1 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7 2018-03-19 Jim Wilson <jimw@sifive.com>
10 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12 (riscv_first_stack_step): Likewise.
13 (riscv_option_override): Use STACK_BOUNDARY instead of
15 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17 (BIGGEST_ALIGNMENT): Set to 128.
18 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
22 2018-03-19 Richard Biener <rguenther@suse.de>
24 PR tree-optimization/84933
25 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
26 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
28 2018-03-19 Richard Biener <rguenther@suse.de>
30 PR tree-optimization/84859
31 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
32 (cond_if_else_store_replacement): Perform sinking operation on
33 single-store BBs regardless of MAX_STORES_TO_SINK setting.
34 Generalize what a BB with a single eligible store is.
36 2018-03-19 Richard Biener <rguenther@suse.de>
38 PR tree-optimization/84929
39 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
40 chrec_is_positive against non-chrec arg.
42 2018-03-19 Tamar Christina <tamar.christina@arm.com>
45 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
47 2018-03-18 Martin Liska <mliska@suse.cz>
49 PR rtl-optimization/84635
50 * regrename.c (build_def_use): Use matches_mode only when
53 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
55 PR tree-optimization/84913
56 * tree-vect-loop.c (vectorizable_reduction): Don't try to
57 vectorize chains of COND_EXPRs.
59 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
61 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
63 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
65 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
67 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
69 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
71 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
72 Kito Cheng <kito.cheng@gmail.com>
74 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
75 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
76 (nds32_adjust_reg_alloc_order): New function.
77 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
79 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
81 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
82 nds32_print_operand, nds32_print_operand_address): Use
83 HOST_WIDE_INT_PRINT_DEC instead.
85 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
87 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
89 2018-03-17 Jakub Jelinek <jakub@redhat.com>
92 * config/i386/i386.c (initial_ix86_tune_features,
93 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
95 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
96 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
97 rather than 1u << ix86_tune. Formatting fix.
98 (ix86_option_override_internal): Change ix86_arch_mask from
99 unsigned int to unsigned HOST_WIDE_INT, initialize to
100 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
101 (ix86_function_specific_restore): Likewise.
103 2018-03-16 Jakub Jelinek <jakub@redhat.com>
106 * postreload.c (reload_combine_recognize_pattern): Perform
107 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
108 truncate_int_for_mode the result for the destination's mode.
111 * hsa-gen.c (mem_type_for_type): Fix comment typo.
112 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
114 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
117 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
120 * lra-assigns.c (lra_split_hard_reg_for): Don't use
121 regno_allocno_class_array and sorted_pseudos.
122 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
123 insns where regno is used.
125 2018-03-16 Martin Liska <mliska@suse.cz>
128 * multiple_target.c (create_dispatcher_calls): Redirect
129 reference in the symbol table.
131 2018-03-16 Martin Liska <mliska@suse.cz>
134 * multiple_target.c (create_dispatcher_calls): Redirect also
137 2018-03-16 Jakub Jelinek <jakub@redhat.com>
141 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
142 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
143 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
145 2018-03-16 Julia Koval <julia.koval@intel.com>
147 * doc/invoke.texi (Skylake Server): Add CLWB.
148 Cannonlake): Remove CLWB.
150 2018-03-16 Jakub Jelinek <jakub@redhat.com>
152 PR tree-optimization/84841
153 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
155 (FLOAT_ONE_CONST_TYPE): Define.
156 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
157 (sort_by_operand_rank): Put entries with higher constant_type last
158 rather than first to match comments.
160 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
162 * config/nios2/nios2.md (movsi_internal): Fix thinko in
165 2018-03-15 Jakub Jelinek <jakub@redhat.com>
168 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
169 check and use address of target always.
171 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
174 * config/i386/i386.c (indirect_thunk_needed): Update comments.
175 (indirect_thunk_bnd_needed): Likewise.
176 (indirect_thunks_used): Likewise.
177 (indirect_thunks_bnd_used): Likewise.
178 (indirect_return_needed): New.
179 (indirect_return_bnd_needed): Likewise.
180 (output_indirect_thunk_function): Add a bool argument for
182 (output_indirect_thunk_function): Don't generate alias for
183 function return thunk.
184 (ix86_code_end): Call output_indirect_thunk_function to generate
185 function return thunks.
186 (ix86_output_function_return): Set indirect_return_bnd_needed
187 and indirect_return_needed instead of indirect_thunk_bnd_needed
188 and indirect_thunk_needed.
190 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
192 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
193 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
194 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
196 2018-03-15 David Malcolm <dmalcolm@redhat.com>
197 Paul Hua <paul.hua.gm@gmail.com>
200 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
202 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
204 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
205 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
208 2018-03-15 Tamar Christina <tamar.christina@arm.com>
211 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
212 instead of GET_MODE_SIZE when comparing Units.
214 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
217 * varasm.c (hash_section): Return an unchangeble hash value
218 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
219 Return !aarch64_can_use_per_function_literal_pools_p ().
221 2018-03-15 Jakub Jelinek <jakub@redhat.com>
224 * optabs.c (emit_conditional_move): Pass address of cmode's copy
225 rather than address of cmode as last argument to prepare_cmp_insn.
227 2018-03-15 Julia Koval <julia.koval@intel.com>
229 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
230 F_AVX512VNNI, F_AVX512BITALG): New.
232 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
235 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
236 insn for floating-point loads and stores.
238 2018-03-14 Carl Love <cel@us.ibm.com>
240 * config/rs6000/rs6000-c.c: Add macro definitions for
241 ALTIVEC_BUILTIN_VEC_PERMXOR.
242 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
243 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
244 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
245 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
247 * config/doc/extend.texi: Add prototypes for vec_permxor.
249 2018-03-14 David Malcolm <dmalcolm@redhat.com>
252 * diagnostic-show-locus.c (class layout_point): Convert m_line
253 from int to linenum_type.
254 (line_span::comparator): Use linenum "compare" function when
255 comparing line numbers.
256 (test_line_span): New function.
257 (layout_range::contains_point): Convert param "row" from int to
259 (layout_range::intersects_line_p): Likewise.
260 (layout::will_show_line_p): Likewise.
261 (layout::print_source_line): Likewise.
262 (layout::should_print_annotation_line_p): Likewise.
263 (layout::print_annotation_line): Likewise.
264 (layout::print_leading_fixits): Likewise.
265 (layout::annotation_line_showed_range_p): Likewise.
266 (struct line_corrections): Likewise for field m_row.
267 (line_corrections::line_corrections): Likewise for param "row".
268 (layout::print_trailing_fixits): Likewise.
269 (layout::get_state_at_point): Likewise.
270 (layout::get_x_bound_for_row): Likewise.
271 (layout::print_line): Likewise.
272 (diagnostic_show_locus): Likewise for locals "last_line" and
274 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
275 * input.c (selftest::test_linenum_comparisons): New function.
276 (selftest::input_c_tests): Call it.
277 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
278 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
279 * selftest.h (ASSERT_GT): New macro.
280 (ASSERT_GT_AT): New macro.
281 (ASSERT_LT): New macro.
282 (ASSERT_LT_AT): New macro.
284 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
286 PR rtl-optimization/84780
287 * combine.c (distribute_links): Don't make a link based on pc_rtx.
289 2018-03-14 Martin Liska <mliska@suse.cz>
291 * tree.c (record_node_allocation_statistics): Use
293 (get_stats_node_kind): New function extracted from
294 record_node_allocation_statistics.
295 (free_node): Use get_stats_node_kind.
297 2018-03-14 Richard Biener <rguenther@suse.de>
299 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
300 that the value-set of ANTIC_IN doesn't grow.
303 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
305 (BB_VISITED_WITH_VISITED_SUCCS): New define.
306 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
308 2018-03-14 Julia Koval <julia.koval@intel.com>
310 * config.gcc (icelake-client, icelake-server): New.
312 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
313 (initial_ix86_arch_features): Ditto.
314 (PTA_SKYLAKE): Add SGX.
315 (PTA_ICELAKE): Remove.
316 (PTA_ICELAKE_CLIENT): New.
317 (PTA_ICELAKE_SERVER): New.
318 (ix86_option_override_internal): Split up icelake on icelake client and
320 (get_builtin_code_for_version): Ditto.
321 (fold_builtin_cpu): Ditto.
322 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
323 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
324 * config/i386/i386.h (processor_type): Ditto.
325 * doc/invoke.texi: Ditto.
327 2018-03-14 Jakub Jelinek <jakub@redhat.com>
330 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
331 INTEGER_CST offset, add it together with bitpos / 8 and
332 sign extend based on POINTER_SIZE.
336 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
339 * config/i386/constraints.md (Yc): New register constraint.
340 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
341 Use Yc constraint for alternative 2 of operand 0. Remove
342 preferred_for_speed attribute.
344 2018-03-14 Richard Biener <rguenther@suse.de>
346 PR tree-optimization/84830
347 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
348 with the old one to avoid oscillations.
350 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
353 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
355 (assign_by_spills): Return a flag of reload assignment failure.
356 Do not process the reload assignment failures. Do not spill other
357 reload pseudos if they has the same reg class. Update n if
359 (lra_assign): Add a return arg. Set up from the result of
360 assign_by_spills call.
361 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
362 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
363 usage_insns if it is not NULL.
364 (spill_hard_reg_in_range): New function.
365 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
366 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
368 (lra_assign): Change prototype.
369 * lra.c (lra): Add code to deal with fails by splitting hard reg
372 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
374 * config/riscv/riscv.opt (mrelax): New option.
375 * config/riscv/riscv.c (riscv_file_start): Emit ".option
376 "norelax" when riscv_mrelax is disabled.
