 comment describing this function.
5183 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5184 and stores that access quad-word aligned addresses and replace
5185 with stvx or lvx instructions when appropriate.
5186 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5187 New function prototype.
5188 (rs6000_quadword_masked_address_p): Likewise.
5189 (rs6000_gen_lvx): Likewise.
5190 (rs6000_gen_stvx): Likewise.
5191 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5192 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5193 when memory address is aligned.
5194 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5195 this split to select lvx instruction when memory address is aligned.
5196 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5197 instruction when memory address is aligned.
5198 (*vsx_le_perm_load_v16qi): Likewise.
5199 (four unnamed splitters): Modify to select the stvx instruction
5200 when memory is aligned.
5202 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5204 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5205 which appears in the queue multiple times.
5207 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5208 Alan Hayward <alan.hayward@arm.com>
5209 David Sherwood <david.sherwood@arm.com>
5211 * tree-vectorizer.h (vec_lower_bound): New structure.
5212 (_loop_vec_info): Add check_nonzero and lower_bounds.
5213 (LOOP_VINFO_CHECK_NONZERO): New macro.
5214 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5215 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5216 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5217 fields. Make seg_len the distance travelled, not including the
5219 (dr_direction_indicator): Declare.
5220 (dr_zero_step_indicator): Likewise.
5221 (dr_known_forward_stride_p): Likewise.
5222 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5224 (runtime_alias_check_p): Allow runtime alias checks with
5226 (operator ==): Compare access_size and align.
5227 (prune_runtime_alias_test_list): Rework for new distinction between
5228 the access_size and seg_len.
5229 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5231 (get_segment_min_max): New function.
5232 (create_intersect_range_checks): Use it.
5233 (dr_step_indicator): New function.
5234 (dr_direction_indicator): Likewise.
5235 (dr_zero_step_indicator): Likewise.
5236 (dr_known_forward_stride_p): Likewise.
5237 * tree-loop-distribution.c (data_ref_segment_size): Return
5238 DR_STEP * (niters - 1).
5239 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5241 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5242 (vect_preserves_scalar_order_p): New function, split out from...
5243 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5244 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5245 (vect_vfa_access_size): New function.
5246 (vect_vfa_align): Likewise.
5247 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5248 (dump_lower_bound): New function.
5249 (vect_check_lower_bound): Likewise.
5250 (vect_small_gap_p): Likewise.
5251 (vectorizable_with_step_bound_p): Likewise.
5252 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5253 depencies if the vectorization factor is 1. Convert the checks
5254 for nonzero steps into checks on the bounds of DR_STEP. Try using
5255 a bunds check for variable steps if the minimum required step is
5256 relatively small. Update calls to the dr_with_seg_len
5257 constructor and to vect_compile_time_alias.
5258 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5260 (vect_loop_versioning): Call it.
5261 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5263 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5266 Alan Hayward <alan.hayward@arm.com>
5267 David Sherwood <david.sherwood@arm.com>
5269 * doc/sourcebuild.texi (vect_scatter_store): Document.
5270 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5272 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5274 * genopinit.c (main): Add supports_vec_scatter_store and
5275 supports_vec_scatter_store_cached to target_optabs.
5276 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5277 IFN_MASK_SCATTER_STORE.
5278 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5280 * internal-fn.h (internal_store_fn_p): Declare.
5281 (internal_fn_stored_value_index): Likewise.
5282 * internal-fn.c (scatter_store_direct): New macro.
5283 (expand_scatter_store_optab_fn): New function.
5284 (direct_scatter_store_optab_supported_p): New macro.
5285 (internal_store_fn_p): New function.
5286 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5287 IFN_MASK_SCATTER_STORE.
5288 (internal_fn_mask_index): Likewise.
5289 (internal_fn_stored_value_index): New function.
5290 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5292 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5293 * optabs-query.c (supports_vec_scatter_store_p): New function.
5294 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5295 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5296 true for scatter stores.
5297 (vect_gather_scatter_fn_p): Handle scatter stores too.
5298 (vect_check_gather_scatter): Consider using scatter stores if
5299 supports_vec_scatter_store_p.
5300 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5302 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5303 internal_fn_stored_value_index.
5304 (check_load_store_masking): Handle scatter stores too.
5305 (vect_get_store_rhs): Make public.
5306 (vectorizable_call): Use internal_store_fn_p.
5307 (vectorizable_store): Handle scatter store internal functions.
5308 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5309 when deciding whether the end of the group has been reached.
5310 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5311 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5312 (mask_scatter_store<mode>): New insns.
5314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5315 Alan Hayward <alan.hayward@arm.com>
5316 David Sherwood <david.sherwood@arm.com>
5318 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5319 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5320 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5322 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5323 Use vect_truncate_gather_scatter_offset if we can't treat the
5324 operation as a normal gather load or scatter store.
5325 (get_group_load_store_type): Take the gather_scatter_info
5326 as argument. Try using a gather load or scatter store for
5327 single-element groups.
5328 (get_load_store_type): Update calls to get_group_load_store_type
5329 and vect_use_strided_gather_scatters_p.
5331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5332 Alan Hayward <alan.hayward@arm.com>
5333 David Sherwood <david.sherwood@arm.com>
5335 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5336 optional tree argument.
5337 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5339 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5340 but continue to use the current value as a fallback.
5341 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5342 to compare the updates.
5343 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5344 (get_load_store_type): Use it when handling a strided access.
5345 (vect_get_strided_load_store_ops): New function.
5346 (vect_get_data_ptr_increment): Likewise.
5347 (vectorizable_load): Handle strided gather loads. Always pass
5348 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5351 Alan Hayward <alan.hayward@arm.com>
5352 David Sherwood <david.sherwood@arm.com>
5354 * doc/md.texi (gather_load@var{m}): Document.
5355 (mask_gather_load@var{m}): Likewise.
5356 * genopinit.c (main): Add supports_vec_gather_load and
5357 supports_vec_gather_load_cached to target_optabs.
5358 * optabs-tree.c (init_tree_optimization_optabs): Use
5359 ggc_cleared_alloc to allocate target_optabs.
5360 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5361 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5363 * internal-fn.h (internal_load_fn_p): Declare.
5364 (internal_gather_scatter_fn_p): Likewise.
5365 (internal_fn_mask_index): Likewise.
5366 (internal_gather_scatter_fn_supported_p): Likewise.
5367 * internal-fn.c (gather_load_direct): New macro.
5368 (expand_gather_load_optab_fn): New function.
5369 (direct_gather_load_optab_supported_p): New macro.
5370 (direct_internal_fn_optab): New function.
5371 (internal_load_fn_p): Likewise.
5372 (internal_gather_scatter_fn_p): Likewise.
5373 (internal_fn_mask_index): Likewise.
5374 (internal_gather_scatter_fn_supported_p): Likewise.
5375 * optabs-query.c (supports_at_least_one_mode_p): New function.
5376 (supports_vec_gather_load_p): Likewise.
5377 * optabs-query.h (supports_vec_gather_load_p): Declare.
5378 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5379 and memory_type field.
5380 (NUM_PATTERNS): Bump to 15.
5381 * tree-vect-data-refs.c: Include internal-fn.h.
5382 (vect_gather_scatter_fn_p): New function.
5383 (vect_describe_gather_scatter_call): Likewise.
5384 (vect_check_gather_scatter): Try using internal functions for
5385 gather loads. Recognize existing calls to a gather load function.
5386 (vect_analyze_data_refs): Consider using gather loads if
5387 supports_vec_gather_load_p.
5388 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5389 (vect_get_gather_scatter_offset_type): Likewise.
5390 (vect_convert_mask_for_vectype): Likewise.
5391 (vect_add_conversion_to_patterm): Likewise.
5392 (vect_try_gather_scatter_pattern): Likewise.
5393 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5394 (vect_vect_recog_func_ptrs): Add it.
5395 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5396 internal_fn_mask_index and internal_gather_scatter_fn_p.
5397 (check_load_store_masking): Take the gather_scatter_info as an
5398 argument and handle gather loads.
5399 (vect_get_gather_scatter_ops): New function.
5400 (vectorizable_call): Check internal_load_fn_p.
5401 (vectorizable_load): Likewise. Handle gather load internal
5403 (vectorizable_store): Update call to check_load_store_masking.
5404 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5405 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5406 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5407 (aarch64_gather_scale_operand_d): New predicates.
5408 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5409 (mask_gather_load<mode>): New insns.
5411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5412 Alan Hayward <alan.hayward@arm.com>
5413 David Sherwood <david.sherwood@arm.com>
5415 * optabs.def (fold_left_plus_optab): New optab.
5416 * doc/md.texi (fold_left_plus_@var{m}): Document.
5417 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5418 * internal-fn.c (fold_left_direct): Define.
5419 (expand_fold_left_optab_fn): Likewise.
5420 (direct_fold_left_optab_supported_p): Likewise.
5421 * fold-const-call.c (fold_const_fold_left): New function.
5422 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5423 * tree-parloops.c (valid_reduction_p): New function.
5424 (gather_scalar_reductions): Use it.
5425 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5426 (vect_finish_replace_stmt): Declare.
5427 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5428 (needs_fold_left_reduction_p): New function, split out from...
5429 (vect_is_simple_reduction): ...here. Accept reductions that
5430 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5431 (vect_force_simple_reduction): Also store the reduction type in
5432 the assignment's STMT_VINFO_REDUC_TYPE.
5433 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5434 (merge_with_identity): New function.
5435 (vect_expand_fold_left): Likewise.
5436 (vectorize_fold_left_reduction): Likewise.
5437 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5438 scalar phi in place for it. Check for target support and reject
5439 cases that would reassociate the operation. Defer the transform
5440 phase to vectorize_fold_left_reduction.
5441 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5442 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5443 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5447 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5448 call to ifc_temp_var.
5450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5451 Alan Hayward <alan.hayward@arm.com>
5452 David Sherwood <david.sherwood@arm.com>
5454 * target.def (legitimize_address_displacement): Take the original
5455 offset as a poly_int.
5456 * targhooks.h (default_legitimize_address_displacement): Update
5458 * targhooks.c (default_legitimize_address_displacement): Likewise.
5459 * doc/tm.texi: Regenerate.
5460 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5461 as an argument, moving assert of ad->disp == ad->disp_term to...
5462 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5463 Try calling targetm.legitimize_address_displacement before expanding
5464 the address rather than afterwards, and adjust for the new interface.
5465 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5466 Match the new hook interface. Handle SVE addresses.
5467 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5472 * Makefile.in (OBJS): Add early-remat.o.
5473 * target.def (select_early_remat_modes): New hook.
5474 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5475 * doc/tm.texi: Regenerate.
5476 * targhooks.h (default_select_early_remat_modes): Declare.
5477 * targhooks.c (default_select_early_remat_modes): New function.
5478 * timevar.def (TV_EARLY_REMAT): New timevar.
5479 * passes.def (pass_early_remat): New pass.
5480 * tree-pass.h (make_pass_early_remat): Declare.
5481 * early-remat.c: New file.
5482 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5484 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5487 Alan Hayward <alan.hayward@arm.com>
5488 David Sherwood <david.sherwood@arm.com>
5490 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5491 vfm1 with a bound_epilog parameter.
5492 (vect_do_peeling): Update calls accordingly, and move the prologue
5493 call earlier in the function. Treat the base bound_epilog as 0 for
5494 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5495 this base when peeling for gaps.
5496 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5497 with fully-masked loops.
5498 (vect_estimate_min_profitable_iters): Handle the single peeled
5499 iteration in that case.
5501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5502 Alan Hayward <alan.hayward@arm.com>
5503 David Sherwood <david.sherwood@arm.com>
5505 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5506 single-element interleaving even if the size is not a power of 2.
5507 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5508 accesses for single-element interleaving if the group size is
5511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5512 Alan Hayward <alan.hayward@arm.com>
5513 David Sherwood <david.sherwood@arm.com>
5515 * doc/md.texi (fold_extract_last_@var{m}): Document.
5516 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5517 * optabs.def (fold_extract_last_optab): New optab.
5518 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5519 * internal-fn.c (fold_extract_direct): New macro.
5520 (expand_fold_extract_optab_fn): Likewise.
5521 (direct_fold_extract_optab_supported_p): Likewise.
5522 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5523 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5524 EXTRACT_LAST_REDUCTION.
5525 (get_initial_def_for_reduction): Do not create an initial vector
5526 for EXTRACT_LAST_REDUCTION reductions.
5527 (vectorizable_reduction): Leave the scalar phi in place for
5528 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5529 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5530 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5531 transform phase to vectorizable_condition.
5532 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5534 (vect_finish_stmt_generation): ...here.
5535 (vect_finish_replace_stmt): New function.
5536 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5537 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5539 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5542 Alan Hayward <alan.hayward@arm.com>
5543 David Sherwood <david.sherwood@arm.com>
5545 * doc/md.texi (extract_last_@var{m}): Document.
5546 * optabs.def (extract_last_optab): New optab.
5547 * internal-fn.def (EXTRACT_LAST): New internal function.
5548 * internal-fn.c (cond_unary_direct): New macro.
5549 (expand_cond_unary_optab_fn): Likewise.
5550 (direct_cond_unary_optab_supported_p): Likewise.
5551 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5552 loops using EXTRACT_LAST.
5553 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5554 (extract_last_<mode>): ...this optab.
5555 (vec_extract<mode><Vel>): Update accordingly.
5557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5558 Alan Hayward <alan.hayward@arm.com>
5559 David Sherwood <david.sherwood@arm.com>
5561 * target.def (empty_mask_is_expensive): New hook.
5562 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5563 * doc/tm.texi: Regenerate.
5564 * targhooks.h (default_empty_mask_is_expensive): Declare.
5565 * targhooks.c (default_empty_mask_is_expensive): New function.
5566 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5567 if the target says that empty masks are expensive.
5568 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5570 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5573 Alan Hayward <alan.hayward@arm.com>
5574 David Sherwood <david.sherwood@arm.com>
5576 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5577 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5578 (vect_use_loop_mask_for_alignment_p): New function.
5579 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5580 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5581 niters_skip argument. Make sure that the first niters_skip elements
5582 of the first iteration are inactive.
5583 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5584 Update call to vect_set_loop_masks_directly.
5585 (get_misalign_in_elems): New function, split out from...
5586 (vect_gen_prolog_loop_niters): ...here.
5587 (vect_update_init_of_dr): Take a code argument that specifies whether
5588 the adjustment should be added or subtracted.
5589 (vect_update_init_of_drs): Likewise.
5590 (vect_prepare_for_masked_peels): New function.
5591 (vect_do_peeling): Skip prologue peeling if we're using a mask
5592 instead. Update call to vect_update_inits_of_drs.
5593 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5595 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5596 alignment. Do not include the number of peeled iterations in
5597 the minimum threshold in that case.
5598 (vectorizable_induction): Adjust the start value down by
5599 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5600 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5601 Take the number of skipped iterations into account when calculating
5603 * tree-vect-stmts.c (vect_gen_while_not): New function.
5605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5606 Alan Hayward <alan.hayward@arm.com>
5607 David Sherwood <david.sherwood@arm.com>
5609 * doc/sourcebuild.texi (vect_fully_masked): Document.
5610 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5612 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5614 (vect_analyze_loop_2): ...here. Don't check the vectorization
5615 factor against the number of loop iterations if the loop is
5618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5619 Alan Hayward <alan.hayward@arm.com>
5620 David Sherwood <david.sherwood@arm.com>
5622 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5623 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5624 (dump_groups): Update accordingly.
5625 (iv_use::mem_type): New member variable.
5626 (address_p): New function.
5627 (record_use): Add a mem_type argument and initialize the new
5629 (record_group_use): Add a mem_type argument. Use address_p.
