1 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4 * dwarf2asm.c: Include fold-const.c.
5 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6 of decl rather than decl itself.
8 PR rtl-optimization/84643
9 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15 calling assemble_variable.
17 2018-03-19 Sudakshina Das <sudi.das@arm.com>
20 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23 2018-03-19 Jim Wilson <jimw@sifive.com>
26 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
27 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
28 (riscv_first_stack_step): Likewise.
29 (riscv_option_override): Use STACK_BOUNDARY instead of
31 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
33 (BIGGEST_ALIGNMENT): Set to 128.
34 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
35 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
38 2018-03-19 Richard Biener <rguenther@suse.de>
40 PR tree-optimization/84933
41 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
42 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
44 2018-03-19 Richard Biener <rguenther@suse.de>
46 PR tree-optimization/84859
47 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
48 (cond_if_else_store_replacement): Perform sinking operation on
49 single-store BBs regardless of MAX_STORES_TO_SINK setting.
50 Generalize what a BB with a single eligible store is.
52 2018-03-19 Richard Biener <rguenther@suse.de>
54 PR tree-optimization/84929
55 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
56 chrec_is_positive against non-chrec arg.
58 2018-03-19 Tamar Christina <tamar.christina@arm.com>
61 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
63 2018-03-18 Martin Liska <mliska@suse.cz>
65 PR rtl-optimization/84635
66 * regrename.c (build_def_use): Use matches_mode only when
69 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
71 PR tree-optimization/84913
72 * tree-vect-loop.c (vectorizable_reduction): Don't try to
73 vectorize chains of COND_EXPRs.
75 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
77 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
79 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
81 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
83 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
85 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
87 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
88 Kito Cheng <kito.cheng@gmail.com>
90 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
91 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
92 (nds32_adjust_reg_alloc_order): New function.
93 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
95 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
97 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
98 nds32_print_operand, nds32_print_operand_address): Use
99 HOST_WIDE_INT_PRINT_DEC instead.
101 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
103 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
105 2018-03-17 Jakub Jelinek <jakub@redhat.com>
108 * config/i386/i386.c (initial_ix86_tune_features,
109 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
111 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
112 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
113 rather than 1u << ix86_tune. Formatting fix.
114 (ix86_option_override_internal): Change ix86_arch_mask from
115 unsigned int to unsigned HOST_WIDE_INT, initialize to
116 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
117 (ix86_function_specific_restore): Likewise.
119 2018-03-16 Jakub Jelinek <jakub@redhat.com>
122 * postreload.c (reload_combine_recognize_pattern): Perform
123 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
124 truncate_int_for_mode the result for the destination's mode.
127 * hsa-gen.c (mem_type_for_type): Fix comment typo.
128 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
130 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
133 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
136 * lra-assigns.c (lra_split_hard_reg_for): Don't use
137 regno_allocno_class_array and sorted_pseudos.
138 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
139 insns where regno is used.
141 2018-03-16 Martin Liska <mliska@suse.cz>
144 * multiple_target.c (create_dispatcher_calls): Redirect
145 reference in the symbol table.
147 2018-03-16 Martin Liska <mliska@suse.cz>
150 * multiple_target.c (create_dispatcher_calls): Redirect also
153 2018-03-16 Jakub Jelinek <jakub@redhat.com>
157 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
158 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
159 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
161 2018-03-16 Julia Koval <julia.koval@intel.com>
163 * doc/invoke.texi (Skylake Server): Add CLWB.
164 Cannonlake): Remove CLWB.
166 2018-03-16 Jakub Jelinek <jakub@redhat.com>
168 PR tree-optimization/84841
169 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
171 (FLOAT_ONE_CONST_TYPE): Define.
172 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
173 (sort_by_operand_rank): Put entries with higher constant_type last
174 rather than first to match comments.
176 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
178 * config/nios2/nios2.md (movsi_internal): Fix thinko in
181 2018-03-15 Jakub Jelinek <jakub@redhat.com>
184 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
185 check and use address of target always.
187 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
190 * config/i386/i386.c (indirect_thunk_needed): Update comments.
191 (indirect_thunk_bnd_needed): Likewise.
192 (indirect_thunks_used): Likewise.
193 (indirect_thunks_bnd_used): Likewise.
194 (indirect_return_needed): New.
195 (indirect_return_bnd_needed): Likewise.
196 (output_indirect_thunk_function): Add a bool argument for
198 (output_indirect_thunk_function): Don't generate alias for
199 function return thunk.
200 (ix86_code_end): Call output_indirect_thunk_function to generate
201 function return thunks.
202 (ix86_output_function_return): Set indirect_return_bnd_needed
203 and indirect_return_needed instead of indirect_thunk_bnd_needed
204 and indirect_thunk_needed.
206 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
208 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
209 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
210 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
212 2018-03-15 David Malcolm <dmalcolm@redhat.com>
213 Paul Hua <paul.hua.gm@gmail.com>
216 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
218 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
220 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
221 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
224 2018-03-15 Tamar Christina <tamar.christina@arm.com>
227 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
228 instead of GET_MODE_SIZE when comparing Units.
230 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
233 * varasm.c (hash_section): Return an unchangeble hash value
234 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
235 Return !aarch64_can_use_per_function_literal_pools_p ().
237 2018-03-15 Jakub Jelinek <jakub@redhat.com>
240 * optabs.c (emit_conditional_move): Pass address of cmode's copy
241 rather than address of cmode as last argument to prepare_cmp_insn.
243 2018-03-15 Julia Koval <julia.koval@intel.com>
245 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
246 F_AVX512VNNI, F_AVX512BITALG): New.
248 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
251 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
252 insn for floating-point loads and stores.
254 2018-03-14 Carl Love <cel@us.ibm.com>
256 * config/rs6000/rs6000-c.c: Add macro definitions for
257 ALTIVEC_BUILTIN_VEC_PERMXOR.
258 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
259 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
260 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
261 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
263 * config/doc/extend.texi: Add prototypes for vec_permxor.
265 2018-03-14 David Malcolm <dmalcolm@redhat.com>
268 * diagnostic-show-locus.c (class layout_point): Convert m_line
269 from int to linenum_type.
270 (line_span::comparator): Use linenum "compare" function when
271 comparing line numbers.
272 (test_line_span): New function.
273 (layout_range::contains_point): Convert param "row" from int to
275 (layout_range::intersects_line_p): Likewise.
276 (layout::will_show_line_p): Likewise.
277 (layout::print_source_line): Likewise.
278 (layout::should_print_annotation_line_p): Likewise.
279 (layout::print_annotation_line): Likewise.
280 (layout::print_leading_fixits): Likewise.
281 (layout::annotation_line_showed_range_p): Likewise.
282 (struct line_corrections): Likewise for field m_row.
283 (line_corrections::line_corrections): Likewise for param "row".
284 (layout::print_trailing_fixits): Likewise.
285 (layout::get_state_at_point): Likewise.
286 (layout::get_x_bound_for_row): Likewise.
287 (layout::print_line): Likewise.
288 (diagnostic_show_locus): Likewise for locals "last_line" and
290 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
291 * input.c (selftest::test_linenum_comparisons): New function.
292 (selftest::input_c_tests): Call it.
293 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
294 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
295 * selftest.h (ASSERT_GT): New macro.
296 (ASSERT_GT_AT): New macro.
297 (ASSERT_LT): New macro.
298 (ASSERT_LT_AT): New macro.
300 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
302 PR rtl-optimization/84780
303 * combine.c (distribute_links): Don't make a link based on pc_rtx.
305 2018-03-14 Martin Liska <mliska@suse.cz>
307 * tree.c (record_node_allocation_statistics): Use
309 (get_stats_node_kind): New function extracted from
310 record_node_allocation_statistics.
311 (free_node): Use get_stats_node_kind.
313 2018-03-14 Richard Biener <rguenther@suse.de>
315 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
316 that the value-set of ANTIC_IN doesn't grow.
319 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
321 (BB_VISITED_WITH_VISITED_SUCCS): New define.
322 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
324 2018-03-14 Julia Koval <julia.koval@intel.com>
326 * config.gcc (icelake-client, icelake-server): New.
328 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
329 (initial_ix86_arch_features): Ditto.
330 (PTA_SKYLAKE): Add SGX.
331 (PTA_ICELAKE): Remove.
332 (PTA_ICELAKE_CLIENT): New.
333 (PTA_ICELAKE_SERVER): New.
334 (ix86_option_override_internal): Split up icelake on icelake client and
336 (get_builtin_code_for_version): Ditto.
337 (fold_builtin_cpu): Ditto.
338 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
339 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
340 * config/i386/i386.h (processor_type): Ditto.
341 * doc/invoke.texi: Ditto.
343 2018-03-14 Jakub Jelinek <jakub@redhat.com>
346 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
347 INTEGER_CST offset, add it together with bitpos / 8 and
348 sign extend based on POINTER_SIZE.
352 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
355 * config/i386/constraints.md (Yc): New register constraint.
356 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
357 Use Yc constraint for alternative 2 of operand 0. Remove
358 preferred_for_speed attribute.
360 2018-03-14 Richard Biener <rguenther@suse.de>
362 PR tree-optimization/84830
363 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
364 with the old one to avoid oscillations.
366 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
369 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
371 (assign_by_spills): Return a flag of reload assignment failure.
372 Do not process the reload assignment failures. Do not spill other
373 reload pseudos if they has the same reg class. Update n if
375 (lra_assign): Add a return arg. Set up from the result of
376 assign_by_spills call.
377 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
378 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
379 usage_insns if it is not NULL.
380 (spill_hard_reg_in_range): New function.
381 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
382 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
384 (lra_assign): Change prototype.
385 * lra.c (lra): Add code to deal with fails by splitting hard reg
388 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
390 * config/riscv/riscv.opt (mrelax): New option.
391 * config/riscv/riscv.c (riscv_file_start): Emit ".option
392 "norelax" when riscv_mrelax is disabled.
393 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
395 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
398 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
399 reassociation for int modes.
401 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
403 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
404 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
406 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
407 * config/aarch64/aarch64-sve.md
408 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
409 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
410 (*extend<mode><Vwide>2): Rename to...
411 (aarch64_sve_extend<mode><Vwide>2): ...this.
412 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
413 renaming the old pattern to...
414 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
416 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
417 define_expand, renaming the old pattern to...
418 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
419 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
420 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
421 account when deciding which SVE instruction the optab should use.
422 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
424 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
426 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
427 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
428 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
429 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
430 (tlsdesc_small_<mode>): Turn a define_expand and use
431 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
432 (tlsdesc_small_advsimd_<mode>): ...this.
433 (tlsdesc_small_sve_<mode>): New pattern.
435 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
437 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
438 (UNSPEC_UMUL_HIGHPART): New constants.
439 (MUL_HIGHPART): New int iteraor.
440 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
441 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
443 (*<su>mul<mode>3_highpart): New define_insn.
445 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
448 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
451 2018-03-13 Martin Liska <mliska@suse.cz>
454 * (sem_item_optimizer::sem_item_optimizer): Initialize new
456 (sem_item_optimizer::~sem_item_optimizer): Release it.
457 (sem_item_optimizer::merge_classes): Register variable aliases.
458 (sem_item_optimizer::fixup_pt_set): New function.
459 (sem_item_optimizer::fixup_points_to_sets): Likewise.
460 * ipa-icf.h: Declare new variables and functions.
462 2018-03-13 Jakub Jelinek <jakub@redhat.com>
465 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
466 integer_pow2p@2 and test integer_pow2p in condition.
467 (A < 0 ? C : 0): Similarly for @1.
470 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
471 characters starting at p contain '\0' character, don't look beyond
475 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
476 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
479 * reg-stack.c (change_stack): Change update_end var from int to
480 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
481 also call set_block_for_insn on the newly added insns and rescan.
484 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
488 * tree.c (stabilize_reference_1): Return save_expr (e) for
489 STATEMENT_LIST even if it doesn't have side-effects.
491 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
493 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
495 2018-03-12 Renlin Li <renlin.li@arm.com>
497 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
498 aarch64_output_scalar_simd_mov_immediate.
500 2018-03-12 Martin Sebor <msebor@redhat.com>
502 PR tree-optimization/83456
503 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
504 for perfectly overlapping calls to memcpy.
505 (gimple_fold_builtin_memory_chk): Same.
506 (gimple_fold_builtin_strcpy): Handle no-warning.
507 (gimple_fold_builtin_stxcpy_chk): Same.
508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
510 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
512 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
513 parameter. Use it for SFmode.
514 (rs6000_function_arg_advance_1): Adjust.
515 (rs6000_function_arg): Adjust.
516 (rs6000_gimplify_va_arg): Pass false for that new parameter.
518 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
520 PR rtl-optimization/84169
521 PR rtl-optimization/84780
522 * combine.c (can_combine_p): Check for a 2-insn combination whether
523 the destination register is used between the two insns, too.
525 2018-03-12 Richard Biener <rguenther@suse.de>
527 PR tree-optimization/84803
528 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
529 for refs DR analysis didn't process.
531 2018-03-12 Richard Biener <rguenther@suse.de>
533 PR tree-optimization/84777
534 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
535 force-vectorize loops ignore whether we are optimizing for size.
537 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
539 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
540 (TARGET_MD_ASM_ADJUST): Define.
542 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
543 Kito Cheng <kito.cheng@gmail.com>
544 Chung-Ju Wu <jasonwucj@gmail.com>
546 * config/nds32/nds32.c (nds32_compute_stack_frame,
547 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
548 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
549 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
550 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
551 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
552 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
553 * config/nds32/nds32.md (prologue, epilogue): Use macro
554 NDS32_V3PUSH_AVAILABLE_P to do checking.
556 2018-03-11 Jakub Jelinek <jakub@redhat.com>
559 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
560 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
561 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
562 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
563 addition of most attributes on !orig_type_die or the attribute not
564 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
566 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
567 Chung-Ju Wu <jasonwucj@gmail.com>
569 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
571 * config/nds32/nds32.opt (mvh): New option.
573 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
574 Chung-Ju Wu <jasonwucj@gmail.com>
576 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
578 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
579 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
582 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
583 Chung-Ju Wu <jasonwucj@gmail.com>
585 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
587 * config/nds32/nds32-multiple.md (strlensi): New pattern.
588 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
590 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
591 Kito Cheng <kito.cheng@gmail.com>
592 Chung-Ju Wu <jasonwucj@gmail.com>
594 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
595 UNSPEC_FFMISM and UNSPEC_FLMISM.
596 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
597 for ffb, ffmism and flmism.
598 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
599 (unspec_ffmism): Ditto.
600 (unspec_flmism): Ditto.
601 (nds32_expand_builtin_impl): Check if string extension is available.
602 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
603 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
605 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
608 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
611 * lra-assigns.c (assign_by_spills): Return a flag of reload
612 assignment failure. Do not process the reload assignment
613 failures. Do not spill other reload pseudos if they has the same
615 (lra_assign): Add a return arg. Set up from the result of
616 assign_by_spills call.
617 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
618 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
619 usage_insns if it is not NULL.
620 (spill_hard_reg_in_range): New function.
621 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
622 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
624 (lra_assign): Change prototype.
625 * lra.c (lra): Add code to deal with fails by splitting hard reg
628 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
631 * config/i386/i386.opt: Replace Enforcment with Enforcement.
633 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
636 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
637 (dw_val_node): Add val_symbolic_view.
638 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
639 (symview_upper_bound): New.
640 (new_line_info_table): Initialize symviews_since_reset.
641 (dwarf2out_source_line): Count symviews_since_reset and set
643 (dw_val_equal_p): Handle symview.
644 (add_AT_symview): New.
645 (print_dw_val): Handle symview.
646 (attr_checksum, attr_checksum_ordered): Likewise.
647 (same_dw_val_p, size_of_die): Likewise.
648 (value_format, output_die): Likewise.
649 (add_high_low_attributes): Use add_AT_symview for entry_view.
650 (dwarf2out_finish): Reset symview_upper_bound, clear
653 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
656 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
657 Add strict argument and use it.
658 (rs6000_split_multireg_move): Update for new strict argument.
659 (mem_operand_gpr): Disallow all non-offsettable addresses.
660 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
662 2018-03-09 Jakub Jelinek <jakub@redhat.com>
665 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
666 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
667 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
670 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
671 decl, use remap_type if we want to use the type.
673 2018-03-09 Martin Sebor <msebor@redhat.com>
675 PR tree-optimization/84526
676 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
678 (builtin_access::generic_overlap): Be prepared to handle non-array
681 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
683 PR rtl-optimization/84682
684 * lra-constraints.c (process_address_1): Check is_address flag
685 for address constraints.
686 (process_alt_operands): Likewise.
687 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
688 preprocess_constraints.
689 * recog.h (preprocess_constraints): Add oploc parameter.
691 * recog.c (preprocess_constraints): Test address_operand for
692 CT_ADDRESS constraints.
694 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
697 * lra-assigns.c (assign_by_spills): Return a flag of reload
698 assignment failure. Do not process the reload assignment
699 failures. Do not spill other reload pseudos if they has the same
701 (lra_assign): Add a return arg. Set up from the result of
702 assign_by_spills call.
703 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
704 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
705 usage_insns if it is not NULL.
706 (spill_hard_reg_in_range): New function.
707 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
708 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
710 (lra_assign): Change prototype.
711 * lra.c (lra): Add code to deal with fails by splitting hard reg
714 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
717 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
718 Accept complain bool parameter. Only emit errors if it is true.
719 (arm_parse_cpu_option_name): Likewise.
720 (arm_target_thumb_only): Adjust callers of the above.
721 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
722 prototype to take a default true bool parameter.
723 (arm_parse_arch_option_name): Likewise.
725 2018-03-09 David Malcolm <dmalcolm@redhat.com>
726 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
730 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
731 * configure: Regenerate.
732 * configure.ac ("linker --version-script option"): New.
733 ("linker soname option"): New.
735 2018-03-09 Richard Biener <rguenther@suse.de>
737 PR tree-optimization/84775
738 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
739 immediate uses of predicate stmts and mark them modified.
742 PR tree-optimization/84178
743 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
745 (version_loop_for_if_conversion): Delay update_ssa call.
746 (tree_if_conversion): Delay update_ssa until after predicate
749 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
752 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
753 when the function accesses prior frames.
755 2018-03-08 Jakub Jelinek <jakub@redhat.com>
758 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
759 gen_llsym, otherwise call maybe_gen_llsym.
762 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
763 has ',' character inside of it.
765 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
768 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
769 as clobbering CC_REGNUM.
771 2018-03-08 Richard Biener <rguenther@suse.de>
774 * tree-scalar-evolution.c: Include tree-into-ssa.h.
775 (follow_copies_to_constant): Do not follow SSA names registered
778 2018-03-08 Richard Biener <rguenther@suse.de>
780 PR tree-optimization/84178
781 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
783 (version_loop_for_if_conversion): Delay update_ssa call.
784 (tree_if_conversion): Delay update_ssa until after predicate
787 2018-03-08 David Malcolm <dmalcolm@redhat.com>
789 PR tree-optimization/84178
790 * tree-if-conv.c (release_bb_predicate): Remove the
791 the assertion that the stmts have NULL use_ops.
792 Discard the statements, asserting that they haven't
793 yet been added to a BB.
795 2018-03-08 Richard Biener <rguenther@suse.de>
797 PR tree-optimization/84746
798 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
799 (phi_translate): Pass in destination ANTIC_OUT set.
800 (phi_translate_1): Likewise. For a simplified result lookup
801 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
802 (phi_translate_set): Adjust.
803 (do_pre_regular_insertion): Likewise.
804 (do_pre_partial_partial_insertion): Likewise.
806 2018-03-08 Martin Liska <mliska@suse.cz>
808 PR gcov-profile/84735
809 * doc/gcov.texi: Document usage of profile files.
810 * gcov-io.h: Document changes in the format.
812 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
816 * dwarf2out.c (struct dw_line_info_table): Update comments for
818 (FORCE_RESET_NEXT_VIEW): New.
819 (FORCE_RESETTING_VIEW_P): New.
820 (RESETTING_VIEW_P): Check for -1 too.
821 (ZERO_VIEW_P): Likewise.
822 (new_line_info_table): Force-reset next view.
823 (dwarf2out_begin_function): Likewise.
824 (dwarf2out_source_line): Simplify zero_view_p initialization.
825 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
826 view directly. Omit view when omitting .loc at line 0.
828 2018-03-08 Jakub Jelinek <jakub@redhat.com>
830 PR tree-optimization/84740
831 * tree-switch-conversion.c (process_switch): Call build_constructors
832 only if info.phi_count is non-zero.
834 PR tree-optimization/84739
835 * tree-tailcall.c (find_tail_calls): Check call arguments against
836 DECL_ARGUMENTS (current_function_decl) rather than
837 DECL_ARGUMENTS (func) when checking for tail recursion.
839 2018-03-07 Jakub Jelinek <jakub@redhat.com>
841 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
842 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
843 Volker Reichelt's entry and add entries for people that perform
844 GCC fuzzy testing and report numerous bugs.
846 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
849 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
850 readonly data in sdata, if that is disabled.
851 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
852 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
853 -mreadonly-in-sdata option.
855 2018-03-07 Martin Sebor <msebor@redhat.com>
857 PR tree-optimization/84468
858 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
859 basic block when looking for nul assignment.
861 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
864 * except.h (output_function_exception_table): Adjust prototype.
865 * except.c (output_function_exception_table): Remove FNNAME parameter
866 and add SECTION parameter. Ouput one part of the table at a time.
867 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
868 the first part of the exception table and emit unwind directives.
869 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
870 (i386_pe_seh_cold_init): Likewise.
871 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
872 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
873 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
874 (ix86_output_call_insn): Emit a nop in one more case for SEH.
875 * config/i386/winnt.c: Include except.h.
876 (struct seh_frame_state): Add reg_offset, after_prologue and
877 in_cold_section fields.
878 (i386_pe_seh_end_prologue): Set seh->after_prologue.
879 (i386_pe_seh_cold_init): New function.
880 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
881 to seh->in_cold_section.
882 (seh_emit_push): Record the offset of the push.
883 (seh_emit_save): Record the offet of the save.
884 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
885 Test seh->after_prologue to disregard the epilogue.
886 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
887 (i386_pe_end_cold_function): New function.
889 2018-03-07 Jakub Jelinek <jakub@redhat.com>
892 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
893 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
896 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
898 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
899 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
902 * multiple_target.c: Include tree-inline.h and intl.h.
903 (expand_target_clones): Diagnose and fail if node->definition and
904 !tree_versionable_function_p (node->decl).
906 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
908 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
910 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
911 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
912 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
914 2018-03-06 Jakub Jelinek <jakub@redhat.com>
917 * combine.c (try_combine): Use reg_or_subregno instead of handling
918 just paradoxical SUBREGs and REGs.
920 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
922 * config/arc/arc.c (arc_finalize_pic): Remove function.
923 (arc_must_save_register): We use single base PIC register, remove
924 checks to save/restore the PIC register.
925 (arc_expand_prologue): Likewise.
926 * config/arc/arc-protos.h (arc_set_default_type_attributes):
928 (arc_verify_short): Likewise.
929 (arc_attr_type): Likewise.
930 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
931 (walk_stores): Likewise.
932 (arc_address_cost): Make it static.
933 (arc_verify_short): Likewise.
934 (branch_dest): Likewise.
935 (arc_attr_type): Likewise.
936 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
937 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
938 (arc_final_prescan_insn): Remove inserting the nops due to
939 hardware hazards. It is done in reorg step.
940 (insn_length_variant_t): Remove.
941 (insn_length_parameters_t): Likewise.
942 (arc_insn_length_parameters): Likewise.
943 (arc_get_insn_variants): Likewise.
944 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
946 2018-03-06 Jakub Jelinek <jakub@redhat.com>
949 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
952 PR tree-optimization/84687
953 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
955 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
957 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
959 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
960 Rename to ppc_speculation_barrier.
961 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
962 __builtin_ppc_speculation_barrier.
964 2018-03-05 Jakub Jelinek <jakub@redhat.com>
967 * combine.c (combine_simplify_rtx): Don't try to simplify if
968 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
971 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
973 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
974 to 32 bytes when compiling for POWER9.
976 2018-03-05 Jakub Jelinek <jakub@redhat.com>
979 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
980 regparm >= 3 with no arg reg available also for calls with
981 flag_force_indirect_call. Pass decl to ix86_function_regparm.
984 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
986 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
988 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
991 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
993 2018-03-05 Richard Biener <rguenther@suse.de>
995 PR tree-optimization/84486
996 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
997 When inserting a __builtin_assume_aligned call set the LHS
998 SSA name alignment info accordingly.
1000 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1002 PR tree-optimization/84114
1003 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1004 Avoid reassociation of FLOAT_MODE addition.
1006 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1008 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1009 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1010 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1011 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1012 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1013 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1014 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1016 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1017 __builtin_ia32_wbinvd): New builtins.
1018 (SPECIAL_ARGS2): New.
1019 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1020 (SPECIAL_ARGS2): New.
1021 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1022 (ix86_valid_target_attribute_inner_p): Ditto.
1023 (ix86_init_mmx_sse_builtins): Add special_args2.
1024 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1025 TARGET_WBNOINVD_P): New.
1026 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1027 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1028 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1029 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1030 * config/i386/pconfigintrin.h: New file.
1031 * config/i386/wbnoinvdintrin.h: Ditto.
1032 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1033 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1035 2018-03-05 Richard Biener <rguenther@suse.de>
1037 PR tree-optimization/84670
1038 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1040 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1041 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1042 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1043 doesn't grow if all successors (recursively) were visited at least
1046 2018-03-05 Richard Biener <rguenther@suse.de>
1048 PR tree-optimization/84650
1049 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1050 if executed in the loop pipeline.
1052 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1054 * doc/configfiles.texi (Configuration Files): Move info about
1055 conditionalizing $target-protos.h to...
1056 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1057 differs from $target-protos.h.
1059 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1060 Chung-Ju Wu <jasonwucj@gmail.com>
1062 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1063 * config/nds32/nds32-multiple.md (setmemsi): Define.
1064 * config/nds32/nds32-memory-manipulation.c
1065 (nds32_gen_dup_4_byte_to_word_value): New.
1066 (emit_setmem_word_loop): New.
1067 (emit_setmem_byte_loop): New.
1068 (nds32_expand_setmem_loop): New.
1069 (nds32_expand_setmem_loop_v3m): New.
1070 (nds32_expand_setmem_unroll): New.
1071 (nds32_expand_setmem): New.
1073 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1074 Chung-Ju Wu <jasonwucj@gmail.com>
1076 * config/nds32/nds32-memory-manipulation.c
1077 (nds32_emit_load_store): New.
1078 (nds32_emit_post_inc_load_store): New.
1079 (nds32_emit_mem_move): New.
1080 (nds32_emit_mem_move_block): New.
1081 (nds32_expand_movmemsi_loop_unknown_size): New.
1082 (nds32_expand_movmemsi_loop_known_size): New.
1083 (nds32_expand_movmemsi_loop): New.
1084 (nds32_expand_movmemsi_unroll): New.
1085 (nds32_expand_movmemqi): Rename ...
1086 (nds32_expand_movmemsi): ... to this.
1087 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1088 (movmemsi): ... to this.
1089 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1090 (nds32_expand_movmemsi): ... to this.
1092 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1093 Monk Chiang <sh.chiang04@gmail.com>
1094 Chung-Ju Wu <jasonwucj@gmail.com>
1096 * config/nds32/nds32-protos.h
1097 (nds32_expand_load_multiple): New arguments.
1098 (nds32_expand_store_multiple): Ditto.
1099 (nds32_valid_multiple_load_store): Rename ...
1100 (nds32_valid_multiple_load_store_p): ... to this.
1101 * config/nds32/nds32-memory-manipulation.c
1102 (nds32_expand_load_multiple): Refine implementation.
1103 (nds32_expand_store_multiple): Ditto.
1104 * config/nds32/nds32-multiple.md
1105 (load_multiple): Update nds32_expand_load_multiple interface.
1106 (store_multiple): Update nds32_expand_store_multiple interface.
1107 * config/nds32/nds32-predicates.c
1108 (nds32_valid_multiple_load_store): Rename ...
1109 (nds32_valid_multiple_load_store_p): ... to this and refine
1111 * config/nds32/predicates.md
1112 (nds32_load_multiple_and_update_address_operation): New predicate.
1113 (nds32_store_multiple_and_update_address_operation): New predicate.
1115 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1116 Chung-Ju Wu <jasonwucj@gmail.com>
1118 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1119 (combo): New attribute.
1120 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1122 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1124 * config/nds32/nds32.opt: Change -mcmodel= default value.
1126 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1127 Monk Chiang <sh.chiang04@gmail.com>
1128 Chung-Ju Wu <jasonwucj@gmail.com>
1130 * config/nds32/constants.md (unspec_element): New enum.
1131 * config/nds32/constraints.md (Umw): New constraint.
1132 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1133 * config/nds32/nds32-intrinsic.md: Likewise.
1134 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1135 (nds32_valid_smw_lwm_base_p): New.
1136 (nds32_output_smw_single_word): New.
1137 (nds32_output_lmw_single_word): New.
1138 (nds32_expand_unaligned_load): New.
1139 (nds32_expand_unaligned_store): New.
1140 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1141 (nds32_output_smw_single_word): Declare.
1142 (nds32_output_lmw_single_word): Declare.
1143 (nds32_expand_unaligned_load): Declare.
1144 (nds32_expand_unaligned_store): Declare.
1145 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1146 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1147 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1148 NDS32_BUILTIN_UASTORE_DW.
1149 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1152 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1153 Kito Cheng <kito.cheng@gmail.com>
1154 Chung-Ju Wu <jasonwucj@gmail.com>
1156 * config/nds32/nds32-intrinsic.c
1157 (nds32_expand_builtin_null_ftype_reg): Delete.
1158 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1159 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1160 (nds32_read_argument): New.
1161 (nds32_legitimize_target): Ditto.
1162 (nds32_legitimize_argument): Ditto.
1163 (nds32_check_constant_argument): Ditto.
1164 (nds32_expand_unop_builtin): Ditto.
1165 (nds32_expand_unopimm_builtin): Ditto.
1166 (nds32_expand_binop_builtin): Ditto.
1167 (nds32_builtin_decl_impl): Ditto.
1168 (builtin_description): Ditto.
1169 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1170 (nds32_init_builtins_impl): Ditto.
1171 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1172 (nds32_builtin_decl): New.
1173 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1174 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1176 2018-03-02 Jeff Law <law@redhat.com>
1178 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1179 (redundant_insn, fill_simple_delay_slots): Likewise.
1180 (fill_slots_from_thread): Likewise.
1181 * resource.c (mark_referenced_resources): Likewise.
1182 (mark_set_resources, find_dead_or_set_registers): Likewise.
1184 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1186 * substring-locations.h (format_warning_va): Formatting fix for
1188 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1190 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1191 * substring-locations.c: Include intl.h.
1192 (format_warning_va): Turned into small wrapper around
1193 format_warning_n_va, renamed to ...
1194 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1195 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1197 (format_warning_at_substring_n): New function.
1198 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1199 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1200 format_warning_at_substring with just a shorter name instead of
1201 const function pointer.
1202 (fmtwarn_n): New function.