377 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
379 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
382 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
383 reassociation for int modes.
385 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
387 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
388 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
390 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
391 * config/aarch64/aarch64-sve.md
392 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
393 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
394 (*extend<mode><Vwide>2): Rename to...
395 (aarch64_sve_extend<mode><Vwide>2): ...this.
396 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
397 renaming the old pattern to...
398 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
400 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
401 define_expand, renaming the old pattern to...
402 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
403 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
404 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
405 account when deciding which SVE instruction the optab should use.
406 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
408 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
410 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
411 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
412 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
413 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
414 (tlsdesc_small_<mode>): Turn a define_expand and use
415 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
416 (tlsdesc_small_advsimd_<mode>): ...this.
417 (tlsdesc_small_sve_<mode>): New pattern.
419 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
421 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
422 (UNSPEC_UMUL_HIGHPART): New constants.
423 (MUL_HIGHPART): New int iteraor.
424 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
425 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
427 (*<su>mul<mode>3_highpart): New define_insn.
429 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
432 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
435 2018-03-13 Martin Liska <mliska@suse.cz>
438 * (sem_item_optimizer::sem_item_optimizer): Initialize new
440 (sem_item_optimizer::~sem_item_optimizer): Release it.
441 (sem_item_optimizer::merge_classes): Register variable aliases.
442 (sem_item_optimizer::fixup_pt_set): New function.
443 (sem_item_optimizer::fixup_points_to_sets): Likewise.
444 * ipa-icf.h: Declare new variables and functions.
446 2018-03-13 Jakub Jelinek <jakub@redhat.com>
449 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
450 integer_pow2p@2 and test integer_pow2p in condition.
451 (A < 0 ? C : 0): Similarly for @1.
454 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
455 characters starting at p contain '\0' character, don't look beyond
459 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
460 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
463 * reg-stack.c (change_stack): Change update_end var from int to
464 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
465 also call set_block_for_insn on the newly added insns and rescan.
468 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
472 * tree.c (stabilize_reference_1): Return save_expr (e) for
473 STATEMENT_LIST even if it doesn't have side-effects.
475 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
477 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
479 2018-03-12 Renlin Li <renlin.li@arm.com>
481 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
482 aarch64_output_scalar_simd_mov_immediate.
484 2018-03-12 Martin Sebor <msebor@redhat.com>
486 PR tree-optimization/83456
487 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
488 for perfectly overlapping calls to memcpy.
489 (gimple_fold_builtin_memory_chk): Same.
490 (gimple_fold_builtin_strcpy): Handle no-warning.
491 (gimple_fold_builtin_stxcpy_chk): Same.
492 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
494 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
496 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
497 parameter. Use it for SFmode.
498 (rs6000_function_arg_advance_1): Adjust.
499 (rs6000_function_arg): Adjust.
500 (rs6000_gimplify_va_arg): Pass false for that new parameter.
502 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
504 PR rtl-optimization/84169
505 PR rtl-optimization/84780
506 * combine.c (can_combine_p): Check for a 2-insn combination whether
507 the destination register is used between the two insns, too.
509 2018-03-12 Richard Biener <rguenther@suse.de>
511 PR tree-optimization/84803
512 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
513 for refs DR analysis didn't process.
515 2018-03-12 Richard Biener <rguenther@suse.de>
517 PR tree-optimization/84777
518 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
519 force-vectorize loops ignore whether we are optimizing for size.
521 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
523 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
524 (TARGET_MD_ASM_ADJUST): Define.
526 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
527 Kito Cheng <kito.cheng@gmail.com>
528 Chung-Ju Wu <jasonwucj@gmail.com>
530 * config/nds32/nds32.c (nds32_compute_stack_frame,
531 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
532 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
533 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
534 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
535 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
536 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
537 * config/nds32/nds32.md (prologue, epilogue): Use macro
538 NDS32_V3PUSH_AVAILABLE_P to do checking.
540 2018-03-11 Jakub Jelinek <jakub@redhat.com>
543 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
544 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
545 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
546 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
547 addition of most attributes on !orig_type_die or the attribute not
548 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
550 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
551 Chung-Ju Wu <jasonwucj@gmail.com>
553 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
555 * config/nds32/nds32.opt (mvh): New option.
557 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
558 Chung-Ju Wu <jasonwucj@gmail.com>
560 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
562 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
563 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
566 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
567 Chung-Ju Wu <jasonwucj@gmail.com>
569 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
571 * config/nds32/nds32-multiple.md (strlensi): New pattern.
572 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
574 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
575 Kito Cheng <kito.cheng@gmail.com>
576 Chung-Ju Wu <jasonwucj@gmail.com>
578 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
579 UNSPEC_FFMISM and UNSPEC_FLMISM.
580 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
581 for ffb, ffmism and flmism.
582 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
583 (unspec_ffmism): Ditto.
584 (unspec_flmism): Ditto.
585 (nds32_expand_builtin_impl): Check if string extension is available.
586 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
587 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
589 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
592 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
595 * lra-assigns.c (assign_by_spills): Return a flag of reload
596 assignment failure. Do not process the reload assignment
597 failures. Do not spill other reload pseudos if they has the same
599 (lra_assign): Add a return arg. Set up from the result of
600 assign_by_spills call.
601 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
602 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
603 usage_insns if it is not NULL.
604 (spill_hard_reg_in_range): New function.
605 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
606 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
608 (lra_assign): Change prototype.
609 * lra.c (lra): Add code to deal with fails by splitting hard reg
612 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
615 * config/i386/i386.opt: Replace Enforcment with Enforcement.
617 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
620 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
621 (dw_val_node): Add val_symbolic_view.
622 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
623 (symview_upper_bound): New.
624 (new_line_info_table): Initialize symviews_since_reset.
625 (dwarf2out_source_line): Count symviews_since_reset and set
627 (dw_val_equal_p): Handle symview.
628 (add_AT_symview): New.
629 (print_dw_val): Handle symview.
630 (attr_checksum, attr_checksum_ordered): Likewise.
631 (same_dw_val_p, size_of_die): Likewise.
632 (value_format, output_die): Likewise.
633 (add_high_low_attributes): Use add_AT_symview for entry_view.
634 (dwarf2out_finish): Reset symview_upper_bound, clear
637 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
640 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
641 Add strict argument and use it.
642 (rs6000_split_multireg_move): Update for new strict argument.
643 (mem_operand_gpr): Disallow all non-offsettable addresses.
644 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
646 2018-03-09 Jakub Jelinek <jakub@redhat.com>
649 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
650 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
651 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
654 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
655 decl, use remap_type if we want to use the type.
657 2018-03-09 Martin Sebor <msebor@redhat.com>
659 PR tree-optimization/84526
660 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
662 (builtin_access::generic_overlap): Be prepared to handle non-array
665 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
667 PR rtl-optimization/84682
668 * lra-constraints.c (process_address_1): Check is_address flag
669 for address constraints.
670 (process_alt_operands): Likewise.
671 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
672 preprocess_constraints.
673 * recog.h (preprocess_constraints): Add oploc parameter.
675 * recog.c (preprocess_constraints): Test address_operand for
676 CT_ADDRESS constraints.
678 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
681 * lra-assigns.c (assign_by_spills): Return a flag of reload
682 assignment failure. Do not process the reload assignment
683 failures. Do not spill other reload pseudos if they has the same
685 (lra_assign): Add a return arg. Set up from the result of
686 assign_by_spills call.
687 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
688 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
689 usage_insns if it is not NULL.
690 (spill_hard_reg_in_range): New function.
691 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
692 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
694 (lra_assign): Change prototype.
695 * lra.c (lra): Add code to deal with fails by splitting hard reg
698 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
701 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
702 Accept complain bool parameter. Only emit errors if it is true.
703 (arm_parse_cpu_option_name): Likewise.
704 (arm_target_thumb_only): Adjust callers of the above.
705 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
706 prototype to take a default true bool parameter.
707 (arm_parse_arch_option_name): Likewise.
709 2018-03-09 David Malcolm <dmalcolm@redhat.com>
710 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
714 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
715 * configure: Regenerate.
716 * configure.ac ("linker --version-script option"): New.
717 ("linker soname option"): New.
719 2018-03-09 Richard Biener <rguenther@suse.de>
721 PR tree-optimization/84775
722 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
723 immediate uses of predicate stmts and mark them modified.
726 PR tree-optimization/84178
727 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
729 (version_loop_for_if_conversion): Delay update_ssa call.
730 (tree_if_conversion): Delay update_ssa until after predicate
733 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
736 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
737 when the function accesses prior frames.
739 2018-03-08 Jakub Jelinek <jakub@redhat.com>
742 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
743 gen_llsym, otherwise call maybe_gen_llsym.
746 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
747 has ',' character inside of it.
749 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
752 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
753 as clobbering CC_REGNUM.
755 2018-03-08 Richard Biener <rguenther@suse.de>
758 * tree-scalar-evolution.c: Include tree-into-ssa.h.
759 (follow_copies_to_constant): Do not follow SSA names registered
762 2018-03-08 Richard Biener <rguenther@suse.de>
764 PR tree-optimization/84178
765 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
767 (version_loop_for_if_conversion): Delay update_ssa call.
768 (tree_if_conversion): Delay update_ssa until after predicate
771 2018-03-08 David Malcolm <dmalcolm@redhat.com>
773 PR tree-optimization/84178
774 * tree-if-conv.c (release_bb_predicate): Remove the
775 the assertion that the stmts have NULL use_ops.
776 Discard the statements, asserting that they haven't
777 yet been added to a BB.
779 2018-03-08 Richard Biener <rguenther@suse.de>
781 PR tree-optimization/84746
782 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
783 (phi_translate): Pass in destination ANTIC_OUT set.