5630 Remove obsolete null checks of base_object. Update call to record_use.
5631 (find_interesting_uses_op): Update call to record_group_use.
5632 (find_interesting_uses_cond): Likewise.
5633 (find_interesting_uses_address): Likewise.
5634 (get_mem_type_for_internal_fn): New function.
5635 (find_address_like_use): Likewise.
5636 (find_interesting_uses_stmt): Try find_address_like_use before
5637 calling find_interesting_uses_op.
5638 (addr_offset_valid_p): Use the iv mem_type field as the type
5639 of the addressed memory.
5640 (add_autoinc_candidates): Likewise.
5641 (get_address_cost): Likewise.
5642 (split_small_address_groups_p): Use address_p.
5643 (split_address_groups): Likewise.
5644 (add_iv_candidate_for_use): Likewise.
5645 (autoinc_possible_for_pair): Likewise.
5646 (rewrite_groups): Likewise.
5647 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5648 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5649 (get_alias_ptr_type_for_ptr_address): New function.
5650 (rewrite_use_address): Rewrite address uses in calls that were
5651 identified by find_address_like_use.
5653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5654 Alan Hayward <alan.hayward@arm.com>
5655 David Sherwood <david.sherwood@arm.com>
5657 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5659 * gimple-expr.h (is_gimple_addressable: Likewise.
5660 * gimple-expr.c (is_gimple_address): Likewise.
5661 * internal-fn.c (expand_call_mem_ref): New function.
5662 (expand_mask_load_optab_fn): Use it.
5663 (expand_mask_store_optab_fn): Likewise.
5665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5666 Alan Hayward <alan.hayward@arm.com>
5667 David Sherwood <david.sherwood@arm.com>
5669 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5670 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5671 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5672 (cond_umax@var{mode}): Document.
5673 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5674 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5675 (cond_umin_optab, cond_umax_optab): New optabs.
5676 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5677 (COND_IOR, COND_XOR): New internal functions.
5678 * internal-fn.h (get_conditional_internal_fn): Declare.
5679 * internal-fn.c (cond_binary_direct): New macro.
5680 (expand_cond_binary_optab_fn): Likewise.
5681 (direct_cond_binary_optab_supported_p): Likewise.
5682 (get_conditional_internal_fn): New function.
5683 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5684 Cope with reduction statements that are vectorized as calls rather
5686 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5687 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5688 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5689 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5690 (UNSPEC_COND_EOR): New unspecs.
5691 (optab): Add mappings for them.
5692 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5693 (sve_int_op, sve_fp_op): New int attributes.
5695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5696 Alan Hayward <alan.hayward@arm.com>
5697 David Sherwood <david.sherwood@arm.com>
5699 * optabs.def (while_ult_optab): New optab.
5700 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5701 * internal-fn.def (WHILE_ULT): New internal function.
5702 * internal-fn.h (direct_internal_fn_supported_p): New override
5703 that takes two types as argument.
5704 * internal-fn.c (while_direct): New macro.
5705 (expand_while_optab_fn): New function.
5706 (convert_optab_supported_p): Likewise.
5707 (direct_while_optab_supported_p): New macro.
5708 * wide-int.h (wi::udiv_ceil): New function.
5709 * tree-vectorizer.h (rgroup_masks): New structure.
5710 (vec_loop_masks): New typedef.
5711 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5713 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5714 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5715 (vect_max_vf): New function.
5716 (slpeel_make_loop_iterate_ntimes): Delete.
5717 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5718 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5719 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5720 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5721 internal-fn.h, stor-layout.h and optabs-query.h.
5722 (vect_set_loop_mask): New function.
5723 (add_preheader_seq): Likewise.
5724 (add_header_seq): Likewise.
5725 (interleave_supported_p): Likewise.
5726 (vect_maybe_permute_loop_masks): Likewise.
5727 (vect_set_loop_masks_directly): Likewise.
5728 (vect_set_loop_condition_masked): Likewise.
5729 (vect_set_loop_condition_unmasked): New function, split out from
5730 slpeel_make_loop_iterate_ntimes.
5731 (slpeel_make_loop_iterate_ntimes): Rename to..
5732 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5733 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5734 (vect_do_peeling): Update call accordingly.
5735 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5737 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5738 mask_compare_type, can_fully_mask_p and fully_masked_p.
5739 (release_vec_loop_masks): New function.
5740 (_loop_vec_info): Use it to free the loop masks.
5741 (can_produce_all_loop_masks_p): New function.
5742 (vect_get_max_nscalars_per_iter): Likewise.
5743 (vect_verify_full_masking): Likewise.
5744 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5745 retries, and free the mask rgroups before retrying. Check loop-wide
5746 reasons for disallowing fully-masked loops. Make the final decision
5747 about whether use a fully-masked loop or not.
5748 (vect_estimate_min_profitable_iters): Do not assume that peeling
5749 for the number of iterations will be needed for fully-masked loops.
5750 (vectorizable_reduction): Disable fully-masked loops.
5751 (vectorizable_live_operation): Likewise.
5752 (vect_halve_mask_nunits): New function.
5753 (vect_double_mask_nunits): Likewise.
5754 (vect_record_loop_mask): Likewise.
5755 (vect_get_loop_mask): Likewise.
5756 (vect_transform_loop): Handle the case in which the final loop
5757 iteration might handle a partial vector. Call vect_set_loop_condition
5758 instead of slpeel_make_loop_iterate_ntimes.
5759 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5760 (check_load_store_masking): New function.
5761 (prepare_load_store_mask): Likewise.
5762 (vectorizable_store): Handle fully-masked loops.
5763 (vectorizable_load): Likewise.
5764 (supportable_widening_operation): Use vect_halve_mask_nunits for
5766 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5767 (vect_gen_while): New function.
5768 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5769 (aarch64_uqdec<mode>): New insn.
5771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5772 Alan Hayward <alan.hayward@arm.com>
5773 David Sherwood <david.sherwood@arm.com>
5775 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5776 (reduc_xor_scal_optab): New optabs.
5777 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5778 (reduc_xor_scal_@var{m}): Document.
5779 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5780 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5782 * fold-const-call.c (fold_const_call): Handle them.
5783 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5784 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5785 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5786 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5787 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5788 (UNSPEC_XORV): New unspecs.
5789 (optab): Add entries for them.
5790 (BITWISEV): New int iterator.
5791 (bit_reduc_op): New int attributes.
5793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5794 Alan Hayward <alan.hayward@arm.com>
5795 David Sherwood <david.sherwood@arm.com>
5797 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5798 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5799 * optabs.def (vec_shl_insert_optab): New optab.
5800 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5801 (duplicate_and_interleave): Likewise.
5802 * tree-vect-loop.c: Include internal-fn.h.
5803 (neutral_op_for_slp_reduction): New function, split out from
5804 get_initial_defs_for_reduction.
5805 (get_initial_def_for_reduction): Handle option 2 for variable-length
5806 vectors by loading the neutral value into a vector and then shifting
5807 the initial value into element 0.
5808 (get_initial_defs_for_reduction): Replace the code argument with
5809 the neutral value calculated by neutral_op_for_slp_reduction.
5810 Use gimple_build_vector for constant-length vectors.
5811 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5812 but the first group_size elements have a neutral value.
5813 Use duplicate_and_interleave otherwise.
5814 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5815 Update call to get_initial_defs_for_reduction. Handle SLP
5816 reductions for variable-length vectors by creating one vector
5817 result for each scalar result, with the elements associated
5818 with other scalar results stubbed out with the neutral value.
5819 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5820 Require IFN_VEC_SHL_INSERT for double reductions on
5821 variable-length vectors, or SLP reductions that have
5822 a neutral value. Require can_duplicate_and_interleave_p
5823 support for variable-length unchained SLP reductions if there
5824 is no neutral value, such as for MIN/MAX reductions. Also require
5825 the number of vector elements to be a multiple of the number of
5826 SLP statements when doing variable-length unchained SLP reductions.
5827 Update call to vect_create_epilog_for_reduction.
5828 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5829 and remove initial values.
5830 (duplicate_and_interleave): Make public.
5831 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5832 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5835 Alan Hayward <alan.hayward@arm.com>
5836 David Sherwood <david.sherwood@arm.com>
5838 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5839 (can_duplicate_and_interleave_p): New function.
5840 (vect_get_and_check_slp_defs): Take the vector of statements
5841 rather than just the current one. Remove excess parentheses.
5842 Restriction rejectinon of vect_constant_def and vect_external_def
5843 for variable-length vectors to boolean types, or types for which
5844 can_duplicate_and_interleave_p is false.
5845 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5846 (duplicate_and_interleave): New function.
5847 (vect_get_constant_vectors): Use gimple_build_vector for
5848 constant-length vectors and suitable variable-length constant
5849 vectors. Use duplicate_and_interleave for other variable-length
5850 vectors. Don't defer the update when inserting new statements.
5852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5853 Alan Hayward <alan.hayward@arm.com>
5854 David Sherwood <david.sherwood@arm.com>
5856 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5857 min_profitable_iters doesn't go negative.
5859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5860 Alan Hayward <alan.hayward@arm.com>
5861 David Sherwood <david.sherwood@arm.com>
5863 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5864 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5865 * optabs.def (vec_mask_load_lanes_optab): New optab.
5866 (vec_mask_store_lanes_optab): Likewise.
5867 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5868 (MASK_STORE_LANES): Likewise.
5869 * internal-fn.c (mask_load_lanes_direct): New macro.
5870 (mask_store_lanes_direct): Likewise.
5871 (expand_mask_load_optab_fn): Handle masked operations.
5872 (expand_mask_load_lanes_optab_fn): New macro.
5873 (expand_mask_store_optab_fn): Handle masked operations.
5874 (expand_mask_store_lanes_optab_fn): New macro.
5875 (direct_mask_load_lanes_optab_supported_p): Likewise.
5876 (direct_mask_store_lanes_optab_supported_p): Likewise.
5877 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5879 (vect_load_lanes_supported): Likewise.
5880 * tree-vect-data-refs.c (strip_conversion): New function.
5881 (can_group_stmts_p): Likewise.
5882 (vect_analyze_data_ref_accesses): Use it instead of checking
5883 for a pair of assignments.
5884 (vect_store_lanes_supported): Take a masked_p parameter.
5885 (vect_load_lanes_supported): Likewise.
5886 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5887 vect_store_lanes_supported and vect_load_lanes_supported.
5888 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5889 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5890 parameter. Don't allow gaps for masked accesses.
5891 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5892 and vect_load_lanes_supported.
5893 (get_load_store_type): Take a masked_p parameter and update
5894 call to get_group_load_store_type.
5895 (vectorizable_store): Update call to get_load_store_type.
5896 Handle IFN_MASK_STORE_LANES.
5897 (vectorizable_load): Update call to get_load_store_type.
5898 Handle IFN_MASK_LOAD_LANES.
5900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5901 Alan Hayward <alan.hayward@arm.com>
5902 David Sherwood <david.sherwood@arm.com>
5904 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5906 * config/aarch64/aarch64-protos.h
5907 (aarch64_sve_struct_memory_operand_p): Declare.
5908 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5909 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5910 (VPRED, vpred): Handle SVE structure modes.
5911 * config/aarch64/constraints.md (Utx): New constraint.
5912 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5913 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5914 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5915 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5916 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5917 structure modes. Split into pieces after RA.
5918 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5919 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5921 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5922 SVE structure modes.
5923 (aarch64_classify_address): Likewise.
5924 (sizetochar): Move earlier in file.
5925 (aarch64_print_operand): Handle SVE register lists.
5926 (aarch64_array_mode): New function.
5927 (aarch64_sve_struct_memory_operand_p): Likewise.
5928 (TARGET_ARRAY_MODE): Redefine.
5930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5931 Alan Hayward <alan.hayward@arm.com>
5932 David Sherwood <david.sherwood@arm.com>
5934 * target.def (array_mode): New target hook.
5935 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5936 * doc/tm.texi: Regenerate.
5937 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5938 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5939 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5941 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5945 Alan Hayward <alan.hayward@arm.com>
5946 David Sherwood <david.sherwood@arm.com>
5948 * fold-const.c (fold_binary_loc): Check the argument types
5949 rather than the result type when testing for a vector operation.
5951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5953 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5954 * doc/tm.texi: Regenerate.
5956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5957 Alan Hayward <alan.hayward@arm.com>
5958 David Sherwood <david.sherwood@arm.com>
5960 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5961 (sve): Document new AArch64 extension.
5962 * doc/md.texi (w): Extend the description of the AArch64
5963 constraint to include SVE vectors.
5964 (Upl, Upa): Document new AArch64 predicate constraints.
5965 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5967 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5968 (msve-vector-bits=): New option.
5969 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5970 SVE when these are disabled.
5971 (sve): New extension.
5972 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5973 modes. Adjust their number of units based on aarch64_sve_vg.
5974 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5975 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5976 aarch64_addr_query_type.
5977 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5978 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5979 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5980 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5981 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5982 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5983 (aarch64_simd_imm_zero_p): Delete.
5984 (aarch64_check_zero_based_sve_index_immediate): Declare.
5985 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5986 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5987 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5988 (aarch64_sve_float_mul_immediate_p): Likewise.
5989 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5991 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5992 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5993 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5994 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5995 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5996 (aarch64_regmode_natural_size): Likewise.
5997 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5998 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6000 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6001 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6002 for VG and the SVE predicate registers.
6003 (V_ALIASES): Add a "z"-prefixed alias.
6004 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6005 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6006 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6007 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6008 (REG_CLASS_NAMES): Add entries for them.
6009 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6010 and the predicate registers.
6011 (aarch64_sve_vg): Declare.
6012 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6013 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6014 (REGMODE_NATURAL_SIZE): Define.
6015 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6017 * config/aarch64/aarch64.c: Include cfgrtl.h.
6018 (simd_immediate_info): Add a constructor for series vectors,
6019 and an associated step field.
6020 (aarch64_sve_vg): New variable.
6021 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6022 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6023 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6024 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6025 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6026 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6027 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6028 (aarch64_get_mask_mode): New functions.
6029 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6030 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6031 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6032 predicate modes and predicate registers. Explicitly restrict
6033 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6034 to store a vector mode if it is recognized by
6035 aarch64_classify_vector_mode.
6036 (aarch64_regmode_natural_size): New function.
6037 (aarch64_hard_regno_caller_save_mode): Return the original mode
6039 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6040 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6041 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6042 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6044 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6045 does not overlap dest if the function is frame-related. Handle
6047 (aarch64_split_add_offset): New function.
6048 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6049 them aarch64_add_offset.
6050 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6051 and update call to aarch64_sub_sp.
6052 (aarch64_add_cfa_expression): New function.
6053 (aarch64_expand_prologue): Pass extra temporary registers to the
6054 functions above. Handle the case in which we need to emit new
6055 DW_CFA_expressions for registers that were originally saved
6056 relative to the stack pointer, but now have to be expressed
6057 relative to the frame pointer.
6058 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6060 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6061 IP0 and IP1 values for SVE frames.
6062 (aarch64_expand_vec_series): New function.
6063 (aarch64_expand_sve_widened_duplicate): Likewise.
6064 (aarch64_expand_sve_const_vector): Likewise.
6065 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6066 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6067 into the register, rather than emitting a SET directly.
6068 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6069 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6070 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6071 (offset_9bit_signed_scaled_p): New functions.
6072 (aarch64_replicate_bitmask_imm): New function.
6073 (aarch64_bitmask_imm): Use it.
6074 (aarch64_cannot_force_const_mem): Reject expressions involving
6075 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6076 (aarch64_classify_index): Handle SVE indices, by requiring
6077 a plain register index with a scale that matches the element size.