1203 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1204 appropriate, get rid of all the fmtstr temporaries, move conditionals
1205 with G_() wrapped string literals directly into fmtwarn arguments,
1206 cast dir.len to (int), formatting fixes.
1208 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1210 * doc/invoke.texi: Remove "Cilk Plus" references.
1212 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1213 Richard Biener <rguenther@suse.de>
1216 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1217 for error or warning attributes if CALL_FROM_THUNK_P is set.
1220 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1223 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1224 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1227 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1228 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1230 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1231 instead of -1U in last predictors element's probability member.
1233 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1236 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1237 arguments if they are comparable.
1239 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1241 PR tree-optimization/84634
1242 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1243 masks and masked_loop_p with a single loop_masks, making sure it's
1244 null for bb vectorization.
1246 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1248 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1249 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1250 loop->force_vectorize to check whether there is no alias.
1252 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1255 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1257 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1259 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1260 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1261 instead of a loop around prev_real_insn.
1262 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1266 * config/i386/i386.c (ix86_print_operand): Use conditional
1267 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1270 2018-03-02 Richard Biener <rguenther@suse.de>
1272 PR tree-optimization/84427
1273 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1274 (bitmap_set_subtract_values): Rewrite to handle multiple
1277 (prune_clobbered_mems): Likewise.
1278 (phi_translate): Take edge instead of pred/phiblock.
1279 (phi_translate_1): Likewise.
1280 (phi_translate_set): Likewise. Insert all translated
1281 exprs for a value into the set, keeping possibly multiple
1282 expressions per value.
1283 (compute_antic_aux): Adjust for phi_translate changes.
1284 When intersecting union the expressions and prune those
1285 not in the final value set, keeping possibly multiple
1286 expressions per value. Do not use value-insertion
1287 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1288 all expressions. Add verification that the value-sets
1289 only shrink during iteration.
1290 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1291 (do_pre_regular_insertion): Likewise.
1292 (do_pre_partial_partial_insertion): Likewise.
1294 2018-03-02 Richard Biener <rguenther@suse.de>
1297 * config/darwin.c (saved_debug_info_level): New static global.
1298 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1299 (darwin_asm_lto_end): Restore debug info generation settings.
1301 2018-03-01 Martin Liska <mliska@suse.cz>
1304 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1307 2018-03-01 Richard Biener <rguenther@suse.de>
1310 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1311 type annotation with LTO when debug was disabled at compile-time.
1313 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1315 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1317 (mips_final_postscan_insn): Likewise.
1319 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1321 PR rtl-optimization/84528
1322 * alias.c (init_alias_target): Add commentary.
1323 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1324 a unique base value if the frame pointer is not eliminated
1325 to the stack pointer.
1327 2018-03-01 Tom de Vries <tom@codesourcery.com>
1329 PR rtl-optimization/83327
1330 * lra-int.h (hard_regs_spilled_into): Declare.
1331 * lra.c (hard_regs_spilled_into): Define.
1332 (init_reg_info): Init hard_regs_spilled_into.
1333 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1334 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1335 (process_bb_lives): Handle hard_regs_spilled_into.
1336 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1337 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1339 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1341 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1342 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1343 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1344 * config/rs6000/aix72.h: New file.
1346 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1348 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1349 instead of warning_at with conditional singular and plural messages
1353 * stor-layout.c (update_alignment_for_field): For
1354 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1355 && !DECL_PACKED (field), do the alignment update, just use
1356 only desired_align instead of MAX (type_align, desired_align)
1358 (place_field): Don't do known_align < desired_align handling
1359 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1360 is non-NULL, instead do it after rli->prev_field handling and
1361 only if not within a bitfield word. For DECL_PACKED (field)
1362 use type_align of BITS_PER_UNIT.
1364 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1366 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1367 superfluous parentheses and trailing spaces.
1369 2018-02-28 Richard Biener <rguenther@suse.de>
1371 PR tree-optimization/84584
1372 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1373 SCoPs with fake exit edge.
1375 2018-02-28 Martin Liska <mliska@suse.cz>
1378 * timevar.c (timer::print): Fix format to properly print 100%
1381 2018-02-28 Richard Biener <rguenther@suse.de>
1384 * genmatch.c (capture_info::walk_match): Do not mark
1385 captured expressions without operands as expr_p given
1386 they act more like predicates and should be subject to
1387 "lost tail" side-effect preserving.
1389 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1391 PR rtl-optimization/81611
1392 * auto-inc-dec.c (attempt_change): Move dead note from
1393 mem_insn if it's the next use of regno
1394 (find_address): Take address use of reg holding
1395 non-incremented value. Add parm to limit search to the named
1397 (merge_in_block): Attempt to use a mem insn that is the next
1398 use of the original regno.
1400 2018-02-27 Martin Sebor <msebor@redhat.com>
1403 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1404 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1406 2018-02-27 Martin Sebor <msebor@redhat.com>
1408 PR translation/84207
1409 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1410 n argument to unsigned HOST_WIDE_INT.
1411 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1412 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1413 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1416 2018-02-27 Richard Biener <rguenther@suse.de>
1418 PR tree-optimization/84512
1419 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1420 Do not use the estimate returned from record_stmt_cost for
1421 the scalar iteration cost but sum properly using add_stmt_cost.
1423 2018-02-27 Richard Biener <rguenther@suse.de>
1425 PR tree-optimization/84466
1426 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1427 Adjust last change to less strictly validate use operands.
1429 2018-02-27 Martin Liska <mliska@suse.cz>
1431 PR gcov-profile/84548
1432 * gcov.c (process_file): Allow partial overlap and consider it
1433 also as group functions.
1434 (output_lines): Properly calculate range of lines for a group.
1436 2018-02-27 Martin Liska <mliska@suse.cz>
1438 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1439 'ggc' suffixes. Change first column width.
1440 (timer::print): Fix formatting of the column.
1442 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1444 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1445 preserve inline entry blocks for the sake of debug inline
1446 entry point markers alone.
1447 (remove_unused_locals): Suggest in comments a better place to
1448 force the preservation of inline entry blocks that are
1449 otherwise unused, but do not preserve them.
1451 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1453 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1455 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1458 * config/i386/constraints.md (Bs): Replace
1459 ix86_indirect_branch_register with
1460 TARGET_INDIRECT_BRANCH_REGISTER.
1462 * config/i386/i386.md (indirect_jump): Likewise.
1463 (tablejump): Likewise.
1464 (*sibcall_memory): Likewise.
1465 (*sibcall_value_memory): Likewise.
1466 Peepholes of indirect call and jump via memory: Likewise.
1467 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1468 (*sibcall_value_GOT_32): Likewise.
1469 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1470 (GOT_memory_operand): Likewise.
1471 (call_insn_operand): Likewise.
1472 (sibcall_insn_operand): Likewise.
1473 (GOT32_symbol_operand): Likewise.
1474 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1476 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1478 PR rtl-optimization/83496
1479 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1480 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1481 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1482 redundant insn, if any.
1483 (relax_delay_slots): Likewise.
1484 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1486 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1488 PR tree-optimization/83965
1489 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1490 that grouped statements are part of a reduction chain. Return
1491 true if the statement is not marked as a reduction itself but
1493 (vect_recog_dot_prod_pattern): Don't check whether the statement
1494 is part of a group here.
1495 (vect_recog_sad_pattern): Likewise.
1496 (vect_recog_widen_sum_pattern): Likewise.
1498 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1501 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1503 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1506 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1508 (ix86_output_indirect_function_return): New prototype.
1509 (ix86_split_simple_return_pop_internal): Likewise.
1510 * config/i386/i386.c (indirect_return_via_cx): New.
1511 (indirect_return_via_cx_bnd): Likewise.
1512 (indirect_thunk_name): Handle return va CX_REG.
1513 (output_indirect_thunk_function): Create alias for
1514 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1515 (ix86_output_indirect_jmp): Remove the bool argument.
1516 (ix86_output_indirect_function_return): New function.
1517 (ix86_split_simple_return_pop_internal): Likewise.
1518 * config/i386/i386.md (*indirect_jump): Don't pass false
1519 to ix86_output_indirect_jmp.
1520 (*tablejump_1): Likewise.
1521 (simple_return_pop_internal): Change it to define_insn_and_split.
1522 Call ix86_split_simple_return_pop_internal to split it for
1524 (simple_return_indirect_internal): Call
1525 ix86_output_indirect_function_return instead of
1526 ix86_output_indirect_jmp.
1528 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1531 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1532 memset and value initialization afterwards.
1534 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1536 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1538 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1541 * common/config/aarch64/aarch64-common.c
1542 (aarch_option_optimization_table[]): Switch
1543 off fomit-frame-pointer
1545 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1546 Chung-Ju Wu <jasonwucj@gmail.com>
1548 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1550 (store_multiple): Ditto.
1552 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1554 * config.gcc: Add --with-cpu support for nds32 target.
1555 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1556 * config/nds32/nds32.opt: Add -mcpu= option.
1558 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1560 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1561 isel=yes): Warn for these deprecated options.
1563 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1565 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1566 ISA_2_5_MASKS_EMBEDDED.
1568 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1570 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1571 p->max as pointers rather than using iterative_hash_expr.
1573 2018-02-23 Carl Love <cel@us.ibm.com>
1575 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1576 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1578 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1579 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1580 P8V_BUILTIN_VEC_VUNSIGNED2.
1582 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1585 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1586 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1588 * lra-constraints.c (curr_insn_transform): Set up
1589 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1591 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1592 * lra-eliminations.c (spill_pseudos): Ditto.
1593 (process_insn_for_elimination): Ditto.
1594 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1595 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1596 LRA_NON_CLOBBERED_ALT.
1598 2018-02-22 Martin Sebor <msebor@redhat.com>
1600 PR tree-optimization/84480
1601 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1602 to maybe_diag_stxncpy_trunc. Call it.
1603 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1604 from gimple_fold_builtin_strcpy. Print inlining stack.
1605 (handle_builtin_stxncpy): Print inlining stack.
1606 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1608 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1611 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1612 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1613 and -fcheck-pointer-bounds are used together.
1614 (indirect_thunk_prefix): New enum.
1615 (indirect_thunk_need_prefix): New function.
1616 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1617 "_nt" instead of "_bnd" for NOTRACK prefix.
1618 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1619 (output_indirect_thunk_function): Likewise.
1621 (ix86_code_end): Update output_indirect_thunk_function calls.
1622 (ix86_output_indirect_branch_via_reg): Replace
1623 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1624 (ix86_output_indirect_branch_via_push): Likewise.
1625 (ix86_output_function_return): Likewise.
1626 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1627 incompatible with -fcf-protection=branch and
1628 -fcheck-pointer-bounds.
1630 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1633 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1634 Change gcc_assert call to output_operand_lossage.
1636 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1638 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1640 2018-02-22 DJ Delorie <dj@redhat.com>
1641 Sebastian Perta <sebastian.perta@renesas.com>
1642 Oleg Endo <olegendo@gcc.gnu.org>
1644 * config/rx/rx.c (rx_rtx_costs): New function.
1645 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1647 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1649 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1651 2018-02-22 Martin Liska <mliska@suse.cz>
1654 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1655 Add "native" as a possible value.
1657 2018-02-22 Martin Liska <mliska@suse.cz>
1660 * config/i386/i386.c (ix86_option_override_internal):
1661 Add "native" as a possible value for -march and -mtune.
1663 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1666 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1667 to all type variants.
1669 PR tree-optimization/84503
1670 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1671 width as info->bitpos + info->bitsize - start.
1672 (merged_store_group::merge_overlapping): Simplify width computation.
1673 (check_no_overlap): New function.
1674 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1675 start + width and last_order of the group, fail if check_no_overlap
1677 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1678 to group if check_no_overlap fails.
1680 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1682 * config/rs6000/altivec.md: Delete contraint arguments to
1683 define_expand, define_split, and define_peephole2, and in
1684 define_insn_and_split if always unused.
1685 * config/rs6000/darwin.md: Ditto.
1686 * config/rs6000/dfp.md: Ditto.
1687 * config/rs6000/rs6000.md: Ditto.
1688 * config/rs6000/sync.md: Ditto.
1689 * config/rs6000/vector.md: Ditto.
1690 * config/rs6000/vsx.md: Ditto.
1692 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1694 * config/rs6000/altivec.md: Write output control strings as braced
1695 blocks instead of double-quoted strings.
1696 * config/rs6000/darwin.md: Ditto.
1697 * config/rs6000/rs6000.md: Ditto.
1698 * config/rs6000/vector.md: Ditto.
1699 * config/rs6000/vsx.md: Ditto.
1701 2018-02-21 Jason Merrill <jason@redhat.com>
1703 PR c++/84314 - ICE with templates and fastcall attribute.
1704 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1706 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1708 * ipa-cp.c (determine_versionability): Fix comment typos.
1710 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1713 * ipa-cp.c (determine_versionability): Do not version functions caling
1716 2018-02-21 Martin Liska <mliska@suse.cz>
1719 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1720 Add "native" as a possible value.
1721 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1722 the macro when native cpu detection is available.
1724 2018-02-21 Martin Liska <mliska@suse.cz>
1727 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1728 Add "native" as a possible value.
1729 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1730 when native cpu detection is available.
1732 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1733 Martin Sebor <msebor@redhat.com>
1735 PR tree-optimization/84478
1736 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1738 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1739 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1740 support which is conservatively correct, for 2 only stay conservative
1741 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1742 argument to the 2 argument get_range_strlen, adjust 6 arg
1743 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1745 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1746 (gimple_fold_builtin_strlen): Pass true as last argument to
1749 2018-02-20 Martin Sebor <msebor@redhat.com>
1752 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1753 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1754 (builtin_memref::builtin_memref): Factor out parts into
1755 set_base_and_offset and call it.
1757 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1760 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1761 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1762 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1763 search at the associated MODE_INT.
1765 2018-02-20 Jeff Law <law@redhat.com>
1768 PR tree-optimization/81592
1770 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1771 for range data rather than using global data.
1772 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1773 range data rather than using global data.
1774 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1775 pass it to children as needed.
1776 (struct directive::fmtresult): Similarly.
1777 (struct directive::set_width): Similarly.
1778 (struct directive::set_precision): Similarly.
1779 (format_integer, format_directive, parse_directive): Similarly.
1780 (format_none): Accept unnamed vr_values parameter.
1781 (format_percent, format_floating, format_character): Similarly.
1782 (format_string, format_plain): Similarly.
1783 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1784 the EVRP range analyzer for range data rather than using global data.
1785 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1786 gimple-ssa-evrp-analyze.h
1787 (class sprintf_dom_walker): Add after_dom_children member function.
1788 Add evrp_range_analyzer member.
1789 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1790 range analyzer as needed.
1791 (sprintf_dom_walker::after_dom_children): New member function.
1792 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1794 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1795 (evrp_range_analyzer::pop_to_marker): Likewise.
1797 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1799 PR tree-optimization/84419
1800 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1801 with the required type if its current type is compatible but
1804 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1807 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1808 after vectorization.
1810 2018-02-20 Martin Liska <mliska@suse.cz>
1813 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1814 possible values if we don't have a hint.
1816 2018-02-20 Martin Liska <mliska@suse.cz>
1820 * final.c (shorten_branches): Build align_tab array with one
1822 * opts.c (finish_options): Add alignment option limit check.
1823 (MAX_CODE_ALIGN): Likewise.
1824 (MAX_CODE_ALIGN_VALUE): Likewise.
1825 * doc/invoke.texi: Document maximum allowed option value for
1826 all -falign-* options.
1828 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1831 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1832 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1833 * var-tracking.c (emit_note_insn_var_location): Remove all references
1834 to NOTE_INSN_CALL_ARG_LOCATION.
1835 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1836 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1837 Use copy_rtx_if_shared.
1838 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1839 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1840 (dwarf2out_var_location): Remove handling of
1841 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1843 * final.c (final_scan_insn): Remove all references to
1844 NOTE_INSN_CALL_ARG_LOCATION.
1845 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1846 before dumping final insns.
1847 * except.c (emit_note_eh_region_end): Remove all references to
1848 NOTE_INSN_CALL_ARG_LOCATION.
1849 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1850 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1851 * config/arc/arc.c (hwloop_optimize): Likewise.
1852 * config/arm/arm.c (create_fix_barrier): Likewise.
1853 * config/s390/s390.c (s390_chunkify_start): Likewise.
1854 * config/sh/sh.c (find_barrier): Likewise.
1855 * config/i386/i386.c (rest_of_insert_endbranch,
1856 ix86_seh_fixup_eh_fallthru): Likewise.
1857 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1858 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1859 * config/frv/frv.c (frv_function_prologue): Likewise.
1860 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1862 (note_outside_basic_block_p): Remove all references to
1863 NOTE_INSN_CALL_ARG_LOCATION.
1864 * gengtype.c (adjust_field_rtx_def): Likewise.
1865 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1867 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1868 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1871 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1874 PR tree-optimization/84452
1875 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1876 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1879 2018-02-19 Martin Liska <mliska@suse.cz>
1882 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1884 2018-02-19 Martin Liska <mliska@suse.cz>
1885 Richard Sandiford <richard.sandiford@linaro.org>
1887 PR tree-optimization/82491
1888 * gimple-fold.c (get_base_constructor): Make earlier bail out
1891 2018-02-19 Carl Love <cel@us.ibm.com>
1893 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1894 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1896 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1897 P8V_BUILTIN_VEC_NEG.
1899 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1901 * config/rl78/rl78.md (movdf): New define expand.
1903 2018-02-19 Martin Liska <mliska@suse.cz>
1906 * doc/invoke.texi: Fix typo.
1907 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1909 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1911 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1912 handle rs6000_single_float and rs6000_double_float specially for
1915 2018-02-16 Jeff Law <law@redhat.com>
1917 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1920 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1923 * ipa-inline.c (inline_small_functions): Fix a typo.
1925 2018-02-16 Nathan Sidwell <nathan@acm.org>
1927 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1929 2018-02-16 Carl Love <cel@us.ibm.com>
1931 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1932 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1933 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1934 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1935 expansion to P8V_BUILTIN_VEC_FLOAT2.
1937 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1939 PR rtl-optimization/70023
1940 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1941 src_regno into account.
1943 2018-02-16 Carl Love <cel@us.ibm.com>
1945 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1946 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1947 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1948 * config/rs6000/rs6000.c: Remove case statements for
1949 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1950 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1951 and P9V_BUILTIN_VEC_VINSERT4B.
1952 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1953 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1954 * config/rs6000/vsx.md:
1955 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1958 2018-02-16 Carl Love <cel@us.ibm.com>
1960 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1962 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1964 * config/rs6000/rs6000-c.c: Add the definitions for
1965 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1966 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1967 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1968 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1969 definition for insert4b and define insn *insert3b_internal.
1970 * doc/extend.texi: Add documentation for vec_extract4b.
1972 2018-02-16 Nathan Sidwell <nathan@acm.org>
1974 * doc/extend.texi (Backwards Compatibility): Mention friend
1975 injection. Note for-scope is deprecated.
1976 * doc/invoke.texi (-ffriend-injection): Deprecate.
1978 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1980 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1981 that moved to I2, also allow destinations that are a paradoxical
1982 subreg (instead of a normal reg).
1984 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1987 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1990 2018-02-16 Richard Biener <rguenther@suse.de>
1992 PR tree-optimization/84037
1993 PR tree-optimization/84016
1995 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1996 Adjust vec_construct for the fact we need additional higher latency
1997 128bit inserts for AVX256 and AVX512 vector builds.
1998 (ix86_add_stmt_cost): Scale vector construction cost for
2001 2018-02-16 Richard Biener <rguenther@suse.de>
2003 PR tree-optimization/84417
2004 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2005 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2006 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2008 2018-02-16 Martin Liska <mliska@suse.cz>
2011 * internal-fn.def (ASAN_CHECK): Set proper flags.
2012 (ASAN_MARK): Likewise.
2014 2018-02-16 Julia Koval <julia.koval@intel.com>
2016 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2017 from PTA_CANNONLAKE.
2019 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2022 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2023 Use ++iter rather than iter++ for std::list iterators.
2024 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2025 defer deleting them until all nodes in the forest are processed. Do
2026 free even leaf nodes. Change to_process into auto_vec.
2029 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2030 * vec.h (vec_default_construct): Use memset instead of placement new
2031 if BROKEN_VALUE_INITIALIZATION is defined.
2032 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2033 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2036 PR rtl-optimization/83723
2037 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2038 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2039 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2041 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2043 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2045 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2047 PR rtl-optimization/81443
2048 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2049 from inner REGs to paradoxical SUBREGs.
2051 2018-02-16 Richard Biener <rguenther@suse.de>
2053 PR tree-optimization/84399
2054 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2055 For operands we can analyze at their definition make sure we can
2056 analyze them at each use as well.
2058 2018-02-16 Richard Biener <rguenther@suse.de>
2060 PR tree-optimization/84190
2061 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2062 volatile accesses if the decl isn't volatile.
2064 2018-02-15 Jason Merrill <jason@redhat.com>
2066 PR c++/84314 - ICE with templates and fastcall attribute.
2067 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2068 TYPE_CANONICAL on an existing type.
2070 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2072 PR tree-optimization/84383
2073 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2074 dstoff nor call operand_equal_p if dstbase is NULL.
2076 PR tree-optimization/84334
2077 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2078 also a CONSTANT_CLASS_P, punt.
2080 2018-02-14 Jim Wilson <jimw@sifive.com>
2082 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2083 first SMALL_OPERAND check. New local min_second_step. Move assert
2084 to where locals are set. Add TARGET_RVC support.
2085 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2087 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2089 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2091 2018-02-14 Martin Sebor <msebor@redhat.com>
2093 PR tree-optimization/83698
2094 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2095 arrays constrain the offset range to their bounds.
2096 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2097 (builtin_access::overlap): Avoid setting the size of overlap if it's
2099 (maybe_diag_overlap): Also consider arrays when deciding what values
2100 of offsets to include in diagnostics.
2102 2018-02-14 Martin Sebor <msebor@redhat.com>
2105 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2106 that correspond to the kind of a declaration.
2108 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2111 * config/pa/pa.md: Load address of PIC label using the linkage table
2112 if the label is nonlocal.
2114 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2116 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2117 warning message if user requests -maltivec=be.
2118 * doc/invoke.texi: Document deprecation of -maltivec=be.
2120 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2123 * config/rs6000/rs6000-c.c: Update definitions for
2124 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2125 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2127 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2130 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2131 add _get_ssp intrinsics. Remove argument from
2132 __builtin_ia32_rdssp[d|q].
2133 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2134 * config/i386/i386-builtin.def: Remove argument from
2135 __builtin_ia32_rdssp[d|q].
2136 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2137 ix86_expand_special_args_builtin for _rdssp[d|q].
2138 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2139 Clear register before usage.
2140 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2141 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2143 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2145 PR tree-optimization/84357
2146 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2147 operand 1 of an ARRAY_REF too.
2149 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2152 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2153 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2155 (set_of_reg): New struct.
2156 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2157 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2158 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2160 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2161 Split into bitclr, bitset, bitinvert patterns if appropriate.
2162 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2163 use rx_fuse_in_memory_bitop.
2164 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2165 to named insn, correct maximum insn length.
2167 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2170 * machmode.def: Define a complex mode for PARTIAL_INT.
2171 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2173 * doc/rtl.texi: Document CSPImode.
2174 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2176 (msp430_hard_regno_nregs_with_padding): Likewise.
2178 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2181 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2183 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2185 PR rtl-optimization/84169
2186 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2187 we generated a parallel as new i3 and we split that to new i2 and i3
2188 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2189 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2190 those to i2, not i1. Partially rewrite this scan code.
2192 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2195 * stor-layout.c (place_field): For variable length fields, adjust
2196 offset_align afterwards not just based on the field's alignment,
2197 but also on the size.
2200 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2201 of exps and logs in the use_exp2 case.
2203 2018-02-13 Jeff Law <law@redhat.com>
2205 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2208 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2211 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2215 * common.opt (gas-loc-support, gas-locview-support): New.
2216 (ginline-points, ginternal-reset-location-views): New.
2217 * doc/invoke.texi: Document them. Use @itemx where intended.
2218 (gvariable-location-views): Adjust.
2219 * target.def (reset_location_view): New.
2220 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2221 (TARGET_RESET_LOCATION_VIEW): New.
2222 * doc/tm.texi: Rebuilt.
2223 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2224 (dwarf2out_default_as_locview_support): New.
2225 (output_asm_line_debug_info): Use option variables.
2226 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2227 (output_loc_list): Likewise.
2228 (add_high_low_attributes): Check option variables.
2229 Don't output entry view attribute in strict mode.
2230 (gen_inlined_subroutine_die): Check option variables.
2231 (dwarf2out_inline_entry): Likewise.
2232 (init_sections_and_labels): Likewise.
2233 (dwarf2out_early_finish): Likewise.
2234 (maybe_reset_location_view): New, from...
2235 (dwarf2out_var_location): ... here. Call it.
2236 * debug.h (dwarf2out_default_as_loc_support): Declare.
2237 (dwarf2out_default_as_locview_support): Declare.
2238 * hooks.c (hook_int_rtx_insn_0): New.
2239 * hooks.h (hook_int_rtx_insn_0): Declare.
2240 * toplev.c (process_options): Take -gas-loc-support and
2241 -gas-locview-support from dwarf2out. Enable
2242 -gvariable-location-views by default only with locview
2243 assembler support. Enable -ginternal-reset-location-views by
2244 default only if the target defines the corresponding hook.
2245 Enable -ginline-points by default if location views are
2246 enabled; force it disabled if statement frontiers are
2248 * tree-inline.c (expand_call_inline): Check option variables.
2249 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2251 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2253 PR tree-optimization/84321
2254 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2255 handling. Also check whether the anti-range contains any values
2256 that satisfy the mask; switch to a VR_RANGE if not.
2258 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2261 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2263 2018-02-13 Martin Jambor <mjambor@suse.cz>
2266 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2267 of call statements, also set location of a load to a temporary.
2269 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2271 * config/rl78/rl78.c (add_vector_labels): New function.
2272 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2273 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2274 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2275 which checks that no arguments are passed.
2276 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2277 * doc/extend.texi: Documentation for the new attribute.
2279 2018-02-13 Andreas Schwab <schwab@suse.de>
2281 * config/riscv/linux.h (CPP_SPEC): Define.
2283 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2286 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2287 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2288 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2289 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2290 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2291 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2292 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2293 temporarily for AES and PCLMUL builtins.
2295 PR tree-optimization/84339
2296 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2297 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2301 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2302 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2303 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2305 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2307 * omp-simd-clone.h: New file.
2308 * omp-simd-clone.c: Include omp-simd-clone.h.
2309 (expand_simd_clones): No longer static.
2310 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2311 cgraph.h and omp-simd-clone.h.
2312 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2313 (vect_recog_widen_shift_pattern): Formatting fix.
2314 (vect_pattern_recog_1): Don't check optab for calls.
2317 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2318 operands[2] into a REG before using gen_lowpart on it.
2320 2018-02-12 Jeff Law <law@redhat.com>
2323 * config/sh/sh.c (find_barrier): Consider a sibling call
2326 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2327 successfully back substituting a reg.
2329 2018-02-12 Richard Biener <rguenther@suse.de>
2331 PR tree-optimization/84037
2332 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2333 parameter, move visited init to caller.
2334 (vect_slp_analyze_operations): Separate cost from validity
2335 check, initialize visited once for all instances.
2336 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2338 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2340 (vect_model_promotion_demotion_cost): Likewise.
2341 (vectorizable_bswap): Guard cost modeling with !slp_node
2342 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2344 (vectorizable_call): Likewise.
2345 (vectorizable_conversion): Likewise.
2346 (vectorizable_assignment): Likewise.
2347 (vectorizable_shift): Likewise.
2348 (vectorizable_operation): Likewise.
2349 (vectorizable_store): Likewise.
2350 (vectorizable_load): Likewise.
2351 (vectorizable_condition): Likewise.
2352 (vectorizable_comparison): Likewise.
2354 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2357 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2358 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2359 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2361 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2364 * params.def (inline-min-speedup): Increase from 8 to 15.
2365 (max-inline-insns-auto): Decrease from 40 to 30.
2366 * ipa-split.c (consider_split): Add some buffer for function to
2367 be considered inlining candidate.
2368 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2371 2018-02-12 Richard Biener <rguenther@suse.de>
2373 PR tree-optimization/84037
2374 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2375 matched stmts if we cannot swap the non-matched ones.
2377 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2379 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2380 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2381 _mm_maskz_scalef_round_ss): New intrinsics.
2382 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2383 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2384 __builtin_ia32_scalefss_round): Remove.
2385 (__builtin_ia32_scalefsd_mask_round,
2386 __builtin_ia32_scalefss_mask_round): New intrinsics.
2387 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2388 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2389 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2390 "<round_constraint>")): Changed to ...
2391 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2392 "<round_scalar_constraint>")): ... this.
2393 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2394 %0, %1, %2<round_op3>}"): Changed to ...
2395 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2396 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2397 %2<round_scalar_mask_op3>}"): ... this.
2398 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2400 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2402 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2403 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2404 (_mm_maskz_sqrt_round_ss): New intrinsics.
2405 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2406 (__builtin_ia32_sqrtsd_mask_round)
2407 (__builtin_ia32_sqrtss_mask_round): New builtins.
2408 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2409 (__builtin_ia32_sqrtss_round): Remove.
2410 (__builtin_ia32_sqrtsd_mask_round)
2411 (__builtin_ia32_sqrtss_mask_round): New builtins.
2412 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2413 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2414 ((match_operand:VF_128 1 "vector_operand"
2415 "xBm,<round_constraint>")): Changed to ...
2416 ((match_operand:VF_128 1 "vector_operand"
2417 "xBm,<round_scalar_constraint>")): ... this.
2418 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2419 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2420 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2421 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2422 %<iptr>1<round_scalar_mask_op3>}): ... this.
2423 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2424 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2426 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2429 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2430 Cast vec_cmpeq result to correct type.
2431 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2432 Cast vec_cmpgt result to correct type.
2434 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2436 * final.c (final_scan_insn_1): Renamed from...
2437 (final_scan_insn): ... this. New wrapper, to recover
2438 seen from the outermost call in recursive ones.
2439 * config/sparc/sparc.c (output_return): Drop seen from call.
2440 (output_sibcall): Likewise.
2441 * config/visium/visium.c (output_branch): Likewise.
2443 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2445 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2448 2018-02-10 Alan Modra <amodra@gmail.com>
2451 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2452 Specify LR as an input.
2454 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2457 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2458 remove_member_access_dummy_vars): New functions.
2459 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2460 lower_omp_1, execute_lower_omp): Use them.
2462 PR rtl-optimization/84308
2463 * shrink-wrap.c (spread_components): Release todo vector.
2465 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2467 PR rtl-optimization/57193
2468 * ira-color.c (struct allocno_color_data): Add member
2469 conflict_allocno_hard_prefs.
2470 (update_conflict_allocno_hard_prefs): New.
2471 (bucket_allocno_compare_func): Add a preference based on
2472 conflict_allocno_hard_prefs.
2473 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2474 (color_allocnos): Remove a dead code. Initiate
2475 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2477 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2480 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2481 constraint from =wa to wa. Avoid a subreg on the output operand,
2482 instead use a pseudo and subreg it in a move.
2483 (p9_xxbrd_<mode>): Changed to ...