784 (phi_translate_1): Likewise. For a simplified result lookup
785 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
786 (phi_translate_set): Adjust.
787 (do_pre_regular_insertion): Likewise.
788 (do_pre_partial_partial_insertion): Likewise.
790 2018-03-08 Martin Liska <mliska@suse.cz>
792 PR gcov-profile/84735
793 * doc/gcov.texi: Document usage of profile files.
794 * gcov-io.h: Document changes in the format.
796 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
800 * dwarf2out.c (struct dw_line_info_table): Update comments for
802 (FORCE_RESET_NEXT_VIEW): New.
803 (FORCE_RESETTING_VIEW_P): New.
804 (RESETTING_VIEW_P): Check for -1 too.
805 (ZERO_VIEW_P): Likewise.
806 (new_line_info_table): Force-reset next view.
807 (dwarf2out_begin_function): Likewise.
808 (dwarf2out_source_line): Simplify zero_view_p initialization.
809 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
810 view directly. Omit view when omitting .loc at line 0.
812 2018-03-08 Jakub Jelinek <jakub@redhat.com>
814 PR tree-optimization/84740
815 * tree-switch-conversion.c (process_switch): Call build_constructors
816 only if info.phi_count is non-zero.
818 PR tree-optimization/84739
819 * tree-tailcall.c (find_tail_calls): Check call arguments against
820 DECL_ARGUMENTS (current_function_decl) rather than
821 DECL_ARGUMENTS (func) when checking for tail recursion.
823 2018-03-07 Jakub Jelinek <jakub@redhat.com>
825 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
826 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
827 Volker Reichelt's entry and add entries for people that perform
828 GCC fuzzy testing and report numerous bugs.
830 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
833 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
834 readonly data in sdata, if that is disabled.
835 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
836 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
837 -mreadonly-in-sdata option.
839 2018-03-07 Martin Sebor <msebor@redhat.com>
841 PR tree-optimization/84468
842 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
843 basic block when looking for nul assignment.
845 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
848 * except.h (output_function_exception_table): Adjust prototype.
849 * except.c (output_function_exception_table): Remove FNNAME parameter
850 and add SECTION parameter. Ouput one part of the table at a time.
851 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
852 the first part of the exception table and emit unwind directives.
853 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
854 (i386_pe_seh_cold_init): Likewise.
855 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
856 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
857 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
858 (ix86_output_call_insn): Emit a nop in one more case for SEH.
859 * config/i386/winnt.c: Include except.h.
860 (struct seh_frame_state): Add reg_offset, after_prologue and
861 in_cold_section fields.
862 (i386_pe_seh_end_prologue): Set seh->after_prologue.
863 (i386_pe_seh_cold_init): New function.
864 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
865 to seh->in_cold_section.
866 (seh_emit_push): Record the offset of the push.
867 (seh_emit_save): Record the offet of the save.
868 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
869 Test seh->after_prologue to disregard the epilogue.
870 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
871 (i386_pe_end_cold_function): New function.
873 2018-03-07 Jakub Jelinek <jakub@redhat.com>
876 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
877 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
880 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
882 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
883 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
886 * multiple_target.c: Include tree-inline.h and intl.h.
887 (expand_target_clones): Diagnose and fail if node->definition and
888 !tree_versionable_function_p (node->decl).
890 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
892 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
894 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
895 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
896 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
898 2018-03-06 Jakub Jelinek <jakub@redhat.com>
901 * combine.c (try_combine): Use reg_or_subregno instead of handling
902 just paradoxical SUBREGs and REGs.
904 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
906 * config/arc/arc.c (arc_finalize_pic): Remove function.
907 (arc_must_save_register): We use single base PIC register, remove
908 checks to save/restore the PIC register.
909 (arc_expand_prologue): Likewise.
910 * config/arc/arc-protos.h (arc_set_default_type_attributes):
912 (arc_verify_short): Likewise.
913 (arc_attr_type): Likewise.
914 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
915 (walk_stores): Likewise.
916 (arc_address_cost): Make it static.
917 (arc_verify_short): Likewise.
918 (branch_dest): Likewise.
919 (arc_attr_type): Likewise.
920 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
921 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
922 (arc_final_prescan_insn): Remove inserting the nops due to
923 hardware hazards. It is done in reorg step.
924 (insn_length_variant_t): Remove.
925 (insn_length_parameters_t): Likewise.
926 (arc_insn_length_parameters): Likewise.
927 (arc_get_insn_variants): Likewise.
928 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
930 2018-03-06 Jakub Jelinek <jakub@redhat.com>
933 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
936 PR tree-optimization/84687
937 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
939 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
941 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
943 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
944 Rename to ppc_speculation_barrier.
945 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
946 __builtin_ppc_speculation_barrier.
948 2018-03-05 Jakub Jelinek <jakub@redhat.com>
951 * combine.c (combine_simplify_rtx): Don't try to simplify if
952 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
955 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
957 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
958 to 32 bytes when compiling for POWER9.
960 2018-03-05 Jakub Jelinek <jakub@redhat.com>
963 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
964 regparm >= 3 with no arg reg available also for calls with
965 flag_force_indirect_call. Pass decl to ix86_function_regparm.
968 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
970 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
972 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
975 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
977 2018-03-05 Richard Biener <rguenther@suse.de>
979 PR tree-optimization/84486
980 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
981 When inserting a __builtin_assume_aligned call set the LHS
982 SSA name alignment info accordingly.
984 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
986 PR tree-optimization/84114
987 * config/aarch64/aarch64.c (aarch64_reassociation_width)
988 Avoid reassociation of FLOAT_MODE addition.
990 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
992 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
993 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
994 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
995 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
996 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
997 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
998 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1000 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1001 __builtin_ia32_wbinvd): New builtins.
1002 (SPECIAL_ARGS2): New.
1003 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1004 (SPECIAL_ARGS2): New.
1005 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1006 (ix86_valid_target_attribute_inner_p): Ditto.
1007 (ix86_init_mmx_sse_builtins): Add special_args2.
1008 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1009 TARGET_WBNOINVD_P): New.
1010 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1011 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1012 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1013 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1014 * config/i386/pconfigintrin.h: New file.
1015 * config/i386/wbnoinvdintrin.h: Ditto.
1016 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1017 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1019 2018-03-05 Richard Biener <rguenther@suse.de>
1021 PR tree-optimization/84670
1022 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1024 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1025 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1026 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1027 doesn't grow if all successors (recursively) were visited at least
1030 2018-03-05 Richard Biener <rguenther@suse.de>
1032 PR tree-optimization/84650
1033 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1034 if executed in the loop pipeline.
1036 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1038 * doc/configfiles.texi (Configuration Files): Move info about
1039 conditionalizing $target-protos.h to...
1040 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1041 differs from $target-protos.h.
1043 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1044 Chung-Ju Wu <jasonwucj@gmail.com>
1046 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1047 * config/nds32/nds32-multiple.md (setmemsi): Define.
1048 * config/nds32/nds32-memory-manipulation.c
1049 (nds32_gen_dup_4_byte_to_word_value): New.
1050 (emit_setmem_word_loop): New.
1051 (emit_setmem_byte_loop): New.
1052 (nds32_expand_setmem_loop): New.
1053 (nds32_expand_setmem_loop_v3m): New.
1054 (nds32_expand_setmem_unroll): New.
1055 (nds32_expand_setmem): New.
1057 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1058 Chung-Ju Wu <jasonwucj@gmail.com>
1060 * config/nds32/nds32-memory-manipulation.c
1061 (nds32_emit_load_store): New.
1062 (nds32_emit_post_inc_load_store): New.
1063 (nds32_emit_mem_move): New.
1064 (nds32_emit_mem_move_block): New.
1065 (nds32_expand_movmemsi_loop_unknown_size): New.
1066 (nds32_expand_movmemsi_loop_known_size): New.
1067 (nds32_expand_movmemsi_loop): New.
1068 (nds32_expand_movmemsi_unroll): New.
1069 (nds32_expand_movmemqi): Rename ...
1070 (nds32_expand_movmemsi): ... to this.
1071 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1072 (movmemsi): ... to this.
1073 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1074 (nds32_expand_movmemsi): ... to this.
1076 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1077 Monk Chiang <sh.chiang04@gmail.com>
1078 Chung-Ju Wu <jasonwucj@gmail.com>
1080 * config/nds32/nds32-protos.h
1081 (nds32_expand_load_multiple): New arguments.
1082 (nds32_expand_store_multiple): Ditto.
1083 (nds32_valid_multiple_load_store): Rename ...
1084 (nds32_valid_multiple_load_store_p): ... to this.
1085 * config/nds32/nds32-memory-manipulation.c
1086 (nds32_expand_load_multiple): Refine implementation.
1087 (nds32_expand_store_multiple): Ditto.
1088 * config/nds32/nds32-multiple.md
1089 (load_multiple): Update nds32_expand_load_multiple interface.
1090 (store_multiple): Update nds32_expand_store_multiple interface.
1091 * config/nds32/nds32-predicates.c
1092 (nds32_valid_multiple_load_store): Rename ...
1093 (nds32_valid_multiple_load_store_p): ... to this and refine
1095 * config/nds32/predicates.md
1096 (nds32_load_multiple_and_update_address_operation): New predicate.
1097 (nds32_store_multiple_and_update_address_operation): New predicate.
1099 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1100 Chung-Ju Wu <jasonwucj@gmail.com>
1102 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1103 (combo): New attribute.
1104 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1106 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1108 * config/nds32/nds32.opt: Change -mcmodel= default value.