6078 (aarch64_classify_address): Handle SVE addresses. Assert that
6079 the mode of the address is VOIDmode or an integer mode.
6080 Update call to aarch64_classify_symbol.
6081 (aarch64_classify_symbolic_expression): Update call to
6082 aarch64_classify_symbol.
6083 (aarch64_const_vec_all_in_range_p): New function.
6084 (aarch64_print_vector_float_operand): Likewise.
6085 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6086 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6087 and the FP immediates 1.0 and 0.5.
6088 (aarch64_print_address_internal): Handle SVE addresses.
6089 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6090 (aarch64_regno_regclass): Handle predicate registers.
6091 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6093 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6094 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6095 (aarch64_convert_sve_vector_bits): New function.
6096 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6097 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6099 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6100 Handle SVE vector and predicate modes. Accept VL-based constants
6101 that need only one temporary register, and VL offsets that require
6102 no temporary registers.
6103 (aarch64_conditional_register_usage): Mark the predicate registers
6104 as fixed if SVE isn't available.
6105 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6106 Return true for SVE vector and predicate modes.
6107 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6108 rather than an unsigned int. Handle SVE modes.
6109 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6111 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6113 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6114 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6115 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6116 (aarch64_sve_float_mul_immediate_p): New functions.
6117 (aarch64_sve_valid_immediate): New function.
6118 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6119 Explicitly reject structure modes. Check for INDEX constants.
6120 Handle PTRUE and PFALSE constants.
6121 (aarch64_check_zero_based_sve_index_immediate): New function.
6122 (aarch64_simd_imm_zero_p): Delete.
6123 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6124 vector modes. Accept constants in the range of CNT[BHWD].
6125 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6126 ask for an Advanced SIMD mode.
6127 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6128 (aarch64_simd_vector_alignment): Handle SVE predicates.
6129 (aarch64_vectorize_preferred_vector_alignment): New function.
6130 (aarch64_simd_vector_alignment_reachable): Use it instead of
6132 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6133 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6135 (MAX_VECT_LEN): Delete.
6136 (expand_vec_perm_d): Add a vec_flags field.
6137 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6138 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6139 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6141 (aarch64_evpc_rev): Rename to...
6142 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6143 (aarch64_evpc_rev_global): New function.
6144 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6145 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6147 (aarch64_evpc_sve_tbl): New function.
6148 (aarch64_expand_vec_perm_const_1): Update after rename of
6149 aarch64_evpc_rev. Handle SVE permutes too, trying
6150 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6151 than aarch64_evpc_tbl.
6152 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6153 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6154 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6155 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6156 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6157 (aarch64_expand_sve_vcond): New functions.
6158 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6159 of aarch64_vector_mode_p.
6160 (aarch64_dwarf_poly_indeterminate_value): New function.
6161 (aarch64_compute_pressure_classes): Likewise.
6162 (aarch64_can_change_mode_class): Likewise.
6163 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6164 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6165 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6166 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6167 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6168 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6169 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6170 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6172 (Dn, Dl, Dr): Accept const as well as const_vector.
6173 (Dz): Likewise. Compare against CONST0_RTX.
6174 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6175 of "vector" where appropriate.
6176 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6177 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6178 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6179 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6180 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6181 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6182 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6183 (v_int_equiv): Extend to SVE modes.
6184 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6186 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6187 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6188 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6189 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6190 (SVE_COND_FP_CMP): New int iterators.
6191 (perm_hilo): Handle the new unpack unspecs.
6192 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6194 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6195 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6196 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6197 (aarch64_equality_operator, aarch64_constant_vector_operand)
6198 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6199 (aarch64_sve_nonimmediate_operand): Likewise.
6200 (aarch64_sve_general_operand): Likewise.
6201 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6202 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6203 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6204 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6205 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6206 (aarch64_sve_float_arith_immediate): Likewise.
6207 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6208 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6209 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6210 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6211 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6212 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6213 (aarch64_sve_float_arith_operand): Likewise.
6214 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6215 (aarch64_sve_float_mul_operand): Likewise.
6216 (aarch64_sve_vec_perm_operand): Likewise.
6217 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6218 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6219 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6220 as well as const_vector.
6221 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6222 in file. Use CONST0_RTX and CONSTM1_RTX.
6223 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6224 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6225 Use aarch64_simd_imm_zero.
6226 * config/aarch64/aarch64-sve.md: New file.
6227 * config/aarch64/aarch64.md: Include it.
6228 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6229 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6230 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6231 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6232 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6233 (sve): New attribute.
6234 (enabled): Disable instructions with the sve attribute unless
6236 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6237 aarch64_expand_mov_immediate.
6238 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6239 CNT[BHSD] immediates.
6240 (movti): Split CONST_POLY_INT moves into two halves.
6241 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6242 Split additions that need a temporary here if the destination
6243 is the stack pointer.
6244 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6245 (*add<mode>3_poly_1): New instruction.
6246 (set_clobber_cc): New expander.
6248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6250 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6251 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6252 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6253 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6254 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6255 Change innermode from fixed_mode_size to machine_mode.
6256 (simplify_subreg): Update call accordingly. Handle a constant-sized
6257 subreg of a variable-length CONST_VECTOR.
6259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6260 Alan Hayward <alan.hayward@arm.com>
6261 David Sherwood <david.sherwood@arm.com>
6263 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6264 (add_offset_to_base): New function, split out from...
6265 (create_mem_ref): ...here. When handling a scale other than 1,
6266 check first whether the address is valid without the offset.
6267 Add it into the base if so, leaving the index and scale as-is.
6269 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6272 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6273 fold_for_warn before checking if arg2 is INTEGER_CST.
6275 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6277 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6278 (store_multiple_operation): Delete.
6279 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6280 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6281 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6282 guarded by TARGET_STRING.
6283 (rs6000_output_load_multiple): Delete.
6284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6285 OPTION_MASK_STRING / TARGET_STRING handling.
6286 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6287 (const rs6000_opt_masks) <"string">: Change mask to 0.
6288 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6289 (MASK_STRING): Delete.
6290 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6292 (load_multiple): Delete.
6299 (store_multiple): Delete.
6306 (movmemsi_8reg): Delete.
6307 (corresponding unnamed define_insn): Delete.
6308 (movmemsi_6reg): Delete.
6309 (corresponding unnamed define_insn): Delete.
6310 (movmemsi_4reg): Delete.
6311 (corresponding unnamed define_insn): Delete.
6312 (movmemsi_2reg): Delete.
6313 (corresponding unnamed define_insn): Delete.
6314 (movmemsi_1reg): Delete.
6315 (corresponding unnamed define_insn): Delete.
6316 * config/rs6000/rs6000.opt (mno-string): New.
6317 (mstring): Replace by deprecation warning stub.
6318 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6320 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6322 * regrename.c (regrename_do_replace): If replacing the same
6323 reg multiple times, try to reuse last created gen_raw_REG.
6326 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6327 main to workaround a bug in GDB.
6329 2018-01-12 Tom de Vries <tom@codesourcery.com>
6332 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6334 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6336 PR rtl-optimization/80481
6337 * ira-color.c (get_cap_member): New function.
6338 (allocnos_conflict_by_live_ranges_p): Use it.
6339 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6340 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6342 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6345 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6346 (*saddl_se_1): Ditto.
6348 (*ssubl_se_1): Ditto.
6350 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6352 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6353 rather than wi::to_widest for DR_INITs.
6354 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6355 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6356 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6358 (vect_analyze_group_access_1): Note that here.
6360 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6362 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6363 polynomial type sizes.
6365 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6367 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6368 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6369 (gimple_add_tmp_var): Likewise.
6371 2018-01-12 Martin Liska <mliska@suse.cz>
6373 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6374 (gimple_alloc_sizes): Likewise.
6375 (dump_gimple_statistics): Use PRIu64 in printf format.
6376 * gimple.h: Change uint64_t to int.
6378 2018-01-12 Martin Liska <mliska@suse.cz>
6380 * tree-core.h: Use uint64_t instead of int.
6381 * tree.c (tree_node_counts): Likewise.
6382 (tree_node_sizes): Likewise.
6383 (dump_tree_statistics): Use PRIu64 in printf format.
6385 2018-01-12 Martin Liska <mliska@suse.cz>
6387 * Makefile.in: As qsort_chk is implemented in vec.c, add
6388 vec.o to linkage of gencfn-macros.
6389 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6390 passing the info to record_node_allocation_statistics.
6391 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6393 * ggc-common.c (struct ggc_usage): Add operator== and use
6394 it in operator< and compare function.
6395 * mem-stats.h (struct mem_usage): Likewise.
6396 * vec.c (struct vec_usage): Remove operator< and compare
6397 function. Can be simply inherited.
6399 2018-01-12 Martin Jambor <mjambor@suse.cz>
6402 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6403 * tree-ssa-math-opts.c: Include domwalk.h.
6404 (convert_mult_to_fma_1): New function.
6405 (fma_transformation_info): New type.
6406 (fma_deferring_state): Likewise.
6407 (cancel_fma_deferring): New function.
6408 (result_of_phi): Likewise.
6409 (last_fma_candidate_feeds_initial_phi): Likewise.
6410 (convert_mult_to_fma): Added deferring logic, split actual
6411 transformation to convert_mult_to_fma_1.
6412 (math_opts_dom_walker): New type.
6413 (math_opts_dom_walker::after_dom_children): New method, body moved
6414 here from pass_optimize_widening_mul::execute, added deferring logic
6416 (pass_optimize_widening_mul::execute): Moved most of code to
6417 math_opts_dom_walker::after_dom_children.
6418 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6419 * config/i386/i386.c (ix86_option_override_internal): Added
6420 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6422 2018-01-12 Richard Biener <rguenther@suse.de>
6425 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6426 inline instance vars.
6428 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6431 * config/rx/rx.c (rx_is_restricted_memory_address):
6434 2018-01-12 Richard Biener <rguenther@suse.de>
6436 PR tree-optimization/80846
6437 * target.def (split_reduction): New target hook.
6438 * targhooks.c (default_split_reduction): New function.
6439 * targhooks.h (default_split_reduction): Declare.
6440 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6441 target requests first reduce vectors by combining low and high
6443 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6444 (get_vectype_for_scalar_type_and_size): Export.
6445 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6446 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6447 * doc/tm.texi: Regenerate.
6448 * config/i386/i386.c (ix86_split_reduction): Implement
6449 TARGET_VECTORIZE_SPLIT_REDUCTION.
6451 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6454 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6455 in PIC mode except for TARGET_VXWORKS_RTP.
6456 * config/sparc/sparc.c: Include cfgrtl.h.
6457 (TARGET_INIT_PIC_REG): Define.
6458 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6459 (sparc_pic_register_p): New predicate.
6460 (sparc_legitimate_address_p): Use it.
6461 (sparc_legitimize_pic_address): Likewise.
6462 (sparc_delegitimize_address): Likewise.
6463 (sparc_mode_dependent_address_p): Likewise.
6464 (gen_load_pcrel_sym): Remove 4th parameter.
6465 (load_got_register): Adjust call to above. Remove obsolete stuff.
6466 (sparc_expand_prologue): Do not call load_got_register here.
6467 (sparc_flat_expand_prologue): Likewise.
6468 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6469 (sparc_use_pseudo_pic_reg): New function.
6470 (sparc_init_pic_reg): Likewise.
6471 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6472 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6474 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6476 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6477 Add item for branch_cost.
6479 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6481 PR rtl-optimization/83565
6482 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6483 not extend the result to a larger mode for rotate operations.
6484 (num_sign_bit_copies1): Likewise.
6486 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6489 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6491 Use values-Xc.o for -pedantic.
6492 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6494 2018-01-12 Martin Liska <mliska@suse.cz>
6497 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6499 (possible_polymorphic_call_targets): Use it.
6500 (ipa_devirt): Likewise.
6502 2018-01-12 Martin Liska <mliska@suse.cz>
6504 * profile-count.h (enum profile_quality): Use 0 as invalid
6505 enum value of profile_quality.
6507 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6509 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6510 -mext-string options.
6512 2018-01-12 Richard Biener <rguenther@suse.de>
6514 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6515 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6516 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6518 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6520 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6522 * configure.ac (--with-long-double-format): Add support for the
6523 configuration option to change the default long double format on
6525 * config.gcc (powerpc*-linux*-*): Likewise.
6526 * configure: Regenerate.
6527 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6528 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6529 used without modification.
6531 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6533 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6534 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6535 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6536 MISC_BUILTIN_SPEC_BARRIER.
6537 (rs6000_init_builtins): Likewise.
6538 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6540 (speculation_barrier): New define_insn.
6541 * doc/extend.texi: Document __builtin_speculation_barrier.
6543 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6546 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6547 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6548 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6550 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6551 integral modes instead of "ss" and "sd".
6552 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6553 vectors with 32-bit and 64-bit elements.
6554 (vecdupssescalarmodesuffix): New mode attribute.
6555 (vec_dup<mode>): Use it.
6557 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6560 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6561 frame if argument is passed on stack.
6563 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6566 * ree.c (combine_reaching_defs): Optimize also
6567 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6568 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6570 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6573 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6575 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6578 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6579 after they are computed.
6581 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6583 PR tree-optimization/83695
6584 * gimple-loop-linterchange.cc
6585 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6586 reset cached scev information after interchange.
6587 (pass_linterchange::execute): Remove call to scev_reset_htab.
6589 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6591 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6592 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6593 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6594 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6595 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6596 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6597 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6598 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6599 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6600 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6601 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6602 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6603 (V_lane_reg): Likewise.
6604 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6606 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6607 (vfmal_lane_low<mode>_intrinsic,
6608 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6609 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6610 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6611 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6612 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6613 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6615 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6617 * config/arm/arm-cpus.in (fp16fml): New feature.
6618 (ALL_SIMD): Add fp16fml.
6619 (armv8.2-a): Add fp16fml as an option.
6620 (armv8.3-a): Likewise.
6621 (armv8.4-a): Add fp16fml as part of fp16.
6622 * config/arm/arm.h (TARGET_FP16FML): Define.
6623 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6625 * config/arm/arm-modes.def (V2HF): Define.
6626 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6627 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6628 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6629 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6630 vfmsl_low, vfmsl_high): New set of builtins.
6631 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6632 (vfml_op): New code attribute.
6633 (VFMLHALVES): New int iterator.
6634 (VFML, VFMLSEL): New mode attributes.
6635 (V_reg): Define mapping for V2HF.
6636 (V_hi, V_lo): New mode attributes.
6637 (VF_constraint): Likewise.
6638 (vfml_half, vfml_half_selector): New int attributes.
6639 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6641 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6642 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6644 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6645 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6646 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6647 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6649 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6650 Document new effective target and option set.
6652 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6654 * config/arm/arm-cpus.in (armv8_4): New feature.
6655 (ARMv8_4a): New fgroup.
6656 (armv8.4-a): New arch.
6657 * config/arm/arm-tables.opt: Regenerate.
6658 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6659 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6660 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6661 Add matching rules for -march=armv8.4-a and extensions.
6662 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6664 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6667 * config/rx/rx.md (BW): New mode attribute.
6668 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6670 2018-01-11 Richard Biener <rguenther@suse.de>
6672 PR tree-optimization/83435
6673 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6674 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6675 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6677 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6678 Alan Hayward <alan.hayward@arm.com>
6679 David Sherwood <david.sherwood@arm.com>
6681 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6683 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6684 (aarch64_print_address_internal): Use it to check for a zero offset.
6686 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6687 Alan Hayward <alan.hayward@arm.com>
6688 David Sherwood <david.sherwood@arm.com>
6690 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6691 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6692 Return a poly_int64 rather than a HOST_WIDE_INT.
6693 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6694 rather than a HOST_WIDE_INT.