2484 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2485 (p9_xxbrd_v2df): New expander.
2486 (p9_xxbrw_<mode>): Changed to ...
2487 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2488 (p9_xxbrw_v4sf): New expander.
2490 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2492 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2494 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2497 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2498 multiply in 32-bit mode.
2499 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2500 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2503 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2505 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2506 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2507 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2508 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2510 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2513 * dwarf2out.c (is_trivial_indirect_ref): New function.
2514 (dwarf2out_late_global_decl): Do not generate a location
2515 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2516 and that are not defined in the current unit.
2518 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2520 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2521 instead of a libcall for UNORDERED.
2523 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2526 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2527 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2529 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2532 * config/s390/s390.c (s390_set_current_function): Invoke
2533 s390_indirect_branch_settings also if fndecl didn't change.
2535 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2537 * config/rs6000/rs6000.md (blockage): Set length to zero.
2539 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2541 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2543 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2546 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2547 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2551 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2552 PARALLEL incoming that failed vt_get_decl_and_offset check.
2555 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2556 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2558 (get_variable_section): For decls in named .bss* sections pass true as
2559 second argument to bss_initializer_p.
2561 2018-02-09 Marek Polacek <polacek@redhat.com>
2562 Jakub Jelinek <jakub@redhat.com>
2565 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2566 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2567 Sync some changes from cxx_fold_indirect_ref.
2569 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2571 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2573 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2574 (BLOCK_INLINE_ENTRY_LABEL): New.
2575 (dwarf2out_var_location): Disregard inline entry markers.
2576 (inline_entry_data): New struct.
2577 (inline_entry_data_hasher): New hashtable type.
2578 (inline_entry_data_hasher::hash): New.
2579 (inline_entry_data_hasher::equal): New.
2580 (inline_entry_data_table): New variable.
2581 (add_high_low_attributes): Add DW_AT_entry_pc and
2582 DW_AT_GNU_entry_view attributes if a pending entry is found
2583 in inline_entry_data_table. Add old entry_pc attribute only
2584 if debug nonbinding markers are disabled.
2585 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2586 markers are enabled.
2587 (block_within_block_p, dwarf2out_inline_entry): New.
2588 (dwarf2out_finish): Check that no entries remained in
2589 inline_entry_data_table.
2590 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2591 (final_scan_insn, notice_source_line): Likewise.
2592 (rest_of_clean_state): Skip inline entry markers.
2593 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2595 * gimple.c (gimple_build_debug_inline_entry): New.
2596 * gimple.h (enum gimple_debug_subcode): Add
2597 GIMPLE_DEBUG_INLINE_ENTRY.
2598 (gimple_build_debug_inline_entry): Declare.
2599 (gimple_debug_inline_entry_p): New.
2600 (gimple_debug_nonbind_marker_p): Adjust.
2601 * insn-notes.def (INLINE_ENTRY): New.
2602 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2603 inline entry marker notes.
2604 (print_insn): Likewise.
2605 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2606 (INSN_DEBUG_MARKER_KIND): Likewise.
2607 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2608 * tree-inline.c (expand_call_inline): Build and insert
2609 debug_inline_entry stmt.
2610 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2611 inline entry blocks early, if nonbind markers are enabled.
2612 (dump_scope_block): Dump fragment info.
2613 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2614 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2615 (gimple_build_debug_inline_entry): New.
2616 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2617 Enable/disable inline entry points too.
2618 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2619 (DEBUG_INSN): Describe inline entry markers.
2621 * common.opt (gvariable-location-views): New.
2622 (gvariable-location-views=incompat5): New.
2623 * config.in: Rebuilt.
2624 * configure: Rebuilt.
2625 * configure.ac: Test assembler for view support.
2626 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2627 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2628 * dwarf2out.c (var_loc_view): New typedef.
2629 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2630 (dwarf2out_locviews_in_attribute): New.
2631 (dwarf2out_locviews_in_loclist): New.
2632 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2633 (enum dw_line_info_opcode): Add LI_adv_address.
2634 (struct dw_line_info_table): Add view.
2635 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2636 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2637 (zero_view_p): New variable.
2638 (ZERO_VIEW_P): New macro.
2639 (output_asm_line_debug_info): New.
2640 (struct var_loc_node): Add view.
2641 (add_AT_view_list, AT_loc_list): New.
2642 (add_var_loc_to_decl): Add view param. Test it against last.
2643 (new_loc_list): Add view params. Record them.
2644 (AT_loc_list_ptr): Handle loc and view lists.
2645 (view_list_to_loc_list_val_node): New.
2646 (print_dw_val): Handle dw_val_class_view_list.
2647 (size_of_die): Likewise.
2648 (value_format): Likewise.
2649 (loc_list_has_views): New.
2650 (gen_llsym): Set vl_symbol too.
2651 (maybe_gen_llsym, skip_loc_list_entry): New.
2652 (dwarf2out_maybe_output_loclist_view_pair): New.
2653 (output_loc_list): Output view list or entries too.
2654 (output_view_list_offset): New.
2655 (output_die): Handle dw_val_class_view_list.
2656 (output_dwarf_version): New.
2657 (output_compilation_unit_header): Use it.
2658 (output_skeleton_debug_sections): Likewise.
2659 (output_rnglists, output_line_info): Likewise.
2660 (output_pubnames, output_aranges): Update version comments.
2661 (output_one_line_info_table): Output view numbers in asm comments.
2662 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2663 Call maybe_gen_llsym.
2664 (loc_list_from_tree_1): Adjust.
2665 (add_AT_location_description): Create view list attribute if
2666 needed, check it's absent otherwise.
2667 (convert_cfa_to_fb_loc_list): Adjust.
2668 (maybe_emit_file): Call output_asm_line_debug_info for test.
2669 (dwarf2out_var_location): Reset views as needed. Precompute
2670 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2671 attribute. Set view.
2672 (new_line_info_table): Reset next view.
2673 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2674 (dwarf2out_source_line): Likewise. Output view resets and labels to
2675 the assembler, or select appropriate line info opcodes.
2676 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2677 (optimize_string_length): Catch it. Adjust.
2678 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2679 dw_val_class_view_list, and remove it if no longer needed.
2680 (hash_loc_list): Hash view numbers.
2681 (loc_list_hasher::equal): Compare them.
2682 (optimize_location_lists): Check whether a view list symbol is
2683 needed, and whether the locview attribute is present, and
2684 whether they match. Remove the locview attribute if no longer
2686 (index_location_lists): Call skip_loc_list_entry for test.
2687 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2688 Use output_dwarf_version.
2689 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2690 (struct dw_val_node): Add val_view_list.
2691 * final.c (SEEN_NEXT_VIEW): New.
2692 (set_next_view_needed): New.
2693 (clear_next_view_needed): New.
2694 (maybe_output_next_view): New.
2695 (final_start_function): Rename to...
2696 (final_start_function_1): ... this. Take pointer to FIRST,
2697 add SEEN parameter. Emit param bindings in the initial view.
2698 (final_start_function): Reintroduce SEEN-less interface.
2699 (final): Rename to...
2700 (final_1): ... this. Take SEEN parameter. Output final pending
2701 next view at the end.
2702 (final): Reintroduce seen-less interface.
2703 (final_scan_insn): Output pending next view before switching
2704 sections or ending a block. Mark the next view as needed when
2705 outputting variable locations. Notify debug backend of section
2706 changes, and of location view changes.
2707 (rest_of_handle_final): Adjust.
2708 * toplev.c (process_options): Autodetect value for debug variable
2709 location views option. Warn on incompat5 without -gdwarf-5.
2710 * doc/invoke.texi (gvariable-location-views): New.
2711 (gvariable-location-views=incompat5): New.
2712 (gno-variable-location-views): New.
2714 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2716 PR tree-optimization/84136
2717 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2718 that the result of find_edge is non-NULL.
2720 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2723 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2724 storing integer register in SImode. Fix cost of 256 and 512
2725 byte aligned SSE register store.
2727 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2729 * config/i386/i386.c (ix86_multiplication_cost): Fix
2730 multiplication cost for TARGET_AVX512DQ.
2732 2018-02-08 Marek Polacek <polacek@redhat.com>
2734 PR tree-optimization/84238
2735 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2738 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2740 PR tree-optimization/84265
2741 * tree-vect-stmts.c (vectorizable_store): Don't treat
2742 VMAT_CONTIGUOUS accesses as grouped.
2743 (vectorizable_load): Likewise.
2745 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2747 PR tree-optimization/81635
2748 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2749 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2750 (test_round_for_mask): New functions.
2751 (wide_int_cc_tests): Call test_round_for_mask.
2752 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2753 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2754 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2755 range returned by get_range_info.
2757 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2760 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2761 * symtab.c: Include builtins.h
2762 (symtab_node::output_to_lto_symbol_table_p): Move here
2763 from lto-streamer-out.c:output_symbol_p.
2764 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2765 (output_symbol_p): Move all logic to symtab.c
2766 (produce_symtab): Update.
2768 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2770 * config/s390/s390-opts.h (enum indirect_branch): Define.
2771 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2772 (s390_indirect_branch_via_thunk)
2773 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2774 (enum s390_indirect_branch_type): Define.
2775 * config/s390/s390.c (struct s390_frame_layout, struct
2776 machine_function): Remove.
2777 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2778 (indirect_branch_table_label_no, indirect_branch_table_name):
2780 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2781 (enum s390_indirect_branch_option): Define.
2782 (s390_return_addr_from_memory): New function.
2783 (s390_handle_string_attribute): New function.
2784 (s390_attribute_table): Add new attribute handler.
2785 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2786 (s390_indirect_branch_via_thunk): New function.
2787 (s390_indirect_branch_via_inline_thunk): New function.
2788 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2789 sibling call optimization for non z10 compiles.
2790 (s390_emit_call): Force indirect branch target to be a single
2791 register. Add r1 clobber for non-z10 compiles.
2792 (s390_emit_epilogue): Emit return jump via return_use expander.
2793 (s390_reorg): Handle JUMP_INSNs as execute targets.
2794 (s390_option_override_internal): Perform validity checks for the
2795 new command line options.
2796 (s390_indirect_branch_attrvalue): New function.
2797 (s390_indirect_branch_settings): New function.
2798 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2799 (s390_output_indirect_thunk_function): New function.
2800 (s390_code_end): Implement target hook.
2801 (s390_case_values_threshold): Implement target hook.
2802 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2804 * config/s390/s390.h (struct s390_frame_layout)
2805 (struct machine_function): Move here from s390.c.
2806 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2807 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2808 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2809 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2810 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2811 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2812 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2813 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2814 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2815 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2816 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2817 (mnemonic attribute): Add values which aren't recognized
2819 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2820 pattern for branch conversion. Fix mnemonic attribute.
2821 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2822 indirect branch via thunk if requested.
2823 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2824 ("*indirect_jump"): Disable for branch conversion using out of
2826 ("indirect_jump_via_thunk<mode>_z10")
2827 ("indirect_jump_via_thunk<mode>")
2828 ("indirect_jump_via_inlinethunk<mode>_z10")
2829 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2830 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2831 ("casesi_jump_via_inlinethunk<mode>_z10")
2832 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2833 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2834 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2835 ("*indirect2_jump"): Disable for branch conversion.
2836 ("casesi_jump"): Turn into expander and expand patterns for branch
2838 ("return_use"): New expander.
2839 ("*return"): Emit return via thunk and rename it to ...
2840 ("*return<mode>"): ... this one.
2841 * config/s390/s390.opt: Add new options and and enum for the
2844 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2846 * lra-constraints.c (match_reload): Unconditionally use
2847 gen_lowpart_SUBREG, rather than selecting between that
2848 and equivalent gen_rtx_SUBREG code.
2850 2018-02-08 Richard Biener <rguenther@suse.de>
2852 PR tree-optimization/84233
2853 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2854 changed flag instead of boguously re-using phi_inserted.
2856 2018-02-08 Martin Jambor <mjambor@suse.cz>
2858 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2859 static local variables.
2861 2018-02-08 Richard Biener <rguenther@suse.de>
2863 PR tree-optimization/84278
2864 * tree-vect-stmts.c (vectorizable_store): When looking for
2865 smaller vector types to perform grouped strided loads/stores
2866 make sure the mode is supported by the target.
2867 (vectorizable_load): Likewise.
2869 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2871 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2872 Increase LDP/STP opportunities by adding adjacent callee-saves.
2874 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2876 PR rtl-optimization/84068
2877 PR rtl-optimization/83459
2878 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2880 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2882 PR tree-optimization/84224
2883 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2884 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2887 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2890 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2891 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2892 count, remove one USE.
2894 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2896 * doc/install.texi (Configuration): Document the
2897 --with-long-double-format={ibm,ieee} PowerPC configuration
2901 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2902 Convert from define_expand to be define_insn_and_split. Rework
2903 float/double/_Float128 conversions to QI/HI/SImode to work with
2904 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2905 conversions to QI/HImode types did a store and then a load to
2906 truncate the value. For conversions to VSX registers, don't split
2907 the insn, instead emit the code directly. Use the code iterator
2908 any_fix to combine signed and unsigned conversions.
2909 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2910 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2911 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2912 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2913 (fix_<mode>di2_hw): Likewise.
2914 (fixuns_<mode>di2_hw): Likewise.
2915 (fix_<mode>si2_hw): Likewise.
2916 (fixuns_<mode>si2_hw): Likewise.
2917 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2918 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2919 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2920 fix<uns>_trunc<SFDF:mode>si2_p8.
2921 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2923 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2924 (fix<uns>_<mode>_mem): Likewise.
2925 (fctiw<u>z_<mode>_mem): Likewise.
2926 (fix<uns>_<mode>_mem): Likewise.
2927 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2928 the register allocator from doing a direct move to the GPRs to do
2929 a store, and instead use the ISA 3.0 store byte/half-word from
2930 vector register instruction. For IEEE 128-bit floating point,
2931 also optimize stores of 32-bit ints.
2932 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2934 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2936 * genextract.c (push_pathstr_operand): New function to support
2938 (walk_rtx): Call push_pathstr_operand.
2939 (print_path): Support [a-zA-Z].
2941 2018-02-07 Richard Biener <rguenther@suse.de>
2943 PR tree-optimization/84037
2944 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2945 (cse_and_gimplify_to_preheader): Declare.
2946 (vect_get_place_in_interleaving_chain): Likewise.
2947 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2949 (_loop_vec_info::~_loop_vec_info): Delete it.
2950 (cse_and_gimplify_to_preheader): New function.
2951 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2952 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2953 (vectorizable_load): Likewise. For grouped stores always base
2954 the IV on the first element.
2955 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2956 condition before gimplifying.
2958 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2960 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2961 *DIV_EXPR and *MOD_EXPR.
2963 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2966 * config/i386/i386.c (ix86_option_override_internal): Mask out
2967 the CF_SET bit when checking -fcf-protection.
2969 2018-02-07 Tom de Vries <tom@codesourcery.com>
2972 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2975 2018-02-07 Richard Biener <rguenther@suse.de>
2977 PR tree-optimization/84204
2978 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2981 PR tree-optimization/84205
2982 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2983 special-case isl_ast_op_zdiv_r.
2985 PR tree-optimization/84223
2986 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2987 Only add conditions from within the region.
2988 (gather_bbs::after_dom_children): Adjust.
2990 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2993 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2994 * config/avr/avr.md: Only post-reload split REG-REG moves if
2995 either register is GENERAL_REG_P.
2997 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2999 PR tree-optimization/84235
3000 * tree-ssa-scopedtables.c
3001 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3002 if the subtraction is performed in floating point type where NaNs are
3003 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3004 build 1. Formatting fix.
3006 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3009 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3010 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3011 and skip it regardless of bb boundaries. Use CALL_P macro,
3012 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3013 unnecessarily, formatting fix.
3015 2018-02-06 Michael Collison <michael.collison@arm.com>
3017 * config/arm/thumb2.md:
3018 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3019 (*thumb_mov_notscc): Ditto.
3021 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3024 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3025 unsigned_fix, not "s".
3027 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3029 * configure.ac (gcc_fn_eh_frame_ro): New function.
3030 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3031 correct .eh_frame permissions.
3032 * configure: Regenerate.
3034 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3036 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3039 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3041 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3042 Display warning message for -mno-speculate-indirect-jumps.
3044 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3046 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3048 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3050 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3052 PR tree-optimization/84225
3053 * tree-eh.c (find_trapping_overflow): Only call
3054 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3056 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3059 * config/i386/i386.c: Reimplement the check of possible options
3060 -mibt/-mshstk conbination. Change error messages.
3061 * doc/invoke.texi: Fix a typo: remove extra '='.
3063 2018-02-06 Marek Polacek <polacek@redhat.com>
3065 PR tree-optimization/84228
3066 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3068 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3071 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3072 emitted arch directives.
3073 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3074 __ARM_FEATURE_COPROC before changing architectures.
3076 2018-02-06 Richard Biener <rguenther@suse.de>
3078 * config/i386/i386.c (print_reg): Fix typo.
3079 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3081 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3083 * configure: Regenerate.
3085 2018-02-05 Martin Sebor <msebor@redhat.com>
3087 PR tree-optimization/83369
3088 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3091 2018-02-05 Martin Liska <mliska@suse.cz>
3093 * doc/invoke.texi: Cherry-pick upstream r323995.
3095 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3097 * ira.c (ira_init_register_move_cost): Adjust comment.
3099 2018-02-05 Martin Liska <mliska@suse.cz>
3101 PR gcov-profile/84137
3102 * doc/gcov.texi: Fix typo in documentation.
3104 2018-02-05 Martin Liska <mliska@suse.cz>
3106 PR gcov-profile/83879
3107 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3108 using dlopen functionality.
3110 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3112 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3113 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3114 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3115 _mm_maskz_range_round_ss): New intrinsics.
3116 (__builtin_ia32_rangesd128_round)
3117 (__builtin_ia32_rangess128_round): Remove.
3118 (__builtin_ia32_rangesd128_mask_round,
3119 __builtin_ia32_rangess128_mask_round): New builtins.
3120 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3121 __builtin_ia32_rangess128_round): Remove.
3122 (__builtin_ia32_rangesd128_mask_round,
3123 __builtin_ia32_rangess128_mask_round): New builtins.
3124 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3125 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3126 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3127 "<round_saeonly_constraint>")): Changed to ...
3128 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3129 "<round_saeonly_scalar_constraint>")): ... this.
3130 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3131 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3132 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3133 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3134 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3136 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3138 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3140 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3141 Remove all values except native, 8540 and 8548.
3143 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3145 * config/i386/i386.c (ix86_output_function_return): Pass
3146 INVALID_REGNUM, instead of -1, as invalid register number to
3147 indirect_thunk_name and output_indirect_thunk.
3149 2018-02-02 Julia Koval <julia.koval@intel.com>
3151 * config.gcc: Add -march=icelake.
3152 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3153 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3154 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3155 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3156 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3157 (processor_target_table): Add icelake.
3158 (ix86_option_override_internal): Handle new PTAs.
3159 (get_builtin_code_for_version): Handle icelake.
3160 (M_INTEL_COREI7_ICELAKE): New.
3161 (fold_builtin_cpu): Handle icelake.
3162 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3163 * doc/invoke.texi: Add -march=icelake.
3165 2018-02-02 Julia Koval <julia.koval@intel.com>
3167 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3168 to wide_int_bitmask.
3169 * wide-int-bitmask.h: New.
3171 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3174 * config/i386/i386.md: Replace Pmode with word_mode in
3175 builtin_setjmp_setup and builtin_longjmp to support x32.
3177 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3181 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3183 (rs6000_supported_cpu_names): New static variable.
3184 (linux_cpu_translation_table): Likewise.
3185 (elf_platform) <cpu>: Define new static variable and use it.
3186 Translate kernel AT_PLATFORM name to canonical name if needed.
3187 Error if platform name is unknown.
3189 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3192 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3194 2018-02-01 Jeff Law <law@redhat.com>
3197 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3198 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3199 the scratch if RELEASE_VIA_POP is false.
3200 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3201 If we have to save a temporary register, decrement SIZE appropriately.
3202 Pass new arguments to release_scratch_register_on_entry.
3203 (ix86_adjust_stack_and_probe): Likewise.
3204 (ix86_emit_probe_stack_range): Pass new arguments to
3205 release_scratch_register_on_entry.
3207 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3209 PR rtl-optimization/84157
3210 * combine.c (change_zero_ext): Use REG_P predicate in
3211 front of HARD_REGISTER_P predicate.
3213 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3215 * config/avr/avr.c (avr_option_override): Move disabling of
3216 -fdelete-null-pointer-checks to...
3217 * common/config/avr/avr-common.c (avr_option_optimization_table):
3220 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3222 PR tree-optimization/81635
3223 * tree-data-ref.c (split_constant_offset_1): For types that
3224 wrap on overflow, try to use range info to prove that wrapping
3227 2018-02-01 Renlin Li <renlin.li@arm.com>
3230 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3232 (aarch64_register_move_cost): Likewise.
3233 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3235 (REG_CLASS_NAMES): Likewise.
3236 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3237 TAILCALL_ADDR_REGS. Remove IP registers.
3238 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3240 2018-02-01 Richard Biener <rguenther@suse.de>
3242 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3243 for specifying RPO order and allow NULL for that.
3244 * domwalk.c (dom_walker::dom_walker): Likewise.
3245 (dom_walker::walk): Handle NULL RPO order.
3246 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3248 (rewrite_update_dom_walker): Likewise.
3249 (mark_def_dom_walker): Likewise.
3251 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3253 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3254 (aarch64_maybe_expand_sve_subreg_move): Declare.
3255 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3256 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3258 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3259 that are semantically a reverse operation.
3260 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3261 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3262 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3264 (aarch64_can_change_mode_class): For big-endian, forbid changes
3265 between two SVE modes if they have different element sizes.
3267 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3269 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3270 the TImode handling for big-endian targets.
3272 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3274 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3275 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3277 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3278 Remove BSWAP handing for big-endian targets and use the form of
3279 LD1RQ appropariate for the mode.
3281 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3283 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3284 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3287 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3290 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3291 check for operands that need to go through aarch64_sve_reload_be.
3293 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3295 PR tree-optimization/81661
3296 PR tree-optimization/84117
3297 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3298 * tree-eh.c: Include gimplify.h.
3299 (find_trapping_overflow, replace_trapping_overflow,
3300 rewrite_to_non_trapping_overflow): New functions.
3301 * tree-vect-loop.c: Include tree-eh.h.
3302 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3303 * tree-data-ref.c: Include tree-eh.h.
3304 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3306 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3308 PR rtl-optimization/84123
3309 * combine.c (change_zero_ext): Check if hard register satisfies
3310 can_change_dest_mode before calling gen_lowpart_SUBREG.
3312 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3315 * ira.c (ira_init_register_move_cost): Remove assert.
3317 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3319 PR rtl-optimization/84071
3320 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3321 * doc/tm.texi: Regenerate.
3323 2018-01-31 Richard Biener <rguenther@suse.de>
3325 PR tree-optimization/84132
3326 * tree-data-ref.c (analyze_miv_subscript): Properly
3327 check whether evolution_function_is_affine_multivariate_p
3328 before calling gcd_of_steps_may_divide_p.
3330 2018-01-31 Julia Koval <julia.koval@intel.com>
3333 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3334 * config/i386/i386.md (rdpid_rex64) New.
3335 (rdpid): Make 32bit only.
3337 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3340 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3341 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3343 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3346 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3348 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3350 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3352 PR rtl-optimization/84071
3353 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3354 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3356 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3358 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3359 (arc_attribute_table): Add 'aux' attribute.
3360 (arc_in_small_data_p): Consider aux like variables.
3361 (arc_is_aux_reg_p): New function.
3362 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3363 (arc_get_aux_arg): New function.
3364 (prepare_move_operands): Handle aux-register access.
3365 (arc_handle_aux_attribute): New function.
3366 * doc/extend.texi (ARC Variable attributes): Add subsection.
3368 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3370 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3371 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3372 (arc_attribute_table): Add 'uncached' attribute.
3373 (arc_print_operand): Print '.di' flag for uncached memory
3375 (arc_in_small_data_p): Do not consider for small data the uncached
3377 (arc_is_uncached_mem_p): New function.
3378 * config/arc/predicates.md (compact_store_memory_operand): Check
3379 for uncached memory accesses.
3380 (nonvol_nonimm_operand): Likewise.
3381 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3383 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3386 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3387 falign-loops=): Add Optimization flag.
3389 2018-01-30 Jeff Law <law@redhat.com>
3392 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3393 INT_REGISTERS_SAVED. Check it prior to calling
3394 get_scratch_register_on_entry.
3395 (ix86_adjust_stack_and_probe): Similarly.
3396 (ix86_emit_probe_stack_range): Similarly.
3397 (ix86_expand_prologue): Corresponding changes.
3399 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3402 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3403 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3405 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3408 * lra-constraints.c (curr_insn_transform): Process AND in the
3411 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3413 PR rtl-optimization/83986
3414 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3415 dependence against last_pending_memory_flush in addition to
3418 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3420 PR tree-optimization/81611
3421 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3424 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3427 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3430 2018-01-30 Richard Biener <rguenther@suse.de>
3431 Jakub Jelinek <jakub@redhat.com>
3433 PR tree-optimization/84111
3434 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3435 inner loops added during recursion, as they don't have up-to-date
3438 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3441 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3442 (can_inline_edge_by_limits_p): ... here.
3443 (can_early_inline_edge_p, check_callers,
3444 update_caller_keys, update_callee_keys, recursive_inlining,
3445 add_new_edges_to_heap, speculation_useful_p,
3446 inline_small_functions,
3447 inline_small_functions, flatten_function,
3448 inline_to_all_callers_1): Update.
3450 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3452 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3455 2018-01-30 Richard Biener <rguenther@suse.de>
3457 PR tree-optimization/83008
3458 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3459 invariant and constant vector uses in stmts when they need
3462 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3465 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3466 * configure: Regenerate.
3468 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3470 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3472 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3473 Use gen_rtx_REG rather than gen_lowpart.
3475 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3477 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3478 rather than 0 when creating partial subregs.
3480 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3482 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3485 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3488 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3489 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3490 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3491 flags. This restores the settings used before the 2017-07-24.
3492 Turning off pre increment/decrement/modify allows IVOPTS to
3493 optimize DF/SF loops where the index is an int.
3495 2018-01-29 Richard Biener <rguenther@suse.de>
3496 Kelvin Nilsen <kelvin@gcc.gnu.org>
3499 * tree-vect-stmts.c (vectorizable_call): Don't call
3500 targetm.vectorize_builtin_md_vectorized_function if callee is
3503 2018-01-22 Carl Love <cel@us.ibm.com>
3505 * doc/extend.tex: Fix typo in second arg in
3506 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3508 2018-01-29 Richard Biener <rguenther@suse.de>
3510 PR tree-optimization/84086
3511 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3512 (flush_ssaname_freelist): When SSA names were released reset
3513 the SCEV hash table.
3515 2018-01-29 Richard Biener <rguenther@suse.de>
3517 PR tree-optimization/84057
3518 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3519 removed paths when removing edges.
3521 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3523 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3524 -mfunction-return=@var{choice}.
3526 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3529 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3531 (layout::print_source_line): Likewise.
3532 (test_get_line_width_without_trailing_whitespace): Add test cases.
3534 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3537 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3540 2018-01-26 Jim Wilson <jimw@sifive.com>
3542 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3544 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3547 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3549 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3550 and CMP + SUB-immediate -> SUBS.
3552 2018-01-26 Martin Sebor <msebor@redhat.com>
3554 PR tree-optimization/83896
3555 * tree-ssa-strlen.c (get_string_len): Rename...
3556 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3557 Avoid assuming length is constant.
3558 (handle_char_store): Use HOST_WIDE_INT for string length.
3560 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3563 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3564 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3566 2018-01-26 Richard Biener <rguenther@suse.de>
3568 PR rtl-optimization/84003
3569 * dse.c (record_store): Only record redundant stores when
3570 the earlier store aliases at least all accesses the later one does.
3572 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3574 PR rtl-optimization/83985
3575 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3576 REG_CFA_RESTORE insns.
3577 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3578 REG_CFA_RESTORE insns here.
3581 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3582 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3584 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3586 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3587 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3588 (arc_init): Likewise.
3589 (arc_override_options): Likewise.
3590 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3592 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3594 * config/arc/arc.h (TARGET_DBNZ): Define.
3595 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3596 properly set the tune attribute.
3597 (dbnz): Use TARGET_DBNZ guard.
3598 * config/arc/arc.opt (mtune): Add core3 option.
3600 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3602 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3603 recognize new pic like addresses.
3604 (arc_delegitimize_address): Clean up.
3606 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3608 * config/arc/arc-arches.def: Option mrf16 valid for all
3610 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3611 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3612 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3613 * config/arc/arc-tables.opt: Regenerate.
3614 * config/arc/arc.c (arc_conditional_register_usage): Handle
3615 reduced register file case.
3616 (arc_file_start): Set must have build attributes.
3617 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3619 * config/arc/arc.opt (mrf16): Add new option.
3620 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3621 * config/arc/genmultilib.awk: Handle new mrf16 option.
3622 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3623 * config/arc/t-multilib: Regenerate.
3624 * doc/invoke.texi (ARC Options): Document mrf16 option.
3626 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3628 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3629 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3630 (arc_attribute_table): Add 'secure_call' attribute.
3631 (arc_print_operand): Print secure call operand.
3632 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3634 (arc_is_secure_call_p): New function. * config/arc/arc.md
3635 (call_i): Add support for sjli instruction.
3636 (call_value_i): Likewise.
3637 * config/arc/constraints.md (Csc): New constraint.
3639 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3640 John Eric Martin <John.Martin@emmicro-us.com>
3642 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3643 * config/arc/arc.c (_arc_jli_section): New struct.
3644 (arc_jli_section): New type.
3645 (rc_jli_sections): New static variable.
3646 (arc_handle_jli_attribute): New function.
3647 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3648 (arc_file_end): New function.
3649 (TARGET_ASM_FILE_END): Define.
3650 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3651 (arc_add_jli_section): New function.
3652 (jli_call_scan): Likewise.
3653 (arc_reorg): Call jli_call_scan.
3654 (arc_output_addsi): Remove 'S' from printing asm operand.
3655 (arc_is_jli_call_p): New function.
3656 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3658 (movhi_insn): Likewise.
3659 (movsi_insn): Likewise.
3660 (movsi_set_cc_insn): Likewise.
3661 (loadqi_update): Likewise.
3662 (load_zeroextendqisi_update): Likewise.
3663 (load_signextendqisi_update): Likewise.
3664 (loadhi_update): Likewise.
3665 (load_zeroextendhisi_update): Likewise.
3666 (load_signextendhisi_update): Likewise.
3667 (loadsi_update): Likewise.
3668 (loadsf_update): Likewise.
3669 (movsicc_insn): Likewise.
3670 (bset_insn): Likewise.
3671 (bxor_insn): Likewise.
3672 (bclr_insn): Likewise.
3673 (bmsk_insn): Likewise.
3674 (bicsi3_insn): Likewise.
3675 (cmpsi_cc_c_insn): Likewise.
3676 (movsi_ne): Likewise.
3677 (movsi_cond_exec): Likewise.
3678 (clrsbsi2): Likewise.