1110 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1111 Monk Chiang <sh.chiang04@gmail.com>
1112 Chung-Ju Wu <jasonwucj@gmail.com>
1114 * config/nds32/constants.md (unspec_element): New enum.
1115 * config/nds32/constraints.md (Umw): New constraint.
1116 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1117 * config/nds32/nds32-intrinsic.md: Likewise.
1118 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1119 (nds32_valid_smw_lwm_base_p): New.
1120 (nds32_output_smw_single_word): New.
1121 (nds32_output_lmw_single_word): New.
1122 (nds32_expand_unaligned_load): New.
1123 (nds32_expand_unaligned_store): New.
1124 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1125 (nds32_output_smw_single_word): Declare.
1126 (nds32_output_lmw_single_word): Declare.
1127 (nds32_expand_unaligned_load): Declare.
1128 (nds32_expand_unaligned_store): Declare.
1129 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1130 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1131 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1132 NDS32_BUILTIN_UASTORE_DW.
1133 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1136 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1137 Kito Cheng <kito.cheng@gmail.com>
1138 Chung-Ju Wu <jasonwucj@gmail.com>
1140 * config/nds32/nds32-intrinsic.c
1141 (nds32_expand_builtin_null_ftype_reg): Delete.
1142 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1143 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1144 (nds32_read_argument): New.
1145 (nds32_legitimize_target): Ditto.
1146 (nds32_legitimize_argument): Ditto.
1147 (nds32_check_constant_argument): Ditto.
1148 (nds32_expand_unop_builtin): Ditto.
1149 (nds32_expand_unopimm_builtin): Ditto.
1150 (nds32_expand_binop_builtin): Ditto.
1151 (nds32_builtin_decl_impl): Ditto.
1152 (builtin_description): Ditto.
1153 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1154 (nds32_init_builtins_impl): Ditto.
1155 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1156 (nds32_builtin_decl): New.
1157 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1158 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1160 2018-03-02 Jeff Law <law@redhat.com>
1162 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1163 (redundant_insn, fill_simple_delay_slots): Likewise.
1164 (fill_slots_from_thread): Likewise.
1165 * resource.c (mark_referenced_resources): Likewise.
1166 (mark_set_resources, find_dead_or_set_registers): Likewise.
1168 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1170 * substring-locations.h (format_warning_va): Formatting fix for
1172 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1174 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1175 * substring-locations.c: Include intl.h.
1176 (format_warning_va): Turned into small wrapper around
1177 format_warning_n_va, renamed to ...
1178 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1179 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1181 (format_warning_at_substring_n): New function.
1182 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1183 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1184 format_warning_at_substring with just a shorter name instead of
1185 const function pointer.
1186 (fmtwarn_n): New function.
1187 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1188 appropriate, get rid of all the fmtstr temporaries, move conditionals
1189 with G_() wrapped string literals directly into fmtwarn arguments,
1190 cast dir.len to (int), formatting fixes.
1192 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1194 * doc/invoke.texi: Remove "Cilk Plus" references.
1196 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1197 Richard Biener <rguenther@suse.de>
1200 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1201 for error or warning attributes if CALL_FROM_THUNK_P is set.
1204 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1207 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1208 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1211 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1212 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1214 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1215 instead of -1U in last predictors element's probability member.
1217 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1220 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1221 arguments if they are comparable.
1223 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1225 PR tree-optimization/84634
1226 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1227 masks and masked_loop_p with a single loop_masks, making sure it's
1228 null for bb vectorization.
1230 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1232 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1233 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1234 loop->force_vectorize to check whether there is no alias.
1236 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1239 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1241 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1243 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1244 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1245 instead of a loop around prev_real_insn.
1246 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1250 * config/i386/i386.c (ix86_print_operand): Use conditional
1251 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1254 2018-03-02 Richard Biener <rguenther@suse.de>
1256 PR tree-optimization/84427
1257 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1258 (bitmap_set_subtract_values): Rewrite to handle multiple
1261 (prune_clobbered_mems): Likewise.
1262 (phi_translate): Take edge instead of pred/phiblock.
1263 (phi_translate_1): Likewise.
1264 (phi_translate_set): Likewise. Insert all translated
1265 exprs for a value into the set, keeping possibly multiple
1266 expressions per value.
1267 (compute_antic_aux): Adjust for phi_translate changes.
1268 When intersecting union the expressions and prune those
1269 not in the final value set, keeping possibly multiple
1270 expressions per value. Do not use value-insertion
1271 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1272 all expressions. Add verification that the value-sets
1273 only shrink during iteration.
1274 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1275 (do_pre_regular_insertion): Likewise.
1276 (do_pre_partial_partial_insertion): Likewise.
1278 2018-03-02 Richard Biener <rguenther@suse.de>
1281 * config/darwin.c (saved_debug_info_level): New static global.
1282 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1283 (darwin_asm_lto_end): Restore debug info generation settings.
1285 2018-03-01 Martin Liska <mliska@suse.cz>
1288 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1291 2018-03-01 Richard Biener <rguenther@suse.de>
1294 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1295 type annotation with LTO when debug was disabled at compile-time.
1297 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1299 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1301 (mips_final_postscan_insn): Likewise.
1303 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1305 PR rtl-optimization/84528
1306 * alias.c (init_alias_target): Add commentary.
1307 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1308 a unique base value if the frame pointer is not eliminated
1309 to the stack pointer.
1311 2018-03-01 Tom de Vries <tom@codesourcery.com>
1313 PR rtl-optimization/83327
1314 * lra-int.h (hard_regs_spilled_into): Declare.
1315 * lra.c (hard_regs_spilled_into): Define.
1316 (init_reg_info): Init hard_regs_spilled_into.
1317 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1318 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1319 (process_bb_lives): Handle hard_regs_spilled_into.
1320 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1321 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1323 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1325 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1326 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1327 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1328 * config/rs6000/aix72.h: New file.
1330 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1332 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1333 instead of warning_at with conditional singular and plural messages
1337 * stor-layout.c (update_alignment_for_field): For
1338 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1339 && !DECL_PACKED (field), do the alignment update, just use
1340 only desired_align instead of MAX (type_align, desired_align)
1342 (place_field): Don't do known_align < desired_align handling
1343 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1344 is non-NULL, instead do it after rli->prev_field handling and
1345 only if not within a bitfield word. For DECL_PACKED (field)
1346 use type_align of BITS_PER_UNIT.
1348 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1350 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1351 superfluous parentheses and trailing spaces.
1353 2018-02-28 Richard Biener <rguenther@suse.de>
1355 PR tree-optimization/84584
1356 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1357 SCoPs with fake exit edge.
1359 2018-02-28 Martin Liska <mliska@suse.cz>
1362 * timevar.c (timer::print): Fix format to properly print 100%
1365 2018-02-28 Richard Biener <rguenther@suse.de>
1368 * genmatch.c (capture_info::walk_match): Do not mark
1369 captured expressions without operands as expr_p given
1370 they act more like predicates and should be subject to
1371 "lost tail" side-effect preserving.
1373 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1375 PR rtl-optimization/81611
1376 * auto-inc-dec.c (attempt_change): Move dead note from
1377 mem_insn if it's the next use of regno
1378 (find_address): Take address use of reg holding
1379 non-incremented value. Add parm to limit search to the named
1381 (merge_in_block): Attempt to use a mem insn that is the next
1382 use of the original regno.
1384 2018-02-27 Martin Sebor <msebor@redhat.com>
1387 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1388 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1390 2018-02-27 Martin Sebor <msebor@redhat.com>
1392 PR translation/84207
1393 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1394 n argument to unsigned HOST_WIDE_INT.
1395 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1396 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1397 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1398 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1400 2018-02-27 Richard Biener <rguenther@suse.de>
1402 PR tree-optimization/84512
1403 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1404 Do not use the estimate returned from record_stmt_cost for
1405 the scalar iteration cost but sum properly using add_stmt_cost.
1407 2018-02-27 Richard Biener <rguenther@suse.de>
1409 PR tree-optimization/84466
1410 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1411 Adjust last change to less strictly validate use operands.
1413 2018-02-27 Martin Liska <mliska@suse.cz>
1415 PR gcov-profile/84548
1416 * gcov.c (process_file): Allow partial overlap and consider it
1417 also as group functions.
1418 (output_lines): Properly calculate range of lines for a group.
1420 2018-02-27 Martin Liska <mliska@suse.cz>
1422 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1423 'ggc' suffixes. Change first column width.
1424 (timer::print): Fix formatting of the column.
1426 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1428 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1429 preserve inline entry blocks for the sake of debug inline
1430 entry point markers alone.
1431 (remove_unused_locals): Suggest in comments a better place to
1432 force the preservation of inline entry blocks that are
1433 otherwise unused, but do not preserve them.
1435 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1437 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1439 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1442 * config/i386/constraints.md (Bs): Replace
1443 ix86_indirect_branch_register with
1444 TARGET_INDIRECT_BRANCH_REGISTER.
1446 * config/i386/i386.md (indirect_jump): Likewise.
1447 (tablejump): Likewise.
1448 (*sibcall_memory): Likewise.
1449 (*sibcall_value_memory): Likewise.
1450 Peepholes of indirect call and jump via memory: Likewise.
1451 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1452 (*sibcall_value_GOT_32): Likewise.
1453 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1454 (GOT_memory_operand): Likewise.
1455 (call_insn_operand): Likewise.
1456 (sibcall_insn_operand): Likewise.
1457 (GOT32_symbol_operand): Likewise.
1458 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1460 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1462 PR rtl-optimization/83496
1463 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1464 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1465 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1466 redundant insn, if any.
1467 (relax_delay_slots): Likewise.