6695 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6696 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6697 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6698 final_offset from HOST_WIDE_INT to poly_int64.
6699 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6700 to_constant when getting the number of units in an Advanced SIMD
6702 (aarch64_builtin_vectorized_function): Check for a constant number
6704 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6706 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6707 attribute instead of GET_MODE_NUNITS.
6708 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6709 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6710 GET_MODE_SIZE for fixed-size registers.
6711 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6712 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6713 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6714 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6715 (aarch64_print_operand, aarch64_print_address_internal)
6716 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6717 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6718 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6719 Handle polynomial GET_MODE_SIZE.
6720 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6721 wider than SImode without modification.
6722 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6723 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6724 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6725 passing and returning SVE modes.
6726 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6727 rather than GEN_INT.
6728 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6729 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6730 (aarch64_allocate_and_probe_stack_space): Likewise.
6731 (aarch64_layout_frame): Cope with polynomial offsets.
6732 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6733 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6735 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6736 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6737 poly_int64 rather than a HOST_WIDE_INT.
6738 (aarch64_get_separate_components, aarch64_process_components)
6739 (aarch64_expand_prologue, aarch64_expand_epilogue)
6740 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6741 (aarch64_anchor_offset): New function, split out from...
6742 (aarch64_legitimize_address): ...here.
6743 (aarch64_builtin_vectorization_cost): Handle polynomial
6744 TYPE_VECTOR_SUBPARTS.
6745 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6747 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6748 number of elements from the PARALLEL rather than the mode.
6749 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6750 rather than GET_MODE_BITSIZE.
6751 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6752 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6753 (aarch64_expand_vec_perm_const_1): Handle polynomial
6754 d->perm.length () and d->perm elements.
6755 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6756 Apply to_constant to d->perm elements.
6757 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6758 polynomial CONST_VECTOR_NUNITS.
6759 (aarch64_move_pointer): Take amount as a poly_int64 rather
6761 (aarch64_progress_pointer): Avoid temporary variable.
6762 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6763 the mode attribute instead of GET_MODE.
6765 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6766 Alan Hayward <alan.hayward@arm.com>
6767 David Sherwood <david.sherwood@arm.com>
6769 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6770 x exists before using it.
6771 (aarch64_add_constant_internal): Rename to...
6772 (aarch64_add_offset_1): ...this. Replace regnum with separate
6773 src and dest rtxes. Handle the case in which they're different,
6774 including when the offset is zero. Replace scratchreg with an rtx.
6775 Use 2 additions if there is no spare register into which we can
6776 move a 16-bit constant.
6777 (aarch64_add_constant): Delete.
6778 (aarch64_add_offset): Replace reg with separate src and dest
6779 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6780 Use aarch64_add_offset_1.
6781 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6782 an rtx rather than an int. Take the delta as a poly_int64
6783 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6784 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6785 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6786 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6787 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6789 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6790 aarch64_add_constant.
6792 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6794 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6795 Use scalar_float_mode.
6797 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6799 * config/aarch64/aarch64-simd.md
6800 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6801 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6802 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6803 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6804 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6805 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6806 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6807 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6808 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6809 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6811 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6814 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6815 targ_options->x_arm_arch_string is non NULL.
6817 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6819 * config/aarch64/aarch64.h
6820 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6822 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6825 * expmed.c (emit_store_flag_force): Swap if const op0
6826 and change VOIDmode to mode of op0.
6828 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6830 PR rtl-optimization/83761
6831 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6832 than bytes to mode_for_size.
6834 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6837 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6838 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6841 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6844 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6845 when in layout mode.
6846 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6847 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6850 2018-01-10 Michael Collison <michael.collison@arm.com>
6852 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6853 * config/aarch64/aarch64-option-extension.def: Add
6854 AARCH64_OPT_EXTENSION of 'fp16fml'.
6855 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6856 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6857 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6858 * config/aarch64/constraints.md (Ui7): New constraint.
6859 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6860 (VFMLA_SEL_W): Ditto.
6863 (VFMLA16_LOW): New int iterator.
6864 (VFMLA16_HIGH): Ditto.
6865 (UNSPEC_FMLAL): New unspec.
6866 (UNSPEC_FMLSL): Ditto.
6867 (UNSPEC_FMLAL2): Ditto.
6868 (UNSPEC_FMLSL2): Ditto.
6869 (f16mac): New code attribute.
6870 * config/aarch64/aarch64-simd-builtins.def
6871 (aarch64_fmlal_lowv2sf): Ditto.
6872 (aarch64_fmlsl_lowv2sf): Ditto.
6873 (aarch64_fmlalq_lowv4sf): Ditto.
6874 (aarch64_fmlslq_lowv4sf): Ditto.
6875 (aarch64_fmlal_highv2sf): Ditto.
6876 (aarch64_fmlsl_highv2sf): Ditto.
6877 (aarch64_fmlalq_highv4sf): Ditto.
6878 (aarch64_fmlslq_highv4sf): Ditto.
6879 (aarch64_fmlal_lane_lowv2sf): Ditto.
6880 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6881 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6882 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6883 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6884 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6885 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6886 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6887 (aarch64_fmlal_lane_highv2sf): Ditto.
6888 (aarch64_fmlsl_lane_highv2sf): Ditto.
6889 (aarch64_fmlal_laneq_highv2sf): Ditto.
6890 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6891 (aarch64_fmlalq_lane_highv4sf): Ditto.
6892 (aarch64_fmlsl_lane_highv4sf): Ditto.
6893 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6894 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6895 * config/aarch64/aarch64-simd.md:
6896 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6897 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6898 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6899 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6900 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6901 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6902 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6903 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6904 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6905 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6906 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6907 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6908 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6909 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6910 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6911 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6912 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6913 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6914 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6915 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6916 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6917 (vfmlsl_low_u32): Ditto.
6918 (vfmlalq_low_u32): Ditto.
6919 (vfmlslq_low_u32): Ditto.
6920 (vfmlal_high_u32): Ditto.
6921 (vfmlsl_high_u32): Ditto.
6922 (vfmlalq_high_u32): Ditto.
6923 (vfmlslq_high_u32): Ditto.
6924 (vfmlal_lane_low_u32): Ditto.
6925 (vfmlsl_lane_low_u32): Ditto.
6926 (vfmlal_laneq_low_u32): Ditto.
6927 (vfmlsl_laneq_low_u32): Ditto.
6928 (vfmlalq_lane_low_u32): Ditto.
6929 (vfmlslq_lane_low_u32): Ditto.
6930 (vfmlalq_laneq_low_u32): Ditto.
6931 (vfmlslq_laneq_low_u32): Ditto.
6932 (vfmlal_lane_high_u32): Ditto.
6933 (vfmlsl_lane_high_u32): Ditto.
6934 (vfmlal_laneq_high_u32): Ditto.
6935 (vfmlsl_laneq_high_u32): Ditto.
6936 (vfmlalq_lane_high_u32): Ditto.
6937 (vfmlslq_lane_high_u32): Ditto.
6938 (vfmlalq_laneq_high_u32): Ditto.
6939 (vfmlslq_laneq_high_u32): Ditto.
6940 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6941 (AARCH64_FL_FOR_ARCH8_4): New.
6942 (AARCH64_ISA_F16FML): New ISA flag.
6943 (TARGET_F16FML): New feature flag for fp16fml.
6944 (doc/invoke.texi): Document new fp16fml option.
6946 2018-01-10 Michael Collison <michael.collison@arm.com>
6948 * config/aarch64/aarch64-builtins.c:
6949 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6950 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6951 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6952 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6953 (AARCH64_ISA_SHA3): New ISA flag.
6954 (TARGET_SHA3): New feature flag for sha3.
6955 * config/aarch64/iterators.md (sha512_op): New int attribute.
6956 (CRYPTO_SHA512): New int iterator.
6957 (UNSPEC_SHA512H): New unspec.
6958 (UNSPEC_SHA512H2): Ditto.
6959 (UNSPEC_SHA512SU0): Ditto.
6960 (UNSPEC_SHA512SU1): Ditto.
6961 * config/aarch64/aarch64-simd-builtins.def
6962 (aarch64_crypto_sha512hqv2di): New builtin.
6963 (aarch64_crypto_sha512h2qv2di): Ditto.
6964 (aarch64_crypto_sha512su0qv2di): Ditto.
6965 (aarch64_crypto_sha512su1qv2di): Ditto.
6966 (aarch64_eor3qv8hi): Ditto.
6967 (aarch64_rax1qv2di): Ditto.
6968 (aarch64_xarqv2di): Ditto.
6969 (aarch64_bcaxqv8hi): Ditto.
6970 * config/aarch64/aarch64-simd.md:
6971 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6972 (aarch64_crypto_sha512su0qv2di): Ditto.
6973 (aarch64_crypto_sha512su1qv2di): Ditto.
6974 (aarch64_eor3qv8hi): Ditto.
6975 (aarch64_rax1qv2di): Ditto.
6976 (aarch64_xarqv2di): Ditto.
6977 (aarch64_bcaxqv8hi): Ditto.
6978 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6979 (vsha512h2q_u64): Ditto.
6980 (vsha512su0q_u64): Ditto.
6981 (vsha512su1q_u64): Ditto.
6982 (veor3q_u16): Ditto.
6983 (vrax1q_u64): Ditto.
6985 (vbcaxq_u16): Ditto.
6986 * config/arm/types.md (crypto_sha512): New type attribute.
6987 (crypto_sha3): Ditto.
6988 (doc/invoke.texi): Document new sha3 option.
6990 2018-01-10 Michael Collison <michael.collison@arm.com>
6992 * config/aarch64/aarch64-builtins.c:
6993 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6994 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6995 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6996 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6997 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6998 (AARCH64_ISA_SM4): New ISA flag.
6999 (TARGET_SM4): New feature flag for sm4.
7000 * config/aarch64/aarch64-simd-builtins.def
7001 (aarch64_sm3ss1qv4si): Ditto.
7002 (aarch64_sm3tt1aq4si): Ditto.
7003 (aarch64_sm3tt1bq4si): Ditto.
7004 (aarch64_sm3tt2aq4si): Ditto.
7005 (aarch64_sm3tt2bq4si): Ditto.
7006 (aarch64_sm3partw1qv4si): Ditto.
7007 (aarch64_sm3partw2qv4si): Ditto.
7008 (aarch64_sm4eqv4si): Ditto.
7009 (aarch64_sm4ekeyqv4si): Ditto.
7010 * config/aarch64/aarch64-simd.md:
7011 (aarch64_sm3ss1qv4si): Ditto.
7012 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7013 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7014 (aarch64_sm4eqv4si): Ditto.
7015 (aarch64_sm4ekeyqv4si): Ditto.
7016 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7017 (sm3part_op): Ditto.
7018 (CRYPTO_SM3TT): Ditto.
7019 (CRYPTO_SM3PART): Ditto.
7020 (UNSPEC_SM3SS1): New unspec.
7021 (UNSPEC_SM3TT1A): Ditto.
7022 (UNSPEC_SM3TT1B): Ditto.
7023 (UNSPEC_SM3TT2A): Ditto.
7024 (UNSPEC_SM3TT2B): Ditto.
7025 (UNSPEC_SM3PARTW1): Ditto.
7026 (UNSPEC_SM3PARTW2): Ditto.
7027 (UNSPEC_SM4E): Ditto.
7028 (UNSPEC_SM4EKEY): Ditto.
7029 * config/aarch64/constraints.md (Ui2): New constraint.
7030 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7031 * config/arm/types.md (crypto_sm3): New type attribute.
7032 (crypto_sm4): Ditto.
7033 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7034 (vsm3tt1aq_u32): Ditto.
7035 (vsm3tt1bq_u32): Ditto.
7036 (vsm3tt2aq_u32): Ditto.
7037 (vsm3tt2bq_u32): Ditto.
7038 (vsm3partw1q_u32): Ditto.
7039 (vsm3partw2q_u32): Ditto.
7040 (vsm4eq_u32): Ditto.
7041 (vsm4ekeyq_u32): Ditto.
7042 (doc/invoke.texi): Document new sm4 option.
7044 2018-01-10 Michael Collison <michael.collison@arm.com>
7046 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7047 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7048 (AARCH64_FL_FOR_ARCH8_4): New.
7049 (AARCH64_FL_V8_4): New flag.
7050 (doc/invoke.texi): Document new armv8.4-a option.
7052 2018-01-10 Michael Collison <michael.collison@arm.com>
7054 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7055 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7056 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7057 * config/aarch64/aarch64-option-extension.def: Add
7058 AARCH64_OPT_EXTENSION of 'sha2'.
7059 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7060 (crypto): Disable sha2 and aes if crypto disabled.
7061 (crypto): Enable aes and sha2 if enabled.
7062 (simd): Disable sha2 and aes if simd disabled.
7063 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7065 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7066 (TARGET_SHA2): New feature flag for sha2.
7067 (TARGET_AES): New feature flag for aes.
7068 * config/aarch64/aarch64-simd.md:
7069 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7070 conditional on TARGET_AES.
7071 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7072 (aarch64_crypto_sha1hsi): Make pattern conditional
7074 (aarch64_crypto_sha1hv4si): Ditto.
7075 (aarch64_be_crypto_sha1hv4si): Ditto.
7076 (aarch64_crypto_sha1su1v4si): Ditto.
7077 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7078 (aarch64_crypto_sha1su0v4si): Ditto.
7079 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7080 (aarch64_crypto_sha256su0v4si): Ditto.
7081 (aarch64_crypto_sha256su1v4si): Ditto.
7082 (doc/invoke.texi): Document new aes and sha2 options.
7084 2018-01-10 Martin Sebor <msebor@redhat.com>
7086 PR tree-optimization/83781
7087 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7090 2018-01-11 Martin Sebor <msebor@gmail.com>
7091 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7093 PR tree-optimization/83501
7094 PR tree-optimization/81703
7096 * tree-ssa-strlen.c (get_string_cst): Rename...
7097 (get_string_len): ...to this. Handle global constants.
7098 (handle_char_store): Adjust.
7100 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7101 Jim Wilson <jimw@sifive.com>
7103 * config/riscv/riscv-protos.h (riscv_output_return): New.
7104 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7105 (riscv_attribute_table, riscv_output_return),
7106 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7107 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7108 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7110 (riscv_expand_prologue): Add early return for naked function.
7111 (riscv_expand_epilogue): Likewise.
7112 (riscv_function_ok_for_sibcall): Return false for naked function.
7113 (riscv_set_current_function): New.
7114 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7115 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7116 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7117 * doc/extend.texi (RISC-V Function Attributes): New.
7119 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7121 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7122 check for 128-bit long double before checking TCmode.
7123 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7124 128-bit long doubles before checking TFmode or TCmode.
7125 (FLOAT128_IBM_P): Likewise.
7127 2018-01-10 Martin Sebor <msebor@redhat.com>
7129 PR tree-optimization/83671
7130 * builtins.c (c_strlen): Unconditionally return zero for the empty
7132 Use -Warray-bounds for warnings.
7133 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7134 for non-constant array indices with COMPONENT_REF, arrays of
7135 arrays, and pointers to arrays.
7136 (gimple_fold_builtin_strlen): Determine and set length range for
7137 non-constant character arrays.
7139 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7142 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7145 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7147 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7149 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7152 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7153 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7154 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7155 indexed_or_indirect_operand predicate.
7156 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7157 (*vsx_le_perm_load_v8hi): Likewise.
7158 (*vsx_le_perm_load_v16qi): Likewise.
7159 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7160 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7161 (*vsx_le_perm_store_v8hi): Likewise.
7162 (*vsx_le_perm_store_v16qi): Likewise.
7163 (eight unnamed splitters): Likewise.
7165 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7167 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7168 * config/rs6000/emmintrin.h: Likewise.