3689 (call_i): Remove 'S' asm letter, add jli instruction.
3690 (call_value_i): Likewise.
3691 * config/arc/arc.op (mjli-always): New option.
3692 * config/arc/constraints.md (Cji): New constraint.
3693 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3695 (subsf3_fpx): Likewise.
3696 (mulsf3_fpx): Likewise.
3697 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3699 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3701 * doc/invoke.texi (ARC): Document mjli-always option.
3703 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3705 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3706 avoid addition with 0 and use incw and decw where possible.
3708 2018-01-26 Richard Biener <rguenther@suse.de>
3710 PR tree-optimization/81082
3711 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3712 association if it requires casting to unsigned.
3713 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3714 from fold_plusminus_mult_expr to catch important cases late when
3715 range info is available.
3717 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3719 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3720 * configure.ac (hidden_linkonce): New test.
3721 * configure: Regenerate.
3722 * config.in: Regenerate.
3724 2018-01-26 Julia Koval <julia.koval@intel.com>
3726 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3727 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3728 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3729 _mm_mask_bitshuffle_epi64_mask): Fix type.
3730 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3731 USI_FTYPE_V4DI_V4DI_USI): Remove.
3732 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3733 __builtin_ia32_vpshufbitqmb256_mask,
3734 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3735 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3736 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3738 2018-01-26 Alan Modra <amodra@gmail.com>
3741 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3742 UNSPEC_VBPERMQ. Sort other unspecs.
3744 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3746 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3748 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3751 * predict.c (drop_profile): Do not push/pop cfun; update also
3753 (handle_missing_profiles): Fix logic looking for zero profiles.
3755 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3758 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3759 on functions with #pragma omp declare simd or functions with simd
3761 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3762 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3763 Remove trailing \n from warning_at calls.
3765 2018-01-25 Tom de Vries <tom@codesourcery.com>
3768 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3769 for neutered workers.
3771 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3774 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3775 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3777 2018-01-24 Jeff Law <law@redhat.com>
3780 * i386.c (get_probe_interval): Move to earlier point.
3781 (ix86_compute_frame_layout): If -fstack-clash-protection and
3782 the frame is larger than the probe interval, then use pushes
3783 to save registers rather than reg->mem moves.
3784 (ix86_expand_prologue): Remove conditional for int_registers_saved
3787 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3790 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3791 min/max for never referenced object.
3793 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3796 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3798 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3799 attributes from DECL_ATTRIBUTES (decl) without affecting
3800 DECL_ATTRIBUTES (current_function_decl).
3801 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3802 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3804 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3806 PR tree-optimization/83979
3807 * fold-const.c (fold_comparison): Use constant_boolean_node
3808 instead of boolean_{true,false}_node.
3810 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3812 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3815 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3817 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3818 Simplify the clause that sets the length attribute.
3819 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3820 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3821 clause that sets the length attribute.
3822 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3824 2018-01-24 Tom de Vries <tom@codesourcery.com>
3827 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3828 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3829 Add strict parameter.
3830 (prevent_branch_around_nothing): Insert dummy insn between branch to
3831 label and label with no ptx insn inbetween.
3832 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3834 2018-01-24 Tom de Vries <tom@codesourcery.com>
3837 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3838 for neutered threads in warp.
3839 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3841 2018-01-24 Richard Biener <rguenther@suse.de>
3843 PR tree-optimization/83176
3844 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3847 2018-01-24 Richard Biener <rguenther@suse.de>
3849 PR tree-optimization/82819
3850 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3851 code generating pluses that are no-ops in the target precision.
3853 2018-01-24 Richard Biener <rguenther@suse.de>
3856 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3858 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3860 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3861 to merge probabilities.
3862 * predict.c (probably_never_executed): Also mark as cold functions
3863 with global 0 profile and guessed local profile.
3864 * profile-count.c (profile_probability::combine_with_count): New
3866 * profile-count.h (profile_probability::operator*,
3867 profile_probability::operator*=, profile_probability::operator/,
3868 profile_probability::operator/=): Reduce precision to adjusted
3869 and set value to guessed on contradictory divisions.
3870 (profile_probability::combine_with_freq): Remove.
3871 (profile_probability::combine_wiht_count): Declare.
3872 (profile_count::force_nonzero):: Set to adjusted.
3873 (profile_count::probability_in):: Set quality to adjusted.
3874 * tree-ssa-tail-merge.c (replace_block_by): Use
3877 2018-01-23 Andrew Waterman <andrew@sifive.com>
3878 Jim Wilson <jimw@sifive.com>
3880 * config/riscv/riscv.c (riscv_stack_boundary): New.
3881 (riscv_option_override): Set riscv_stack_boundary. Handle
3882 riscv_preferred_stack_boundary_arg.
3883 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3884 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3885 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3886 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3887 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3888 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3890 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3893 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3894 of struct ix86_frame.
3895 (ix86_expand_epilogue): Likewise. Add a local variable for
3896 the reg_save_offset field in struct ix86_frame.
3898 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3900 PR tree-optimization/82604
3901 * tree-loop-distribution.c (enum partition_kind): New enum item
3902 PKIND_PARTIAL_MEMSET.
3903 (partition_builtin_p): Support above new enum item.
3904 (generate_code_for_partition): Ditto.
3905 (compute_access_range): Differentiate cases that equality can be
3906 proven at all loops, the innermost loops or no loops.
3907 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3908 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3909 (finalize_partitions, distribute_loop): Don't fuse partition of
3910 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3911 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3914 2018-01-23 Martin Liska <mliska@suse.cz>
3916 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3917 order to ignore the predictor.
3918 (PRED_POLYMORPHIC_CALL): Likewise.
3919 (PRED_RECURSIVE_CALL): Likewise.
3921 2018-01-23 Martin Liska <mliska@suse.cz>
3923 * tree-profile.c (tree_profiling): Print function header to
3924 aware reader which function we are working on.
3925 * value-prof.c (gimple_find_values_to_profile): Do not print
3926 not interesting value histograms.
3928 2018-01-23 Martin Liska <mliska@suse.cz>
3930 * profile-count.h (enum profile_quality): Add
3931 profile_uninitialized as the first value. Do not number values
3932 as they are zero based.
3933 (profile_count::verify): Update sanity check.
3934 (profile_probability::verify): Likewise.
3936 2018-01-23 Nathan Sidwell <nathan@acm.org>
3938 * doc/invoke.texi (ffor-scope): Deprecate.
3940 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3942 PR tree-optimization/83510
3943 * domwalk.c (set_all_edges_as_executable): New function.
3944 (dom_walker::dom_walker): Convert bool param
3945 "skip_unreachable_blocks" to enum reachability. Move setup of
3946 edge flags to set_all_edges_as_executable and only do it when
3947 reachability is REACHABLE_BLOCKS.
3948 * domwalk.h (enum dom_walker::reachability): New enum.
3949 (dom_walker::dom_walker): Convert bool param
3950 "skip_unreachable_blocks" to enum reachability.
3951 (set_all_edges_as_executable): New decl.
3952 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3953 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3955 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3956 but converting true to REACHABLE_BLOCKS.
3957 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3959 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3960 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3961 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3963 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3964 if check_all_array_refs will be called.
3966 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3968 * tree.c (selftest::test_location_wrappers): Add more test
3971 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3973 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3974 (selftest::test_bit_in_range): Likewise.
3976 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3979 * doc/sourcebuild.texi (vect_float): Say that the selector
3980 only describes the situation when -funsafe-math-optimizations is on.
3981 (vect_float_strict): Document.
3983 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3985 PR tree-optimization/83965
3986 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3987 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3988 instead of checking only for a reduction.
3989 (vect_recog_widen_sum_pattern): Likewise.
3991 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3993 * predict.c (probably_never_executed): Only use precise profile info.
3994 (compute_function_frequency): Skip after inlining hack since we now
3995 have quality checking.
3997 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3999 * profile-count.h (profile_probability::very_unlikely,
4000 profile_probability::unlikely, profile_probability::even): Set
4001 precision to guessed.
4003 2018-01-23 Richard Biener <rguenther@suse.de>
4005 PR tree-optimization/83963
4006 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4007 Properly terminate dominator walk when crossing the exit edge not
4008 when visiting its source block.
4010 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4013 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4014 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4016 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4018 PR tree-optimization/83957
4019 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4020 semicolon after for body surrounded by braces.
4022 PR tree-optimization/83081
4023 * profile-count.h (profile_probability::split): New method.
4024 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4025 Use profile_probability::split.
4026 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4027 when splitting a single conditional jump into 2.
4029 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4031 PR tree-optimization/69452
4032 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4035 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4037 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4038 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4039 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4041 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4043 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4044 * config/rl78/rl78.md (movdi): New define_expand.
4045 * config/rl78/rl78.c (rl78_split_movdi): New function.
4047 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4050 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4052 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4053 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4054 128-bit to produce an UNSPEC move to get the double word with the
4055 signbit and then a shift directly to do signbit.
4056 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4057 implementation with a new version that just does either a direct
4058 move or a regular move. Move memory interface to separate insns.
4059 Move insns so they are next to the expander.
4060 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4061 with signbit move. Split big and little endian case.
4062 (signbit<mode>2_dm_mem_le): Likewise.
4063 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4064 (signbit<mode>2_dm2): Likewise.
4066 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4068 * config/rl78/rl78.md (anddi3): New define_expand.
4070 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4072 * config/rl78/rl78.md (umindi3): New define_expand.
4074 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4076 * config/rl78/rl78.md (smindi3): New define_expand.
4078 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4080 * config/rl78/rl78.md (smaxdi3): New define_expand.
4082 2018-01-22 Carl Love <cel@us.ibm.com>
4084 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4085 LVX_V1TI): Add macro expansion.
4086 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4087 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4088 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4089 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4090 Change check to determine if the instruction is a byte reversing
4091 entry. Fix typo in comment.
4092 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4093 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4094 Add def_builtin calls for new builtins.
4095 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4096 Add define_insn expansion.
4098 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4100 * config/rl78/rl78.md (umaxdi3): New define_expand.
4102 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4104 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4105 for non-QImode registers.
4107 2018-01-22 Richard Biener <rguenther@suse.de>
4109 PR tree-optimization/83963
4110 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4111 including the loop exit block.
4112 (scop_detection::merge_sese): Likewise.
4113 (scop_detection::add_scop): Do it here instead.
4115 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4117 * doc/sourcebuild.texi (arm_softfloat): Document.
4119 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4122 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4123 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4124 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4126 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4127 David Edelsohn <dje.gcc@gmail.com>
4130 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4131 Change "crset eq" to "crset 2".
4132 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4133 (*call_indirect_aix<mode>_nospec): Likewise.
4134 (*call_value_indirect_aix<mode>_nospec): Likewise.
4135 (*call_indirect_elfv2<mode>_nospec): Likewise.
4136 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4137 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4138 change assembly output from . to $.
4139 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4140 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4141 (*tablejump<mode>_internal1_nospec): Likewise.
4143 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4146 * config/sh/sh_optimize_sett_clrt.cc:
4147 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4149 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4151 PR tree-optimization/83940
4152 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4153 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4154 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4155 use it to pass back the definition type.
4156 (vect_check_store_rhs): Likewise rhs_dt_out.
4157 (vect_build_gather_load_calls): Add a mask_dt argument and use
4158 it instead of a call to vect_is_simple_use.
4159 (vectorizable_store): Update calls to vect_check_load_store_mask
4160 and vect_check_store_rhs. Use the dt returned by the latter instead
4161 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4162 instead of calls to vect_is_simple_use. Pass the scalar rather
4163 than the vector operand to vect_is_simple_use when handling
4164 second and subsequent copies of an rhs value.
4165 (vectorizable_load): Update calls to vect_check_load_store_mask
4166 and vect_build_gather_load_calls. Use the cached mask_dt and
4167 gs_info.offset_dt instead of calls to vect_is_simple_use.
4169 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4172 * tree-emutls.c: Include gimplify.h.
4173 (lower_emutls_2): New function.
4174 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4175 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4176 it before further processing.
4179 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4180 UINTVAL (trueop1) instead of INTVAL (op1).
4182 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4186 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4187 INCOMING_FRAME_SP_OFFSET if not defined.
4188 (scan_trace): Add ENTRY argument. If true and
4189 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4190 emit a note to adjust the CFA offset.
4191 (create_cfi_notes): Adjust scan_trace callers.
4192 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4193 INCOMING_FRAME_SP_OFFSET in the CIE.
4194 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4195 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4197 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4198 * doc/tm.texi: Regenerated.
4200 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4202 PR rtl-optimization/83147
4203 * lra-constraints.c (remove_inheritance_pseudos): Use
4204 lra_substitute_pseudo_within_insn.
4206 2018-01-19 Tom de Vries <tom@codesourcery.com>
4207 Cesar Philippidis <cesar@codesourcery.com>
4210 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4212 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4215 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4216 spaces for function labels.
4218 2018-01-19 Martin Liska <mliska@suse.cz>
4220 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4221 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4222 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4223 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4224 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4225 (PRED_CONST_RETURN): Change from 69 to 65.
4226 (PRED_NULL_RETURN): Change from 91 to 71.
4227 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4228 (PRED_LOOP_GUARD): Change from 66 to 73.
4230 2018-01-19 Martin Liska <mliska@suse.cz>
4232 * predict.c (predict_insn_def): Add new assert.
4233 (struct branch_predictor): Change type to signed integer.
4234 (test_prediction_value_range): Amend test to cover
4236 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4237 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4238 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4239 (PRED_LOOP_IV_COMPARE): Likewise.
4240 * predict.h (PROB_UNINITIALIZED): Define new constant.
4242 2018-01-19 Martin Liska <mliska@suse.cz>
4244 * predict.c (dump_prediction): Add new format for
4245 analyze_brprob.py script which is enabled with -details
4247 * profile-count.h (precise_p): New function.
4249 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4251 PR tree-optimization/83922
4252 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4253 there are no statements that need masking.
4254 (vect_active_double_reduction_p): New function.
4255 (vect_analyze_loop_operations): Use it when handling phis that
4256 are not in the loop header.
4258 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4260 PR tree-optimization/83914
4261 * tree-vect-loop.c (vectorizable_induction): Don't convert
4262 init_expr or apply the peeling adjustment for inductions
4263 that are nested within the vectorized loop.
4265 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4267 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4270 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4274 * function.h (gimplify_parameters): Add gimple_seq * argument.
4275 * function.c: Include gimple.h and options.h.
4276 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4277 for the added local temporaries if needed.
4278 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4279 if there are any parameter cleanups, wrap whole body into a
4280 try/finally with the cleanups.
4282 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4285 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4286 Use GET_MODE_CLASS for scalar floating point.
4288 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4292 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4293 Fix call of call_cgraph_insertion_hooks.
4295 2018-01-18 Martin Sebor <msebor@redhat.com>
4297 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4299 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4302 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4305 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4308 * common.opt: (-ffile-prefix-map): New option.
4309 * opts.c (common_handle_option): Defer it.
4310 * opts-global.c (handle_common_deferred_options): Handle it.
4311 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4312 * file-prefix-map.h: New file.
4313 (remap_debug_filename, add_debug_prefix_map): ...here.
4314 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4315 * final.c (debug_prefix_map, add_debug_prefix_map
4316 remap_debug_filename): Move to...
4317 * file-prefix-map.c: New file.
4318 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4319 generalize, get rid of alloca(), use strrchr() instead of strchr().
4320 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4321 Implement in terms of add_prefix_map().
4322 (remap_macro_filename, remap_debug_filename): Implement in term of
4324 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4325 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4326 * dbxout.c: Include file-prefix-map.h.
4327 * varasm.c: Likewise.
4328 * vmsdbgout.c: Likewise.
4329 * xcoffout.c: Likewise.
4330 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4331 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4332 * doc/invoke.texi (-ffile-prefix-map): Document.
4333 (-fdebug-prefix-map): Update description.
4335 2018-01-18 Martin Liska <mliska@suse.cz>
4337 * config/i386/i386.c (indirect_thunk_name): Document that also
4339 (output_indirect_thunk): Document why both instructions
4340 (pause and lfence) are generated.
4342 2018-01-18 Richard Biener <rguenther@suse.de>
4344 PR tree-optimization/83887
4345 * graphite-scop-detection.c
4346 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4347 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4348 (scop_detection::merge_sese): Re-implement with a flood-fill
4349 algorithm that properly finds a SESE region if it exists.
4351 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4354 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4355 pointer_diff optimizations use view_convert instead of convert.
4357 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4359 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4360 Generate different code for -mno-speculate-indirect-jumps.
4361 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4362 (*call_indirect_aix<mode>): Disable for
4363 -mno-speculate-indirect-jumps.
4364 (*call_indirect_aix<mode>_nospec): New define_insn.
4365 (*call_value_indirect_aix<mode>): Disable for
4366 -mno-speculate-indirect-jumps.
4367 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4368 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4369 -mno-speculate-indirect-jumps.
4370 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4372 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4374 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4375 long double type, set the flags for noting the default long double
4376 type, even if we don't pass or return a long double type.
4378 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4381 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4384 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4386 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4387 support for merge[hl].
4388 (fold_mergehl_helper): New helper function.
4389 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4390 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4391 (altivec_vmrglw_direct): Add xxmrglw insn.
4393 2018-01-17 Andrew Waterman <andrew@sifive.com>
4395 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4396 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4398 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4401 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4402 call the lto_location_cache before reading the
4403 DECL_SOURCE_LOCATION of the types.
4405 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4406 Richard Sandiford <richard.sandiford@linaro.org>
4408 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4409 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4410 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4411 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4412 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4414 * config/aarch64/constraints.md (aarch64_movti_operand):
4416 * config/aarch64/predicates.md (Uti): Add new constraint.
4418 2018-01-17 Carl Love <cel@us.ibm.com>
4420 * config/rs6000/vsx.md (define_expand xl_len_r,
4421 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4422 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4424 (define_expand, define_insn): Move the shift left from the
4425 define_insn to the define_expand for lxvl and stxvl instructions.
4426 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4427 and XL_LEN_R definitions to PURE.
4429 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4431 * config/i386/i386.c (indirect_thunk_name): Declare regno
4432 as unsigned int. Compare regno with INVALID_REGNUM.
4433 (output_indirect_thunk): Ditto.
4434 (output_indirect_thunk_function): Ditto.
4435 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4436 in the call to output_indirect_thunk_function.
4438 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4441 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4442 rather than the size of inner_type to determine the stack slot size
4443 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4445 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4448 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4451 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4453 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4454 endian Linux systems to optionally enable multilibs for selecting
4455 the long double type if the user configured an explicit type.
4456 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4457 have no long double multilibs if not defined.
4458 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4459 warn if the user used -mabi={ieee,ibm}longdouble and we built
4460 multilibs for long double.
4461 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4462 appropriate multilib option.
4463 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4465 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4466 for building long double multilibs.
4467 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4469 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4471 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4474 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4476 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4479 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4482 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4485 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4487 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4488 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4490 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4492 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4493 different rtl trees depending on TARGET_64BIT.
4494 (rs6000_gen_lvx): Likewise.
4496 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4498 * config/visium/visium.md (nop): Tweak comment.
4499 (hazard_nop): Likewise.
4501 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4503 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4504 -mspeculate-indirect-jumps.
4505 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4506 for -mno-speculate-indirect-jumps.
4507 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4508 (*call_value_indirect_elfv2<mode>): Disable for
4509 -mno-speculate-indirect-jumps.
4510 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4511 (indirect_jump): Emit different RTL for
4512 -mno-speculate-indirect-jumps.
4513 (*indirect_jump<mode>): Disable for
4514 -mno-speculate-indirect-jumps.
4515 (*indirect_jump<mode>_nospec): New define_insn.
4516 (tablejump): Emit different RTL for
4517 -mno-speculate-indirect-jumps.
4518 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4519 (tablejumpsi_nospec): New define_expand.
4520 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4521 (tablejumpdi_nospec): New define_expand.
4522 (*tablejump<mode>_internal1): Disable for
4523 -mno-speculate-indirect-jumps.
4524 (*tablejump<mode>_internal1_nospec): New define_insn.
4525 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4528 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4530 * caller-save.c (insert_save): Drop unnecessary parameter. All
4533 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4534 Richard Biener <rguenth@suse.de>
4537 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4538 return early, inline manually is_gimple_sizepos. Make sure if we
4539 call gimplify_expr we don't end up with a gimple constant.
4540 * tree.c (variably_modified_type_p): Don't return true for
4541 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4542 * gimplify.h (is_gimple_sizepos): Remove.
4544 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4546 PR tree-optimization/83857
4547 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4548 vectorizable_live_operation for pure SLP statements.
4549 (vectorizable_live_operation): Handle PHIs.
4551 2018-01-16 Richard Biener <rguenther@suse.de>
4553 PR tree-optimization/83867
4554 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4555 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4557 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4560 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4561 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4562 If off is not INTEGER_CST, issue a may not be aligned warning
4563 rather than isn't aligned. Use isn%'t rather than isn't.
4564 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4566 <case MULT_EXPR>: Improve the case when bottom and one of the
4567 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4568 operand, in that case check if the other operand is multiple of
4569 bottom divided by the INTEGER_CST operand.
4571 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4574 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4575 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4576 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4577 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4578 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4579 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4580 (pa_function_arg_size): New function.
4582 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4584 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4585 in a separate statement.
4587 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4589 PR tree-optimization/83847
4590 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4591 group gathers and scatters.
4593 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4595 PR rtl-optimization/86620
4596 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4598 PR rtl-optimization/83213
4599 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4600 to last if both are JUMP_INSNs.
4602 PR tree-optimization/83843
4603 * gimple-ssa-store-merging.c
4604 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4605 store_immediate_info for bswap/nop orig_stores.
4607 2018-01-15 Andrew Waterman <andrew@sifive.com>
4609 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4611 <UDIV>: Increase cost if !TARGET_DIV.
4613 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4615 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4616 (define_attr "cr_logical_3op"): New.
4617 (cceq_ior_compare): Adjust.
4618 (cceq_ior_compare_complement): Adjust.
4619 (*cceq_rev_compare): Adjust.
4620 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4621 (is_cracked_insn): Adjust.
4622 (insn_must_be_first_in_group): Adjust.
4623 * config/rs6000/40x.md: Adjust.
4624 * config/rs6000/440.md: Adjust.
4625 * config/rs6000/476.md: Adjust.
4626 * config/rs6000/601.md: Adjust.
4627 * config/rs6000/603.md: Adjust.
4628 * config/rs6000/6xx.md: Adjust.
4629 * config/rs6000/7450.md: Adjust.
4630 * config/rs6000/7xx.md: Adjust.
4631 * config/rs6000/8540.md: Adjust.
4632 * config/rs6000/cell.md: Adjust.
4633 * config/rs6000/e300c2c3.md: Adjust.
4634 * config/rs6000/e500mc.md: Adjust.
4635 * config/rs6000/e500mc64.md: Adjust.
4636 * config/rs6000/e5500.md: Adjust.
4637 * config/rs6000/e6500.md: Adjust.
4638 * config/rs6000/mpc.md: Adjust.
4639 * config/rs6000/power4.md: Adjust.
4640 * config/rs6000/power5.md: Adjust.
4641 * config/rs6000/power6.md: Adjust.
4642 * config/rs6000/power7.md: Adjust.
4643 * config/rs6000/power8.md: Adjust.
4644 * config/rs6000/power9.md: Adjust.
4645 * config/rs6000/rs64.md: Adjust.
4646 * config/rs6000/titan.md: Adjust.
4648 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4650 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4651 ix86_indirect_branch_register logic.
4653 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4655 * config/i386/constraints.md (Bs): Update
4656 ix86_indirect_branch_register check. Don't check
4657 ix86_indirect_branch_register with GOT_memory_operand.
4659 * config/i386/predicates.md (GOT_memory_operand): Don't check
4660 ix86_indirect_branch_register here.
4661 (GOT32_symbol_operand): Likewise.
4663 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4665 * config/i386/predicates.md (constant_call_address_operand):
4666 Rewrite ix86_indirect_branch_register logic.
4667 (sibcall_insn_operand): Likewise.
4669 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4671 * config/i386/constraints.md (Bs): Replace
4672 ix86_indirect_branch_thunk_register with
4673 ix86_indirect_branch_register.
4675 * config/i386/i386.md (indirect_jump): Likewise.
4676 (tablejump): Likewise.
4677 (*sibcall_memory): Likewise.
4678 (*sibcall_value_memory): Likewise.
4679 Peepholes of indirect call and jump via memory: Likewise.
4680 * config/i386/i386.opt: Likewise.
4681 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4682 (GOT_memory_operand): Likewise.
4683 (call_insn_operand): Likewise.
4684 (sibcall_insn_operand): Likewise.
4685 (GOT32_symbol_operand): Likewise.
4687 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4690 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4691 type rather than type addr's type points to.
4692 (expand_omp_atomic_mutex): Likewise.
4693 (expand_omp_atomic): Likewise.
4695 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4698 * config/i386/i386.c (output_indirect_thunk_function): Use
4699 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4700 for __x86_return_thunk.
4702 2018-01-15 Richard Biener <rguenther@suse.de>
4705 * expmed.c (extract_bit_field_1): Fix typo.
4707 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4710 * config/arm/iterators.md (VF): New mode iterator.
4711 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4712 Remove integer-related logic from pattern.
4713 (neon_vabd<mode>_3): Likewise.
4715 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4718 * common.opt (fstrict-overflow): No longer an alias.
4719 (fwrapv-pointer): New option.
4720 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4721 also for pointer types based on flag_wrapv_pointer.
4722 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4723 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4724 opts->x_flag_wrapv got set.
4725 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4726 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4727 POINTER_TYPE_OVERFLOW_UNDEFINED.
4728 * match.pd: Likewise in address comparison pattern.
4729 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4731 2018-01-15 Richard Biener <rguenther@suse.de>
4734 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4735 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4736 Reset type names to their identifier if their TYPE_DECL doesn't
4737 have linkage (and thus is used for ODR and devirt).
4738 (save_debug_info_for_decl): Remove.
4739 (save_debug_info_for_type): Likewise.
4740 (add_tree_to_fld_list): Adjust.
4741 * tree-pretty-print.c (dump_generic_node): Make dumping of
4742 type names more robust.
4744 2018-01-15 Richard Biener <rguenther@suse.de>
4746 * BASE-VER: Bump to 8.0.1.
4748 2018-01-14 Martin Sebor <msebor@redhat.com>
4751 * builtins.c (check_access): Avoid warning when the no-warning bit
4754 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4756 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4757 * ira-color (allocno_hard_regs_compare): Likewise.
4759 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4762 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4763 Use .pushsection/.popsection.
4765 2018-01-14 Martin Sebor <msebor@redhat.com>
4768 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4770 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4772 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4773 entry from extra_headers.
4774 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4775 extra_headers, make the list bitwise identical to the i?86-*-* one.
4777 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4779 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4780 -mcmodel=large with -mindirect-branch=thunk,
4781 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4782 -mfunction-return=thunk-extern.
4783 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4784 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4785 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4787 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4789 * config/i386/i386.c (print_reg): Print the name of the full
4790 integer register without '%'.
4791 (ix86_print_operand): Handle 'V'.
4792 * doc/extend.texi: Document 'V' modifier.
4794 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4796 * config/i386/constraints.md (Bs): Disallow memory operand for
4797 -mindirect-branch-register.
4799 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4800 (GOT_memory_operand): Likewise.
4801 (call_insn_operand): Likewise.
4802 (sibcall_insn_operand): Likewise.
4803 (GOT32_symbol_operand): Likewise.
4804 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4805 for -mindirect-branch-register.
4806 (tablejump): Likewise.
4807 (*sibcall_memory): Likewise.
4808 (*sibcall_value_memory): Likewise.
4809 Disallow peepholes of indirect call and jump via memory for
4810 -mindirect-branch-register.
4811 (*call_pop): Replace m with Bw.
4812 (*call_value_pop): Likewise.
4813 (*sibcall_pop_memory): Replace m with Bs.
4814 * config/i386/i386.opt (mindirect-branch-register): New option.
4815 * doc/invoke.texi: Document -mindirect-branch-register option.
4817 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4819 * config/i386/i386-protos.h (ix86_output_function_return): New.
4820 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4821 set function_return_type.
4822 (indirect_thunk_name): Add ret_p to indicate thunk for function
4824 (output_indirect_thunk_function): Pass false to
4825 indirect_thunk_name.
4826 (ix86_output_indirect_branch_via_reg): Likewise.
4827 (ix86_output_indirect_branch_via_push): Likewise.
4828 (output_indirect_thunk_function): Create alias for function
4829 return thunk if regno < 0.
4830 (ix86_output_function_return): New function.
4831 (ix86_handle_fndecl_attribute): Handle function_return.
4832 (ix86_attribute_table): Add function_return.
4833 * config/i386/i386.h (machine_function): Add
4834 function_return_type.
4835 * config/i386/i386.md (simple_return_internal): Use
4836 ix86_output_function_return.
4837 (simple_return_internal_long): Likewise.
4838 * config/i386/i386.opt (mfunction-return=): New option.
4839 (indirect_branch): Mention -mfunction-return=.
4840 * doc/extend.texi: Document function_return function attribute.
4841 * doc/invoke.texi: Document -mfunction-return= option.
4843 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4845 * config/i386/i386-opts.h (indirect_branch): New.
4846 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4847 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4848 with local indirect jump when converting indirect call and jump.
4849 (ix86_set_indirect_branch_type): New.
4850 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4851 (indirectlabelno): New.
4852 (indirect_thunk_needed): Likewise.
4853 (indirect_thunk_bnd_needed): Likewise.
4854 (indirect_thunks_used): Likewise.
4855 (indirect_thunks_bnd_used): Likewise.
4856 (INDIRECT_LABEL): Likewise.
4857 (indirect_thunk_name): Likewise.
4858 (output_indirect_thunk): Likewise.
4859 (output_indirect_thunk_function): Likewise.
4860 (ix86_output_indirect_branch_via_reg): Likewise.
4861 (ix86_output_indirect_branch_via_push): Likewise.
4862 (ix86_output_indirect_branch): Likewise.
4863 (ix86_output_indirect_jmp): Likewise.
4864 (ix86_code_end): Call output_indirect_thunk_function if needed.
4865 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4867 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4868 (ix86_attribute_table): Add indirect_branch.
4869 * config/i386/i386.h (machine_function): Add indirect_branch_type
4870 and has_local_indirect_jump.
4871 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4873 (tablejump): Likewise.
4874 (*indirect_jump): Use ix86_output_indirect_jmp.
4875 (*tablejump_1): Likewise.
4876 (simple_return_indirect_internal): Likewise.
4877 * config/i386/i386.opt (mindirect-branch=): New option.
4878 (indirect_branch): New.
4881 (thunk-inline): Likewise.
4882 (thunk-extern): Likewise.
4883 * doc/extend.texi: Document indirect_branch function attribute.
4884 * doc/invoke.texi: Document -mindirect-branch= option.
4886 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4889 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4891 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4893 * ipa-inline.c (want_inline_small_function_p): Return false if
4894 inlining has already failed with CIF_FINAL_ERROR.
4895 (update_caller_keys): Call want_inline_small_function_p before
4897 (update_callee_keys): Likewise.
4899 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4901 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4903 (rs6000_quadword_masked_address_p): Likewise.