1468 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1470 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1472 PR tree-optimization/83965
1473 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1474 that grouped statements are part of a reduction chain. Return
1475 true if the statement is not marked as a reduction itself but
1477 (vect_recog_dot_prod_pattern): Don't check whether the statement
1478 is part of a group here.
1479 (vect_recog_sad_pattern): Likewise.
1480 (vect_recog_widen_sum_pattern): Likewise.
1482 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1485 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1487 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1490 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1492 (ix86_output_indirect_function_return): New prototype.
1493 (ix86_split_simple_return_pop_internal): Likewise.
1494 * config/i386/i386.c (indirect_return_via_cx): New.
1495 (indirect_return_via_cx_bnd): Likewise.
1496 (indirect_thunk_name): Handle return va CX_REG.
1497 (output_indirect_thunk_function): Create alias for
1498 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1499 (ix86_output_indirect_jmp): Remove the bool argument.
1500 (ix86_output_indirect_function_return): New function.
1501 (ix86_split_simple_return_pop_internal): Likewise.
1502 * config/i386/i386.md (*indirect_jump): Don't pass false
1503 to ix86_output_indirect_jmp.
1504 (*tablejump_1): Likewise.
1505 (simple_return_pop_internal): Change it to define_insn_and_split.
1506 Call ix86_split_simple_return_pop_internal to split it for
1508 (simple_return_indirect_internal): Call
1509 ix86_output_indirect_function_return instead of
1510 ix86_output_indirect_jmp.
1512 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1515 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1516 memset and value initialization afterwards.
1518 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1520 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1522 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1525 * common/config/aarch64/aarch64-common.c
1526 (aarch_option_optimization_table[]): Switch
1527 off fomit-frame-pointer
1529 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1530 Chung-Ju Wu <jasonwucj@gmail.com>
1532 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1534 (store_multiple): Ditto.
1536 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1538 * config.gcc: Add --with-cpu support for nds32 target.
1539 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1540 * config/nds32/nds32.opt: Add -mcpu= option.
1542 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1544 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1545 isel=yes): Warn for these deprecated options.
1547 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1549 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1550 ISA_2_5_MASKS_EMBEDDED.
1552 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1554 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1555 p->max as pointers rather than using iterative_hash_expr.
1557 2018-02-23 Carl Love <cel@us.ibm.com>
1559 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1560 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1562 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1563 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1564 P8V_BUILTIN_VEC_VUNSIGNED2.
1566 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1569 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1570 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1572 * lra-constraints.c (curr_insn_transform): Set up
1573 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1575 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1576 * lra-eliminations.c (spill_pseudos): Ditto.
1577 (process_insn_for_elimination): Ditto.
1578 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1579 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1580 LRA_NON_CLOBBERED_ALT.
1582 2018-02-22 Martin Sebor <msebor@redhat.com>
1584 PR tree-optimization/84480
1585 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1586 to maybe_diag_stxncpy_trunc. Call it.
1587 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1588 from gimple_fold_builtin_strcpy. Print inlining stack.
1589 (handle_builtin_stxncpy): Print inlining stack.
1590 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1592 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1595 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1596 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1597 and -fcheck-pointer-bounds are used together.
1598 (indirect_thunk_prefix): New enum.
1599 (indirect_thunk_need_prefix): New function.
1600 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1601 "_nt" instead of "_bnd" for NOTRACK prefix.
1602 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1603 (output_indirect_thunk_function): Likewise.
1605 (ix86_code_end): Update output_indirect_thunk_function calls.
1606 (ix86_output_indirect_branch_via_reg): Replace
1607 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1608 (ix86_output_indirect_branch_via_push): Likewise.
1609 (ix86_output_function_return): Likewise.
1610 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1611 incompatible with -fcf-protection=branch and
1612 -fcheck-pointer-bounds.
1614 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1617 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1618 Change gcc_assert call to output_operand_lossage.
1620 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1622 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1624 2018-02-22 DJ Delorie <dj@redhat.com>
1625 Sebastian Perta <sebastian.perta@renesas.com>
1626 Oleg Endo <olegendo@gcc.gnu.org>
1628 * config/rx/rx.c (rx_rtx_costs): New function.
1629 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1631 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1633 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1635 2018-02-22 Martin Liska <mliska@suse.cz>
1638 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1639 Add "native" as a possible value.
1641 2018-02-22 Martin Liska <mliska@suse.cz>
1644 * config/i386/i386.c (ix86_option_override_internal):
1645 Add "native" as a possible value for -march and -mtune.
1647 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1650 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1651 to all type variants.
1653 PR tree-optimization/84503
1654 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1655 width as info->bitpos + info->bitsize - start.
1656 (merged_store_group::merge_overlapping): Simplify width computation.
1657 (check_no_overlap): New function.
1658 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1659 start + width and last_order of the group, fail if check_no_overlap
1661 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1662 to group if check_no_overlap fails.
1664 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1666 * config/rs6000/altivec.md: Delete contraint arguments to
1667 define_expand, define_split, and define_peephole2, and in
1668 define_insn_and_split if always unused.
1669 * config/rs6000/darwin.md: Ditto.
1670 * config/rs6000/dfp.md: Ditto.
1671 * config/rs6000/rs6000.md: Ditto.
1672 * config/rs6000/sync.md: Ditto.
1673 * config/rs6000/vector.md: Ditto.
1674 * config/rs6000/vsx.md: Ditto.
1676 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1678 * config/rs6000/altivec.md: Write output control strings as braced
1679 blocks instead of double-quoted strings.
1680 * config/rs6000/darwin.md: Ditto.
1681 * config/rs6000/rs6000.md: Ditto.
1682 * config/rs6000/vector.md: Ditto.
1683 * config/rs6000/vsx.md: Ditto.
1685 2018-02-21 Jason Merrill <jason@redhat.com>
1687 PR c++/84314 - ICE with templates and fastcall attribute.
1688 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1690 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1692 * ipa-cp.c (determine_versionability): Fix comment typos.
1694 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1697 * ipa-cp.c (determine_versionability): Do not version functions caling
1700 2018-02-21 Martin Liska <mliska@suse.cz>
1703 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1704 Add "native" as a possible value.
1705 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1706 the macro when native cpu detection is available.
1708 2018-02-21 Martin Liska <mliska@suse.cz>
1711 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1712 Add "native" as a possible value.
1713 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1714 when native cpu detection is available.
1716 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1717 Martin Sebor <msebor@redhat.com>
1719 PR tree-optimization/84478
1720 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1722 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1723 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1724 support which is conservatively correct, for 2 only stay conservative
1725 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1726 argument to the 2 argument get_range_strlen, adjust 6 arg
1727 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1729 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1730 (gimple_fold_builtin_strlen): Pass true as last argument to
1733 2018-02-20 Martin Sebor <msebor@redhat.com>
1736 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1737 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1738 (builtin_memref::builtin_memref): Factor out parts into
1739 set_base_and_offset and call it.
1741 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1744 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1745 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1746 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1747 search at the associated MODE_INT.
1749 2018-02-20 Jeff Law <law@redhat.com>
1752 PR tree-optimization/81592
1754 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1755 for range data rather than using global data.
1756 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1757 range data rather than using global data.
1758 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1759 pass it to children as needed.
1760 (struct directive::fmtresult): Similarly.
1761 (struct directive::set_width): Similarly.
1762 (struct directive::set_precision): Similarly.
1763 (format_integer, format_directive, parse_directive): Similarly.
1764 (format_none): Accept unnamed vr_values parameter.
1765 (format_percent, format_floating, format_character): Similarly.
1766 (format_string, format_plain): Similarly.
1767 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1768 the EVRP range analyzer for range data rather than using global data.
1769 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1770 gimple-ssa-evrp-analyze.h
1771 (class sprintf_dom_walker): Add after_dom_children member function.
1772 Add evrp_range_analyzer member.
1773 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1774 range analyzer as needed.
1775 (sprintf_dom_walker::after_dom_children): New member function.
1776 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1778 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1779 (evrp_range_analyzer::pop_to_marker): Likewise.
1781 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1783 PR tree-optimization/84419
1784 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1785 with the required type if its current type is compatible but
1788 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1791 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1792 after vectorization.
1794 2018-02-20 Martin Liska <mliska@suse.cz>
1797 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1798 possible values if we don't have a hint.
1800 2018-02-20 Martin Liska <mliska@suse.cz>
1804 * final.c (shorten_branches): Build align_tab array with one
1806 * opts.c (finish_options): Add alignment option limit check.
1807 (MAX_CODE_ALIGN): Likewise.
1808 (MAX_CODE_ALIGN_VALUE): Likewise.
1809 * doc/invoke.texi: Document maximum allowed option value for
1810 all -falign-* options.
1812 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1815 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1816 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1817 * var-tracking.c (emit_note_insn_var_location): Remove all references
1818 to NOTE_INSN_CALL_ARG_LOCATION.
1819 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1820 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1821 Use copy_rtx_if_shared.
1822 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1823 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1824 (dwarf2out_var_location): Remove handling of
1825 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1827 * final.c (final_scan_insn): Remove all references to
1828 NOTE_INSN_CALL_ARG_LOCATION.
1829 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1830 before dumping final insns.
1831 * except.c (emit_note_eh_region_end): Remove all references to
1832 NOTE_INSN_CALL_ARG_LOCATION.
1833 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1834 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1835 * config/arc/arc.c (hwloop_optimize): Likewise.
1836 * config/arm/arm.c (create_fix_barrier): Likewise.
1837 * config/s390/s390.c (s390_chunkify_start): Likewise.
1838 * config/sh/sh.c (find_barrier): Likewise.