7169 * config/rs6000/mmintrin.h: Likewise.
7170 * config/rs6000/xmmintrin.h: Likewise.
7172 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7175 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7177 * tree.c (tree_nop_conversion): Return true for location wrapper
7179 (maybe_wrap_with_location): New function.
7180 (selftest::check_strip_nops): New function.
7181 (selftest::test_location_wrappers): New function.
7182 (selftest::tree_c_tests): Call it.
7183 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7184 (maybe_wrap_with_location): New decl.
7185 (EXPR_LOCATION_WRAPPER_P): New macro.
7186 (location_wrapper_p): New inline function.
7187 (tree_strip_any_location_wrapper): New inline function.
7189 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7192 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7193 stack_realign_offset for the largest alignment of stack slot
7195 (ix86_find_max_used_stack_alignment): New function.
7196 (ix86_finalize_stack_frame_flags): Use it. Set
7197 max_used_stack_alignment if we don't realign stack.
7198 * config/i386/i386.h (machine_function): Add
7199 max_used_stack_alignment.
7201 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7203 * config/arm/arm.opt (-mbranch-cost): New option.
7204 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7207 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7210 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7211 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7213 2018-01-10 Richard Biener <rguenther@suse.de>
7216 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7217 early out so it also covers the case where we have a non-NULL
7220 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7222 PR tree-optimization/83753
7223 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7224 for non-strided grouped accesses if the number of elements is 1.
7226 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7229 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7230 * i386.h (TARGET_USE_GATHER): Define.
7231 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7233 2018-01-10 Martin Liska <mliska@suse.cz>
7236 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7237 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7239 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7240 CLEANUP_NO_PARTITIONING is not set.
7242 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7244 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7245 for vectors, as a partial revert of r254296.
7246 * rtl.h (const_vec_p): Delete.
7247 (const_vec_duplicate_p): Don't test for vector CONSTs.
7248 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7249 * expmed.c (make_tree): Likewise.
7252 * common.md (E, F): Use CONSTANT_P instead of checking for
7254 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7255 checking for CONST_VECTOR.
7257 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7260 * predict.c (force_edge_cold): Handle in more sane way edges
7263 2018-01-09 Carl Love <cel@us.ibm.com>
7265 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7267 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7268 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7269 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7270 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7271 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7272 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7273 * config/rs6000/rs6000-protos.h: Add extern defition for
7274 rs6000_generate_float2_double_code.
7275 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7277 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7278 (float2_v2df): Add define_expand.
7280 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7283 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7284 op_mode in the force_to_mode call.
7286 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7288 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7289 instead of checking each element individually.
7290 (aarch64_evpc_uzp): Likewise.
7291 (aarch64_evpc_zip): Likewise.
7292 (aarch64_evpc_ext): Likewise.
7293 (aarch64_evpc_rev): Likewise.
7294 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7295 instead of checking each element individually. Return true without
7297 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7298 whether all selected elements come from the same input, instead of
7299 checking each element individually. Remove calls to gen_rtx_REG,
7300 start_sequence and end_sequence and instead assert that no rtl is
7303 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7305 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7306 order of HIGH and CONST checks.
7308 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7310 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7311 if the destination isn't an SSA_NAME.
7313 2018-01-09 Richard Biener <rguenther@suse.de>
7315 PR tree-optimization/83668
7316 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7318 (canonicalize_loop_form): ... here, renamed from ...
7319 (canonicalize_loop_closed_ssa_form): ... this and amended to
7320 swap successor edges for loop exit blocks to make us use
7321 the RPO order we need for initial schedule generation.
7323 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7325 PR tree-optimization/64811
7326 * match.pd: When optimizing comparisons with Inf, avoid
7327 introducing or losing exceptions from comparisons with NaN.
7329 2018-01-09 Martin Liska <mliska@suse.cz>
7332 * asan.c (shadow_mem_size): Add gcc_assert.
7334 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7336 Don't save registers in main().
7339 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7340 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7341 * config/avr/avr.c (avr_set_current_function): Don't error if
7342 naked, OS_task or OS_main are specified at the same time.
7343 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7345 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7347 * common/config/avr/avr-common.c (avr_option_optimization_table):
7348 Switch on -mmain-is-OS_task for optimizing compilations.
7350 2018-01-09 Richard Biener <rguenther@suse.de>
7352 PR tree-optimization/83572
7353 * graphite.c: Include cfganal.h.
7354 (graphite_transform_loops): Connect infinite loops to exit
7355 and remove fake edges at the end.
7357 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7359 * ipa-inline.c (edge_badness): Revert accidental checkin.
7361 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7364 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7365 symbols; not inline clones.
7367 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7370 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7371 hard registers. Formatting fixes.
7373 PR preprocessor/83722
7374 * gcc.c (try_generate_repro): Pass
7375 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7376 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7379 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7380 Kito Cheng <kito.cheng@gmail.com>
7382 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7383 (riscv_leaf_function_p): Delete.
7384 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7386 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7388 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7390 (do_ifelse): New function.
7391 (do_isel): New function.
7392 (do_sub3): New function.
7393 (do_add3): New function.
7394 (do_load_mask_compare): New function.
7395 (do_overlap_load_compare): New function.
7396 (expand_compare_loop): New function.
7397 (expand_block_compare): Call expand_compare_loop() when appropriate.
7398 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7400 (-mblock-compare-inline-loop-limit): New option.
7402 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7405 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7406 Reverse order of second and third operands in first alternative.
7407 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7408 of first and second elements in UNSPEC_VPERMR vector.
7409 (altivec_expand_vec_perm_le): Likewise.
7411 2018-01-08 Jeff Law <law@redhat.com>
7413 PR rtl-optimizatin/81308
7414 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7415 (process_switch): If group_case_labels makes a change, then set
7417 (pass_convert_switch::execute): If a switch is converted, then
7418 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7420 PR rtl-optimization/81308
7421 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7424 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7426 PR target/83663 - Revert r255946
7427 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7428 generation for cases where splatting a value is not useful.
7429 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7430 across a vec_duplicate and a paradoxical subreg forming a vector
7431 mode to a vec_concat.
7433 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7435 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7436 -march=armv8.3-a variants.
7437 * config/arm/t-multilib: Likewise.
7438 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7440 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7442 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7444 (cceq_ior_compare_complement): Give it a name so I can use it, and
7445 change boolean_or_operator predicate to boolean_operator so it can
7446 be used to generate a crand.
7447 (eqne): New code iterator.
7448 (bd/bd_neg): New code_attrs.
7449 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7450 a single define_insn.
7451 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7452 decrement (bdnzt/bdnzf/bdzt/bdzf).
7453 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7454 with the new names of the branch decrement patterns, and added the
7455 names of the branch decrement conditional patterns.
7457 2018-01-08 Richard Biener <rguenther@suse.de>
7459 PR tree-optimization/83563
7460 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7463 2018-01-08 Richard Biener <rguenther@suse.de>
7466 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7468 2018-01-08 Richard Biener <rguenther@suse.de>
7470 PR tree-optimization/83685
7471 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7472 references to abnormals.
7474 2018-01-08 Richard Biener <rguenther@suse.de>
7477 * dwarf2out.c (output_indirect_strings): Handle empty
7478 skeleton_debug_str_hash.
7479 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7481 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7483 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7484 (emit_store_direct): Likewise.
7485 (arc_trampoline_adjust_address): Likewise.
7486 (arc_asm_trampoline_template): New function.
7487 (arc_initialize_trampoline): Use asm_trampoline_template.
7488 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7489 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7490 * config/arc/arc.md (flush_icache): Delete pattern.
7492 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7494 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7495 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7498 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7501 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7502 by not USED_FOR_TARGET.
7503 (make_pass_resolve_sw_modes): Likewise.
7505 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7507 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7510 2018-01-08 Richard Biener <rguenther@suse.de>
7513 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7515 2018-01-08 Richard Biener <rguenther@suse.de>
7518 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7520 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7523 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7524 basic blocks with a small number of successors.
7525 (convert_control_dep_chain_into_preds): Improve handling of
7527 (dump_predicates): Split apart into...
7528 (dump_pred_chain): ...here...
7529 (dump_pred_info): ...and here.
7530 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7531 (can_chain_union_be_invalidated_p): Improve check for invalidation
7533 (uninit_uses_cannot_happen): Avoid unnecessary if
7534 convert_control_dep_chain_into_preds yielded nothing.
7536 2018-01-06 Martin Sebor <msebor@redhat.com>
7538 PR tree-optimization/83640
7539 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7540 subtracting negative offset from size.
7541 (builtin_access::overlap): Adjust offset bounds of the access to fall
7542 within the size of the object if possible.
7544 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7546 PR rtl-optimization/83699
7547 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7548 extract_bit_field_as_subreg to cases in which the extracted
7549 value is also a vector.
7551 * lra-constraints.c (process_alt_operands): Test for the equivalence
7552 substitutions when detecting a possible reload cycle.
7554 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7557 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7558 by default if flag_selective_schedling{,2}. Formatting fixes.
7560 PR rtl-optimization/83682
7561 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7562 if it has non-VECTOR_MODE element mode.
7563 (vec_duplicate_p): Likewise.
7566 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7567 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7569 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7572 * config/i386/i386-builtin.def
7573 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7574 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7575 Require also OPTION_MASK_ISA_AVX512F in addition to
7576 OPTION_MASK_ISA_GFNI.
7577 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7578 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7579 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7580 to OPTION_MASK_ISA_GFNI.
7581 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7582 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7583 OPTION_MASK_ISA_AVX512BW.
7584 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7585 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7586 addition to OPTION_MASK_ISA_GFNI.
7587 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7588 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7589 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7590 to OPTION_MASK_ISA_GFNI.
7591 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7592 a requirement for all ISAs rather than any of them with a few
7594 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7596 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7597 bitmasks to be enabled with 3 exceptions, instead of requiring any
7598 enabled ISA with lots of exceptions.
7599 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7600 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7601 Change avx512bw in isa attribute to avx512f.
7602 * config/i386/sgxintrin.h: Add license boilerplate.
7603 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7604 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7605 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7606 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7608 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7609 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7610 temporarily sse2 rather than sse if not enabled already.
7613 * config/i386/sse.md (VI248_VLBW): Rename to ...
7614 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7615 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7616 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7617 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7618 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7619 mode iterator instead of VI248_VLBW.
7621 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7623 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7624 (record_modified): Skip clobbers; add debug output.
7625 (param_change_prob): Use sreal frequencies.
7627 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7629 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7630 punt for user-aligned variables.
7632 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7634 * tree-chrec.c (chrec_contains_symbols): Return true for
7637 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7640 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7641 of (x|y) == x for BICS pattern.
7643 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7645 PR tree-optimization/83605
7646 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7647 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7650 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7652 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7653 * config/epiphany/rtems.h: New file.
7655 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7656 Uros Bizjak <ubizjak@gmail.com>
7659 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7660 QIreg_operand instead of register_operand predicate.
7661 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7662 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7663 comments instead of -fmitigate[-_]rop.
7665 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7668 * cgraphunit.c (symbol_table::compile): Switch to text_section
7669 before calling assembly_start debug hook.
7670 * run-rtl-passes.c (run_rtl_passes): Likewise.
7673 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7675 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7676 range_int_cst_p rather than !symbolic_range_p before calling
7677 extract_range_from_multiplicative_op_1.
7679 2018-01-04 Jeff Law <law@redhat.com>
7681 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7682 redundant test in assertion.
7684 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7686 * doc/rtl.texi: Document machine_mode wrapper classes.
7688 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7690 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7693 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7695 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7696 the VEC_PERM_EXPR fold to fail.
7698 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7701 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7702 to switched_sections.
7704 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7707 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7710 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7713 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7714 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7716 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7719 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7720 is BLKmode and bitpos not zero or mode change is needed.
7722 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7725 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7728 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7731 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7732 instead of MULT rtx. Update all corresponding splitters.
7734 (*ssub<modesuffix>): Ditto.
7736 (*cmp_sadd_di): Update split patterns.
7737 (*cmp_sadd_si): Ditto.
7738 (*cmp_sadd_sidi): Ditto.
7739 (*cmp_ssub_di): Ditto.
7740 (*cmp_ssub_si): Ditto.
7741 (*cmp_ssub_sidi): Ditto.
7742 * config/alpha/predicates.md (const23_operand): New predicate.
7743 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7744 Look for ASHIFT, not MULT inner operand.
7745 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7747 2018-01-04 Martin Liska <mliska@suse.cz>
7749 PR gcov-profile/83669
7750 * gcov.c (output_intermediate_file): Add version to intermediate
7752 * doc/gcov.texi: Document new field 'version' in intermediate
7753 file format. Fix location of '-k' option of gcov command.
7755 2018-01-04 Martin Liska <mliska@suse.cz>
7758 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7760 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7762 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7764 2018-01-03 Martin Sebor <msebor@redhat.com>
7766 PR tree-optimization/83655
7767 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7768 checking calls with invalid arguments.
7770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7772 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7773 (vectorizable_mask_load_store): Delete.
7774 (vectorizable_call): Return false for masked loads and stores.
7775 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7776 instead of gimple_assign_rhs1.
7777 (vectorizable_load): Handle IFN_MASK_LOAD.
7778 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7782 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7784 (vectorizable_mask_load_store): ...here.
7785 (vectorizable_load): ...and here.
7787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7789 * tree-vect-stmts.c (vect_build_all_ones_mask)
7790 (vect_build_zero_merge_argument): New functions, split out from...
7791 (vectorizable_load): ...here.
7793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7795 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7797 (vectorizable_mask_load_store): ...here.
7798 (vectorizable_store): ...and here.
7800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7802 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7804 (vectorizable_mask_load_store): ...here.
7806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7808 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7809 (vect_model_store_cost): Take a vec_load_store_type instead of a
7811 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7812 (vect_model_store_cost): Take a vec_load_store_type instead of a
7814 (vectorizable_mask_load_store): Update accordingly.
7815 (vectorizable_store): Likewise.
7816 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7820 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7821 IFN_MASK_LOAD calls here rather than...
7822 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7825 Alan Hayward <alan.hayward@arm.com>
7826 David Sherwood <david.sherwood@arm.com>
7828 * expmed.c (extract_bit_field_1): For vector extracts,
7829 fall back to extract_bit_field_as_subreg if vec_extract
7832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7833 Alan Hayward <alan.hayward@arm.com>
7834 David Sherwood <david.sherwood@arm.com>
7836 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7837 they are variable or constant sized.
7838 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7839 slots for constant-sized data.
7841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7842 Alan Hayward <alan.hayward@arm.com>
7843 David Sherwood <david.sherwood@arm.com>
7845 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7846 handling COND_EXPRs with boolean comparisons, try to find a better
7847 basis for the mask type than the boolean itself.
7849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7851 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7852 is calculated and how it can be overridden.
7853 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7854 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7856 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7860 Alan Hayward <alan.hayward@arm.com>
7861 David Sherwood <david.sherwood@arm.com>
7863 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7864 Remove the mode argument.
7865 (aarch64_simd_valid_immediate): Remove the mode and inverse
7867 * config/aarch64/iterators.md (bitsize): New iterator.
7868 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7869 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7870 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7871 aarch64_simd_valid_immediate.
7872 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7873 (aarch64_reg_or_bic_imm): Likewise.
7874 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7875 with an insn_type enum and msl with a modifier_type enum.
7876 Replace element_width with a scalar_mode. Change the shift
7877 to unsigned int. Add constructors for scalar_float_mode and
7878 scalar_int_mode elements.
7879 (aarch64_vect_float_const_representable_p): Delete.
7880 (aarch64_can_const_movi_rtx_p)
7881 (aarch64_simd_scalar_immediate_valid_for_move)
7882 (aarch64_simd_make_constant): Update call to
7883 aarch64_simd_valid_immediate.