4904 (quad_aligned_load_p): Likewise.
4905 (quad_aligned_store_p): Likewise.
4906 (const_load_sequence_p): Add comment to describe the outer-most loop.
4907 (mimic_memory_attributes_and_flags): New function.
4908 (rs6000_gen_stvx): Likewise.
4909 (replace_swapped_aligned_store): Likewise.
4910 (rs6000_gen_lvx): Likewise.
4911 (replace_swapped_aligned_load): Likewise.
4912 (replace_swapped_load_constant): Capitalize argument name in
4913 comment describing this function.
4914 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4915 and stores that access quad-word aligned addresses and replace
4916 with stvx or lvx instructions when appropriate.
4917 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4918 New function prototype.
4919 (rs6000_quadword_masked_address_p): Likewise.
4920 (rs6000_gen_lvx): Likewise.
4921 (rs6000_gen_stvx): Likewise.
4922 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4923 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4924 when memory address is aligned.
4925 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4926 this split to select lvx instruction when memory address is aligned.
4927 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4928 instruction when memory address is aligned.
4929 (*vsx_le_perm_load_v16qi): Likewise.
4930 (four unnamed splitters): Modify to select the stvx instruction
4931 when memory is aligned.
4933 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4935 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4936 which appears in the queue multiple times.
4938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4939 Alan Hayward <alan.hayward@arm.com>
4940 David Sherwood <david.sherwood@arm.com>
4942 * tree-vectorizer.h (vec_lower_bound): New structure.
4943 (_loop_vec_info): Add check_nonzero and lower_bounds.
4944 (LOOP_VINFO_CHECK_NONZERO): New macro.
4945 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4946 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4947 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4948 fields. Make seg_len the distance travelled, not including the
4950 (dr_direction_indicator): Declare.
4951 (dr_zero_step_indicator): Likewise.
4952 (dr_known_forward_stride_p): Likewise.
4953 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4955 (runtime_alias_check_p): Allow runtime alias checks with
4957 (operator ==): Compare access_size and align.
4958 (prune_runtime_alias_test_list): Rework for new distinction between
4959 the access_size and seg_len.
4960 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4962 (get_segment_min_max): New function.
4963 (create_intersect_range_checks): Use it.
4964 (dr_step_indicator): New function.
4965 (dr_direction_indicator): Likewise.
4966 (dr_zero_step_indicator): Likewise.
4967 (dr_known_forward_stride_p): Likewise.
4968 * tree-loop-distribution.c (data_ref_segment_size): Return
4969 DR_STEP * (niters - 1).
4970 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4972 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4973 (vect_preserves_scalar_order_p): New function, split out from...
4974 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4975 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4976 (vect_vfa_access_size): New function.
4977 (vect_vfa_align): Likewise.
4978 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4979 (dump_lower_bound): New function.
4980 (vect_check_lower_bound): Likewise.
4981 (vect_small_gap_p): Likewise.
4982 (vectorizable_with_step_bound_p): Likewise.
4983 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4984 depencies if the vectorization factor is 1. Convert the checks
4985 for nonzero steps into checks on the bounds of DR_STEP. Try using
4986 a bunds check for variable steps if the minimum required step is
4987 relatively small. Update calls to the dr_with_seg_len
4988 constructor and to vect_compile_time_alias.
4989 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4991 (vect_loop_versioning): Call it.
4992 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4994 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4997 Alan Hayward <alan.hayward@arm.com>
4998 David Sherwood <david.sherwood@arm.com>
5000 * doc/sourcebuild.texi (vect_scatter_store): Document.
5001 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5003 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5005 * genopinit.c (main): Add supports_vec_scatter_store and
5006 supports_vec_scatter_store_cached to target_optabs.
5007 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5008 IFN_MASK_SCATTER_STORE.
5009 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5011 * internal-fn.h (internal_store_fn_p): Declare.
5012 (internal_fn_stored_value_index): Likewise.
5013 * internal-fn.c (scatter_store_direct): New macro.
5014 (expand_scatter_store_optab_fn): New function.
5015 (direct_scatter_store_optab_supported_p): New macro.
5016 (internal_store_fn_p): New function.
5017 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5018 IFN_MASK_SCATTER_STORE.
5019 (internal_fn_mask_index): Likewise.
5020 (internal_fn_stored_value_index): New function.
5021 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5023 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5024 * optabs-query.c (supports_vec_scatter_store_p): New function.
5025 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5026 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5027 true for scatter stores.
5028 (vect_gather_scatter_fn_p): Handle scatter stores too.
5029 (vect_check_gather_scatter): Consider using scatter stores if
5030 supports_vec_scatter_store_p.
5031 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5033 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5034 internal_fn_stored_value_index.
5035 (check_load_store_masking): Handle scatter stores too.
5036 (vect_get_store_rhs): Make public.
5037 (vectorizable_call): Use internal_store_fn_p.
5038 (vectorizable_store): Handle scatter store internal functions.
5039 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5040 when deciding whether the end of the group has been reached.
5041 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5042 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5043 (mask_scatter_store<mode>): New insns.
5045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5046 Alan Hayward <alan.hayward@arm.com>
5047 David Sherwood <david.sherwood@arm.com>
5049 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5050 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5051 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5053 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5054 Use vect_truncate_gather_scatter_offset if we can't treat the
5055 operation as a normal gather load or scatter store.
5056 (get_group_load_store_type): Take the gather_scatter_info
5057 as argument. Try using a gather load or scatter store for
5058 single-element groups.
5059 (get_load_store_type): Update calls to get_group_load_store_type
5060 and vect_use_strided_gather_scatters_p.
5062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5063 Alan Hayward <alan.hayward@arm.com>
5064 David Sherwood <david.sherwood@arm.com>
5066 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5067 optional tree argument.
5068 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5070 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5071 but continue to use the current value as a fallback.
5072 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5073 to compare the updates.
5074 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5075 (get_load_store_type): Use it when handling a strided access.
5076 (vect_get_strided_load_store_ops): New function.
5077 (vect_get_data_ptr_increment): Likewise.
5078 (vectorizable_load): Handle strided gather loads. Always pass
5079 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5082 Alan Hayward <alan.hayward@arm.com>
5083 David Sherwood <david.sherwood@arm.com>
5085 * doc/md.texi (gather_load@var{m}): Document.
5086 (mask_gather_load@var{m}): Likewise.
5087 * genopinit.c (main): Add supports_vec_gather_load and
5088 supports_vec_gather_load_cached to target_optabs.
5089 * optabs-tree.c (init_tree_optimization_optabs): Use
5090 ggc_cleared_alloc to allocate target_optabs.
5091 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5092 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5094 * internal-fn.h (internal_load_fn_p): Declare.
5095 (internal_gather_scatter_fn_p): Likewise.
5096 (internal_fn_mask_index): Likewise.
5097 (internal_gather_scatter_fn_supported_p): Likewise.
5098 * internal-fn.c (gather_load_direct): New macro.
5099 (expand_gather_load_optab_fn): New function.
5100 (direct_gather_load_optab_supported_p): New macro.
5101 (direct_internal_fn_optab): New function.
5102 (internal_load_fn_p): Likewise.
5103 (internal_gather_scatter_fn_p): Likewise.
5104 (internal_fn_mask_index): Likewise.
5105 (internal_gather_scatter_fn_supported_p): Likewise.
5106 * optabs-query.c (supports_at_least_one_mode_p): New function.
5107 (supports_vec_gather_load_p): Likewise.
5108 * optabs-query.h (supports_vec_gather_load_p): Declare.
5109 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5110 and memory_type field.
5111 (NUM_PATTERNS): Bump to 15.
5112 * tree-vect-data-refs.c: Include internal-fn.h.
5113 (vect_gather_scatter_fn_p): New function.
5114 (vect_describe_gather_scatter_call): Likewise.
5115 (vect_check_gather_scatter): Try using internal functions for
5116 gather loads. Recognize existing calls to a gather load function.
5117 (vect_analyze_data_refs): Consider using gather loads if
5118 supports_vec_gather_load_p.
5119 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5120 (vect_get_gather_scatter_offset_type): Likewise.
5121 (vect_convert_mask_for_vectype): Likewise.
5122 (vect_add_conversion_to_patterm): Likewise.
5123 (vect_try_gather_scatter_pattern): Likewise.
5124 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5125 (vect_vect_recog_func_ptrs): Add it.
5126 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5127 internal_fn_mask_index and internal_gather_scatter_fn_p.
5128 (check_load_store_masking): Take the gather_scatter_info as an
5129 argument and handle gather loads.
5130 (vect_get_gather_scatter_ops): New function.
5131 (vectorizable_call): Check internal_load_fn_p.
5132 (vectorizable_load): Likewise. Handle gather load internal
5134 (vectorizable_store): Update call to check_load_store_masking.
5135 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5136 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5137 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5138 (aarch64_gather_scale_operand_d): New predicates.
5139 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5140 (mask_gather_load<mode>): New insns.
5142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5143 Alan Hayward <alan.hayward@arm.com>
5144 David Sherwood <david.sherwood@arm.com>
5146 * optabs.def (fold_left_plus_optab): New optab.
5147 * doc/md.texi (fold_left_plus_@var{m}): Document.
5148 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5149 * internal-fn.c (fold_left_direct): Define.
5150 (expand_fold_left_optab_fn): Likewise.
5151 (direct_fold_left_optab_supported_p): Likewise.
5152 * fold-const-call.c (fold_const_fold_left): New function.
5153 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5154 * tree-parloops.c (valid_reduction_p): New function.
5155 (gather_scalar_reductions): Use it.
5156 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5157 (vect_finish_replace_stmt): Declare.
5158 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5159 (needs_fold_left_reduction_p): New function, split out from...
5160 (vect_is_simple_reduction): ...here. Accept reductions that
5161 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5162 (vect_force_simple_reduction): Also store the reduction type in
5163 the assignment's STMT_VINFO_REDUC_TYPE.
5164 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5165 (merge_with_identity): New function.
5166 (vect_expand_fold_left): Likewise.
5167 (vectorize_fold_left_reduction): Likewise.
5168 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5169 scalar phi in place for it. Check for target support and reject
5170 cases that would reassociate the operation. Defer the transform
5171 phase to vectorize_fold_left_reduction.
5172 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5173 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5174 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5176 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5178 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5179 call to ifc_temp_var.
5181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5182 Alan Hayward <alan.hayward@arm.com>
5183 David Sherwood <david.sherwood@arm.com>
5185 * target.def (legitimize_address_displacement): Take the original
5186 offset as a poly_int.
5187 * targhooks.h (default_legitimize_address_displacement): Update
5189 * targhooks.c (default_legitimize_address_displacement): Likewise.
5190 * doc/tm.texi: Regenerate.
5191 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5192 as an argument, moving assert of ad->disp == ad->disp_term to...
5193 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5194 Try calling targetm.legitimize_address_displacement before expanding
5195 the address rather than afterwards, and adjust for the new interface.
5196 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5197 Match the new hook interface. Handle SVE addresses.
5198 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5203 * Makefile.in (OBJS): Add early-remat.o.
5204 * target.def (select_early_remat_modes): New hook.
5205 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5206 * doc/tm.texi: Regenerate.
5207 * targhooks.h (default_select_early_remat_modes): Declare.
5208 * targhooks.c (default_select_early_remat_modes): New function.
5209 * timevar.def (TV_EARLY_REMAT): New timevar.
5210 * passes.def (pass_early_remat): New pass.
5211 * tree-pass.h (make_pass_early_remat): Declare.
5212 * early-remat.c: New file.
5213 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5215 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5218 Alan Hayward <alan.hayward@arm.com>
5219 David Sherwood <david.sherwood@arm.com>
5221 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5222 vfm1 with a bound_epilog parameter.
5223 (vect_do_peeling): Update calls accordingly, and move the prologue
5224 call earlier in the function. Treat the base bound_epilog as 0 for
5225 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5226 this base when peeling for gaps.
5227 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5228 with fully-masked loops.
5229 (vect_estimate_min_profitable_iters): Handle the single peeled
5230 iteration in that case.
5232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5233 Alan Hayward <alan.hayward@arm.com>
5234 David Sherwood <david.sherwood@arm.com>
5236 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5237 single-element interleaving even if the size is not a power of 2.
5238 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5239 accesses for single-element interleaving if the group size is
5242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5243 Alan Hayward <alan.hayward@arm.com>
5244 David Sherwood <david.sherwood@arm.com>
5246 * doc/md.texi (fold_extract_last_@var{m}): Document.
5247 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5248 * optabs.def (fold_extract_last_optab): New optab.
5249 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5250 * internal-fn.c (fold_extract_direct): New macro.
5251 (expand_fold_extract_optab_fn): Likewise.
5252 (direct_fold_extract_optab_supported_p): Likewise.
5253 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5254 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5255 EXTRACT_LAST_REDUCTION.
5256 (get_initial_def_for_reduction): Do not create an initial vector
5257 for EXTRACT_LAST_REDUCTION reductions.
5258 (vectorizable_reduction): Leave the scalar phi in place for
5259 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5260 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5261 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5262 transform phase to vectorizable_condition.
5263 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5265 (vect_finish_stmt_generation): ...here.
5266 (vect_finish_replace_stmt): New function.
5267 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5268 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5270 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5273 Alan Hayward <alan.hayward@arm.com>
5274 David Sherwood <david.sherwood@arm.com>
5276 * doc/md.texi (extract_last_@var{m}): Document.
5277 * optabs.def (extract_last_optab): New optab.
5278 * internal-fn.def (EXTRACT_LAST): New internal function.
5279 * internal-fn.c (cond_unary_direct): New macro.
5280 (expand_cond_unary_optab_fn): Likewise.
5281 (direct_cond_unary_optab_supported_p): Likewise.
5282 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5283 loops using EXTRACT_LAST.
5284 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5285 (extract_last_<mode>): ...this optab.
5286 (vec_extract<mode><Vel>): Update accordingly.
5288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5289 Alan Hayward <alan.hayward@arm.com>
5290 David Sherwood <david.sherwood@arm.com>
5292 * target.def (empty_mask_is_expensive): New hook.
5293 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5294 * doc/tm.texi: Regenerate.
5295 * targhooks.h (default_empty_mask_is_expensive): Declare.
5296 * targhooks.c (default_empty_mask_is_expensive): New function.
5297 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5298 if the target says that empty masks are expensive.
5299 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5301 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5304 Alan Hayward <alan.hayward@arm.com>
5305 David Sherwood <david.sherwood@arm.com>
5307 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5308 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5309 (vect_use_loop_mask_for_alignment_p): New function.
5310 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5311 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5312 niters_skip argument. Make sure that the first niters_skip elements
5313 of the first iteration are inactive.
5314 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5315 Update call to vect_set_loop_masks_directly.
5316 (get_misalign_in_elems): New function, split out from...
5317 (vect_gen_prolog_loop_niters): ...here.
5318 (vect_update_init_of_dr): Take a code argument that specifies whether
5319 the adjustment should be added or subtracted.
5320 (vect_update_init_of_drs): Likewise.
5321 (vect_prepare_for_masked_peels): New function.
5322 (vect_do_peeling): Skip prologue peeling if we're using a mask
5323 instead. Update call to vect_update_inits_of_drs.
5324 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5326 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5327 alignment. Do not include the number of peeled iterations in
5328 the minimum threshold in that case.
5329 (vectorizable_induction): Adjust the start value down by
5330 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5331 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5332 Take the number of skipped iterations into account when calculating
5334 * tree-vect-stmts.c (vect_gen_while_not): New function.
5336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5337 Alan Hayward <alan.hayward@arm.com>
5338 David Sherwood <david.sherwood@arm.com>
5340 * doc/sourcebuild.texi (vect_fully_masked): Document.
5341 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5343 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5345 (vect_analyze_loop_2): ...here. Don't check the vectorization
5346 factor against the number of loop iterations if the loop is
5349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5350 Alan Hayward <alan.hayward@arm.com>
5351 David Sherwood <david.sherwood@arm.com>
5353 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5354 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5355 (dump_groups): Update accordingly.
5356 (iv_use::mem_type): New member variable.
5357 (address_p): New function.
5358 (record_use): Add a mem_type argument and initialize the new
5360 (record_group_use): Add a mem_type argument. Use address_p.
5361 Remove obsolete null checks of base_object. Update call to record_use.
5362 (find_interesting_uses_op): Update call to record_group_use.
5363 (find_interesting_uses_cond): Likewise.
5364 (find_interesting_uses_address): Likewise.
5365 (get_mem_type_for_internal_fn): New function.
5366 (find_address_like_use): Likewise.
5367 (find_interesting_uses_stmt): Try find_address_like_use before
5368 calling find_interesting_uses_op.
5369 (addr_offset_valid_p): Use the iv mem_type field as the type
5370 of the addressed memory.
5371 (add_autoinc_candidates): Likewise.
5372 (get_address_cost): Likewise.
5373 (split_small_address_groups_p): Use address_p.
5374 (split_address_groups): Likewise.
5375 (add_iv_candidate_for_use): Likewise.
5376 (autoinc_possible_for_pair): Likewise.
5377 (rewrite_groups): Likewise.
5378 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5379 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5380 (get_alias_ptr_type_for_ptr_address): New function.
5381 (rewrite_use_address): Rewrite address uses in calls that were
5382 identified by find_address_like_use.
5384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5385 Alan Hayward <alan.hayward@arm.com>
5386 David Sherwood <david.sherwood@arm.com>
5388 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5390 * gimple-expr.h (is_gimple_addressable: Likewise.
5391 * gimple-expr.c (is_gimple_address): Likewise.
5392 * internal-fn.c (expand_call_mem_ref): New function.
5393 (expand_mask_load_optab_fn): Use it.
5394 (expand_mask_store_optab_fn): Likewise.
5396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5397 Alan Hayward <alan.hayward@arm.com>
5398 David Sherwood <david.sherwood@arm.com>
5400 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5401 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5402 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5403 (cond_umax@var{mode}): Document.
5404 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5405 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5406 (cond_umin_optab, cond_umax_optab): New optabs.
5407 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5408 (COND_IOR, COND_XOR): New internal functions.
5409 * internal-fn.h (get_conditional_internal_fn): Declare.
5410 * internal-fn.c (cond_binary_direct): New macro.
5411 (expand_cond_binary_optab_fn): Likewise.
5412 (direct_cond_binary_optab_supported_p): Likewise.
5413 (get_conditional_internal_fn): New function.
5414 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5415 Cope with reduction statements that are vectorized as calls rather
5417 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5418 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5419 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5420 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5421 (UNSPEC_COND_EOR): New unspecs.
5422 (optab): Add mappings for them.
5423 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5424 (sve_int_op, sve_fp_op): New int attributes.
5426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5427 Alan Hayward <alan.hayward@arm.com>
5428 David Sherwood <david.sherwood@arm.com>
5430 * optabs.def (while_ult_optab): New optab.
5431 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5432 * internal-fn.def (WHILE_ULT): New internal function.
5433 * internal-fn.h (direct_internal_fn_supported_p): New override
5434 that takes two types as argument.
5435 * internal-fn.c (while_direct): New macro.
5436 (expand_while_optab_fn): New function.
5437 (convert_optab_supported_p): Likewise.
5438 (direct_while_optab_supported_p): New macro.
5439 * wide-int.h (wi::udiv_ceil): New function.
5440 * tree-vectorizer.h (rgroup_masks): New structure.
5441 (vec_loop_masks): New typedef.
5442 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5444 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5445 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5446 (vect_max_vf): New function.
5447 (slpeel_make_loop_iterate_ntimes): Delete.
5448 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5449 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5450 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5451 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5452 internal-fn.h, stor-layout.h and optabs-query.h.
5453 (vect_set_loop_mask): New function.
5454 (add_preheader_seq): Likewise.
5455 (add_header_seq): Likewise.
5456 (interleave_supported_p): Likewise.
5457 (vect_maybe_permute_loop_masks): Likewise.
5458 (vect_set_loop_masks_directly): Likewise.
5459 (vect_set_loop_condition_masked): Likewise.
5460 (vect_set_loop_condition_unmasked): New function, split out from
5461 slpeel_make_loop_iterate_ntimes.
5462 (slpeel_make_loop_iterate_ntimes): Rename to..
5463 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5464 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5465 (vect_do_peeling): Update call accordingly.
5466 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5468 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5469 mask_compare_type, can_fully_mask_p and fully_masked_p.
5470 (release_vec_loop_masks): New function.
5471 (_loop_vec_info): Use it to free the loop masks.
5472 (can_produce_all_loop_masks_p): New function.
5473 (vect_get_max_nscalars_per_iter): Likewise.
5474 (vect_verify_full_masking): Likewise.
5475 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5476 retries, and free the mask rgroups before retrying. Check loop-wide
5477 reasons for disallowing fully-masked loops. Make the final decision
5478 about whether use a fully-masked loop or not.
5479 (vect_estimate_min_profitable_iters): Do not assume that peeling
5480 for the number of iterations will be needed for fully-masked loops.
5481 (vectorizable_reduction): Disable fully-masked loops.
5482 (vectorizable_live_operation): Likewise.
5483 (vect_halve_mask_nunits): New function.
5484 (vect_double_mask_nunits): Likewise.
5485 (vect_record_loop_mask): Likewise.
5486 (vect_get_loop_mask): Likewise.
5487 (vect_transform_loop): Handle the case in which the final loop
5488 iteration might handle a partial vector. Call vect_set_loop_condition
5489 instead of slpeel_make_loop_iterate_ntimes.
5490 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5491 (check_load_store_masking): New function.
5492 (prepare_load_store_mask): Likewise.
5493 (vectorizable_store): Handle fully-masked loops.
5494 (vectorizable_load): Likewise.
5495 (supportable_widening_operation): Use vect_halve_mask_nunits for
5497 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5498 (vect_gen_while): New function.
5499 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5500 (aarch64_uqdec<mode>): New insn.
5502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5503 Alan Hayward <alan.hayward@arm.com>
5504 David Sherwood <david.sherwood@arm.com>
5506 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5507 (reduc_xor_scal_optab): New optabs.
5508 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5509 (reduc_xor_scal_@var{m}): Document.
5510 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5511 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5513 * fold-const-call.c (fold_const_call): Handle them.
5514 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5515 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5516 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5517 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5518 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5519 (UNSPEC_XORV): New unspecs.
5520 (optab): Add entries for them.
5521 (BITWISEV): New int iterator.
5522 (bit_reduc_op): New int attributes.
5524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5525 Alan Hayward <alan.hayward@arm.com>
5526 David Sherwood <david.sherwood@arm.com>
5528 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5529 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5530 * optabs.def (vec_shl_insert_optab): New optab.
5531 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5532 (duplicate_and_interleave): Likewise.
5533 * tree-vect-loop.c: Include internal-fn.h.
5534 (neutral_op_for_slp_reduction): New function, split out from
5535 get_initial_defs_for_reduction.
5536 (get_initial_def_for_reduction): Handle option 2 for variable-length
5537 vectors by loading the neutral value into a vector and then shifting
5538 the initial value into element 0.
5539 (get_initial_defs_for_reduction): Replace the code argument with
5540 the neutral value calculated by neutral_op_for_slp_reduction.
5541 Use gimple_build_vector for constant-length vectors.
5542 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5543 but the first group_size elements have a neutral value.
5544 Use duplicate_and_interleave otherwise.
5545 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5546 Update call to get_initial_defs_for_reduction. Handle SLP
5547 reductions for variable-length vectors by creating one vector
5548 result for each scalar result, with the elements associated
5549 with other scalar results stubbed out with the neutral value.
5550 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5551 Require IFN_VEC_SHL_INSERT for double reductions on
5552 variable-length vectors, or SLP reductions that have
5553 a neutral value. Require can_duplicate_and_interleave_p
5554 support for variable-length unchained SLP reductions if there
5555 is no neutral value, such as for MIN/MAX reductions. Also require
5556 the number of vector elements to be a multiple of the number of
5557 SLP statements when doing variable-length unchained SLP reductions.
5558 Update call to vect_create_epilog_for_reduction.
5559 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5560 and remove initial values.
5561 (duplicate_and_interleave): Make public.
5562 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5563 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5566 Alan Hayward <alan.hayward@arm.com>
5567 David Sherwood <david.sherwood@arm.com>
5569 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5570 (can_duplicate_and_interleave_p): New function.
5571 (vect_get_and_check_slp_defs): Take the vector of statements
5572 rather than just the current one. Remove excess parentheses.
5573 Restriction rejectinon of vect_constant_def and vect_external_def
5574 for variable-length vectors to boolean types, or types for which
5575 can_duplicate_and_interleave_p is false.
5576 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5577 (duplicate_and_interleave): New function.
5578 (vect_get_constant_vectors): Use gimple_build_vector for
5579 constant-length vectors and suitable variable-length constant
5580 vectors. Use duplicate_and_interleave for other variable-length
5581 vectors. Don't defer the update when inserting new statements.
5583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5584 Alan Hayward <alan.hayward@arm.com>
5585 David Sherwood <david.sherwood@arm.com>
5587 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5588 min_profitable_iters doesn't go negative.
5590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5591 Alan Hayward <alan.hayward@arm.com>
5592 David Sherwood <david.sherwood@arm.com>
5594 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5595 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5596 * optabs.def (vec_mask_load_lanes_optab): New optab.
5597 (vec_mask_store_lanes_optab): Likewise.
5598 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5599 (MASK_STORE_LANES): Likewise.
5600 * internal-fn.c (mask_load_lanes_direct): New macro.
5601 (mask_store_lanes_direct): Likewise.
5602 (expand_mask_load_optab_fn): Handle masked operations.
5603 (expand_mask_load_lanes_optab_fn): New macro.
5604 (expand_mask_store_optab_fn): Handle masked operations.
5605 (expand_mask_store_lanes_optab_fn): New macro.
5606 (direct_mask_load_lanes_optab_supported_p): Likewise.
5607 (direct_mask_store_lanes_optab_supported_p): Likewise.
5608 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5610 (vect_load_lanes_supported): Likewise.
5611 * tree-vect-data-refs.c (strip_conversion): New function.
5612 (can_group_stmts_p): Likewise.
5613 (vect_analyze_data_ref_accesses): Use it instead of checking
5614 for a pair of assignments.
5615 (vect_store_lanes_supported): Take a masked_p parameter.
5616 (vect_load_lanes_supported): Likewise.
5617 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5618 vect_store_lanes_supported and vect_load_lanes_supported.
5619 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5620 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5621 parameter. Don't allow gaps for masked accesses.
5622 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5623 and vect_load_lanes_supported.
5624 (get_load_store_type): Take a masked_p parameter and update
5625 call to get_group_load_store_type.
5626 (vectorizable_store): Update call to get_load_store_type.
5627 Handle IFN_MASK_STORE_LANES.
5628 (vectorizable_load): Update call to get_load_store_type.
5629 Handle IFN_MASK_LOAD_LANES.
5631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5632 Alan Hayward <alan.hayward@arm.com>
5633 David Sherwood <david.sherwood@arm.com>
5635 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5637 * config/aarch64/aarch64-protos.h
5638 (aarch64_sve_struct_memory_operand_p): Declare.
5639 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5640 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5641 (VPRED, vpred): Handle SVE structure modes.
5642 * config/aarch64/constraints.md (Utx): New constraint.
5643 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5644 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5645 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5646 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5647 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5648 structure modes. Split into pieces after RA.
5649 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5650 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5652 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5653 SVE structure modes.
5654 (aarch64_classify_address): Likewise.
5655 (sizetochar): Move earlier in file.
5656 (aarch64_print_operand): Handle SVE register lists.
5657 (aarch64_array_mode): New function.
5658 (aarch64_sve_struct_memory_operand_p): Likewise.
5659 (TARGET_ARRAY_MODE): Redefine.
5661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5662 Alan Hayward <alan.hayward@arm.com>
5663 David Sherwood <david.sherwood@arm.com>
5665 * target.def (array_mode): New target hook.
5666 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5667 * doc/tm.texi: Regenerate.
5668 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5669 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5670 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5672 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5676 Alan Hayward <alan.hayward@arm.com>
5677 David Sherwood <david.sherwood@arm.com>
5679 * fold-const.c (fold_binary_loc): Check the argument types
5680 rather than the result type when testing for a vector operation.
5682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5684 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5685 * doc/tm.texi: Regenerate.
5687 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5688 Alan Hayward <alan.hayward@arm.com>
5689 David Sherwood <david.sherwood@arm.com>
5691 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5692 (sve): Document new AArch64 extension.
5693 * doc/md.texi (w): Extend the description of the AArch64
5694 constraint to include SVE vectors.
5695 (Upl, Upa): Document new AArch64 predicate constraints.
5696 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5698 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5699 (msve-vector-bits=): New option.
5700 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5701 SVE when these are disabled.
5702 (sve): New extension.
5703 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5704 modes. Adjust their number of units based on aarch64_sve_vg.
5705 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5706 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5707 aarch64_addr_query_type.
5708 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5709 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5710 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5711 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5712 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5713 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5714 (aarch64_simd_imm_zero_p): Delete.
5715 (aarch64_check_zero_based_sve_index_immediate): Declare.
5716 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5717 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5718 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5719 (aarch64_sve_float_mul_immediate_p): Likewise.
5720 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5722 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5723 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5724 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5725 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5726 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5727 (aarch64_regmode_natural_size): Likewise.
5728 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5729 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5731 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5732 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5733 for VG and the SVE predicate registers.
5734 (V_ALIASES): Add a "z"-prefixed alias.
5735 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5736 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5737 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5738 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5739 (REG_CLASS_NAMES): Add entries for them.
5740 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5741 and the predicate registers.
5742 (aarch64_sve_vg): Declare.
5743 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5744 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5745 (REGMODE_NATURAL_SIZE): Define.
5746 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5748 * config/aarch64/aarch64.c: Include cfgrtl.h.
5749 (simd_immediate_info): Add a constructor for series vectors,
5750 and an associated step field.
5751 (aarch64_sve_vg): New variable.
5752 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5753 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5754 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5755 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5756 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5757 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5758 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5759 (aarch64_get_mask_mode): New functions.
5760 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5761 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5762 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5763 predicate modes and predicate registers. Explicitly restrict
5764 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5765 to store a vector mode if it is recognized by
5766 aarch64_classify_vector_mode.
5767 (aarch64_regmode_natural_size): New function.
5768 (aarch64_hard_regno_caller_save_mode): Return the original mode
5770 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5771 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5772 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5773 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5775 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5776 does not overlap dest if the function is frame-related. Handle
5778 (aarch64_split_add_offset): New function.
5779 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5780 them aarch64_add_offset.
5781 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5782 and update call to aarch64_sub_sp.
5783 (aarch64_add_cfa_expression): New function.
5784 (aarch64_expand_prologue): Pass extra temporary registers to the
5785 functions above. Handle the case in which we need to emit new
5786 DW_CFA_expressions for registers that were originally saved
5787 relative to the stack pointer, but now have to be expressed
5788 relative to the frame pointer.
5789 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5791 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5792 IP0 and IP1 values for SVE frames.
5793 (aarch64_expand_vec_series): New function.
5794 (aarch64_expand_sve_widened_duplicate): Likewise.
5795 (aarch64_expand_sve_const_vector): Likewise.
5796 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5797 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5798 into the register, rather than emitting a SET directly.
5799 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5800 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5801 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5802 (offset_9bit_signed_scaled_p): New functions.