1839 * config/i386/i386.c (rest_of_insert_endbranch,
1840 ix86_seh_fixup_eh_fallthru): Likewise.
1841 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1842 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1843 * config/frv/frv.c (frv_function_prologue): Likewise.
1844 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1846 (note_outside_basic_block_p): Remove all references to
1847 NOTE_INSN_CALL_ARG_LOCATION.
1848 * gengtype.c (adjust_field_rtx_def): Likewise.
1849 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1851 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1852 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1855 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1858 PR tree-optimization/84452
1859 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1860 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1863 2018-02-19 Martin Liska <mliska@suse.cz>
1866 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1868 2018-02-19 Martin Liska <mliska@suse.cz>
1869 Richard Sandiford <richard.sandiford@linaro.org>
1871 PR tree-optimization/82491
1872 * gimple-fold.c (get_base_constructor): Make earlier bail out
1875 2018-02-19 Carl Love <cel@us.ibm.com>
1877 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1878 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1880 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1881 P8V_BUILTIN_VEC_NEG.
1883 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1885 * config/rl78/rl78.md (movdf): New define expand.
1887 2018-02-19 Martin Liska <mliska@suse.cz>
1890 * doc/invoke.texi: Fix typo.
1891 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1893 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1895 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1896 handle rs6000_single_float and rs6000_double_float specially for
1899 2018-02-16 Jeff Law <law@redhat.com>
1901 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1904 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1907 * ipa-inline.c (inline_small_functions): Fix a typo.
1909 2018-02-16 Nathan Sidwell <nathan@acm.org>
1911 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1913 2018-02-16 Carl Love <cel@us.ibm.com>
1915 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1916 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1917 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1918 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1919 expansion to P8V_BUILTIN_VEC_FLOAT2.
1921 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1923 PR rtl-optimization/70023
1924 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1925 src_regno into account.
1927 2018-02-16 Carl Love <cel@us.ibm.com>
1929 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1930 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1931 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1932 * config/rs6000/rs6000.c: Remove case statements for
1933 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1934 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1935 and P9V_BUILTIN_VEC_VINSERT4B.
1936 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1937 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1938 * config/rs6000/vsx.md:
1939 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1942 2018-02-16 Carl Love <cel@us.ibm.com>
1944 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1946 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1948 * config/rs6000/rs6000-c.c: Add the definitions for
1949 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1950 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1951 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1952 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1953 definition for insert4b and define insn *insert3b_internal.
1954 * doc/extend.texi: Add documentation for vec_extract4b.
1956 2018-02-16 Nathan Sidwell <nathan@acm.org>
1958 * doc/extend.texi (Backwards Compatibility): Mention friend
1959 injection. Note for-scope is deprecated.
1960 * doc/invoke.texi (-ffriend-injection): Deprecate.
1962 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1964 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1965 that moved to I2, also allow destinations that are a paradoxical
1966 subreg (instead of a normal reg).
1968 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1971 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1974 2018-02-16 Richard Biener <rguenther@suse.de>
1976 PR tree-optimization/84037
1977 PR tree-optimization/84016
1979 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1980 Adjust vec_construct for the fact we need additional higher latency
1981 128bit inserts for AVX256 and AVX512 vector builds.
1982 (ix86_add_stmt_cost): Scale vector construction cost for
1985 2018-02-16 Richard Biener <rguenther@suse.de>
1987 PR tree-optimization/84417
1988 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1989 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1990 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1992 2018-02-16 Martin Liska <mliska@suse.cz>
1995 * internal-fn.def (ASAN_CHECK): Set proper flags.
1996 (ASAN_MARK): Likewise.
1998 2018-02-16 Julia Koval <julia.koval@intel.com>
2000 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2001 from PTA_CANNONLAKE.
2003 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2006 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2007 Use ++iter rather than iter++ for std::list iterators.
2008 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2009 defer deleting them until all nodes in the forest are processed. Do
2010 free even leaf nodes. Change to_process into auto_vec.
2013 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2014 * vec.h (vec_default_construct): Use memset instead of placement new
2015 if BROKEN_VALUE_INITIALIZATION is defined.
2016 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2017 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2020 PR rtl-optimization/83723
2021 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2022 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2023 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2025 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2027 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2029 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2031 PR rtl-optimization/81443
2032 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2033 from inner REGs to paradoxical SUBREGs.
2035 2018-02-16 Richard Biener <rguenther@suse.de>
2037 PR tree-optimization/84399
2038 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2039 For operands we can analyze at their definition make sure we can
2040 analyze them at each use as well.
2042 2018-02-16 Richard Biener <rguenther@suse.de>
2044 PR tree-optimization/84190
2045 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2046 volatile accesses if the decl isn't volatile.
2048 2018-02-15 Jason Merrill <jason@redhat.com>
2050 PR c++/84314 - ICE with templates and fastcall attribute.
2051 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2052 TYPE_CANONICAL on an existing type.
2054 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2056 PR tree-optimization/84383
2057 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2058 dstoff nor call operand_equal_p if dstbase is NULL.
2060 PR tree-optimization/84334
2061 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2062 also a CONSTANT_CLASS_P, punt.
2064 2018-02-14 Jim Wilson <jimw@sifive.com>
2066 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2067 first SMALL_OPERAND check. New local min_second_step. Move assert
2068 to where locals are set. Add TARGET_RVC support.
2069 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2071 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2073 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2075 2018-02-14 Martin Sebor <msebor@redhat.com>
2077 PR tree-optimization/83698
2078 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2079 arrays constrain the offset range to their bounds.
2080 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2081 (builtin_access::overlap): Avoid setting the size of overlap if it's
2083 (maybe_diag_overlap): Also consider arrays when deciding what values
2084 of offsets to include in diagnostics.
2086 2018-02-14 Martin Sebor <msebor@redhat.com>
2089 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2090 that correspond to the kind of a declaration.
2092 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2095 * config/pa/pa.md: Load address of PIC label using the linkage table
2096 if the label is nonlocal.
2098 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2100 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2101 warning message if user requests -maltivec=be.
2102 * doc/invoke.texi: Document deprecation of -maltivec=be.
2104 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2107 * config/rs6000/rs6000-c.c: Update definitions for
2108 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2109 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2111 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2114 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2115 add _get_ssp intrinsics. Remove argument from
2116 __builtin_ia32_rdssp[d|q].
2117 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2118 * config/i386/i386-builtin.def: Remove argument from
2119 __builtin_ia32_rdssp[d|q].
2120 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2121 ix86_expand_special_args_builtin for _rdssp[d|q].
2122 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2123 Clear register before usage.
2124 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2125 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2127 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2129 PR tree-optimization/84357
2130 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2131 operand 1 of an ARRAY_REF too.
2133 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2136 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2137 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2139 (set_of_reg): New struct.
2140 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2141 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2142 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2144 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2145 Split into bitclr, bitset, bitinvert patterns if appropriate.
2146 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2147 use rx_fuse_in_memory_bitop.
2148 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2149 to named insn, correct maximum insn length.
2151 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2154 * machmode.def: Define a complex mode for PARTIAL_INT.
2155 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2157 * doc/rtl.texi: Document CSPImode.
2158 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2160 (msp430_hard_regno_nregs_with_padding): Likewise.
2162 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2165 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2167 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2169 PR rtl-optimization/84169
2170 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2171 we generated a parallel as new i3 and we split that to new i2 and i3
2172 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2173 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2174 those to i2, not i1. Partially rewrite this scan code.
2176 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2179 * stor-layout.c (place_field): For variable length fields, adjust
2180 offset_align afterwards not just based on the field's alignment,
2181 but also on the size.
2184 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2185 of exps and logs in the use_exp2 case.
2187 2018-02-13 Jeff Law <law@redhat.com>
2189 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2192 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2195 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2199 * common.opt (gas-loc-support, gas-locview-support): New.
2200 (ginline-points, ginternal-reset-location-views): New.
2201 * doc/invoke.texi: Document them. Use @itemx where intended.
2202 (gvariable-location-views): Adjust.
2203 * target.def (reset_location_view): New.
2204 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2205 (TARGET_RESET_LOCATION_VIEW): New.
2206 * doc/tm.texi: Rebuilt.
2207 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2208 (dwarf2out_default_as_locview_support): New.
2209 (output_asm_line_debug_info): Use option variables.
2210 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2211 (output_loc_list): Likewise.
2212 (add_high_low_attributes): Check option variables.
2213 Don't output entry view attribute in strict mode.
2214 (gen_inlined_subroutine_die): Check option variables.
2215 (dwarf2out_inline_entry): Likewise.
2216 (init_sections_and_labels): Likewise.
2217 (dwarf2out_early_finish): Likewise.
2218 (maybe_reset_location_view): New, from...
2219 (dwarf2out_var_location): ... here. Call it.
2220 * debug.h (dwarf2out_default_as_loc_support): Declare.
2221 (dwarf2out_default_as_locview_support): Declare.
2222 * hooks.c (hook_int_rtx_insn_0): New.
2223 * hooks.h (hook_int_rtx_insn_0): Declare.
2224 * toplev.c (process_options): Take -gas-loc-support and
2225 -gas-locview-support from dwarf2out. Enable
2226 -gvariable-location-views by default only with locview
2227 assembler support. Enable -ginternal-reset-location-views by
2228 default only if the target defines the corresponding hook.
2229 Enable -ginline-points by default if location views are
2230 enabled; force it disabled if statement frontiers are
2232 * tree-inline.c (expand_call_inline): Check option variables.
2233 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2235 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2237 PR tree-optimization/84321
2238 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2239 handling. Also check whether the anti-range contains any values
2240 that satisfy the mask; switch to a VR_RANGE if not.