7884 (aarch64_advsimd_valid_immediate_hs): New function.
7885 (aarch64_advsimd_valid_immediate): Likewise.
7886 (aarch64_simd_valid_immediate): Remove mode and inverse
7887 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7888 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7889 and aarch64_float_const_representable_p on the result.
7890 (aarch64_output_simd_mov_immediate): Remove mode argument.
7891 Update call to aarch64_simd_valid_immediate and use of
7892 simd_immediate_info.
7893 (aarch64_output_scalar_simd_mov_immediate): Update call
7896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7897 Alan Hayward <alan.hayward@arm.com>
7898 David Sherwood <david.sherwood@arm.com>
7900 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7901 (mode_nunits): Likewise CONST_MODE_NUNITS.
7902 * machmode.def (ADJUST_NUNITS): Document.
7903 * genmodes.c (mode_data::need_nunits_adj): New field.
7904 (blank_mode): Update accordingly.
7905 (adj_nunits): New variable.
7906 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7908 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7909 listed in adj_nunits.
7910 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7911 listed in adj_nunits. Don't emit case statements for such modes.
7912 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7913 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7914 nothing if adj_nunits is nonnull.
7915 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7916 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7917 (emit_mode_fbit): Update use of print_maybe_const_decl.
7918 (emit_move_size): Likewise. Treat the array as non-const
7920 (emit_mode_adjustments): Handle adj_nunits.
7922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7924 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7925 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7926 (VECTOR_MODES): Use it.
7927 (make_vector_modes): Take the prefix as an argument.
7929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7930 Alan Hayward <alan.hayward@arm.com>
7931 David Sherwood <david.sherwood@arm.com>
7933 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7934 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7935 for MODE_VECTOR_BOOL.
7936 * machmode.def (VECTOR_BOOL_MODE): Document.
7937 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7938 (make_vector_bool_mode): New function.
7939 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7941 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7942 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7944 * stor-layout.c (int_mode_for_mode): Likewise.
7945 * tree.c (build_vector_type_for_mode): Likewise.
7946 * varasm.c (output_constant_pool_2): Likewise.
7947 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7948 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7949 for MODE_VECTOR_BOOL.
7950 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7951 of mode class checks.
7952 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7953 instead of a list of mode class checks.
7954 (expand_vector_scalar_condition): Likewise.
7955 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7958 Alan Hayward <alan.hayward@arm.com>
7959 David Sherwood <david.sherwood@arm.com>
7961 * machmode.h (mode_size): Change from unsigned short to
7963 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7964 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7965 or if measurement_type is not polynomial.
7966 (fixed_size_mode::includes_p): Check for constant-sized modes.
7967 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7968 return a poly_uint16 rather than an unsigned short.
7969 (emit_mode_size): Change the type of mode_size from unsigned short
7970 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7971 (emit_mode_adjustments): Cope with polynomial vector sizes.
7972 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7974 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7976 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7977 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7978 * caller-save.c (setup_save_areas): Likewise.
7979 (replace_reg_with_saved_mem): Likewise.
7980 * calls.c (emit_library_call_value_1): Likewise.
7981 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7982 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7983 (gen_lowpart_for_combine): Likewise.
7984 * convert.c (convert_to_integer_1): Likewise.
7985 * cse.c (equiv_constant, cse_insn): Likewise.
7986 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7987 (cselib_subst_to_values): Likewise.
7988 * dce.c (word_dce_process_block): Likewise.
7989 * df-problems.c (df_word_lr_mark_ref): Likewise.
7990 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7991 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7992 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7993 (rtl_for_decl_location): Likewise.
7994 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7995 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7996 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7997 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7998 (expand_expr_real_1): Likewise.
7999 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8000 (pad_below): Likewise.
8001 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8002 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8003 * ira.c (get_subreg_tracking_sizes): Likewise.
8004 * ira-build.c (ira_create_allocno_objects): Likewise.
8005 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8006 (ira_sort_regnos_for_alter_reg): Likewise.
8007 * ira-costs.c (record_operand_costs): Likewise.
8008 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8009 (resolve_simple_move): Likewise.
8010 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8011 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8012 (lra_constraints): Likewise.
8013 (CONST_POOL_OK_P): Reject variable-sized modes.
8014 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8015 (add_pseudo_to_slot, lra_spill): Likewise.
8016 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8017 * optabs-query.c (get_best_extraction_insn): Likewise.
8018 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8019 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8020 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8021 * recog.c (offsettable_address_addr_space_p): Likewise.
8022 * regcprop.c (maybe_mode_change): Likewise.
8023 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8024 * regrename.c (build_def_use): Likewise.
8025 * regstat.c (dump_reg_info): Likewise.
8026 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8027 (find_reloads, find_reloads_subreg_address): Likewise.
8028 * reload1.c (eliminate_regs_1): Likewise.
8029 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8030 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8031 (simplify_binary_operation_1, simplify_subreg): Likewise.
8032 * targhooks.c (default_function_arg_padding): Likewise.
8033 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8034 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8035 (verify_gimple_assign_ternary): Likewise.
8036 * tree-inline.c (estimate_move_cost): Likewise.
8037 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8038 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8039 (get_address_cost_ainc): Likewise.
8040 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8041 (vect_supportable_dr_alignment): Likewise.
8042 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8043 (vectorizable_reduction): Likewise.
8044 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8045 (vectorizable_operation, vectorizable_load): Likewise.
8046 * tree.c (build_same_sized_truth_vector_type): Likewise.
8047 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8048 * var-tracking.c (emit_note_insn_var_location): Likewise.
8049 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8050 (ADDR_VEC_ALIGN): Likewise.
8052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8053 Alan Hayward <alan.hayward@arm.com>
8054 David Sherwood <david.sherwood@arm.com>
8056 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8058 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8059 or if measurement_type is polynomial.
8060 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8061 * combine.c (make_extraction): Likewise.
8062 * dse.c (find_shift_sequence): Likewise.
8063 * dwarf2out.c (mem_loc_descriptor): Likewise.
8064 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8065 (extract_bit_field, extract_low_bits): Likewise.
8066 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8067 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8068 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8069 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8070 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8071 * reload.c (find_reloads): Likewise.
8072 * reload1.c (alter_reg): Likewise.
8073 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8074 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8075 * tree-if-conv.c (predicate_mem_writes): Likewise.
8076 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8077 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8078 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8079 * valtrack.c (dead_debug_insert_temp): Likewise.
8080 * varasm.c (mergeable_constant_section): Likewise.
8081 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8084 Alan Hayward <alan.hayward@arm.com>
8085 David Sherwood <david.sherwood@arm.com>
8087 * expr.c (expand_assignment): Cope with polynomial mode sizes
8088 when assigning to a CONCAT.
8090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8091 Alan Hayward <alan.hayward@arm.com>
8092 David Sherwood <david.sherwood@arm.com>
8094 * machmode.h (mode_precision): Change from unsigned short to
8096 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8098 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8099 or if measurement_type is not polynomial.
8100 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8101 in which the mode is already known to be a scalar_int_mode.
8102 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8103 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8105 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8106 for GET_MODE_PRECISION.
8107 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8108 for GET_MODE_PRECISION.
8109 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8111 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8112 (expand_field_assignment, make_extraction): Likewise.
8113 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8114 (get_last_value): Likewise.
8115 * convert.c (convert_to_integer_1): Likewise.
8116 * cse.c (cse_insn): Likewise.
8117 * expr.c (expand_expr_real_1): Likewise.
8118 * lra-constraints.c (simplify_operand_subreg): Likewise.
8119 * optabs-query.c (can_atomic_load_p): Likewise.
8120 * optabs.c (expand_atomic_load): Likewise.
8121 (expand_atomic_store): Likewise.
8122 * ree.c (combine_reaching_defs): Likewise.
8123 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8124 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8125 * tree.h (type_has_mode_precision_p): Likewise.
8126 * ubsan.c (instrument_si_overflow): Likewise.
8128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8129 Alan Hayward <alan.hayward@arm.com>
8130 David Sherwood <david.sherwood@arm.com>
8132 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8133 polynomial numbers of units.
8134 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8135 (valid_vector_subparts_p): New function.
8136 (build_vector_type): Remove temporary shim and take the number
8137 of units as a poly_uint64 rather than an int.
8138 (build_opaque_vector_type): Take the number of units as a
8139 poly_uint64 rather than an int.
8140 * tree.c (build_vector_from_ctor): Handle polynomial
8141 TYPE_VECTOR_SUBPARTS.
8142 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8143 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8144 (build_vector_from_val): If the number of units is variable,
8145 use build_vec_duplicate_cst for constant operands and
8146 VEC_DUPLICATE_EXPR otherwise.
8147 (make_vector_type): Remove temporary is_constant ().
8148 (build_vector_type, build_opaque_vector_type): Take the number of
8149 units as a poly_uint64 rather than an int.
8150 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8152 * cfgexpand.c (expand_debug_expr): Likewise.
8153 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8154 (store_constructor, expand_expr_real_1): Likewise.
8155 (const_scalar_mask_from_tree): Likewise.
8156 * fold-const-call.c (fold_const_reduction): Likewise.
8157 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8158 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8159 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8160 (fold_relational_const): Likewise.
8161 (native_interpret_vector): Likewise. Change the size from an
8162 int to an unsigned int.
8163 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8164 TYPE_VECTOR_SUBPARTS.
8165 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8166 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8167 duplicating a non-constant operand into a variable-length vector.
8168 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8169 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8170 * ipa-icf.c (sem_variable::equals): Likewise.
8171 * match.pd: Likewise.
8172 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8173 * print-tree.c (print_node): Likewise.
8174 * stor-layout.c (layout_type): Likewise.
8175 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8176 * tree-cfg.c (verify_gimple_comparison): Likewise.
8177 (verify_gimple_assign_binary): Likewise.
8178 (verify_gimple_assign_ternary): Likewise.
8179 (verify_gimple_assign_single): Likewise.
8180 * tree-pretty-print.c (dump_generic_node): Likewise.
8181 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8182 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8183 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8184 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8185 (vect_shift_permute_load_chain): Likewise.
8186 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8187 (expand_vector_condition, optimize_vector_constructor): Likewise.
8188 (lower_vec_perm, get_compute_type): Likewise.
8189 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8190 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8191 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8192 (vect_recog_mask_conversion_pattern): Likewise.
8193 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8194 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8195 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8196 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8197 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8198 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8199 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8200 (supportable_widening_operation): Likewise.
8201 (supportable_narrowing_operation): Likewise.
8202 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8204 * varasm.c (output_constant): Likewise.
8206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8207 Alan Hayward <alan.hayward@arm.com>
8208 David Sherwood <david.sherwood@arm.com>
8210 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8211 so that both the length == 3 and length != 3 cases set up their
8212 own permute vectors. Add comments explaining why we know the
8213 number of elements is constant.
8214 (vect_permute_load_chain): Likewise.
8216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8217 Alan Hayward <alan.hayward@arm.com>
8218 David Sherwood <david.sherwood@arm.com>
8220 * machmode.h (mode_nunits): Change from unsigned char to
8222 (ONLY_FIXED_SIZE_MODES): New macro.
8223 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8224 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8225 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8227 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8228 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8229 or if measurement_type is not polynomial.
8230 * genmodes.c (ZERO_COEFFS): New macro.
8231 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8233 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8234 Use ZERO_COEFFS when emitting initializers.
8235 * data-streamer.h (bp_pack_poly_value): New function.
8236 (bp_unpack_poly_value): Likewise.
8237 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8238 for GET_MODE_NUNITS.
8239 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8240 for GET_MODE_NUNITS.
8241 * tree.c (make_vector_type): Remove temporary shim and make
8242 the real function take the number of units as a poly_uint64
8244 (build_vector_type_for_mode): Handle polynomial nunits.
8245 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8246 * emit-rtl.c (const_vec_series_p_1): Likewise.
8247 (gen_rtx_CONST_VECTOR): Likewise.
8248 * fold-const.c (test_vec_duplicate_folding): Likewise.
8249 * genrecog.c (validate_pattern): Likewise.
8250 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8251 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8252 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8253 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8254 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8255 * rtlanal.c (subreg_get_info): Likewise.
8256 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8257 (vect_grouped_load_supported): Likewise.
8258 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8259 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8260 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8261 (simplify_const_unary_operation, simplify_binary_operation_1)
8262 (simplify_const_binary_operation, simplify_ternary_operation)
8263 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8264 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8265 instead of CONST_VECTOR_NUNITS.
8266 * varasm.c (output_constant_pool_2): Likewise.
8267 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8268 explicit-encoded elements in the XVEC for variable-length vectors.
8270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8272 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8275 Alan Hayward <alan.hayward@arm.com>
8276 David Sherwood <david.sherwood@arm.com>
8278 * coretypes.h (fixed_size_mode): Declare.
8279 (fixed_size_mode_pod): New typedef.
8280 * builtins.h (target_builtins::x_apply_args_mode)
8281 (target_builtins::x_apply_result_mode): Change type to
8282 fixed_size_mode_pod.
8283 * builtins.c (apply_args_size, apply_result_size, result_vector)
8284 (expand_builtin_apply_args_1, expand_builtin_apply)
8285 (expand_builtin_return): Update accordingly.
8287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8289 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8290 * cselib.c (cselib_hash_rtx): Likewise.
8291 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8292 CONST_VECTOR encoding.
8294 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8295 Jeff Law <law@redhat.com>
8298 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8299 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8300 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8301 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8304 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8305 explicitly probe *sp in a noreturn function if there were any callee
8306 register saves or frame pointer is needed.
8308 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8311 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8312 BLKmode for ternary, binary or unary expressions.
8315 * var-tracking.c (delete_vta_debug_insn): New inline function.
8316 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8317 insns from get_insns () to NULL instead of each bb separately.
8318 Use delete_vta_debug_insn. No longer static.
8319 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8320 delete_vta_debug_insns callers.
8321 * rtl.h (delete_vta_debug_insns): Declare.
8322 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8323 instead of variable_tracking_main.
8325 2018-01-03 Martin Sebor <msebor@redhat.com>
8327 PR tree-optimization/83603
8328 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8329 arguments past the endof the argument list in functions declared
8330 without a prototype.
8331 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8332 Avoid checking when arguments are null.
8334 2018-01-03 Martin Sebor <msebor@redhat.com>
8337 * doc/extend.texi (attribute const): Fix a typo.
8338 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8339 issuing -Wsuggest-attribute for void functions.
8341 2018-01-03 Martin Sebor <msebor@redhat.com>
8343 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8344 offset_int::from instead of wide_int::to_shwi.
8345 (maybe_diag_overlap): Remove assertion.
8346 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8347 * gimple-ssa-sprintf.c (format_directive): Same.
8348 (parse_directive): Same.
8349 (sprintf_dom_walker::compute_format_length): Same.
8350 (try_substitute_return_value): Same.
8352 2018-01-03 Jeff Law <law@redhat.com>
8355 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8356 non-constant residual for zero at runtime and avoid probing in
8357 that case. Reorganize code for trailing problem to mirror handling
8360 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8362 PR tree-optimization/83501
8363 * tree-ssa-strlen.c (get_string_cst): New.
8364 (handle_char_store): Call get_string_cst.
8366 2018-01-03 Martin Liska <mliska@suse.cz>
8368 PR tree-optimization/83593
8369 * tree-ssa-strlen.c: Include tree-cfg.h.
8370 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8371 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8372 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8374 (strlen_dom_walker::before_dom_children): Call
8375 gimple_purge_dead_eh_edges. Dump tranformation with details
8377 (strlen_dom_walker::before_dom_children): Update call by adding
8378 new argument cleanup_eh.