5803 (aarch64_replicate_bitmask_imm): New function.
5804 (aarch64_bitmask_imm): Use it.
5805 (aarch64_cannot_force_const_mem): Reject expressions involving
5806 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5807 (aarch64_classify_index): Handle SVE indices, by requiring
5808 a plain register index with a scale that matches the element size.
5809 (aarch64_classify_address): Handle SVE addresses. Assert that
5810 the mode of the address is VOIDmode or an integer mode.
5811 Update call to aarch64_classify_symbol.
5812 (aarch64_classify_symbolic_expression): Update call to
5813 aarch64_classify_symbol.
5814 (aarch64_const_vec_all_in_range_p): New function.
5815 (aarch64_print_vector_float_operand): Likewise.
5816 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5817 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5818 and the FP immediates 1.0 and 0.5.
5819 (aarch64_print_address_internal): Handle SVE addresses.
5820 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5821 (aarch64_regno_regclass): Handle predicate registers.
5822 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5824 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5825 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5826 (aarch64_convert_sve_vector_bits): New function.
5827 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5828 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5830 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5831 Handle SVE vector and predicate modes. Accept VL-based constants
5832 that need only one temporary register, and VL offsets that require
5833 no temporary registers.
5834 (aarch64_conditional_register_usage): Mark the predicate registers
5835 as fixed if SVE isn't available.
5836 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5837 Return true for SVE vector and predicate modes.
5838 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5839 rather than an unsigned int. Handle SVE modes.
5840 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5842 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5844 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5845 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5846 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5847 (aarch64_sve_float_mul_immediate_p): New functions.
5848 (aarch64_sve_valid_immediate): New function.
5849 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5850 Explicitly reject structure modes. Check for INDEX constants.
5851 Handle PTRUE and PFALSE constants.
5852 (aarch64_check_zero_based_sve_index_immediate): New function.
5853 (aarch64_simd_imm_zero_p): Delete.
5854 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5855 vector modes. Accept constants in the range of CNT[BHWD].
5856 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5857 ask for an Advanced SIMD mode.
5858 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5859 (aarch64_simd_vector_alignment): Handle SVE predicates.
5860 (aarch64_vectorize_preferred_vector_alignment): New function.
5861 (aarch64_simd_vector_alignment_reachable): Use it instead of
5863 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5864 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5866 (MAX_VECT_LEN): Delete.
5867 (expand_vec_perm_d): Add a vec_flags field.
5868 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5869 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5870 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5872 (aarch64_evpc_rev): Rename to...
5873 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5874 (aarch64_evpc_rev_global): New function.
5875 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5876 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5878 (aarch64_evpc_sve_tbl): New function.
5879 (aarch64_expand_vec_perm_const_1): Update after rename of
5880 aarch64_evpc_rev. Handle SVE permutes too, trying
5881 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5882 than aarch64_evpc_tbl.
5883 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5884 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5885 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5886 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5887 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5888 (aarch64_expand_sve_vcond): New functions.
5889 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5890 of aarch64_vector_mode_p.
5891 (aarch64_dwarf_poly_indeterminate_value): New function.
5892 (aarch64_compute_pressure_classes): Likewise.
5893 (aarch64_can_change_mode_class): Likewise.
5894 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5895 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5896 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5897 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5898 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5899 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5900 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5901 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5903 (Dn, Dl, Dr): Accept const as well as const_vector.
5904 (Dz): Likewise. Compare against CONST0_RTX.
5905 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5906 of "vector" where appropriate.
5907 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5908 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5909 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5910 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5911 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5912 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5913 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5914 (v_int_equiv): Extend to SVE modes.
5915 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5917 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5918 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5919 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5920 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5921 (SVE_COND_FP_CMP): New int iterators.
5922 (perm_hilo): Handle the new unpack unspecs.
5923 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5925 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5926 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5927 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5928 (aarch64_equality_operator, aarch64_constant_vector_operand)
5929 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5930 (aarch64_sve_nonimmediate_operand): Likewise.
5931 (aarch64_sve_general_operand): Likewise.
5932 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5933 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5934 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5935 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5936 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5937 (aarch64_sve_float_arith_immediate): Likewise.
5938 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5939 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5940 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5941 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5942 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5943 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5944 (aarch64_sve_float_arith_operand): Likewise.
5945 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5946 (aarch64_sve_float_mul_operand): Likewise.
5947 (aarch64_sve_vec_perm_operand): Likewise.
5948 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5949 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5950 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5951 as well as const_vector.
5952 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5953 in file. Use CONST0_RTX and CONSTM1_RTX.
5954 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5955 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5956 Use aarch64_simd_imm_zero.
5957 * config/aarch64/aarch64-sve.md: New file.
5958 * config/aarch64/aarch64.md: Include it.
5959 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5960 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5961 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5962 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5963 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5964 (sve): New attribute.
5965 (enabled): Disable instructions with the sve attribute unless
5967 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5968 aarch64_expand_mov_immediate.
5969 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5970 CNT[BHSD] immediates.
5971 (movti): Split CONST_POLY_INT moves into two halves.
5972 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5973 Split additions that need a temporary here if the destination
5974 is the stack pointer.
5975 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5976 (*add<mode>3_poly_1): New instruction.
5977 (set_clobber_cc): New expander.
5979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5981 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5982 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5983 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5984 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5985 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5986 Change innermode from fixed_mode_size to machine_mode.
5987 (simplify_subreg): Update call accordingly. Handle a constant-sized
5988 subreg of a variable-length CONST_VECTOR.
5990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5991 Alan Hayward <alan.hayward@arm.com>
5992 David Sherwood <david.sherwood@arm.com>
5994 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5995 (add_offset_to_base): New function, split out from...
5996 (create_mem_ref): ...here. When handling a scale other than 1,
5997 check first whether the address is valid without the offset.
5998 Add it into the base if so, leaving the index and scale as-is.
6000 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6003 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6004 fold_for_warn before checking if arg2 is INTEGER_CST.
6006 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6008 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6009 (store_multiple_operation): Delete.
6010 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6011 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6012 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6013 guarded by TARGET_STRING.
6014 (rs6000_output_load_multiple): Delete.
6015 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6016 OPTION_MASK_STRING / TARGET_STRING handling.
6017 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6018 (const rs6000_opt_masks) <"string">: Change mask to 0.
6019 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6020 (MASK_STRING): Delete.
6021 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6023 (load_multiple): Delete.
6030 (store_multiple): Delete.
6037 (movmemsi_8reg): Delete.
6038 (corresponding unnamed define_insn): Delete.
6039 (movmemsi_6reg): Delete.
6040 (corresponding unnamed define_insn): Delete.
6041 (movmemsi_4reg): Delete.
6042 (corresponding unnamed define_insn): Delete.
6043 (movmemsi_2reg): Delete.
6044 (corresponding unnamed define_insn): Delete.
6045 (movmemsi_1reg): Delete.
6046 (corresponding unnamed define_insn): Delete.
6047 * config/rs6000/rs6000.opt (mno-string): New.
6048 (mstring): Replace by deprecation warning stub.
6049 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6051 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6053 * regrename.c (regrename_do_replace): If replacing the same
6054 reg multiple times, try to reuse last created gen_raw_REG.
6057 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6058 main to workaround a bug in GDB.
6060 2018-01-12 Tom de Vries <tom@codesourcery.com>
6063 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6065 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6067 PR rtl-optimization/80481
6068 * ira-color.c (get_cap_member): New function.
6069 (allocnos_conflict_by_live_ranges_p): Use it.
6070 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6071 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6073 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6076 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6077 (*saddl_se_1): Ditto.
6079 (*ssubl_se_1): Ditto.
6081 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6083 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6084 rather than wi::to_widest for DR_INITs.
6085 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6086 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6087 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6089 (vect_analyze_group_access_1): Note that here.
6091 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6093 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6094 polynomial type sizes.
6096 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6098 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6099 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6100 (gimple_add_tmp_var): Likewise.
6102 2018-01-12 Martin Liska <mliska@suse.cz>
6104 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6105 (gimple_alloc_sizes): Likewise.
6106 (dump_gimple_statistics): Use PRIu64 in printf format.
6107 * gimple.h: Change uint64_t to int.
6109 2018-01-12 Martin Liska <mliska@suse.cz>
6111 * tree-core.h: Use uint64_t instead of int.
6112 * tree.c (tree_node_counts): Likewise.
6113 (tree_node_sizes): Likewise.
6114 (dump_tree_statistics): Use PRIu64 in printf format.
6116 2018-01-12 Martin Liska <mliska@suse.cz>
6118 * Makefile.in: As qsort_chk is implemented in vec.c, add
6119 vec.o to linkage of gencfn-macros.
6120 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6121 passing the info to record_node_allocation_statistics.
6122 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6124 * ggc-common.c (struct ggc_usage): Add operator== and use
6125 it in operator< and compare function.
6126 * mem-stats.h (struct mem_usage): Likewise.
6127 * vec.c (struct vec_usage): Remove operator< and compare
6128 function. Can be simply inherited.
6130 2018-01-12 Martin Jambor <mjambor@suse.cz>
6133 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6134 * tree-ssa-math-opts.c: Include domwalk.h.
6135 (convert_mult_to_fma_1): New function.
6136 (fma_transformation_info): New type.
6137 (fma_deferring_state): Likewise.
6138 (cancel_fma_deferring): New function.
6139 (result_of_phi): Likewise.
6140 (last_fma_candidate_feeds_initial_phi): Likewise.
6141 (convert_mult_to_fma): Added deferring logic, split actual
6142 transformation to convert_mult_to_fma_1.
6143 (math_opts_dom_walker): New type.
6144 (math_opts_dom_walker::after_dom_children): New method, body moved
6145 here from pass_optimize_widening_mul::execute, added deferring logic
6147 (pass_optimize_widening_mul::execute): Moved most of code to
6148 math_opts_dom_walker::after_dom_children.
6149 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6150 * config/i386/i386.c (ix86_option_override_internal): Added
6151 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6153 2018-01-12 Richard Biener <rguenther@suse.de>
6156 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6157 inline instance vars.
6159 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6162 * config/rx/rx.c (rx_is_restricted_memory_address):
6165 2018-01-12 Richard Biener <rguenther@suse.de>
6167 PR tree-optimization/80846
6168 * target.def (split_reduction): New target hook.
6169 * targhooks.c (default_split_reduction): New function.
6170 * targhooks.h (default_split_reduction): Declare.
6171 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6172 target requests first reduce vectors by combining low and high
6174 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6175 (get_vectype_for_scalar_type_and_size): Export.
6176 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6177 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6178 * doc/tm.texi: Regenerate.
6179 * config/i386/i386.c (ix86_split_reduction): Implement
6180 TARGET_VECTORIZE_SPLIT_REDUCTION.
6182 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6185 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6186 in PIC mode except for TARGET_VXWORKS_RTP.
6187 * config/sparc/sparc.c: Include cfgrtl.h.
6188 (TARGET_INIT_PIC_REG): Define.
6189 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6190 (sparc_pic_register_p): New predicate.
6191 (sparc_legitimate_address_p): Use it.
6192 (sparc_legitimize_pic_address): Likewise.
6193 (sparc_delegitimize_address): Likewise.
6194 (sparc_mode_dependent_address_p): Likewise.
6195 (gen_load_pcrel_sym): Remove 4th parameter.
6196 (load_got_register): Adjust call to above. Remove obsolete stuff.
6197 (sparc_expand_prologue): Do not call load_got_register here.
6198 (sparc_flat_expand_prologue): Likewise.
6199 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6200 (sparc_use_pseudo_pic_reg): New function.
6201 (sparc_init_pic_reg): Likewise.
6202 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6203 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6205 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6207 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6208 Add item for branch_cost.
6210 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6212 PR rtl-optimization/83565
6213 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6214 not extend the result to a larger mode for rotate operations.
6215 (num_sign_bit_copies1): Likewise.
6217 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6220 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6222 Use values-Xc.o for -pedantic.
6223 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6225 2018-01-12 Martin Liska <mliska@suse.cz>
6228 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6230 (possible_polymorphic_call_targets): Use it.
6231 (ipa_devirt): Likewise.
6233 2018-01-12 Martin Liska <mliska@suse.cz>
6235 * profile-count.h (enum profile_quality): Use 0 as invalid
6236 enum value of profile_quality.
6238 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6240 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6241 -mext-string options.
6243 2018-01-12 Richard Biener <rguenther@suse.de>
6245 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6246 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6247 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6249 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6251 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6253 * configure.ac (--with-long-double-format): Add support for the
6254 configuration option to change the default long double format on
6256 * config.gcc (powerpc*-linux*-*): Likewise.
6257 * configure: Regenerate.
6258 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6259 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6260 used without modification.
6262 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6264 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6265 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6266 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6267 MISC_BUILTIN_SPEC_BARRIER.
6268 (rs6000_init_builtins): Likewise.
6269 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6271 (speculation_barrier): New define_insn.
6272 * doc/extend.texi: Document __builtin_speculation_barrier.
6274 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6277 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6278 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6279 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6281 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6282 integral modes instead of "ss" and "sd".
6283 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6284 vectors with 32-bit and 64-bit elements.
6285 (vecdupssescalarmodesuffix): New mode attribute.
6286 (vec_dup<mode>): Use it.
6288 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6291 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6292 frame if argument is passed on stack.
6294 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6297 * ree.c (combine_reaching_defs): Optimize also
6298 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6299 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6301 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6304 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6306 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6309 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6310 after they are computed.
6312 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6314 PR tree-optimization/83695
6315 * gimple-loop-linterchange.cc
6316 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6317 reset cached scev information after interchange.
6318 (pass_linterchange::execute): Remove call to scev_reset_htab.
6320 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6322 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6323 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6324 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6325 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6326 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6327 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6328 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6329 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6330 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6331 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6332 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6333 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6334 (V_lane_reg): Likewise.
6335 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6337 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6338 (vfmal_lane_low<mode>_intrinsic,
6339 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6340 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6341 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6342 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6343 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6344 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6346 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6348 * config/arm/arm-cpus.in (fp16fml): New feature.
6349 (ALL_SIMD): Add fp16fml.
6350 (armv8.2-a): Add fp16fml as an option.
6351 (armv8.3-a): Likewise.
6352 (armv8.4-a): Add fp16fml as part of fp16.
6353 * config/arm/arm.h (TARGET_FP16FML): Define.
6354 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6356 * config/arm/arm-modes.def (V2HF): Define.
6357 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6358 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6359 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6360 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6361 vfmsl_low, vfmsl_high): New set of builtins.
6362 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6363 (vfml_op): New code attribute.
6364 (VFMLHALVES): New int iterator.
6365 (VFML, VFMLSEL): New mode attributes.
6366 (V_reg): Define mapping for V2HF.
6367 (V_hi, V_lo): New mode attributes.
6368 (VF_constraint): Likewise.
6369 (vfml_half, vfml_half_selector): New int attributes.
6370 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6372 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6373 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6375 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6376 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6377 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6378 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6380 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6381 Document new effective target and option set.
6383 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6385 * config/arm/arm-cpus.in (armv8_4): New feature.
6386 (ARMv8_4a): New fgroup.
6387 (armv8.4-a): New arch.
6388 * config/arm/arm-tables.opt: Regenerate.
6389 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6390 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6391 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6392 Add matching rules for -march=armv8.4-a and extensions.
6393 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6395 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6398 * config/rx/rx.md (BW): New mode attribute.
6399 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6401 2018-01-11 Richard Biener <rguenther@suse.de>
6403 PR tree-optimization/83435
6404 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6405 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6406 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6408 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6409 Alan Hayward <alan.hayward@arm.com>
6410 David Sherwood <david.sherwood@arm.com>
6412 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6414 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6415 (aarch64_print_address_internal): Use it to check for a zero offset.
6417 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6418 Alan Hayward <alan.hayward@arm.com>
6419 David Sherwood <david.sherwood@arm.com>
6421 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6422 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6423 Return a poly_int64 rather than a HOST_WIDE_INT.
6424 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6425 rather than a HOST_WIDE_INT.
6426 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6427 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6428 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6429 final_offset from HOST_WIDE_INT to poly_int64.
6430 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6431 to_constant when getting the number of units in an Advanced SIMD
6433 (aarch64_builtin_vectorized_function): Check for a constant number
6435 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6437 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6438 attribute instead of GET_MODE_NUNITS.
6439 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6440 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6441 GET_MODE_SIZE for fixed-size registers.
6442 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6443 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6444 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6445 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6446 (aarch64_print_operand, aarch64_print_address_internal)
6447 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6448 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6449 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6450 Handle polynomial GET_MODE_SIZE.
6451 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6452 wider than SImode without modification.
6453 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6454 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6455 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6456 passing and returning SVE modes.
6457 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6458 rather than GEN_INT.
6459 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6460 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6461 (aarch64_allocate_and_probe_stack_space): Likewise.
6462 (aarch64_layout_frame): Cope with polynomial offsets.
6463 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6464 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6466 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6467 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6468 poly_int64 rather than a HOST_WIDE_INT.
6469 (aarch64_get_separate_components, aarch64_process_components)
6470 (aarch64_expand_prologue, aarch64_expand_epilogue)
6471 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6472 (aarch64_anchor_offset): New function, split out from...
6473 (aarch64_legitimize_address): ...here.
6474 (aarch64_builtin_vectorization_cost): Handle polynomial
6475 TYPE_VECTOR_SUBPARTS.
6476 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6478 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6479 number of elements from the PARALLEL rather than the mode.
6480 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6481 rather than GET_MODE_BITSIZE.
6482 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6483 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6484 (aarch64_expand_vec_perm_const_1): Handle polynomial
6485 d->perm.length () and d->perm elements.
6486 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6487 Apply to_constant to d->perm elements.
6488 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6489 polynomial CONST_VECTOR_NUNITS.
6490 (aarch64_move_pointer): Take amount as a poly_int64 rather
6492 (aarch64_progress_pointer): Avoid temporary variable.
6493 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6494 the mode attribute instead of GET_MODE.
6496 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6497 Alan Hayward <alan.hayward@arm.com>
6498 David Sherwood <david.sherwood@arm.com>
6500 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6501 x exists before using it.
6502 (aarch64_add_constant_internal): Rename to...
6503 (aarch64_add_offset_1): ...this. Replace regnum with separate
6504 src and dest rtxes. Handle the case in which they're different,
6505 including when the offset is zero. Replace scratchreg with an rtx.
6506 Use 2 additions if there is no spare register into which we can
6507 move a 16-bit constant.
6508 (aarch64_add_constant): Delete.
6509 (aarch64_add_offset): Replace reg with separate src and dest
6510 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6511 Use aarch64_add_offset_1.
6512 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6513 an rtx rather than an int. Take the delta as a poly_int64
6514 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6515 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6516 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6517 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6518 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6520 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6521 aarch64_add_constant.
6523 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6525 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6526 Use scalar_float_mode.
6528 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6530 * config/aarch64/aarch64-simd.md
6531 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6532 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6533 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6534 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6535 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6536 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6537 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6538 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6539 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6540 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6542 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6545 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6546 targ_options->x_arm_arch_string is non NULL.
6548 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6550 * config/aarch64/aarch64.h
6551 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6553 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6556 * expmed.c (emit_store_flag_force): Swap if const op0
6557 and change VOIDmode to mode of op0.
6559 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6561 PR rtl-optimization/83761
6562 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6563 than bytes to mode_for_size.
6565 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6568 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6569 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6572 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6575 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6576 when in layout mode.
6577 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6578 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6581 2018-01-10 Michael Collison <michael.collison@arm.com>
6583 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6584 * config/aarch64/aarch64-option-extension.def: Add
6585 AARCH64_OPT_EXTENSION of 'fp16fml'.
6586 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6587 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6588 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6589 * config/aarch64/constraints.md (Ui7): New constraint.
6590 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6591 (VFMLA_SEL_W): Ditto.
6594 (VFMLA16_LOW): New int iterator.
6595 (VFMLA16_HIGH): Ditto.
6596 (UNSPEC_FMLAL): New unspec.
6597 (UNSPEC_FMLSL): Ditto.
6598 (UNSPEC_FMLAL2): Ditto.
6599 (UNSPEC_FMLSL2): Ditto.
6600 (f16mac): New code attribute.
6601 * config/aarch64/aarch64-simd-builtins.def
6602 (aarch64_fmlal_lowv2sf): Ditto.
6603 (aarch64_fmlsl_lowv2sf): Ditto.
6604 (aarch64_fmlalq_lowv4sf): Ditto.
6605 (aarch64_fmlslq_lowv4sf): Ditto.
6606 (aarch64_fmlal_highv2sf): Ditto.
6607 (aarch64_fmlsl_highv2sf): Ditto.
6608 (aarch64_fmlalq_highv4sf): Ditto.
6609 (aarch64_fmlslq_highv4sf): Ditto.
6610 (aarch64_fmlal_lane_lowv2sf): Ditto.
6611 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6612 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6613 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6614 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6615 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6616 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6617 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6618 (aarch64_fmlal_lane_highv2sf): Ditto.
6619 (aarch64_fmlsl_lane_highv2sf): Ditto.
6620 (aarch64_fmlal_laneq_highv2sf): Ditto.
6621 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6622 (aarch64_fmlalq_lane_highv4sf): Ditto.
6623 (aarch64_fmlsl_lane_highv4sf): Ditto.
6624 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6625 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6626 * config/aarch64/aarch64-simd.md:
6627 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6628 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6629 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6630 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6631 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6632 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6633 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6634 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6635 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6636 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6637 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6638 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6639 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6640 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6641 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6642 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6643 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6644 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6645 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6646 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6647 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6648 (vfmlsl_low_u32): Ditto.
6649 (vfmlalq_low_u32): Ditto.
6650 (vfmlslq_low_u32): Ditto.
6651 (vfmlal_high_u32): Ditto.
6652 (vfmlsl_high_u32): Ditto.
6653 (vfmlalq_high_u32): Ditto.
6654 (vfmlslq_high_u32): Ditto.
6655 (vfmlal_lane_low_u32): Ditto.
6656 (vfmlsl_lane_low_u32): Ditto.
6657 (vfmlal_laneq_low_u32): Ditto.
6658 (vfmlsl_laneq_low_u32): Ditto.
6659 (vfmlalq_lane_low_u32): Ditto.
6660 (vfmlslq_lane_low_u32): Ditto.
6661 (vfmlalq_laneq_low_u32): Ditto.
6662 (vfmlslq_laneq_low_u32): Ditto.
6663 (vfmlal_lane_high_u32): Ditto.
6664 (vfmlsl_lane_high_u32): Ditto.
6665 (vfmlal_laneq_high_u32): Ditto.
6666 (vfmlsl_laneq_high_u32): Ditto.
6667 (vfmlalq_lane_high_u32): Ditto.
6668 (vfmlslq_lane_high_u32): Ditto.
6669 (vfmlalq_laneq_high_u32): Ditto.
6670 (vfmlslq_laneq_high_u32): Ditto.
6671 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6672 (AARCH64_FL_FOR_ARCH8_4): New.
6673 (AARCH64_ISA_F16FML): New ISA flag.
6674 (TARGET_F16FML): New feature flag for fp16fml.
6675 (doc/invoke.texi): Document new fp16fml option.
6677 2018-01-10 Michael Collison <michael.collison@arm.com>
6679 * config/aarch64/aarch64-builtins.c:
6680 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6681 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6682 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6683 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6684 (AARCH64_ISA_SHA3): New ISA flag.
6685 (TARGET_SHA3): New feature flag for sha3.
6686 * config/aarch64/iterators.md (sha512_op): New int attribute.
6687 (CRYPTO_SHA512): New int iterator.
6688 (UNSPEC_SHA512H): New unspec.
6689 (UNSPEC_SHA512H2): Ditto.
6690 (UNSPEC_SHA512SU0): Ditto.
6691 (UNSPEC_SHA512SU1): Ditto.
6692 * config/aarch64/aarch64-simd-builtins.def
6693 (aarch64_crypto_sha512hqv2di): New builtin.
6694 (aarch64_crypto_sha512h2qv2di): Ditto.
6695 (aarch64_crypto_sha512su0qv2di): Ditto.
6696 (aarch64_crypto_sha512su1qv2di): Ditto.
6697 (aarch64_eor3qv8hi): Ditto.
6698 (aarch64_rax1qv2di): Ditto.
6699 (aarch64_xarqv2di): Ditto.
6700 (aarch64_bcaxqv8hi): Ditto.
6701 * config/aarch64/aarch64-simd.md:
6702 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6703 (aarch64_crypto_sha512su0qv2di): Ditto.
6704 (aarch64_crypto_sha512su1qv2di): Ditto.
6705 (aarch64_eor3qv8hi): Ditto.
6706 (aarch64_rax1qv2di): Ditto.
6707 (aarch64_xarqv2di): Ditto.
6708 (aarch64_bcaxqv8hi): Ditto.
6709 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6710 (vsha512h2q_u64): Ditto.
6711 (vsha512su0q_u64): Ditto.
6712 (vsha512su1q_u64): Ditto.
6713 (veor3q_u16): Ditto.
6714 (vrax1q_u64): Ditto.
6716 (vbcaxq_u16): Ditto.
6717 * config/arm/types.md (crypto_sha512): New type attribute.
6718 (crypto_sha3): Ditto.
6719 (doc/invoke.texi): Document new sha3 option.
6721 2018-01-10 Michael Collison <michael.collison@arm.com>
6723 * config/aarch64/aarch64-builtins.c:
6724 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6725 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6726 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6727 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6728 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6729 (AARCH64_ISA_SM4): New ISA flag.
6730 (TARGET_SM4): New feature flag for sm4.
6731 * config/aarch64/aarch64-simd-builtins.def
6732 (aarch64_sm3ss1qv4si): Ditto.
6733 (aarch64_sm3tt1aq4si): Ditto.
6734 (aarch64_sm3tt1bq4si): Ditto.
6735 (aarch64_sm3tt2aq4si): Ditto.
6736 (aarch64_sm3tt2bq4si): Ditto.
6737 (aarch64_sm3partw1qv4si): Ditto.
6738 (aarch64_sm3partw2qv4si): Ditto.
6739 (aarch64_sm4eqv4si): Ditto.
6740 (aarch64_sm4ekeyqv4si): Ditto.
6741 * config/aarch64/aarch64-simd.md:
6742 (aarch64_sm3ss1qv4si): Ditto.
6743 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6744 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6745 (aarch64_sm4eqv4si): Ditto.
6746 (aarch64_sm4ekeyqv4si): Ditto.
6747 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6748 (sm3part_op): Ditto.
6749 (CRYPTO_SM3TT): Ditto.
6750 (CRYPTO_SM3PART): Ditto.
6751 (UNSPEC_SM3SS1): New unspec.
6752 (UNSPEC_SM3TT1A): Ditto.
6753 (UNSPEC_SM3TT1B): Ditto.
6754 (UNSPEC_SM3TT2A): Ditto.
6755 (UNSPEC_SM3TT2B): Ditto.
6756 (UNSPEC_SM3PARTW1): Ditto.
6757 (UNSPEC_SM3PARTW2): Ditto.
6758 (UNSPEC_SM4E): Ditto.
6759 (UNSPEC_SM4EKEY): Ditto.
6760 * config/aarch64/constraints.md (Ui2): New constraint.
6761 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6762 * config/arm/types.md (crypto_sm3): New type attribute.
6763 (crypto_sm4): Ditto.
6764 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6765 (vsm3tt1aq_u32): Ditto.
6766 (vsm3tt1bq_u32): Ditto.
6767 (vsm3tt2aq_u32): Ditto.
6768 (vsm3tt2bq_u32): Ditto.
6769 (vsm3partw1q_u32): Ditto.
6770 (vsm3partw2q_u32): Ditto.
6771 (vsm4eq_u32): Ditto.
6772 (vsm4ekeyq_u32): Ditto.
6773 (doc/invoke.texi): Document new sm4 option.
6775 2018-01-10 Michael Collison <michael.collison@arm.com>
6777 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6778 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6779 (AARCH64_FL_FOR_ARCH8_4): New.
6780 (AARCH64_FL_V8_4): New flag.
6781 (doc/invoke.texi): Document new armv8.4-a option.
6783 2018-01-10 Michael Collison <michael.collison@arm.com>
6785 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6786 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6787 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6788 * config/aarch64/aarch64-option-extension.def: Add
6789 AARCH64_OPT_EXTENSION of 'sha2'.
6790 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6791 (crypto): Disable sha2 and aes if crypto disabled.
6792 (crypto): Enable aes and sha2 if enabled.
6793 (simd): Disable sha2 and aes if simd disabled.
6794 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6796 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6797 (TARGET_SHA2): New feature flag for sha2.
6798 (TARGET_AES): New feature flag for aes.
6799 * config/aarch64/aarch64-simd.md:
6800 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6801 conditional on TARGET_AES.
6802 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6803 (aarch64_crypto_sha1hsi): Make pattern conditional
6805 (aarch64_crypto_sha1hv4si): Ditto.
6806 (aarch64_be_crypto_sha1hv4si): Ditto.
6807 (aarch64_crypto_sha1su1v4si): Ditto.
6808 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6809 (aarch64_crypto_sha1su0v4si): Ditto.
6810 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6811 (aarch64_crypto_sha256su0v4si): Ditto.
6812 (aarch64_crypto_sha256su1v4si): Ditto.
6813 (doc/invoke.texi): Document new aes and sha2 options.
6815 2018-01-10 Martin Sebor <msebor@redhat.com>
6817 PR tree-optimization/83781
6818 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6821 2018-01-11 Martin Sebor <msebor@gmail.com>
6822 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6824 PR tree-optimization/83501
6825 PR tree-optimization/81703
6827 * tree-ssa-strlen.c (get_string_cst): Rename...
6828 (get_string_len): ...to this. Handle global constants.
6829 (handle_char_store): Adjust.
6831 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6832 Jim Wilson <jimw@sifive.com>
6834 * config/riscv/riscv-protos.h (riscv_output_return): New.
6835 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6836 (riscv_attribute_table, riscv_output_return),
6837 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6838 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6839 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6841 (riscv_expand_prologue): Add early return for naked function.
6842 (riscv_expand_epilogue): Likewise.
6843 (riscv_function_ok_for_sibcall): Return false for naked function.
6844 (riscv_set_current_function): New.
6845 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6846 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6847 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6848 * doc/extend.texi (RISC-V Function Attributes): New.
6850 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6852 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6853 check for 128-bit long double before checking TCmode.
6854 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6855 128-bit long doubles before checking TFmode or TCmode.
6856 (FLOAT128_IBM_P): Likewise.
6858 2018-01-10 Martin Sebor <msebor@redhat.com>
6860 PR tree-optimization/83671
6861 * builtins.c (c_strlen): Unconditionally return zero for the empty
6863 Use -Warray-bounds for warnings.
6864 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6865 for non-constant array indices with COMPONENT_REF, arrays of
6866 arrays, and pointers to arrays.
6867 (gimple_fold_builtin_strlen): Determine and set length range for
6868 non-constant character arrays.
6870 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6873 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6876 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6878 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6880 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6883 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6884 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6885 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6886 indexed_or_indirect_operand predicate.
6887 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6888 (*vsx_le_perm_load_v8hi): Likewise.
6889 (*vsx_le_perm_load_v16qi): Likewise.