2242 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2245 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2247 2018-02-13 Martin Jambor <mjambor@suse.cz>
2250 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2251 of call statements, also set location of a load to a temporary.
2253 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2255 * config/rl78/rl78.c (add_vector_labels): New function.
2256 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2257 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2258 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2259 which checks that no arguments are passed.
2260 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2261 * doc/extend.texi: Documentation for the new attribute.
2263 2018-02-13 Andreas Schwab <schwab@suse.de>
2265 * config/riscv/linux.h (CPP_SPEC): Define.
2267 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2270 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2271 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2272 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2273 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2274 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2275 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2276 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2277 temporarily for AES and PCLMUL builtins.
2279 PR tree-optimization/84339
2280 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2281 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2285 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2286 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2287 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2289 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2291 * omp-simd-clone.h: New file.
2292 * omp-simd-clone.c: Include omp-simd-clone.h.
2293 (expand_simd_clones): No longer static.
2294 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2295 cgraph.h and omp-simd-clone.h.
2296 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2297 (vect_recog_widen_shift_pattern): Formatting fix.
2298 (vect_pattern_recog_1): Don't check optab for calls.
2301 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2302 operands[2] into a REG before using gen_lowpart on it.
2304 2018-02-12 Jeff Law <law@redhat.com>
2307 * config/sh/sh.c (find_barrier): Consider a sibling call
2310 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2311 successfully back substituting a reg.
2313 2018-02-12 Richard Biener <rguenther@suse.de>
2315 PR tree-optimization/84037
2316 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2317 parameter, move visited init to caller.
2318 (vect_slp_analyze_operations): Separate cost from validity
2319 check, initialize visited once for all instances.
2320 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2322 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2324 (vect_model_promotion_demotion_cost): Likewise.
2325 (vectorizable_bswap): Guard cost modeling with !slp_node
2326 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2328 (vectorizable_call): Likewise.
2329 (vectorizable_conversion): Likewise.
2330 (vectorizable_assignment): Likewise.
2331 (vectorizable_shift): Likewise.
2332 (vectorizable_operation): Likewise.
2333 (vectorizable_store): Likewise.
2334 (vectorizable_load): Likewise.
2335 (vectorizable_condition): Likewise.
2336 (vectorizable_comparison): Likewise.
2338 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2341 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2342 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2343 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2345 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2348 * params.def (inline-min-speedup): Increase from 8 to 15.
2349 (max-inline-insns-auto): Decrease from 40 to 30.
2350 * ipa-split.c (consider_split): Add some buffer for function to
2351 be considered inlining candidate.
2352 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2355 2018-02-12 Richard Biener <rguenther@suse.de>
2357 PR tree-optimization/84037
2358 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2359 matched stmts if we cannot swap the non-matched ones.
2361 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2363 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2364 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2365 _mm_maskz_scalef_round_ss): New intrinsics.
2366 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2367 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2368 __builtin_ia32_scalefss_round): Remove.
2369 (__builtin_ia32_scalefsd_mask_round,
2370 __builtin_ia32_scalefss_mask_round): New intrinsics.
2371 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2372 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2373 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2374 "<round_constraint>")): Changed to ...
2375 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2376 "<round_scalar_constraint>")): ... this.
2377 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2378 %0, %1, %2<round_op3>}"): Changed to ...
2379 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2380 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2381 %2<round_scalar_mask_op3>}"): ... this.
2382 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2384 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2386 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2387 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2388 (_mm_maskz_sqrt_round_ss): New intrinsics.
2389 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2390 (__builtin_ia32_sqrtsd_mask_round)
2391 (__builtin_ia32_sqrtss_mask_round): New builtins.
2392 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2393 (__builtin_ia32_sqrtss_round): Remove.
2394 (__builtin_ia32_sqrtsd_mask_round)
2395 (__builtin_ia32_sqrtss_mask_round): New builtins.
2396 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2397 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2398 ((match_operand:VF_128 1 "vector_operand"
2399 "xBm,<round_constraint>")): Changed to ...
2400 ((match_operand:VF_128 1 "vector_operand"
2401 "xBm,<round_scalar_constraint>")): ... this.
2402 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2403 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2404 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2405 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2406 %<iptr>1<round_scalar_mask_op3>}): ... this.
2407 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2408 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2410 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2413 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2414 Cast vec_cmpeq result to correct type.
2415 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2416 Cast vec_cmpgt result to correct type.
2418 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2420 * final.c (final_scan_insn_1): Renamed from...
2421 (final_scan_insn): ... this. New wrapper, to recover
2422 seen from the outermost call in recursive ones.
2423 * config/sparc/sparc.c (output_return): Drop seen from call.
2424 (output_sibcall): Likewise.
2425 * config/visium/visium.c (output_branch): Likewise.
2427 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2429 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2432 2018-02-10 Alan Modra <amodra@gmail.com>
2435 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2436 Specify LR as an input.
2438 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2441 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2442 remove_member_access_dummy_vars): New functions.
2443 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2444 lower_omp_1, execute_lower_omp): Use them.
2446 PR rtl-optimization/84308
2447 * shrink-wrap.c (spread_components): Release todo vector.
2449 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2451 PR rtl-optimization/57193
2452 * ira-color.c (struct allocno_color_data): Add member
2453 conflict_allocno_hard_prefs.
2454 (update_conflict_allocno_hard_prefs): New.
2455 (bucket_allocno_compare_func): Add a preference based on
2456 conflict_allocno_hard_prefs.
2457 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2458 (color_allocnos): Remove a dead code. Initiate
2459 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2461 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2464 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2465 constraint from =wa to wa. Avoid a subreg on the output operand,
2466 instead use a pseudo and subreg it in a move.
2467 (p9_xxbrd_<mode>): Changed to ...
2468 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2469 (p9_xxbrd_v2df): New expander.
2470 (p9_xxbrw_<mode>): Changed to ...
2471 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2472 (p9_xxbrw_v4sf): New expander.
2474 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2476 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2478 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2481 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2482 multiply in 32-bit mode.
2483 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2484 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2487 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2489 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2490 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2491 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2492 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2494 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2497 * dwarf2out.c (is_trivial_indirect_ref): New function.
2498 (dwarf2out_late_global_decl): Do not generate a location
2499 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2500 and that are not defined in the current unit.
2502 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2504 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2505 instead of a libcall for UNORDERED.
2507 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2510 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2511 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2513 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2516 * config/s390/s390.c (s390_set_current_function): Invoke
2517 s390_indirect_branch_settings also if fndecl didn't change.
2519 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2521 * config/rs6000/rs6000.md (blockage): Set length to zero.
2523 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2525 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2527 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2530 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2531 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2535 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2536 PARALLEL incoming that failed vt_get_decl_and_offset check.
2539 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2540 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2542 (get_variable_section): For decls in named .bss* sections pass true as
2543 second argument to bss_initializer_p.
2545 2018-02-09 Marek Polacek <polacek@redhat.com>
2546 Jakub Jelinek <jakub@redhat.com>
2549 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2550 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2551 Sync some changes from cxx_fold_indirect_ref.
2553 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2555 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2557 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2558 (BLOCK_INLINE_ENTRY_LABEL): New.
2559 (dwarf2out_var_location): Disregard inline entry markers.
2560 (inline_entry_data): New struct.
2561 (inline_entry_data_hasher): New hashtable type.
2562 (inline_entry_data_hasher::hash): New.
2563 (inline_entry_data_hasher::equal): New.
2564 (inline_entry_data_table): New variable.
2565 (add_high_low_attributes): Add DW_AT_entry_pc and
2566 DW_AT_GNU_entry_view attributes if a pending entry is found
2567 in inline_entry_data_table. Add old entry_pc attribute only
2568 if debug nonbinding markers are disabled.
2569 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2570 markers are enabled.
2571 (block_within_block_p, dwarf2out_inline_entry): New.
2572 (dwarf2out_finish): Check that no entries remained in
2573 inline_entry_data_table.
2574 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2575 (final_scan_insn, notice_source_line): Likewise.
2576 (rest_of_clean_state): Skip inline entry markers.
2577 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2579 * gimple.c (gimple_build_debug_inline_entry): New.
2580 * gimple.h (enum gimple_debug_subcode): Add
2581 GIMPLE_DEBUG_INLINE_ENTRY.
2582 (gimple_build_debug_inline_entry): Declare.
2583 (gimple_debug_inline_entry_p): New.
2584 (gimple_debug_nonbind_marker_p): Adjust.
2585 * insn-notes.def (INLINE_ENTRY): New.
2586 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2587 inline entry marker notes.
2588 (print_insn): Likewise.
2589 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2590 (INSN_DEBUG_MARKER_KIND): Likewise.
2591 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2592 * tree-inline.c (expand_call_inline): Build and insert
2593 debug_inline_entry stmt.
2594 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2595 inline entry blocks early, if nonbind markers are enabled.
2596 (dump_scope_block): Dump fragment info.
2597 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2598 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2599 (gimple_build_debug_inline_entry): New.
2600 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2601 Enable/disable inline entry points too.
2602 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2603 (DEBUG_INSN): Describe inline entry markers.
2605 * common.opt (gvariable-location-views): New.
2606 (gvariable-location-views=incompat5): New.
2607 * config.in: Rebuilt.
2608 * configure: Rebuilt.
2609 * configure.ac: Test assembler for view support.
2610 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2611 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2612 * dwarf2out.c (var_loc_view): New typedef.
2613 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2614 (dwarf2out_locviews_in_attribute): New.
2615 (dwarf2out_locviews_in_loclist): New.
2616 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2617 (enum dw_line_info_opcode): Add LI_adv_address.
2618 (struct dw_line_info_table): Add view.