8379 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8381 2018-01-03 Martin Liska <mliska@suse.cz>
8384 * cif-code.def (VARIADIC_THUNK): New enum value.
8385 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8388 2018-01-03 Jan Beulich <jbeulich@suse.com>
8390 * sse.md (mov<mode>_internal): Tighten condition for when to use
8391 vmovdqu<ssescalarsize> for TI and OI modes.
8393 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8395 Update copyright years.
8397 2018-01-03 Martin Liska <mliska@suse.cz>
8400 * ipa-visibility.c (function_and_variable_visibility): Skip
8401 functions with noipa attribure.
8403 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8405 * gcc.c (process_command): Update copyright notice dates.
8406 * gcov-dump.c (print_version): Ditto.
8407 * gcov.c (print_version): Ditto.
8408 * gcov-tool.c (print_version): Ditto.
8409 * gengtype.c (create_file): Ditto.
8410 * doc/cpp.texi: Bump @copying's copyright year.
8411 * doc/cppinternals.texi: Ditto.
8412 * doc/gcc.texi: Ditto.
8413 * doc/gccint.texi: Ditto.
8414 * doc/gcov.texi: Ditto.
8415 * doc/install.texi: Ditto.
8416 * doc/invoke.texi: Ditto.
8418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8420 * vector-builder.h (vector_builder::m_full_nelts): Change from
8421 unsigned int to poly_uint64.
8422 (vector_builder::full_nelts): Update prototype accordingly.
8423 (vector_builder::new_vector): Likewise.
8424 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8425 (vector_builder::operator ==): Likewise.
8426 (vector_builder::finalize): Likewise.
8427 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8428 Take the number of elements as a poly_uint64 rather than an
8430 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8431 from unsigned int to poly_uint64.
8432 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8433 (vec_perm_indices::new_vector): Likewise.
8434 (vec_perm_indices::length): Likewise.
8435 (vec_perm_indices::nelts_per_input): Likewise.
8436 (vec_perm_indices::input_nelts): Likewise.
8437 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8438 number of elements per input as a poly_uint64 rather than an
8439 unsigned int. Use the original encoding for variable-length
8440 vectors, rather than clamping each individual element.
8441 For the second and subsequent elements in each pattern,
8442 clamp the step and base before clamping their sum.
8443 (vec_perm_indices::series_p): Handle polynomial element counts.
8444 (vec_perm_indices::all_in_range_p): Likewise.
8445 (vec_perm_indices_to_tree): Likewise.
8446 (vec_perm_indices_to_rtx): Likewise.
8447 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8448 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8449 (tree_vector_builder::new_binary_operation): Handle polynomial
8450 element counts. Return false if we need to know the number
8451 of elements at compile time.
8452 * fold-const.c (fold_vec_perm): Punt if the number of elements
8453 isn't known at compile time.
8455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8457 * vec-perm-indices.h (vec_perm_builder): Change element type
8458 from HOST_WIDE_INT to poly_int64.
8459 (vec_perm_indices::element_type): Update accordingly.
8460 (vec_perm_indices::clamp): Handle polynomial element_types.
8461 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8462 (vec_perm_indices::all_in_range_p): Likewise.
8463 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8465 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8466 polynomial vec_perm_indices element types.
8467 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8468 * fold-const.c (fold_vec_perm): Likewise.
8469 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8470 * tree-vect-generic.c (lower_vec_perm): Likewise.
8471 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8472 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8473 element type to HOST_WIDE_INT.
8475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8476 Alan Hayward <alan.hayward@arm.com>
8477 David Sherwood <david.sherwood@arm.com>
8479 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8480 rather than an int. Use plus_constant.
8481 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8482 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8485 Alan Hayward <alan.hayward@arm.com>
8486 David Sherwood <david.sherwood@arm.com>
8488 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8489 a HOST_WIDE_INT to a poly_int64.
8491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8492 Alan Hayward <alan.hayward@arm.com>
8493 David Sherwood <david.sherwood@arm.com>
8495 * calls.c (load_register_parameters): Cope with polynomial
8496 mode sizes. Require a constant size for BLKmode parameters
8497 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8498 forces a parameter to be padded at the lsb end in order to
8499 fill a complete number of words, require the parameter size
8500 to be ordered wrt UNITS_PER_WORD.
8502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8503 Alan Hayward <alan.hayward@arm.com>
8504 David Sherwood <david.sherwood@arm.com>
8506 * reload1.c (spill_stack_slot_width): Change element type
8507 from unsigned int to poly_uint64_pod.
8508 (alter_reg): Treat mode sizes as polynomial.
8510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8511 Alan Hayward <alan.hayward@arm.com>
8512 David Sherwood <david.sherwood@arm.com>
8514 * reload.c (complex_word_subreg_p): New function.
8515 (reload_inner_reg_of_subreg, push_reload): Use it.
8517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8518 Alan Hayward <alan.hayward@arm.com>
8519 David Sherwood <david.sherwood@arm.com>
8521 * lra-constraints.c (process_alt_operands): Reject matched
8522 operands whose sizes aren't ordered.
8523 (match_reload): Refer to this check here.
8525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8526 Alan Hayward <alan.hayward@arm.com>
8527 David Sherwood <david.sherwood@arm.com>
8529 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8530 that the mode size is in the set {1, 2, 4, 8, 16}.
8532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8533 Alan Hayward <alan.hayward@arm.com>
8534 David Sherwood <david.sherwood@arm.com>
8536 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8537 Use plus_constant instead of gen_rtx_PLUS.
8539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8540 Alan Hayward <alan.hayward@arm.com>
8541 David Sherwood <david.sherwood@arm.com>
8543 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8544 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8545 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8546 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8547 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8548 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8549 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8550 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8551 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8552 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8554 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8555 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8556 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8557 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8558 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8559 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8560 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8561 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8562 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8563 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8565 * expr.c (emit_move_resolve_push): Treat the input and result
8566 of PUSH_ROUNDING as a poly_int64.
8567 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8568 (emit_push_insn): Likewise.
8569 * lra-eliminations.c (mark_not_eliminable): Likewise.
8570 * recog.c (push_operand): Likewise.
8571 * reload1.c (elimination_effects): Likewise.
8572 * rtlanal.c (nonzero_bits1): Likewise.
8573 * calls.c (store_one_arg): Likewise. Require the padding to be
8574 known at compile time.
8576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8577 Alan Hayward <alan.hayward@arm.com>
8578 David Sherwood <david.sherwood@arm.com>
8580 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8581 Use plus_constant instead of gen_rtx_PLUS.
8583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8584 Alan Hayward <alan.hayward@arm.com>
8585 David Sherwood <david.sherwood@arm.com>
8587 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8591 Alan Hayward <alan.hayward@arm.com>
8592 David Sherwood <david.sherwood@arm.com>
8594 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8595 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8596 via stack temporaries. Treat the mode size as polynomial too.
8598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8599 Alan Hayward <alan.hayward@arm.com>
8600 David Sherwood <david.sherwood@arm.com>
8602 * expr.c (expand_expr_real_2): When handling conversions involving
8603 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8604 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8605 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8614 Alan Hayward <alan.hayward@arm.com>
8615 David Sherwood <david.sherwood@arm.com>
8617 * combine.c (can_change_dest_mode): Handle polynomial
8618 REGMODE_NATURAL_SIZE.
8619 * expmed.c (store_bit_field_1): Likewise.
8620 * expr.c (store_constructor): Likewise.
8621 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8622 and polynomial REGMODE_NATURAL_SIZE.
8623 (gen_lowpart_common): Likewise.
8624 * reginfo.c (record_subregs_of_mode): Likewise.
8625 * rtlanal.c (read_modify_subreg_p): Likewise.
8627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8628 Alan Hayward <alan.hayward@arm.com>
8629 David Sherwood <david.sherwood@arm.com>
8631 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8632 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
8640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8641 Alan Hayward <alan.hayward@arm.com>
8642 David Sherwood <david.sherwood@arm.com>
8644 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8645 in a POINTER_PLUS_EXPR.
8647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8648 Alan Hayward <alan.hayward@arm.com>
8649 David Sherwood <david.sherwood@arm.com>
8651 * omp-simd-clone.c (simd_clone_subparts): New function.
8652 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8653 (ipa_simd_modify_function_body): Likewise.
8655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8656 Alan Hayward <alan.hayward@arm.com>
8657 David Sherwood <david.sherwood@arm.com>
8659 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8660 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8661 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8662 (expand_vector_condition, vector_element): Likewise.
8663 (subparts_gt): New function.
8664 (get_compute_type): Use subparts_gt.
8665 (count_type_subparts): Delete.
8666 (expand_vector_operations_1): Use subparts_gt instead of
8667 count_type_subparts.
8669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8670 Alan Hayward <alan.hayward@arm.com>
8671 David Sherwood <david.sherwood@arm.com>
8673 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8674 (vect_compile_time_alias): ...this new function. Do the calculation
8675 on poly_ints rather than trees.
8676 (vect_prune_runtime_alias_test_list): Update call accordingly.
8678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8679 Alan Hayward <alan.hayward@arm.com>
8680 David Sherwood <david.sherwood@arm.com>
8682 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8684 (vect_schedule_slp_instance): Likewise.
8686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8687 Alan Hayward <alan.hayward@arm.com>
8688 David Sherwood <david.sherwood@arm.com>
8690 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8691 constant and extern definitions for variable-length vectors.
8692 (vect_get_constant_vectors): Note that the number of units
8693 is known to be constant.
8695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8696 Alan Hayward <alan.hayward@arm.com>
8697 David Sherwood <david.sherwood@arm.com>
8699 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8700 of units as polynomial. Choose between WIDE and NARROW based
8703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8704 Alan Hayward <alan.hayward@arm.com>
8705 David Sherwood <david.sherwood@arm.com>
8707 * tree-vect-stmts.c (simd_clone_subparts): New function.
8708 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8711 Alan Hayward <alan.hayward@arm.com>
8712 David Sherwood <david.sherwood@arm.com>
8714 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8715 vectors as polynomial. Use build_index_vector for
8718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8719 Alan Hayward <alan.hayward@arm.com>
8720 David Sherwood <david.sherwood@arm.com>
8722 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8723 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8724 for variable-length vectors.
8725 (vectorizable_mask_load_store): Treat the number of units as
8726 polynomial, asserting that it is constant if the condition has
8727 already been enforced.
8728 (vectorizable_store, vectorizable_load): Likewise.
8730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8731 Alan Hayward <alan.hayward@arm.com>
8732 David Sherwood <david.sherwood@arm.com>
8734 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8735 of units as polynomial. Punt if we can't tell at compile time
8736 which vector contains the final result.
8738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8739 Alan Hayward <alan.hayward@arm.com>
8740 David Sherwood <david.sherwood@arm.com>
8742 * tree-vect-loop.c (vectorizable_induction): Treat the number
8743 of units as polynomial. Punt on SLP inductions. Use an integer
8744 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8745 cast of such a series for variable-length floating-point
8748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8749 Alan Hayward <alan.hayward@arm.com>
8750 David Sherwood <david.sherwood@arm.com>
8752 * tree.h (build_index_vector): Declare.
8753 * tree.c (build_index_vector): New function.
8754 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8755 of units as polynomial, forcibly converting it to a constant if
8756 vectorizable_reduction has already enforced the condition.
8757 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8758 to create a {1,2,3,...} vector.
8759 (vectorizable_reduction): Treat the number of units as polynomial.
8760 Choose vectype_in based on the largest scalar element size rather
8761 than the smallest number of units. Enforce the restrictions
8764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8765 Alan Hayward <alan.hayward@arm.com>
8766 David Sherwood <david.sherwood@arm.com>
8768 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8769 number of units as polynomial.
8771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8772 Alan Hayward <alan.hayward@arm.com>
8773 David Sherwood <david.sherwood@arm.com>
8775 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8776 * target.def (autovectorize_vector_sizes): Return the vector sizes
8777 by pointer, using vector_sizes rather than a bitmask.
8778 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8779 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8780 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8782 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8783 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8784 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8785 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8786 * omp-general.c (omp_max_vf): Likewise.
8787 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8788 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8789 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8790 * tree-vect-slp.c (vect_slp_bb): Likewise.
8791 * doc/tm.texi: Regenerate.
8792 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8794 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8795 the vector size as a poly_uint64 rather than an unsigned int.
8796 (current_vector_size): Change from an unsigned int to a poly_uint64.
8797 (get_vectype_for_scalar_type): Update accordingly.
8798 * tree.h (build_truth_vector_type): Take the size and number of
8799 units as a poly_uint64 rather than an unsigned int.
8800 (build_vector_type): Add a temporary overload that takes
8801 the number of units as a poly_uint64 rather than an unsigned int.
8802 * tree.c (make_vector_type): Likewise.
8803 (build_truth_vector_type): Take the number of units as a poly_uint64
8804 rather than an unsigned int.
8806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8807 Alan Hayward <alan.hayward@arm.com>
8808 David Sherwood <david.sherwood@arm.com>
8810 * target.def (get_mask_mode): Take the number of units and length
8811 as poly_uint64s rather than unsigned ints.
8812 * targhooks.h (default_get_mask_mode): Update accordingly.
8813 * targhooks.c (default_get_mask_mode): Likewise.
8814 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8815 * doc/tm.texi: Regenerate.
8817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8818 Alan Hayward <alan.hayward@arm.com>
8819 David Sherwood <david.sherwood@arm.com>
8821 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8822 * omp-general.c (omp_max_vf): Likewise.
8823 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8824 (expand_omp_simd): Handle polynomial safelen.
8825 * omp-low.c (omplow_simd_context): Add a default constructor.
8826 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8827 (lower_rec_simd_input_clauses): Update accordingly.
8828 (lower_rec_input_clauses): Likewise.
8830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8831 Alan Hayward <alan.hayward@arm.com>
8832 David Sherwood <david.sherwood@arm.com>
8834 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8835 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8836 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8837 (vect_analyze_slp_cost): Likewise.
8838 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8839 (vect_model_load_cost): Likewise.
8841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8842 Alan Hayward <alan.hayward@arm.com>
8843 David Sherwood <david.sherwood@arm.com>
8845 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8846 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8847 from an unsigned int * to a poly_uint64_pod *.
8848 (calculate_unrolling_factor): New function.
8849 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8852 Alan Hayward <alan.hayward@arm.com>
8853 David Sherwood <david.sherwood@arm.com>
8855 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8856 from an unsigned int to a poly_uint64.
8857 (_loop_vec_info::slp_unrolling_factor): Likewise.
8858 (_loop_vec_info::vectorization_factor): Change from an int
8860 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8861 (vect_get_num_vectors): New function.
8862 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8863 (vect_get_num_copies): Use vect_get_num_vectors.
8864 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8865 to an unsigned int *.
8866 (vect_analyze_data_refs): Change min_vf from an int * to a
8868 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8869 than an unsigned HOST_WIDE_INT.
8870 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8871 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8872 to an unsigned int *.
8873 (vect_analyze_data_ref_dependences): Likewise.
8874 (vect_compute_data_ref_alignment): Handle polynomial vf.
8875 (vect_enhance_data_refs_alignment): Likewise.
8876 (vect_prune_runtime_alias_test_list): Likewise.
8877 (vect_shift_permute_load_chain): Likewise.
8878 (vect_supportable_dr_alignment): Likewise.
8879 (dependence_distance_ge_vf): Take the vectorization factor as a
8880 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8881 (vect_analyze_data_refs): Change min_vf from an int * to a
8883 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8884 vfm1 as a poly_uint64 rather than an int. Make the same change
8885 for the returned bound_scalar.
8886 (vect_gen_vector_loop_niters): Handle polynomial vf.