6890 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6891 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6892 (*vsx_le_perm_store_v8hi): Likewise.
6893 (*vsx_le_perm_store_v16qi): Likewise.
6894 (eight unnamed splitters): Likewise.
6896 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6898 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6899 * config/rs6000/emmintrin.h: Likewise.
6900 * config/rs6000/mmintrin.h: Likewise.
6901 * config/rs6000/xmmintrin.h: Likewise.
6903 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6906 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6908 * tree.c (tree_nop_conversion): Return true for location wrapper
6910 (maybe_wrap_with_location): New function.
6911 (selftest::check_strip_nops): New function.
6912 (selftest::test_location_wrappers): New function.
6913 (selftest::tree_c_tests): Call it.
6914 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6915 (maybe_wrap_with_location): New decl.
6916 (EXPR_LOCATION_WRAPPER_P): New macro.
6917 (location_wrapper_p): New inline function.
6918 (tree_strip_any_location_wrapper): New inline function.
6920 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6923 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6924 stack_realign_offset for the largest alignment of stack slot
6926 (ix86_find_max_used_stack_alignment): New function.
6927 (ix86_finalize_stack_frame_flags): Use it. Set
6928 max_used_stack_alignment if we don't realign stack.
6929 * config/i386/i386.h (machine_function): Add
6930 max_used_stack_alignment.
6932 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6934 * config/arm/arm.opt (-mbranch-cost): New option.
6935 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6938 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6941 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6942 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6944 2018-01-10 Richard Biener <rguenther@suse.de>
6947 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6948 early out so it also covers the case where we have a non-NULL
6951 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6953 PR tree-optimization/83753
6954 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6955 for non-strided grouped accesses if the number of elements is 1.
6957 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6960 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6961 * i386.h (TARGET_USE_GATHER): Define.
6962 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6964 2018-01-10 Martin Liska <mliska@suse.cz>
6967 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6968 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6970 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6971 CLEANUP_NO_PARTITIONING is not set.
6973 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6975 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6976 for vectors, as a partial revert of r254296.
6977 * rtl.h (const_vec_p): Delete.
6978 (const_vec_duplicate_p): Don't test for vector CONSTs.
6979 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6980 * expmed.c (make_tree): Likewise.
6983 * common.md (E, F): Use CONSTANT_P instead of checking for
6985 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6986 checking for CONST_VECTOR.
6988 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6991 * predict.c (force_edge_cold): Handle in more sane way edges
6994 2018-01-09 Carl Love <cel@us.ibm.com>
6996 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6998 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6999 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7000 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7001 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7002 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7003 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7004 * config/rs6000/rs6000-protos.h: Add extern defition for
7005 rs6000_generate_float2_double_code.
7006 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7008 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7009 (float2_v2df): Add define_expand.
7011 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7014 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7015 op_mode in the force_to_mode call.
7017 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7019 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7020 instead of checking each element individually.
7021 (aarch64_evpc_uzp): Likewise.
7022 (aarch64_evpc_zip): Likewise.
7023 (aarch64_evpc_ext): Likewise.
7024 (aarch64_evpc_rev): Likewise.
7025 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7026 instead of checking each element individually. Return true without
7028 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7029 whether all selected elements come from the same input, instead of
7030 checking each element individually. Remove calls to gen_rtx_REG,
7031 start_sequence and end_sequence and instead assert that no rtl is
7034 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7036 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7037 order of HIGH and CONST checks.
7039 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7041 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7042 if the destination isn't an SSA_NAME.
7044 2018-01-09 Richard Biener <rguenther@suse.de>
7046 PR tree-optimization/83668
7047 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7049 (canonicalize_loop_form): ... here, renamed from ...
7050 (canonicalize_loop_closed_ssa_form): ... this and amended to
7051 swap successor edges for loop exit blocks to make us use
7052 the RPO order we need for initial schedule generation.
7054 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7056 PR tree-optimization/64811
7057 * match.pd: When optimizing comparisons with Inf, avoid
7058 introducing or losing exceptions from comparisons with NaN.
7060 2018-01-09 Martin Liska <mliska@suse.cz>
7063 * asan.c (shadow_mem_size): Add gcc_assert.
7065 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7067 Don't save registers in main().
7070 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7071 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7072 * config/avr/avr.c (avr_set_current_function): Don't error if
7073 naked, OS_task or OS_main are specified at the same time.
7074 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7076 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7078 * common/config/avr/avr-common.c (avr_option_optimization_table):
7079 Switch on -mmain-is-OS_task for optimizing compilations.
7081 2018-01-09 Richard Biener <rguenther@suse.de>
7083 PR tree-optimization/83572
7084 * graphite.c: Include cfganal.h.
7085 (graphite_transform_loops): Connect infinite loops to exit
7086 and remove fake edges at the end.
7088 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7090 * ipa-inline.c (edge_badness): Revert accidental checkin.
7092 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7095 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7096 symbols; not inline clones.
7098 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7101 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7102 hard registers. Formatting fixes.
7104 PR preprocessor/83722
7105 * gcc.c (try_generate_repro): Pass
7106 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7107 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7110 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7111 Kito Cheng <kito.cheng@gmail.com>
7113 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7114 (riscv_leaf_function_p): Delete.
7115 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7117 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7119 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7121 (do_ifelse): New function.
7122 (do_isel): New function.
7123 (do_sub3): New function.
7124 (do_add3): New function.
7125 (do_load_mask_compare): New function.
7126 (do_overlap_load_compare): New function.
7127 (expand_compare_loop): New function.
7128 (expand_block_compare): Call expand_compare_loop() when appropriate.
7129 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7131 (-mblock-compare-inline-loop-limit): New option.
7133 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7136 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7137 Reverse order of second and third operands in first alternative.
7138 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7139 of first and second elements in UNSPEC_VPERMR vector.
7140 (altivec_expand_vec_perm_le): Likewise.
7142 2018-01-08 Jeff Law <law@redhat.com>
7144 PR rtl-optimizatin/81308
7145 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7146 (process_switch): If group_case_labels makes a change, then set
7148 (pass_convert_switch::execute): If a switch is converted, then
7149 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7151 PR rtl-optimization/81308
7152 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7155 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7157 PR target/83663 - Revert r255946
7158 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7159 generation for cases where splatting a value is not useful.
7160 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7161 across a vec_duplicate and a paradoxical subreg forming a vector
7162 mode to a vec_concat.
7164 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7166 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7167 -march=armv8.3-a variants.
7168 * config/arm/t-multilib: Likewise.
7169 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7171 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7173 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7175 (cceq_ior_compare_complement): Give it a name so I can use it, and
7176 change boolean_or_operator predicate to boolean_operator so it can
7177 be used to generate a crand.
7178 (eqne): New code iterator.
7179 (bd/bd_neg): New code_attrs.
7180 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7181 a single define_insn.
7182 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7183 decrement (bdnzt/bdnzf/bdzt/bdzf).
7184 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7185 with the new names of the branch decrement patterns, and added the
7186 names of the branch decrement conditional patterns.
7188 2018-01-08 Richard Biener <rguenther@suse.de>
7190 PR tree-optimization/83563
7191 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7194 2018-01-08 Richard Biener <rguenther@suse.de>
7197 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7199 2018-01-08 Richard Biener <rguenther@suse.de>
7201 PR tree-optimization/83685
7202 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7203 references to abnormals.
7205 2018-01-08 Richard Biener <rguenther@suse.de>
7208 * dwarf2out.c (output_indirect_strings): Handle empty
7209 skeleton_debug_str_hash.
7210 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7212 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7214 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7215 (emit_store_direct): Likewise.
7216 (arc_trampoline_adjust_address): Likewise.
7217 (arc_asm_trampoline_template): New function.
7218 (arc_initialize_trampoline): Use asm_trampoline_template.
7219 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7220 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7221 * config/arc/arc.md (flush_icache): Delete pattern.
7223 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7225 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7226 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7229 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7232 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7233 by not USED_FOR_TARGET.
7234 (make_pass_resolve_sw_modes): Likewise.
7236 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7238 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7241 2018-01-08 Richard Biener <rguenther@suse.de>
7244 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7246 2018-01-08 Richard Biener <rguenther@suse.de>
7249 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7251 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7254 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7255 basic blocks with a small number of successors.
7256 (convert_control_dep_chain_into_preds): Improve handling of
7258 (dump_predicates): Split apart into...
7259 (dump_pred_chain): ...here...
7260 (dump_pred_info): ...and here.
7261 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7262 (can_chain_union_be_invalidated_p): Improve check for invalidation
7264 (uninit_uses_cannot_happen): Avoid unnecessary if
7265 convert_control_dep_chain_into_preds yielded nothing.
7267 2018-01-06 Martin Sebor <msebor@redhat.com>
7269 PR tree-optimization/83640
7270 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7271 subtracting negative offset from size.
7272 (builtin_access::overlap): Adjust offset bounds of the access to fall
7273 within the size of the object if possible.
7275 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7277 PR rtl-optimization/83699
7278 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7279 extract_bit_field_as_subreg to cases in which the extracted
7280 value is also a vector.
7282 * lra-constraints.c (process_alt_operands): Test for the equivalence
7283 substitutions when detecting a possible reload cycle.
7285 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7288 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7289 by default if flag_selective_schedling{,2}. Formatting fixes.
7291 PR rtl-optimization/83682
7292 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7293 if it has non-VECTOR_MODE element mode.
7294 (vec_duplicate_p): Likewise.
7297 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7298 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7300 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7303 * config/i386/i386-builtin.def
7304 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7305 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7306 Require also OPTION_MASK_ISA_AVX512F in addition to
7307 OPTION_MASK_ISA_GFNI.
7308 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7309 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7310 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7311 to OPTION_MASK_ISA_GFNI.
7312 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7313 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7314 OPTION_MASK_ISA_AVX512BW.
7315 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7316 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7317 addition to OPTION_MASK_ISA_GFNI.
7318 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7319 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7320 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7321 to OPTION_MASK_ISA_GFNI.
7322 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7323 a requirement for all ISAs rather than any of them with a few
7325 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7327 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7328 bitmasks to be enabled with 3 exceptions, instead of requiring any
7329 enabled ISA with lots of exceptions.
7330 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7331 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7332 Change avx512bw in isa attribute to avx512f.
7333 * config/i386/sgxintrin.h: Add license boilerplate.
7334 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7335 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7336 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7337 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7339 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7340 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7341 temporarily sse2 rather than sse if not enabled already.
7344 * config/i386/sse.md (VI248_VLBW): Rename to ...
7345 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7346 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7347 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7348 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7349 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7350 mode iterator instead of VI248_VLBW.
7352 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7354 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7355 (record_modified): Skip clobbers; add debug output.
7356 (param_change_prob): Use sreal frequencies.
7358 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7360 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7361 punt for user-aligned variables.
7363 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7365 * tree-chrec.c (chrec_contains_symbols): Return true for
7368 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7371 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7372 of (x|y) == x for BICS pattern.
7374 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7376 PR tree-optimization/83605
7377 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7378 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7381 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7383 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7384 * config/epiphany/rtems.h: New file.
7386 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7387 Uros Bizjak <ubizjak@gmail.com>
7390 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7391 QIreg_operand instead of register_operand predicate.
7392 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7393 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7394 comments instead of -fmitigate[-_]rop.
7396 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7399 * cgraphunit.c (symbol_table::compile): Switch to text_section
7400 before calling assembly_start debug hook.
7401 * run-rtl-passes.c (run_rtl_passes): Likewise.
7404 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7406 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7407 range_int_cst_p rather than !symbolic_range_p before calling
7408 extract_range_from_multiplicative_op_1.
7410 2018-01-04 Jeff Law <law@redhat.com>
7412 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7413 redundant test in assertion.
7415 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7417 * doc/rtl.texi: Document machine_mode wrapper classes.
7419 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7421 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7424 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7426 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7427 the VEC_PERM_EXPR fold to fail.
7429 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7432 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7433 to switched_sections.
7435 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7438 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7441 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7444 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7445 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7447 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7450 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7451 is BLKmode and bitpos not zero or mode change is needed.
7453 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7456 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7459 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7462 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7463 instead of MULT rtx. Update all corresponding splitters.
7465 (*ssub<modesuffix>): Ditto.
7467 (*cmp_sadd_di): Update split patterns.
7468 (*cmp_sadd_si): Ditto.
7469 (*cmp_sadd_sidi): Ditto.
7470 (*cmp_ssub_di): Ditto.
7471 (*cmp_ssub_si): Ditto.
7472 (*cmp_ssub_sidi): Ditto.
7473 * config/alpha/predicates.md (const23_operand): New predicate.
7474 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7475 Look for ASHIFT, not MULT inner operand.
7476 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7478 2018-01-04 Martin Liska <mliska@suse.cz>
7480 PR gcov-profile/83669
7481 * gcov.c (output_intermediate_file): Add version to intermediate
7483 * doc/gcov.texi: Document new field 'version' in intermediate
7484 file format. Fix location of '-k' option of gcov command.
7486 2018-01-04 Martin Liska <mliska@suse.cz>
7489 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7491 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7493 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7495 2018-01-03 Martin Sebor <msebor@redhat.com>
7497 PR tree-optimization/83655
7498 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7499 checking calls with invalid arguments.
7501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7503 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7504 (vectorizable_mask_load_store): Delete.
7505 (vectorizable_call): Return false for masked loads and stores.
7506 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7507 instead of gimple_assign_rhs1.
7508 (vectorizable_load): Handle IFN_MASK_LOAD.
7509 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7513 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7515 (vectorizable_mask_load_store): ...here.
7516 (vectorizable_load): ...and here.
7518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7520 * tree-vect-stmts.c (vect_build_all_ones_mask)
7521 (vect_build_zero_merge_argument): New functions, split out from...
7522 (vectorizable_load): ...here.
7524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7526 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7528 (vectorizable_mask_load_store): ...here.
7529 (vectorizable_store): ...and here.
7531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7533 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7535 (vectorizable_mask_load_store): ...here.
7537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7539 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7540 (vect_model_store_cost): Take a vec_load_store_type instead of a
7542 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7543 (vect_model_store_cost): Take a vec_load_store_type instead of a
7545 (vectorizable_mask_load_store): Update accordingly.
7546 (vectorizable_store): Likewise.
7547 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7551 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7552 IFN_MASK_LOAD calls here rather than...
7553 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7556 Alan Hayward <alan.hayward@arm.com>
7557 David Sherwood <david.sherwood@arm.com>
7559 * expmed.c (extract_bit_field_1): For vector extracts,
7560 fall back to extract_bit_field_as_subreg if vec_extract
7563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7564 Alan Hayward <alan.hayward@arm.com>
7565 David Sherwood <david.sherwood@arm.com>
7567 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7568 they are variable or constant sized.
7569 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7570 slots for constant-sized data.
7572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7573 Alan Hayward <alan.hayward@arm.com>
7574 David Sherwood <david.sherwood@arm.com>
7576 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7577 handling COND_EXPRs with boolean comparisons, try to find a better
7578 basis for the mask type than the boolean itself.
7580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7582 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7583 is calculated and how it can be overridden.
7584 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7585 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7587 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7591 Alan Hayward <alan.hayward@arm.com>
7592 David Sherwood <david.sherwood@arm.com>
7594 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7595 Remove the mode argument.
7596 (aarch64_simd_valid_immediate): Remove the mode and inverse
7598 * config/aarch64/iterators.md (bitsize): New iterator.
7599 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7600 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7601 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7602 aarch64_simd_valid_immediate.
7603 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7604 (aarch64_reg_or_bic_imm): Likewise.
7605 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7606 with an insn_type enum and msl with a modifier_type enum.
7607 Replace element_width with a scalar_mode. Change the shift
7608 to unsigned int. Add constructors for scalar_float_mode and
7609 scalar_int_mode elements.
7610 (aarch64_vect_float_const_representable_p): Delete.
7611 (aarch64_can_const_movi_rtx_p)
7612 (aarch64_simd_scalar_immediate_valid_for_move)
7613 (aarch64_simd_make_constant): Update call to
7614 aarch64_simd_valid_immediate.
7615 (aarch64_advsimd_valid_immediate_hs): New function.
7616 (aarch64_advsimd_valid_immediate): Likewise.
7617 (aarch64_simd_valid_immediate): Remove mode and inverse
7618 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7619 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7620 and aarch64_float_const_representable_p on the result.
7621 (aarch64_output_simd_mov_immediate): Remove mode argument.
7622 Update call to aarch64_simd_valid_immediate and use of
7623 simd_immediate_info.
7624 (aarch64_output_scalar_simd_mov_immediate): Update call
7627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7628 Alan Hayward <alan.hayward@arm.com>
7629 David Sherwood <david.sherwood@arm.com>
7631 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7632 (mode_nunits): Likewise CONST_MODE_NUNITS.
7633 * machmode.def (ADJUST_NUNITS): Document.
7634 * genmodes.c (mode_data::need_nunits_adj): New field.
7635 (blank_mode): Update accordingly.
7636 (adj_nunits): New variable.
7637 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7639 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7640 listed in adj_nunits.
7641 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7642 listed in adj_nunits. Don't emit case statements for such modes.
7643 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7644 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7645 nothing if adj_nunits is nonnull.
7646 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7647 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7648 (emit_mode_fbit): Update use of print_maybe_const_decl.
7649 (emit_move_size): Likewise. Treat the array as non-const
7651 (emit_mode_adjustments): Handle adj_nunits.
7653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7655 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7656 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7657 (VECTOR_MODES): Use it.
7658 (make_vector_modes): Take the prefix as an argument.
7660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7661 Alan Hayward <alan.hayward@arm.com>
7662 David Sherwood <david.sherwood@arm.com>
7664 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7665 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7666 for MODE_VECTOR_BOOL.
7667 * machmode.def (VECTOR_BOOL_MODE): Document.
7668 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7669 (make_vector_bool_mode): New function.
7670 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7672 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7673 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7675 * stor-layout.c (int_mode_for_mode): Likewise.
7676 * tree.c (build_vector_type_for_mode): Likewise.
7677 * varasm.c (output_constant_pool_2): Likewise.
7678 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7679 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7680 for MODE_VECTOR_BOOL.
7681 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7682 of mode class checks.
7683 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7684 instead of a list of mode class checks.
7685 (expand_vector_scalar_condition): Likewise.
7686 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7689 Alan Hayward <alan.hayward@arm.com>
7690 David Sherwood <david.sherwood@arm.com>
7692 * machmode.h (mode_size): Change from unsigned short to
7694 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7695 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7696 or if measurement_type is not polynomial.
7697 (fixed_size_mode::includes_p): Check for constant-sized modes.
7698 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7699 return a poly_uint16 rather than an unsigned short.
7700 (emit_mode_size): Change the type of mode_size from unsigned short
7701 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7702 (emit_mode_adjustments): Cope with polynomial vector sizes.
7703 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7705 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7707 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7708 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7709 * caller-save.c (setup_save_areas): Likewise.
7710 (replace_reg_with_saved_mem): Likewise.
7711 * calls.c (emit_library_call_value_1): Likewise.
7712 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7713 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7714 (gen_lowpart_for_combine): Likewise.
7715 * convert.c (convert_to_integer_1): Likewise.
7716 * cse.c (equiv_constant, cse_insn): Likewise.
7717 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7718 (cselib_subst_to_values): Likewise.
7719 * dce.c (word_dce_process_block): Likewise.
7720 * df-problems.c (df_word_lr_mark_ref): Likewise.
7721 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7722 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7723 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7724 (rtl_for_decl_location): Likewise.
7725 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7726 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7727 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7728 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7729 (expand_expr_real_1): Likewise.
7730 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7731 (pad_below): Likewise.
7732 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7733 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7734 * ira.c (get_subreg_tracking_sizes): Likewise.
7735 * ira-build.c (ira_create_allocno_objects): Likewise.
7736 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7737 (ira_sort_regnos_for_alter_reg): Likewise.
7738 * ira-costs.c (record_operand_costs): Likewise.
7739 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7740 (resolve_simple_move): Likewise.
7741 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7742 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7743 (lra_constraints): Likewise.
7744 (CONST_POOL_OK_P): Reject variable-sized modes.
7745 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7746 (add_pseudo_to_slot, lra_spill): Likewise.
7747 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7748 * optabs-query.c (get_best_extraction_insn): Likewise.
7749 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7750 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7751 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7752 * recog.c (offsettable_address_addr_space_p): Likewise.
7753 * regcprop.c (maybe_mode_change): Likewise.
7754 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7755 * regrename.c (build_def_use): Likewise.
7756 * regstat.c (dump_reg_info): Likewise.
7757 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7758 (find_reloads, find_reloads_subreg_address): Likewise.
7759 * reload1.c (eliminate_regs_1): Likewise.
7760 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7761 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7762 (simplify_binary_operation_1, simplify_subreg): Likewise.
7763 * targhooks.c (default_function_arg_padding): Likewise.
7764 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7765 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7766 (verify_gimple_assign_ternary): Likewise.
7767 * tree-inline.c (estimate_move_cost): Likewise.
7768 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7769 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7770 (get_address_cost_ainc): Likewise.
7771 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7772 (vect_supportable_dr_alignment): Likewise.
7773 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7774 (vectorizable_reduction): Likewise.
7775 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7776 (vectorizable_operation, vectorizable_load): Likewise.
7777 * tree.c (build_same_sized_truth_vector_type): Likewise.
7778 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7779 * var-tracking.c (emit_note_insn_var_location): Likewise.
7780 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7781 (ADDR_VEC_ALIGN): Likewise.
7783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7784 Alan Hayward <alan.hayward@arm.com>
7785 David Sherwood <david.sherwood@arm.com>
7787 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7789 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7790 or if measurement_type is polynomial.
7791 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7792 * combine.c (make_extraction): Likewise.
7793 * dse.c (find_shift_sequence): Likewise.
7794 * dwarf2out.c (mem_loc_descriptor): Likewise.
7795 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7796 (extract_bit_field, extract_low_bits): Likewise.
7797 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7798 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7799 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7800 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7801 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7802 * reload.c (find_reloads): Likewise.
7803 * reload1.c (alter_reg): Likewise.
7804 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7805 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7806 * tree-if-conv.c (predicate_mem_writes): Likewise.
7807 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7808 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7809 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7810 * valtrack.c (dead_debug_insert_temp): Likewise.
7811 * varasm.c (mergeable_constant_section): Likewise.
7812 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7815 Alan Hayward <alan.hayward@arm.com>
7816 David Sherwood <david.sherwood@arm.com>
7818 * expr.c (expand_assignment): Cope with polynomial mode sizes
7819 when assigning to a CONCAT.
7821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7822 Alan Hayward <alan.hayward@arm.com>
7823 David Sherwood <david.sherwood@arm.com>
7825 * machmode.h (mode_precision): Change from unsigned short to
7827 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7829 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7830 or if measurement_type is not polynomial.
7831 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7832 in which the mode is already known to be a scalar_int_mode.
7833 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7834 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7836 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7837 for GET_MODE_PRECISION.
7838 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7839 for GET_MODE_PRECISION.
7840 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7842 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7843 (expand_field_assignment, make_extraction): Likewise.
7844 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7845 (get_last_value): Likewise.
7846 * convert.c (convert_to_integer_1): Likewise.
7847 * cse.c (cse_insn): Likewise.
7848 * expr.c (expand_expr_real_1): Likewise.
7849 * lra-constraints.c (simplify_operand_subreg): Likewise.
7850 * optabs-query.c (can_atomic_load_p): Likewise.
7851 * optabs.c (expand_atomic_load): Likewise.
7852 (expand_atomic_store): Likewise.
7853 * ree.c (combine_reaching_defs): Likewise.
7854 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7855 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7856 * tree.h (type_has_mode_precision_p): Likewise.
7857 * ubsan.c (instrument_si_overflow): Likewise.
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 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7864 polynomial numbers of units.
7865 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7866 (valid_vector_subparts_p): New function.
7867 (build_vector_type): Remove temporary shim and take the number
7868 of units as a poly_uint64 rather than an int.
7869 (build_opaque_vector_type): Take the number of units as a
7870 poly_uint64 rather than an int.
7871 * tree.c (build_vector_from_ctor): Handle polynomial
7872 TYPE_VECTOR_SUBPARTS.
7873 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7874 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7875 (build_vector_from_val): If the number of units is variable,
7876 use build_vec_duplicate_cst for constant operands and
7877 VEC_DUPLICATE_EXPR otherwise.
7878 (make_vector_type): Remove temporary is_constant ().
7879 (build_vector_type, build_opaque_vector_type): Take the number of
7880 units as a poly_uint64 rather than an int.
7881 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7883 * cfgexpand.c (expand_debug_expr): Likewise.
7884 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7885 (store_constructor, expand_expr_real_1): Likewise.
7886 (const_scalar_mask_from_tree): Likewise.
7887 * fold-const-call.c (fold_const_reduction): Likewise.
7888 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7889 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7890 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7891 (fold_relational_const): Likewise.
7892 (native_interpret_vector): Likewise. Change the size from an
7893 int to an unsigned int.
7894 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7895 TYPE_VECTOR_SUBPARTS.
7896 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7897 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7898 duplicating a non-constant operand into a variable-length vector.
7899 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7900 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7901 * ipa-icf.c (sem_variable::equals): Likewise.
7902 * match.pd: Likewise.
7903 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7904 * print-tree.c (print_node): Likewise.
7905 * stor-layout.c (layout_type): Likewise.
7906 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7907 * tree-cfg.c (verify_gimple_comparison): Likewise.
7908 (verify_gimple_assign_binary): Likewise.
7909 (verify_gimple_assign_ternary): Likewise.
7910 (verify_gimple_assign_single): Likewise.
7911 * tree-pretty-print.c (dump_generic_node): Likewise.
7912 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7913 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7914 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7915 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7916 (vect_shift_permute_load_chain): Likewise.
7917 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7918 (expand_vector_condition, optimize_vector_constructor): Likewise.
7919 (lower_vec_perm, get_compute_type): Likewise.
7920 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7921 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7922 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7923 (vect_recog_mask_conversion_pattern): Likewise.
7924 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7925 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7926 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7927 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7928 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7929 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7930 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7931 (supportable_widening_operation): Likewise.
7932 (supportable_narrowing_operation): Likewise.
7933 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7935 * varasm.c (output_constant): Likewise.
7937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7938 Alan Hayward <alan.hayward@arm.com>
7939 David Sherwood <david.sherwood@arm.com>
7941 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7942 so that both the length == 3 and length != 3 cases set up their
7943 own permute vectors. Add comments explaining why we know the
7944 number of elements is constant.
7945 (vect_permute_load_chain): Likewise.
7947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7948 Alan Hayward <alan.hayward@arm.com>
7949 David Sherwood <david.sherwood@arm.com>
7951 * machmode.h (mode_nunits): Change from unsigned char to
7953 (ONLY_FIXED_SIZE_MODES): New macro.
7954 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7955 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7956 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7958 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7959 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7960 or if measurement_type is not polynomial.
7961 * genmodes.c (ZERO_COEFFS): New macro.
7962 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7964 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7965 Use ZERO_COEFFS when emitting initializers.
7966 * data-streamer.h (bp_pack_poly_value): New function.
7967 (bp_unpack_poly_value): Likewise.
7968 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7969 for GET_MODE_NUNITS.
7970 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7971 for GET_MODE_NUNITS.
7972 * tree.c (make_vector_type): Remove temporary shim and make
7973 the real function take the number of units as a poly_uint64
7975 (build_vector_type_for_mode): Handle polynomial nunits.
7976 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7977 * emit-rtl.c (const_vec_series_p_1): Likewise.
7978 (gen_rtx_CONST_VECTOR): Likewise.
7979 * fold-const.c (test_vec_duplicate_folding): Likewise.
7980 * genrecog.c (validate_pattern): Likewise.
7981 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7982 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7983 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7984 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7985 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7986 * rtlanal.c (subreg_get_info): Likewise.
7987 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7988 (vect_grouped_load_supported): Likewise.
7989 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7990 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7991 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7992 (simplify_const_unary_operation, simplify_binary_operation_1)
7993 (simplify_const_binary_operation, simplify_ternary_operation)
7994 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7995 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7996 instead of CONST_VECTOR_NUNITS.
7997 * varasm.c (output_constant_pool_2): Likewise.
7998 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7999 explicit-encoded elements in the XVEC for variable-length vectors.
8001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8003 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8006 Alan Hayward <alan.hayward@arm.com>
8007 David Sherwood <david.sherwood@arm.com>
8009 * coretypes.h (fixed_size_mode): Declare.
8010 (fixed_size_mode_pod): New typedef.
8011 * builtins.h (target_builtins::x_apply_args_mode)
8012 (target_builtins::x_apply_result_mode): Change type to
8013 fixed_size_mode_pod.
8014 * builtins.c (apply_args_size, apply_result_size, result_vector)
8015 (expand_builtin_apply_args_1, expand_builtin_apply)
8016 (expand_builtin_return): Update accordingly.
8018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8020 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8021 * cselib.c (cselib_hash_rtx): Likewise.
8022 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8023 CONST_VECTOR encoding.
8025 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8026 Jeff Law <law@redhat.com>
8029 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8030 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8031 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8032 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8035 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8036 explicitly probe *sp in a noreturn function if there were any callee
8037 register saves or frame pointer is needed.
8039 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8042 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8043 BLKmode for ternary, binary or unary expressions.
8046 * var-tracking.c (delete_vta_debug_insn): New inline function.
8047 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8048 insns from get_insns () to NULL instead of each bb separately.
8049 Use delete_vta_debug_insn. No longer static.
8050 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8051 delete_vta_debug_insns callers.
8052 * rtl.h (delete_vta_debug_insns): Declare.
8053 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8054 instead of variable_tracking_main.
8056 2018-01-03 Martin Sebor <msebor@redhat.com>
8058 PR tree-optimization/83603
8059 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8060 arguments past the endof the argument list in functions declared
8061 without a prototype.
8062 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8063 Avoid checking when arguments are null.
8065 2018-01-03 Martin Sebor <msebor@redhat.com>
8068 * doc/extend.texi (attribute const): Fix a typo.
8069 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8070 issuing -Wsuggest-attribute for void functions.
8072 2018-01-03 Martin Sebor <msebor@redhat.com>
8074 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8075 offset_int::from instead of wide_int::to_shwi.
8076 (maybe_diag_overlap): Remove assertion.
8077 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8078 * gimple-ssa-sprintf.c (format_directive): Same.
8079 (parse_directive): Same.
8080 (sprintf_dom_walker::compute_format_length): Same.
8081 (try_substitute_return_value): Same.
8083 2018-01-03 Jeff Law <law@redhat.com>
8086 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8087 non-constant residual for zero at runtime and avoid probing in
8088 that case. Reorganize code for trailing problem to mirror handling
8091 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8093 PR tree-optimization/83501
8094 * tree-ssa-strlen.c (get_string_cst): New.
8095 (handle_char_store): Call get_string_cst.
8097 2018-01-03 Martin Liska <mliska@suse.cz>
8099 PR tree-optimization/83593
8100 * tree-ssa-strlen.c: Include tree-cfg.h.
8101 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8102 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8103 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8105 (strlen_dom_walker::before_dom_children): Call
8106 gimple_purge_dead_eh_edges. Dump tranformation with details
8108 (strlen_dom_walker::before_dom_children): Update call by adding
8109 new argument cleanup_eh.