8887 (vect_do_peeling): Likewise. Update call to
8888 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8889 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8891 * tree-vect-loop.c (vect_determine_vectorization_factor)
8892 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8893 (vect_get_known_peeling_cost): Likewise.
8894 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8895 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8896 (vect_transform_loop): Likewise. Use the lowest possible VF when
8897 updating the upper bounds of the loop.
8898 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8900 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8901 polynomial unroll factors.
8902 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8903 (vect_make_slp_decision): Likewise.
8904 (vect_supported_load_permutation_p): Likewise, and polynomial
8906 (vect_analyze_slp_cost): Handle polynomial vf.
8907 (vect_slp_analyze_node_operations): Likewise.
8908 (vect_slp_analyze_bb_1): Likewise.
8909 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8910 than an unsigned HOST_WIDE_INT.
8911 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8912 (vectorizable_load): Handle polynomial vf.
8913 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8915 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8918 Alan Hayward <alan.hayward@arm.com>
8919 David Sherwood <david.sherwood@arm.com>
8921 * match.pd: Handle bit operations involving three constants
8922 and try to fold one pair.
8924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8926 * tree-vect-loop-manip.c: Include gimple-fold.h.
8927 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8928 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8929 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8930 Add a path that uses a step of VF instead of 1, but disable it
8932 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8933 and niters_no_overflow parameters. Update calls to
8934 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8935 Create a new SSA name if the latter choses to use a ste other
8936 than zero, and return it via niters_vector_mult_vf_var.
8937 * tree-vect-loop.c (vect_transform_loop): Update calls to
8938 vect_do_peeling, vect_gen_vector_loop_niters and
8939 slpeel_make_loop_iterate_ntimes.
8940 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8941 (vect_gen_vector_loop_niters): Update declarations after above changes.
8943 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8945 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8946 128-bit round to integer instructions.
8947 (ceil<mode>2): Likewise.
8948 (btrunc<mode>2): Likewise.
8949 (round<mode>2): Likewise.
8951 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8953 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8954 unaligned VSX load/store on P8/P9.
8955 (expand_block_clear): Allow the use of unaligned VSX
8956 load/store on P8/P9.
8958 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8960 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8962 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8963 swap associated with both a load and a store.
8965 2018-01-02 Andrew Waterman <andrew@sifive.com>
8967 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8968 * config/riscv/riscv.md (clear_cache): Use it.
8970 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8972 * web.c: Remove out-of-date comment.
8974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8976 * expr.c (fixup_args_size_notes): Check that any existing
8977 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8978 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8979 (emit_single_push_insn): ...here.
8981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8983 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8984 (const_vector_encoded_nelts): New function.
8985 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8986 (const_vector_int_elt, const_vector_elt): Declare.
8987 * emit-rtl.c (const_vector_int_elt_1): New function.
8988 (const_vector_elt): Likewise.
8989 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8990 of CONST_VECTOR_ELT.
8992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8994 * expr.c: Include rtx-vector-builder.h.
8995 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8996 directly on the tree encoding.
8997 (const_vector_from_tree): Likewise.
8998 * optabs.c: Include rtx-vector-builder.h.
8999 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9000 sequence of "u" values.
9001 * vec-perm-indices.c: Include rtx-vector-builder.h.
9002 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9003 directly on the vec_perm_indices encoding.
9005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9007 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9008 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9009 * rtx-vector-builder.h: New file.
9010 * rtx-vector-builder.c: Likewise.
9011 * rtl.h (rtx_def::u2): Add a const_vector field.
9012 (CONST_VECTOR_NPATTERNS): New macro.
9013 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9014 (CONST_VECTOR_DUPLICATE_P): Likewise.
9015 (CONST_VECTOR_STEPPED_P): Likewise.
9016 (CONST_VECTOR_ENCODED_ELT): Likewise.
9017 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9018 (unwrap_const_vec_duplicate): Likewise.
9019 (const_vec_series_p): Check for a non-duplicated vector encoding.
9020 Say that the function only returns true for integer vectors.
9021 * emit-rtl.c: Include rtx-vector-builder.h.
9022 (gen_const_vec_duplicate_1): Delete.
9023 (gen_const_vector): Call gen_const_vec_duplicate instead of
9024 gen_const_vec_duplicate_1.
9025 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9026 (gen_const_vec_duplicate): Use rtx_vector_builder.
9027 (gen_const_vec_series): Likewise.
9028 (gen_rtx_CONST_VECTOR): Likewise.
9029 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9030 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9031 Build a new vector rather than modifying a CONST_VECTOR in-place.
9032 (handle_special_swappables): Update call accordingly.
9033 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9034 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9035 Build a new vector rather than modifying a CONST_VECTOR in-place.
9036 (handle_special_swappables): Update call accordingly.
9038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9040 * simplify-rtx.c (simplify_const_binary_operation): Use
9041 CONST_VECTOR_ELT instead of XVECEXP.
9043 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9045 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9046 the selector elements to be different from the data elements
9047 if the selector is a VECTOR_CST.
9048 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9049 ssizetype for the selector.
9051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9053 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9054 before testing each element individually.
9055 * tree-vect-generic.c (lower_vec_perm): Likewise.
9057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9059 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9060 * selftest-run-tests.c (selftest::run_tests): Call it.
9061 * vector-builder.h (vector_builder::operator ==): New function.
9062 (vector_builder::operator !=): Likewise.
9063 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9064 (vec_perm_indices::all_from_input_p): New function.
9065 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9066 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9067 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9068 instead of reading the VECTOR_CST directly. Detect whether both
9069 vector inputs are the same before constructing the vec_perm_indices,
9070 and update the number of inputs argument accordingly. Use the
9071 utility functions added above. Only construct sel2 if we need to.
9073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9075 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9076 the broadcast of the low byte.
9077 (expand_mult_highpart): Use an explicit encoding for the permutes.
9078 * optabs-query.c (can_mult_highpart_p): Likewise.
9079 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9080 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9081 (vectorizable_bswap): Likewise.
9082 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9083 explicit encoding for the power-of-2 permutes.
9084 (vect_permute_store_chain): Likewise.
9085 (vect_grouped_load_supported): Likewise.
9086 (vect_permute_load_chain): Likewise.
9088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9090 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9091 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9092 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9093 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9094 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9095 (vect_gen_perm_mask_any): Likewise.
9097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9099 * int-vector-builder.h: New file.
9100 * vec-perm-indices.h: Include int-vector-builder.h.
9101 (vec_perm_indices): Redefine as an int_vector_builder.
9102 (auto_vec_perm_indices): Delete.
9103 (vec_perm_builder): Redefine as a stand-alone class.
9104 (vec_perm_indices::vec_perm_indices): New function.
9105 (vec_perm_indices::clamp): Likewise.
9106 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9107 (vec_perm_indices::new_vector): New function.
9108 (vec_perm_indices::new_expanded_vector): Update for new
9109 vec_perm_indices class.
9110 (vec_perm_indices::rotate_inputs): New function.
9111 (vec_perm_indices::all_in_range_p): Operate directly on the
9112 encoded form, without computing elided elements.
9113 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9114 encoding. Update for new vec_perm_indices class.
9115 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9116 the given vec_perm_builder.
9117 (expand_vec_perm_var): Update vec_perm_builder constructor.
9118 (expand_mult_highpart): Use vec_perm_builder instead of
9119 auto_vec_perm_indices.
9120 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9121 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9122 or double series encoding as appropriate.
9123 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9124 vec_perm_indices instead of auto_vec_perm_indices.
9125 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9126 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9127 (vect_permute_store_chain): Likewise.
9128 (vect_grouped_load_supported): Likewise.
9129 (vect_permute_load_chain): Likewise.
9130 (vect_shift_permute_load_chain): Likewise.
9131 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9132 (vect_transform_slp_perm_load): Likewise.
9133 (vect_schedule_slp_instance): Likewise.
9134 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9135 (vectorizable_mask_load_store): Likewise.
9136 (vectorizable_bswap): Likewise.
9137 (vectorizable_store): Likewise.
9138 (vectorizable_load): Likewise.
9139 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9140 vec_perm_indices instead of auto_vec_perm_indices. Use
9141 tree_to_vec_perm_builder to read the vector from a tree.
9142 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9143 vec_perm_builder instead of a vec_perm_indices.
9144 (have_whole_vector_shift): Use vec_perm_builder and
9145 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9146 truncation to calc_vec_perm_mask_for_shift.
9147 (vect_create_epilog_for_reduction): Likewise.
9148 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9149 from auto_vec_perm_indices to vec_perm_indices.
9150 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9151 instead of changing individual elements.
9152 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9153 the vector in d.perm.
9154 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9155 from auto_vec_perm_indices to vec_perm_indices.
9156 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9157 instead of changing individual elements.
9158 (arm_vectorize_vec_perm_const): Use new_vector to install
9159 the vector in d.perm.
9160 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9161 Update vec_perm_builder constructor.
9162 (rs6000_expand_interleave): Likewise.
9163 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9164 (rs6000_expand_interleave): Likewise.
9166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9168 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9169 to qimode could truncate the indices.
9170 * optabs.c (expand_vec_perm_var): Likewise.
9172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9174 * Makefile.in (OBJS): Add vec-perm-indices.o.
9175 * vec-perm-indices.h: New file.
9176 * vec-perm-indices.c: Likewise.
9177 * target.h (vec_perm_indices): Replace with a forward class
9179 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9180 * optabs.h: Include vec-perm-indices.h.
9181 (expand_vec_perm): Delete.
9182 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9183 (expand_vec_perm_const): Declare.
9184 * target.def (vec_perm_const_ok): Replace with...
9185 (vec_perm_const): ...this new hook.
9186 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9187 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9188 * doc/tm.texi: Regenerate.
9189 * optabs.def (vec_perm_const): Delete.
9190 * doc/md.texi (vec_perm_const): Likewise.
9191 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9192 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9193 expand_vec_perm for constant permutation vectors. Assert that
9194 the mode of variable permutation vectors is the integer equivalent
9195 of the mode that is being permuted.
9196 * optabs-query.h (selector_fits_mode_p): Declare.
9197 * optabs-query.c: Include vec-perm-indices.h.
9198 (selector_fits_mode_p): New function.
9199 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9200 is defined, instead of checking whether the vec_perm_const_optab
9201 exists. Use targetm.vectorize.vec_perm_const instead of
9202 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9203 fit in the vector mode before using a variable permute.
9204 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9205 vec_perm_indices instead of an rtx.
9206 (expand_vec_perm): Replace with...
9207 (expand_vec_perm_const): ...this new function. Take the selector
9208 as a vec_perm_indices rather than an rtx. Also take the mode of
9209 the selector. Update call to shift_amt_for_vec_perm_mask.
9210 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9211 Use vec_perm_indices::new_expanded_vector to expand the original
9212 selector into bytes. Check whether the indices fit in the vector
9213 mode before using a variable permute.
9214 (expand_vec_perm_var): Make global.
9215 (expand_mult_highpart): Use expand_vec_perm_const.
9216 * fold-const.c: Includes vec-perm-indices.h.
9217 * tree-ssa-forwprop.c: Likewise.
9218 * tree-vect-data-refs.c: Likewise.
9219 * tree-vect-generic.c: Likewise.
9220 * tree-vect-loop.c: Likewise.
9221 * tree-vect-slp.c: Likewise.
9222 * tree-vect-stmts.c: Likewise.
9223 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9225 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9226 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9227 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9228 (aarch64_vectorize_vec_perm_const): ...this new function.
9229 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9230 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9231 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9232 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9233 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9234 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9235 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9237 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9238 check for NEON modes.
9239 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9240 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9241 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9242 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9244 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9245 the old VEC_PERM_CONST conditions.
9246 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9247 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9248 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9249 (ia64_vectorize_vec_perm_const_ok): Merge into...
9250 (ia64_vectorize_vec_perm_const): ...this new function.
9251 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9252 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9253 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9254 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9255 * config/mips/mips.c (mips_expand_vec_perm_const)
9256 (mips_vectorize_vec_perm_const_ok): Merge into...
9257 (mips_vectorize_vec_perm_const): ...this new function.
9258 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9259 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9260 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9261 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9262 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9263 (rs6000_expand_vec_perm_const): Delete.
9264 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9266 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9267 (altivec_expand_vec_perm_const_le): Take each operand individually.
9268 Operate on constant selectors rather than rtxes.
9269 (altivec_expand_vec_perm_const): Likewise. Update call to
9270 altivec_expand_vec_perm_const_le.
9271 (rs6000_expand_vec_perm_const): Delete.
9272 (rs6000_vectorize_vec_perm_const_ok): Delete.
9273 (rs6000_vectorize_vec_perm_const): New function.
9274 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9275 an element count and rtx array.
9276 (rs6000_expand_extract_even): Update call accordingly.
9277 (rs6000_expand_interleave): Likewise.
9278 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9279 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9280 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9281 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9282 (rs6000_expand_vec_perm_const): Delete.
9283 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9284 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9285 (altivec_expand_vec_perm_const_le): Take each operand individually.
9286 Operate on constant selectors rather than rtxes.
9287 (altivec_expand_vec_perm_const): Likewise. Update call to
9288 altivec_expand_vec_perm_const_le.
9289 (rs6000_expand_vec_perm_const): Delete.
9290 (rs6000_vectorize_vec_perm_const_ok): Delete.
9291 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9292 reference to the SPE evmerge intructions.
9293 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9294 an element count and rtx array.
9295 (rs6000_expand_extract_even): Update call accordingly.
9296 (rs6000_expand_interleave): Likewise.
9297 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9298 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9300 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9304 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9305 vector mode and that that mode matches the mode of the data
9307 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9308 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9309 directly using expand_vec_perm_1 when forcing selectors into
9311 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9315 * optabs-query.h (can_vec_perm_p): Delete.
9316 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9317 * optabs-query.c (can_vec_perm_p): Split into...
9318 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9319 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9320 particular selector is valid.
9321 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9322 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9323 (vect_grouped_load_supported): Likewise.
9324 (vect_shift_permute_load_chain): Likewise.
9325 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9326 (vect_transform_slp_perm_load): Likewise.
9327 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9328 (vectorizable_bswap): Likewise.
9329 (vect_gen_perm_mask_checked): Likewise.
9330 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9331 implementations of variable permutation vectors into account
9332 when deciding which selector to use.
9333 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9334 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9335 with a false third argument.
9336 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9337 to test whether the constant selector is valid and can_vec_perm_var_p
9338 to test whether a variable selector is valid.
9340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9342 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9343 * optabs-query.c (can_vec_perm_p): Likewise.
9344 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9345 instead of vec_perm_indices.
9346 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9347 (vect_gen_perm_mask_checked): Likewise,
9348 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9349 (vect_gen_perm_mask_checked): Likewise,
9351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9353 * optabs-query.h (qimode_for_vec_perm): Declare.
9354 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9355 (qimode_for_vec_perm): ...this new function.
9356 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9358 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9360 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9361 does not have a conditional at the top.
9363 2018-01-02 Richard Biener <rguenther@suse.de>
9365 * ipa-inline.c (big_speedup_p): Fix expression.
9367 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9370 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9373 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9377 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9378 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9379 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9380 cond_taken_branch_cost 3->4.
9382 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9384 PR tree-optimization/83581
9385 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9386 TODO_cleanup_cfg if any changes have been made.
9389 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9390 convert_modes if target mode has the right side, but different mode
9394 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9395 last argument when extracting from CONCAT. If either from_real or
9396 from_imag is NULL, use expansion through memory. If result is not
9397 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9398 the parts directly to inner mode, if even that fails, use expansion
9402 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9403 check for bswap in mode rather than HImode and use that in expand_unop
9406 Copyright (C) 2018 Free Software Foundation, Inc.
9408 Copying and distribu