8110 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8112 2018-01-03 Martin Liska <mliska@suse.cz>
8115 * cif-code.def (VARIADIC_THUNK): New enum value.
8116 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8119 2018-01-03 Jan Beulich <jbeulich@suse.com>
8121 * sse.md (mov<mode>_internal): Tighten condition for when to use
8122 vmovdqu<ssescalarsize> for TI and OI modes.
8124 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8126 Update copyright years.
8128 2018-01-03 Martin Liska <mliska@suse.cz>
8131 * ipa-visibility.c (function_and_variable_visibility): Skip
8132 functions with noipa attribure.
8134 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8136 * gcc.c (process_command): Update copyright notice dates.
8137 * gcov-dump.c (print_version): Ditto.
8138 * gcov.c (print_version): Ditto.
8139 * gcov-tool.c (print_version): Ditto.
8140 * gengtype.c (create_file): Ditto.
8141 * doc/cpp.texi: Bump @copying's copyright year.
8142 * doc/cppinternals.texi: Ditto.
8143 * doc/gcc.texi: Ditto.
8144 * doc/gccint.texi: Ditto.
8145 * doc/gcov.texi: Ditto.
8146 * doc/install.texi: Ditto.
8147 * doc/invoke.texi: Ditto.
8149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8151 * vector-builder.h (vector_builder::m_full_nelts): Change from
8152 unsigned int to poly_uint64.
8153 (vector_builder::full_nelts): Update prototype accordingly.
8154 (vector_builder::new_vector): Likewise.
8155 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8156 (vector_builder::operator ==): Likewise.
8157 (vector_builder::finalize): Likewise.
8158 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8159 Take the number of elements as a poly_uint64 rather than an
8161 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8162 from unsigned int to poly_uint64.
8163 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8164 (vec_perm_indices::new_vector): Likewise.
8165 (vec_perm_indices::length): Likewise.
8166 (vec_perm_indices::nelts_per_input): Likewise.
8167 (vec_perm_indices::input_nelts): Likewise.
8168 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8169 number of elements per input as a poly_uint64 rather than an
8170 unsigned int. Use the original encoding for variable-length
8171 vectors, rather than clamping each individual element.
8172 For the second and subsequent elements in each pattern,
8173 clamp the step and base before clamping their sum.
8174 (vec_perm_indices::series_p): Handle polynomial element counts.
8175 (vec_perm_indices::all_in_range_p): Likewise.
8176 (vec_perm_indices_to_tree): Likewise.
8177 (vec_perm_indices_to_rtx): Likewise.
8178 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8179 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8180 (tree_vector_builder::new_binary_operation): Handle polynomial
8181 element counts. Return false if we need to know the number
8182 of elements at compile time.
8183 * fold-const.c (fold_vec_perm): Punt if the number of elements
8184 isn't known at compile time.
8186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8188 * vec-perm-indices.h (vec_perm_builder): Change element type
8189 from HOST_WIDE_INT to poly_int64.
8190 (vec_perm_indices::element_type): Update accordingly.
8191 (vec_perm_indices::clamp): Handle polynomial element_types.
8192 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8193 (vec_perm_indices::all_in_range_p): Likewise.
8194 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8196 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8197 polynomial vec_perm_indices element types.
8198 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8199 * fold-const.c (fold_vec_perm): Likewise.
8200 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8201 * tree-vect-generic.c (lower_vec_perm): Likewise.
8202 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8203 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8204 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8211 rather than an int. Use plus_constant.
8212 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8213 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8216 Alan Hayward <alan.hayward@arm.com>
8217 David Sherwood <david.sherwood@arm.com>
8219 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8220 a HOST_WIDE_INT to a poly_int64.
8222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8223 Alan Hayward <alan.hayward@arm.com>
8224 David Sherwood <david.sherwood@arm.com>
8226 * calls.c (load_register_parameters): Cope with polynomial
8227 mode sizes. Require a constant size for BLKmode parameters
8228 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8229 forces a parameter to be padded at the lsb end in order to
8230 fill a complete number of words, require the parameter size
8231 to be ordered wrt UNITS_PER_WORD.
8233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8234 Alan Hayward <alan.hayward@arm.com>
8235 David Sherwood <david.sherwood@arm.com>
8237 * reload1.c (spill_stack_slot_width): Change element type
8238 from unsigned int to poly_uint64_pod.
8239 (alter_reg): Treat mode sizes as polynomial.
8241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8242 Alan Hayward <alan.hayward@arm.com>
8243 David Sherwood <david.sherwood@arm.com>
8245 * reload.c (complex_word_subreg_p): New function.
8246 (reload_inner_reg_of_subreg, push_reload): Use it.
8248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8249 Alan Hayward <alan.hayward@arm.com>
8250 David Sherwood <david.sherwood@arm.com>
8252 * lra-constraints.c (process_alt_operands): Reject matched
8253 operands whose sizes aren't ordered.
8254 (match_reload): Refer to this check here.
8256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8257 Alan Hayward <alan.hayward@arm.com>
8258 David Sherwood <david.sherwood@arm.com>
8260 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8261 that the mode size is in the set {1, 2, 4, 8, 16}.
8263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8264 Alan Hayward <alan.hayward@arm.com>
8265 David Sherwood <david.sherwood@arm.com>
8267 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8268 Use plus_constant instead of gen_rtx_PLUS.
8270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8271 Alan Hayward <alan.hayward@arm.com>
8272 David Sherwood <david.sherwood@arm.com>
8274 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8275 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8276 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8277 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8278 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8279 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8280 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8281 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8282 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8283 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8285 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8286 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8287 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8288 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8289 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8290 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8291 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8292 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8293 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8294 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8296 * expr.c (emit_move_resolve_push): Treat the input and result
8297 of PUSH_ROUNDING as a poly_int64.
8298 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8299 (emit_push_insn): Likewise.
8300 * lra-eliminations.c (mark_not_eliminable): Likewise.
8301 * recog.c (push_operand): Likewise.
8302 * reload1.c (elimination_effects): Likewise.
8303 * rtlanal.c (nonzero_bits1): Likewise.
8304 * calls.c (store_one_arg): Likewise. Require the padding to be
8305 known at compile time.
8307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8308 Alan Hayward <alan.hayward@arm.com>
8309 David Sherwood <david.sherwood@arm.com>
8311 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8312 Use plus_constant instead of gen_rtx_PLUS.
8314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8315 Alan Hayward <alan.hayward@arm.com>
8316 David Sherwood <david.sherwood@arm.com>
8318 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8322 Alan Hayward <alan.hayward@arm.com>
8323 David Sherwood <david.sherwood@arm.com>
8325 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8326 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8327 via stack temporaries. Treat the mode size as polynomial too.
8329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8330 Alan Hayward <alan.hayward@arm.com>
8331 David Sherwood <david.sherwood@arm.com>
8333 * expr.c (expand_expr_real_2): When handling conversions involving
8334 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8335 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8336 as a poly_uint64 too.
8338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8339 Alan Hayward <alan.hayward@arm.com>
8340 David Sherwood <david.sherwood@arm.com>
8342 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8345 Alan Hayward <alan.hayward@arm.com>
8346 David Sherwood <david.sherwood@arm.com>
8348 * combine.c (can_change_dest_mode): Handle polynomial
8349 REGMODE_NATURAL_SIZE.
8350 * expmed.c (store_bit_field_1): Likewise.
8351 * expr.c (store_constructor): Likewise.
8352 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8353 and polynomial REGMODE_NATURAL_SIZE.
8354 (gen_lowpart_common): Likewise.
8355 * reginfo.c (record_subregs_of_mode): Likewise.
8356 * rtlanal.c (read_modify_subreg_p): Likewise.
8358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8359 Alan Hayward <alan.hayward@arm.com>
8360 David Sherwood <david.sherwood@arm.com>
8362 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8363 numbers of elements.
8365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8366 Alan Hayward <alan.hayward@arm.com>
8367 David Sherwood <david.sherwood@arm.com>
8369 * match.pd: Cope with polynomial numbers of vector elements.
8371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8372 Alan Hayward <alan.hayward@arm.com>
8373 David Sherwood <david.sherwood@arm.com>
8375 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8376 in a POINTER_PLUS_EXPR.
8378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8379 Alan Hayward <alan.hayward@arm.com>
8380 David Sherwood <david.sherwood@arm.com>
8382 * omp-simd-clone.c (simd_clone_subparts): New function.
8383 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8384 (ipa_simd_modify_function_body): Likewise.
8386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8387 Alan Hayward <alan.hayward@arm.com>
8388 David Sherwood <david.sherwood@arm.com>
8390 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8391 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8392 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8393 (expand_vector_condition, vector_element): Likewise.
8394 (subparts_gt): New function.
8395 (get_compute_type): Use subparts_gt.
8396 (count_type_subparts): Delete.
8397 (expand_vector_operations_1): Use subparts_gt instead of
8398 count_type_subparts.
8400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8401 Alan Hayward <alan.hayward@arm.com>
8402 David Sherwood <david.sherwood@arm.com>
8404 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8405 (vect_compile_time_alias): ...this new function. Do the calculation
8406 on poly_ints rather than trees.
8407 (vect_prune_runtime_alias_test_list): Update call accordingly.
8409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8410 Alan Hayward <alan.hayward@arm.com>
8411 David Sherwood <david.sherwood@arm.com>
8413 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8415 (vect_schedule_slp_instance): Likewise.
8417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8418 Alan Hayward <alan.hayward@arm.com>
8419 David Sherwood <david.sherwood@arm.com>
8421 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8422 constant and extern definitions for variable-length vectors.
8423 (vect_get_constant_vectors): Note that the number of units
8424 is known to be constant.
8426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8427 Alan Hayward <alan.hayward@arm.com>
8428 David Sherwood <david.sherwood@arm.com>
8430 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8431 of units as polynomial. Choose between WIDE and NARROW based
8434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8435 Alan Hayward <alan.hayward@arm.com>
8436 David Sherwood <david.sherwood@arm.com>
8438 * tree-vect-stmts.c (simd_clone_subparts): New function.
8439 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8442 Alan Hayward <alan.hayward@arm.com>
8443 David Sherwood <david.sherwood@arm.com>
8445 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8446 vectors as polynomial. Use build_index_vector for
8449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8450 Alan Hayward <alan.hayward@arm.com>
8451 David Sherwood <david.sherwood@arm.com>
8453 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8454 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8455 for variable-length vectors.
8456 (vectorizable_mask_load_store): Treat the number of units as
8457 polynomial, asserting that it is constant if the condition has
8458 already been enforced.
8459 (vectorizable_store, vectorizable_load): Likewise.
8461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8462 Alan Hayward <alan.hayward@arm.com>
8463 David Sherwood <david.sherwood@arm.com>
8465 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8466 of units as polynomial. Punt if we can't tell at compile time
8467 which vector contains the final result.
8469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8470 Alan Hayward <alan.hayward@arm.com>
8471 David Sherwood <david.sherwood@arm.com>
8473 * tree-vect-loop.c (vectorizable_induction): Treat the number
8474 of units as polynomial. Punt on SLP inductions. Use an integer
8475 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8476 cast of such a series for variable-length floating-point
8479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8480 Alan Hayward <alan.hayward@arm.com>
8481 David Sherwood <david.sherwood@arm.com>
8483 * tree.h (build_index_vector): Declare.
8484 * tree.c (build_index_vector): New function.
8485 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8486 of units as polynomial, forcibly converting it to a constant if
8487 vectorizable_reduction has already enforced the condition.
8488 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8489 to create a {1,2,3,...} vector.
8490 (vectorizable_reduction): Treat the number of units as polynomial.
8491 Choose vectype_in based on the largest scalar element size rather
8492 than the smallest number of units. Enforce the restrictions
8495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8496 Alan Hayward <alan.hayward@arm.com>
8497 David Sherwood <david.sherwood@arm.com>
8499 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8500 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8507 * target.def (autovectorize_vector_sizes): Return the vector sizes
8508 by pointer, using vector_sizes rather than a bitmask.
8509 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8510 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8511 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8513 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8514 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8515 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8516 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8517 * omp-general.c (omp_max_vf): Likewise.
8518 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8519 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8520 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8521 * tree-vect-slp.c (vect_slp_bb): Likewise.
8522 * doc/tm.texi: Regenerate.
8523 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8525 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8526 the vector size as a poly_uint64 rather than an unsigned int.
8527 (current_vector_size): Change from an unsigned int to a poly_uint64.
8528 (get_vectype_for_scalar_type): Update accordingly.
8529 * tree.h (build_truth_vector_type): Take the size and number of
8530 units as a poly_uint64 rather than an unsigned int.
8531 (build_vector_type): Add a temporary overload that takes
8532 the number of units as a poly_uint64 rather than an unsigned int.
8533 * tree.c (make_vector_type): Likewise.
8534 (build_truth_vector_type): Take the number of units as a poly_uint64
8535 rather than an unsigned int.
8537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8538 Alan Hayward <alan.hayward@arm.com>
8539 David Sherwood <david.sherwood@arm.com>
8541 * target.def (get_mask_mode): Take the number of units and length
8542 as poly_uint64s rather than unsigned ints.
8543 * targhooks.h (default_get_mask_mode): Update accordingly.
8544 * targhooks.c (default_get_mask_mode): Likewise.
8545 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8546 * doc/tm.texi: Regenerate.
8548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8549 Alan Hayward <alan.hayward@arm.com>
8550 David Sherwood <david.sherwood@arm.com>
8552 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8553 * omp-general.c (omp_max_vf): Likewise.
8554 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8555 (expand_omp_simd): Handle polynomial safelen.
8556 * omp-low.c (omplow_simd_context): Add a default constructor.
8557 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8558 (lower_rec_simd_input_clauses): Update accordingly.
8559 (lower_rec_input_clauses): Likewise.
8561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8562 Alan Hayward <alan.hayward@arm.com>
8563 David Sherwood <david.sherwood@arm.com>
8565 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8566 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8567 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8568 (vect_analyze_slp_cost): Likewise.
8569 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8570 (vect_model_load_cost): Likewise.
8572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8573 Alan Hayward <alan.hayward@arm.com>
8574 David Sherwood <david.sherwood@arm.com>
8576 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8577 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8578 from an unsigned int * to a poly_uint64_pod *.
8579 (calculate_unrolling_factor): New function.
8580 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8583 Alan Hayward <alan.hayward@arm.com>
8584 David Sherwood <david.sherwood@arm.com>
8586 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8587 from an unsigned int to a poly_uint64.
8588 (_loop_vec_info::slp_unrolling_factor): Likewise.
8589 (_loop_vec_info::vectorization_factor): Change from an int
8591 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8592 (vect_get_num_vectors): New function.
8593 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8594 (vect_get_num_copies): Use vect_get_num_vectors.
8595 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8596 to an unsigned int *.
8597 (vect_analyze_data_refs): Change min_vf from an int * to a
8599 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8600 than an unsigned HOST_WIDE_INT.
8601 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8602 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8603 to an unsigned int *.
8604 (vect_analyze_data_ref_dependences): Likewise.
8605 (vect_compute_data_ref_alignment): Handle polynomial vf.
8606 (vect_enhance_data_refs_alignment): Likewise.
8607 (vect_prune_runtime_alias_test_list): Likewise.
8608 (vect_shift_permute_load_chain): Likewise.
8609 (vect_supportable_dr_alignment): Likewise.
8610 (dependence_distance_ge_vf): Take the vectorization factor as a
8611 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8612 (vect_analyze_data_refs): Change min_vf from an int * to a
8614 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8615 vfm1 as a poly_uint64 rather than an int. Make the same change
8616 for the returned bound_scalar.
8617 (vect_gen_vector_loop_niters): Handle polynomial vf.
8618 (vect_do_peeling): Likewise. Update call to
8619 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8620 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8622 * tree-vect-loop.c (vect_determine_vectorization_factor)
8623 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8624 (vect_get_known_peeling_cost): Likewise.
8625 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8626 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8627 (vect_transform_loop): Likewise. Use the lowest possible VF when
8628 updating the upper bounds of the loop.
8629 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8631 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8632 polynomial unroll factors.
8633 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8634 (vect_make_slp_decision): Likewise.
8635 (vect_supported_load_permutation_p): Likewise, and polynomial
8637 (vect_analyze_slp_cost): Handle polynomial vf.
8638 (vect_slp_analyze_node_operations): Likewise.
8639 (vect_slp_analyze_bb_1): Likewise.
8640 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8641 than an unsigned HOST_WIDE_INT.
8642 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8643 (vectorizable_load): Handle polynomial vf.
8644 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8646 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8649 Alan Hayward <alan.hayward@arm.com>
8650 David Sherwood <david.sherwood@arm.com>
8652 * match.pd: Handle bit operations involving three constants
8653 and try to fold one pair.
8655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8657 * tree-vect-loop-manip.c: Include gimple-fold.h.
8658 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8659 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8660 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8661 Add a path that uses a step of VF instead of 1, but disable it
8663 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8664 and niters_no_overflow parameters. Update calls to
8665 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8666 Create a new SSA name if the latter choses to use a ste other
8667 than zero, and return it via niters_vector_mult_vf_var.
8668 * tree-vect-loop.c (vect_transform_loop): Update calls to
8669 vect_do_peeling, vect_gen_vector_loop_niters and
8670 slpeel_make_loop_iterate_ntimes.
8671 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8672 (vect_gen_vector_loop_niters): Update declarations after above changes.
8674 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8676 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8677 128-bit round to integer instructions.
8678 (ceil<mode>2): Likewise.
8679 (btrunc<mode>2): Likewise.
8680 (round<mode>2): Likewise.
8682 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8684 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8685 unaligned VSX load/store on P8/P9.
8686 (expand_block_clear): Allow the use of unaligned VSX
8687 load/store on P8/P9.
8689 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8691 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8693 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8694 swap associated with both a load and a store.
8696 2018-01-02 Andrew Waterman <andrew@sifive.com>
8698 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8699 * config/riscv/riscv.md (clear_cache): Use it.
8701 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8703 * web.c: Remove out-of-date comment.
8705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8707 * expr.c (fixup_args_size_notes): Check that any existing
8708 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8709 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8710 (emit_single_push_insn): ...here.
8712 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8714 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8715 (const_vector_encoded_nelts): New function.
8716 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8717 (const_vector_int_elt, const_vector_elt): Declare.
8718 * emit-rtl.c (const_vector_int_elt_1): New function.
8719 (const_vector_elt): Likewise.
8720 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8721 of CONST_VECTOR_ELT.
8723 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8725 * expr.c: Include rtx-vector-builder.h.
8726 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8727 directly on the tree encoding.
8728 (const_vector_from_tree): Likewise.
8729 * optabs.c: Include rtx-vector-builder.h.
8730 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8731 sequence of "u" values.
8732 * vec-perm-indices.c: Include rtx-vector-builder.h.
8733 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8734 directly on the vec_perm_indices encoding.
8736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8738 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8739 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8740 * rtx-vector-builder.h: New file.
8741 * rtx-vector-builder.c: Likewise.
8742 * rtl.h (rtx_def::u2): Add a const_vector field.
8743 (CONST_VECTOR_NPATTERNS): New macro.
8744 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8745 (CONST_VECTOR_DUPLICATE_P): Likewise.
8746 (CONST_VECTOR_STEPPED_P): Likewise.
8747 (CONST_VECTOR_ENCODED_ELT): Likewise.
8748 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8749 (unwrap_const_vec_duplicate): Likewise.
8750 (const_vec_series_p): Check for a non-duplicated vector encoding.
8751 Say that the function only returns true for integer vectors.
8752 * emit-rtl.c: Include rtx-vector-builder.h.
8753 (gen_const_vec_duplicate_1): Delete.
8754 (gen_const_vector): Call gen_const_vec_duplicate instead of
8755 gen_const_vec_duplicate_1.
8756 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8757 (gen_const_vec_duplicate): Use rtx_vector_builder.
8758 (gen_const_vec_series): Likewise.
8759 (gen_rtx_CONST_VECTOR): Likewise.
8760 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8761 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8762 Build a new vector rather than modifying a CONST_VECTOR in-place.
8763 (handle_special_swappables): Update call accordingly.
8764 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8765 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8766 Build a new vector rather than modifying a CONST_VECTOR in-place.
8767 (handle_special_swappables): Update call accordingly.
8769 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8771 * simplify-rtx.c (simplify_const_binary_operation): Use
8772 CONST_VECTOR_ELT instead of XVECEXP.
8774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8776 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8777 the selector elements to be different from the data elements
8778 if the selector is a VECTOR_CST.
8779 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8780 ssizetype for the selector.
8782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8784 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8785 before testing each element individually.
8786 * tree-vect-generic.c (lower_vec_perm): Likewise.
8788 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8790 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8791 * selftest-run-tests.c (selftest::run_tests): Call it.
8792 * vector-builder.h (vector_builder::operator ==): New function.
8793 (vector_builder::operator !=): Likewise.
8794 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8795 (vec_perm_indices::all_from_input_p): New function.
8796 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8797 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8798 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8799 instead of reading the VECTOR_CST directly. Detect whether both
8800 vector inputs are the same before constructing the vec_perm_indices,
8801 and update the number of inputs argument accordingly. Use the
8802 utility functions added above. Only construct sel2 if we need to.
8804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8806 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8807 the broadcast of the low byte.
8808 (expand_mult_highpart): Use an explicit encoding for the permutes.
8809 * optabs-query.c (can_mult_highpart_p): Likewise.
8810 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8811 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8812 (vectorizable_bswap): Likewise.
8813 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8814 explicit encoding for the power-of-2 permutes.
8815 (vect_permute_store_chain): Likewise.
8816 (vect_grouped_load_supported): Likewise.
8817 (vect_permute_load_chain): Likewise.
8819 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8821 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8822 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8823 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8824 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8825 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8826 (vect_gen_perm_mask_any): Likewise.
8828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8830 * int-vector-builder.h: New file.
8831 * vec-perm-indices.h: Include int-vector-builder.h.
8832 (vec_perm_indices): Redefine as an int_vector_builder.
8833 (auto_vec_perm_indices): Delete.
8834 (vec_perm_builder): Redefine as a stand-alone class.
8835 (vec_perm_indices::vec_perm_indices): New function.
8836 (vec_perm_indices::clamp): Likewise.
8837 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8838 (vec_perm_indices::new_vector): New function.
8839 (vec_perm_indices::new_expanded_vector): Update for new
8840 vec_perm_indices class.
8841 (vec_perm_indices::rotate_inputs): New function.
8842 (vec_perm_indices::all_in_range_p): Operate directly on the
8843 encoded form, without computing elided elements.
8844 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8845 encoding. Update for new vec_perm_indices class.
8846 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8847 the given vec_perm_builder.
8848 (expand_vec_perm_var): Update vec_perm_builder constructor.
8849 (expand_mult_highpart): Use vec_perm_builder instead of
8850 auto_vec_perm_indices.
8851 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8852 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8853 or double series encoding as appropriate.
8854 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8855 vec_perm_indices instead of auto_vec_perm_indices.
8856 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8857 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8858 (vect_permute_store_chain): Likewise.
8859 (vect_grouped_load_supported): Likewise.
8860 (vect_permute_load_chain): Likewise.
8861 (vect_shift_permute_load_chain): Likewise.
8862 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8863 (vect_transform_slp_perm_load): Likewise.
8864 (vect_schedule_slp_instance): Likewise.
8865 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8866 (vectorizable_mask_load_store): Likewise.
8867 (vectorizable_bswap): Likewise.
8868 (vectorizable_store): Likewise.
8869 (vectorizable_load): Likewise.
8870 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8871 vec_perm_indices instead of auto_vec_perm_indices. Use
8872 tree_to_vec_perm_builder to read the vector from a tree.
8873 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8874 vec_perm_builder instead of a vec_perm_indices.
8875 (have_whole_vector_shift): Use vec_perm_builder and
8876 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8877 truncation to calc_vec_perm_mask_for_shift.
8878 (vect_create_epilog_for_reduction): Likewise.
8879 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8880 from auto_vec_perm_indices to vec_perm_indices.
8881 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8882 instead of changing individual elements.
8883 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8884 the vector in d.perm.
8885 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8886 from auto_vec_perm_indices to vec_perm_indices.
8887 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8888 instead of changing individual elements.
8889 (arm_vectorize_vec_perm_const): Use new_vector to install
8890 the vector in d.perm.
8891 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8892 Update vec_perm_builder constructor.
8893 (rs6000_expand_interleave): Likewise.
8894 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8895 (rs6000_expand_interleave): Likewise.
8897 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8899 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8900 to qimode could truncate the indices.
8901 * optabs.c (expand_vec_perm_var): Likewise.
8903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8905 * Makefile.in (OBJS): Add vec-perm-indices.o.
8906 * vec-perm-indices.h: New file.
8907 * vec-perm-indices.c: Likewise.
8908 * target.h (vec_perm_indices): Replace with a forward class
8910 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8911 * optabs.h: Include vec-perm-indices.h.
8912 (expand_vec_perm): Delete.
8913 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8914 (expand_vec_perm_const): Declare.
8915 * target.def (vec_perm_const_ok): Replace with...
8916 (vec_perm_const): ...this new hook.
8917 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8918 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8919 * doc/tm.texi: Regenerate.
8920 * optabs.def (vec_perm_const): Delete.
8921 * doc/md.texi (vec_perm_const): Likewise.
8922 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8923 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8924 expand_vec_perm for constant permutation vectors. Assert that
8925 the mode of variable permutation vectors is the integer equivalent
8926 of the mode that is being permuted.
8927 * optabs-query.h (selector_fits_mode_p): Declare.
8928 * optabs-query.c: Include vec-perm-indices.h.
8929 (selector_fits_mode_p): New function.
8930 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8931 is defined, instead of checking whether the vec_perm_const_optab
8932 exists. Use targetm.vectorize.vec_perm_const instead of
8933 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8934 fit in the vector mode before using a variable permute.
8935 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8936 vec_perm_indices instead of an rtx.
8937 (expand_vec_perm): Replace with...
8938 (expand_vec_perm_const): ...this new function. Take the selector
8939 as a vec_perm_indices rather than an rtx. Also take the mode of
8940 the selector. Update call to shift_amt_for_vec_perm_mask.
8941 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8942 Use vec_perm_indices::new_expanded_vector to expand the original
8943 selector into bytes. Check whether the indices fit in the vector
8944 mode before using a variable permute.
8945 (expand_vec_perm_var): Make global.
8946 (expand_mult_highpart): Use expand_vec_perm_const.
8947 * fold-const.c: Includes vec-perm-indices.h.
8948 * tree-ssa-forwprop.c: Likewise.
8949 * tree-vect-data-refs.c: Likewise.
8950 * tree-vect-generic.c: Likewise.
8951 * tree-vect-loop.c: Likewise.
8952 * tree-vect-slp.c: Likewise.
8953 * tree-vect-stmts.c: Likewise.
8954 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8956 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8957 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8958 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8959 (aarch64_vectorize_vec_perm_const): ...this new function.
8960 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8961 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8962 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8963 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8964 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8965 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8966 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8968 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8969 check for NEON modes.
8970 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8971 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8972 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8973 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8975 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8976 the old VEC_PERM_CONST conditions.
8977 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8978 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8979 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8980 (ia64_vectorize_vec_perm_const_ok): Merge into...
8981 (ia64_vectorize_vec_perm_const): ...this new function.
8982 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8983 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8984 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8985 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8986 * config/mips/mips.c (mips_expand_vec_perm_const)
8987 (mips_vectorize_vec_perm_const_ok): Merge into...
8988 (mips_vectorize_vec_perm_const): ...this new function.
8989 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8990 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8991 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8992 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8993 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8994 (rs6000_expand_vec_perm_const): Delete.
8995 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8997 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8998 (altivec_expand_vec_perm_const_le): Take each operand individually.
8999 Operate on constant selectors rather than rtxes.
9000 (altivec_expand_vec_perm_const): Likewise. Update call to
9001 altivec_expand_vec_perm_const_le.
9002 (rs6000_expand_vec_perm_const): Delete.
9003 (rs6000_vectorize_vec_perm_const_ok): Delete.
9004 (rs6000_vectorize_vec_perm_const): New function.
9005 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9006 an element count and rtx array.
9007 (rs6000_expand_extract_even): Update call accordingly.
9008 (rs6000_expand_interleave): Likewise.
9009 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9010 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9011 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9012 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9013 (rs6000_expand_vec_perm_const): Delete.
9014 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9015 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9016 (altivec_expand_vec_perm_const_le): Take each operand individually.
9017 Operate on constant selectors rather than rtxes.
9018 (altivec_expand_vec_perm_const): Likewise. Update call to
9019 altivec_expand_vec_perm_const_le.
9020 (rs6000_expand_vec_perm_const): Delete.
9021 (rs6000_vectorize_vec_perm_const_ok): Delete.
9022 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9023 reference to the SPE evmerge intructions.
9024 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9025 an element count and rtx array.
9026 (rs6000_expand_extract_even): Update call accordingly.
9027 (rs6000_expand_interleave): Likewise.
9028 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9029 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9031 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9033 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9035 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9036 vector mode and that that mode matches the mode of the data
9038 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9039 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9040 directly using expand_vec_perm_1 when forcing selectors into
9042 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9046 * optabs-query.h (can_vec_perm_p): Delete.
9047 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9048 * optabs-query.c (can_vec_perm_p): Split into...
9049 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9050 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9051 particular selector is valid.
9052 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9053 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9054 (vect_grouped_load_supported): Likewise.
9055 (vect_shift_permute_load_chain): Likewise.
9056 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9057 (vect_transform_slp_perm_load): Likewise.
9058 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9059 (vectorizable_bswap): Likewise.
9060 (vect_gen_perm_mask_checked): Likewise.
9061 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9062 implementations of variable permutation vectors into account
9063 when deciding which selector to use.
9064 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9065 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9066 with a false third argument.
9067 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9068 to test whether the constant selector is valid and can_vec_perm_var_p
9069 to test whether a variable selector is valid.
9071 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9073 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9074 * optabs-query.c (can_vec_perm_p): Likewise.
9075 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9076 instead of vec_perm_indices.
9077 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9078 (vect_gen_perm_mask_checked): Likewise,
9079 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9080 (vect_gen_perm_mask_checked): Likewise,
9082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9084 * optabs-query.h (qimode_for_vec_perm): Declare.
9085 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9086 (qimode_for_vec_perm): ...this new function.
9087 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9089 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9091 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9092 does not have a conditional at the top.
9094 2018-01-02 Richard Biener <rguenther@suse.de>
9096 * ipa-inline.c (big_speedup_p): Fix expression.
9098 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9101 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9104 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9108 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9109 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9110 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9111 cond_taken_branch_cost 3->4.
9113 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9115 PR tree-optimization/83581
9116 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9117 TODO_cleanup_cfg if any changes have been made.
9120 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9121 convert_modes if target mode has the right side, but different mode
9125 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9126 last argument when extracting from CONCAT. If either from_real or
9127 from_imag is NULL, use expansion through memory. If result is not
9128 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9129 the parts directly to inner mode, if even that fails, use expansion
9133 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9134 check for bswap in mode rather than HImode and use that in expand_unop
9137 Copyright (C) 2018 Free Software Foundation, Inc.
9139 Copying and distribution of this file, with or without modification,
9140 are permitted in any medium without royalty provided the copyright
9141 notice and this notice are preserved.