2619 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2620 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2621 (zero_view_p): New variable.
2622 (ZERO_VIEW_P): New macro.
2623 (output_asm_line_debug_info): New.
2624 (struct var_loc_node): Add view.
2625 (add_AT_view_list, AT_loc_list): New.
2626 (add_var_loc_to_decl): Add view param. Test it against last.
2627 (new_loc_list): Add view params. Record them.
2628 (AT_loc_list_ptr): Handle loc and view lists.
2629 (view_list_to_loc_list_val_node): New.
2630 (print_dw_val): Handle dw_val_class_view_list.
2631 (size_of_die): Likewise.
2632 (value_format): Likewise.
2633 (loc_list_has_views): New.
2634 (gen_llsym): Set vl_symbol too.
2635 (maybe_gen_llsym, skip_loc_list_entry): New.
2636 (dwarf2out_maybe_output_loclist_view_pair): New.
2637 (output_loc_list): Output view list or entries too.
2638 (output_view_list_offset): New.
2639 (output_die): Handle dw_val_class_view_list.
2640 (output_dwarf_version): New.
2641 (output_compilation_unit_header): Use it.
2642 (output_skeleton_debug_sections): Likewise.
2643 (output_rnglists, output_line_info): Likewise.
2644 (output_pubnames, output_aranges): Update version comments.
2645 (output_one_line_info_table): Output view numbers in asm comments.
2646 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2647 Call maybe_gen_llsym.
2648 (loc_list_from_tree_1): Adjust.
2649 (add_AT_location_description): Create view list attribute if
2650 needed, check it's absent otherwise.
2651 (convert_cfa_to_fb_loc_list): Adjust.
2652 (maybe_emit_file): Call output_asm_line_debug_info for test.
2653 (dwarf2out_var_location): Reset views as needed. Precompute
2654 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2655 attribute. Set view.
2656 (new_line_info_table): Reset next view.
2657 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2658 (dwarf2out_source_line): Likewise. Output view resets and labels to
2659 the assembler, or select appropriate line info opcodes.
2660 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2661 (optimize_string_length): Catch it. Adjust.
2662 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2663 dw_val_class_view_list, and remove it if no longer needed.
2664 (hash_loc_list): Hash view numbers.
2665 (loc_list_hasher::equal): Compare them.
2666 (optimize_location_lists): Check whether a view list symbol is
2667 needed, and whether the locview attribute is present, and
2668 whether they match. Remove the locview attribute if no longer
2670 (index_location_lists): Call skip_loc_list_entry for test.
2671 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2672 Use output_dwarf_version.
2673 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2674 (struct dw_val_node): Add val_view_list.
2675 * final.c (SEEN_NEXT_VIEW): New.
2676 (set_next_view_needed): New.
2677 (clear_next_view_needed): New.
2678 (maybe_output_next_view): New.
2679 (final_start_function): Rename to...
2680 (final_start_function_1): ... this. Take pointer to FIRST,
2681 add SEEN parameter. Emit param bindings in the initial view.
2682 (final_start_function): Reintroduce SEEN-less interface.
2683 (final): Rename to...
2684 (final_1): ... this. Take SEEN parameter. Output final pending
2685 next view at the end.
2686 (final): Reintroduce seen-less interface.
2687 (final_scan_insn): Output pending next view before switching
2688 sections or ending a block. Mark the next view as needed when
2689 outputting variable locations. Notify debug backend of section
2690 changes, and of location view changes.
2691 (rest_of_handle_final): Adjust.
2692 * toplev.c (process_options): Autodetect value for debug variable
2693 location views option. Warn on incompat5 without -gdwarf-5.
2694 * doc/invoke.texi (gvariable-location-views): New.
2695 (gvariable-location-views=incompat5): New.
2696 (gno-variable-location-views): New.
2698 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2700 PR tree-optimization/84136
2701 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2702 that the result of find_edge is non-NULL.
2704 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2707 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2708 storing integer register in SImode. Fix cost of 256 and 512
2709 byte aligned SSE register store.
2711 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2713 * config/i386/i386.c (ix86_multiplication_cost): Fix
2714 multiplication cost for TARGET_AVX512DQ.
2716 2018-02-08 Marek Polacek <polacek@redhat.com>
2718 PR tree-optimization/84238
2719 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2722 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2724 PR tree-optimization/84265
2725 * tree-vect-stmts.c (vectorizable_store): Don't treat
2726 VMAT_CONTIGUOUS accesses as grouped.
2727 (vectorizable_load): Likewise.
2729 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2731 PR tree-optimization/81635
2732 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2733 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2734 (test_round_for_mask): New functions.
2735 (wide_int_cc_tests): Call test_round_for_mask.
2736 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2737 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2738 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2739 range returned by get_range_info.
2741 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2744 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2745 * symtab.c: Include builtins.h
2746 (symtab_node::output_to_lto_symbol_table_p): Move here
2747 from lto-streamer-out.c:output_symbol_p.
2748 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2749 (output_symbol_p): Move all logic to symtab.c
2750 (produce_symtab): Update.
2752 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2754 * config/s390/s390-opts.h (enum indirect_branch): Define.
2755 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2756 (s390_indirect_branch_via_thunk)
2757 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2758 (enum s390_indirect_branch_type): Define.
2759 * config/s390/s390.c (struct s390_frame_layout, struct
2760 machine_function): Remove.
2761 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2762 (indirect_branch_table_label_no, indirect_branch_table_name):
2764 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2765 (enum s390_indirect_branch_option): Define.
2766 (s390_return_addr_from_memory): New function.
2767 (s390_handle_string_attribute): New function.
2768 (s390_attribute_table): Add new attribute handler.
2769 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2770 (s390_indirect_branch_via_thunk): New function.
2771 (s390_indirect_branch_via_inline_thunk): New function.
2772 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2773 sibling call optimization for non z10 compiles.
2774 (s390_emit_call): Force indirect branch target to be a single
2775 register. Add r1 clobber for non-z10 compiles.
2776 (s390_emit_epilogue): Emit return jump via return_use expander.
2777 (s390_reorg): Handle JUMP_INSNs as execute targets.
2778 (s390_option_override_internal): Perform validity checks for the
2779 new command line options.
2780 (s390_indirect_branch_attrvalue): New function.
2781 (s390_indirect_branch_settings): New function.
2782 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2783 (s390_output_indirect_thunk_function): New function.
2784 (s390_code_end): Implement target hook.
2785 (s390_case_values_threshold): Implement target hook.
2786 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2788 * config/s390/s390.h (struct s390_frame_layout)
2789 (struct machine_function): Move here from s390.c.
2790 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2791 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2792 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2793 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2794 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2795 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2796 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2797 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2798 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2799 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2800 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2801 (mnemonic attribute): Add values which aren't recognized
2803 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2804 pattern for branch conversion. Fix mnemonic attribute.
2805 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2806 indirect branch via thunk if requested.
2807 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2808 ("*indirect_jump"): Disable for branch conversion using out of
2810 ("indirect_jump_via_thunk<mode>_z10")
2811 ("indirect_jump_via_thunk<mode>")
2812 ("indirect_jump_via_inlinethunk<mode>_z10")
2813 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2814 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2815 ("casesi_jump_via_inlinethunk<mode>_z10")
2816 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2817 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2818 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2819 ("*indirect2_jump"): Disable for branch conversion.
2820 ("casesi_jump"): Turn into expander and expand patterns for branch
2822 ("return_use"): New expander.
2823 ("*return"): Emit return via thunk and rename it to ...
2824 ("*return<mode>"): ... this one.
2825 * config/s390/s390.opt: Add new options and and enum for the
2828 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2830 * lra-constraints.c (match_reload): Unconditionally use
2831 gen_lowpart_SUBREG, rather than selecting between that
2832 and equivalent gen_rtx_SUBREG code.
2834 2018-02-08 Richard Biener <rguenther@suse.de>
2836 PR tree-optimization/84233
2837 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2838 changed flag instead of boguously re-using phi_inserted.
2840 2018-02-08 Martin Jambor <mjambor@suse.cz>
2842 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2843 static local variables.
2845 2018-02-08 Richard Biener <rguenther@suse.de>
2847 PR tree-optimization/84278
2848 * tree-vect-stmts.c (vectorizable_store): When looking for
2849 smaller vector types to perform grouped strided loads/stores
2850 make sure the mode is supported by the target.
2851 (vectorizable_load): Likewise.
2853 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2855 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2856 Increase LDP/STP opportunities by adding adjacent callee-saves.
2858 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2860 PR rtl-optimization/84068
2861 PR rtl-optimization/83459
2862 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2864 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2866 PR tree-optimization/84224
2867 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2868 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2871 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2874 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2875 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2876 count, remove one USE.
2878 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2880 * doc/install.texi (Configuration): Document the
2881 --with-long-double-format={ibm,ieee} PowerPC configuration
2885 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2886 Convert from define_expand to be define_insn_and_split. Rework
2887 float/double/_Float128 conversions to QI/HI/SImode to work with
2888 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2889 conversions to QI/HImode types did a store and then a load to
2890 truncate the value. For conversions to VSX registers, don't split
2891 the insn, instead emit the code directly. Use the code iterator
2892 any_fix to combine signed and unsigned conversions.
2893 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2894 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2895 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2896 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2897 (fix_<mode>di2_hw): Likewise.
2898 (fixuns_<mode>di2_hw): Likewise.
2899 (fix_<mode>si2_hw): Likewise.
2900 (fixuns_<mode>si2_hw): Likewise.
2901 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2902 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2903 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2904 fix<uns>_trunc<SFDF:mode>si2_p8.
2905 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2907 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2908 (fix<uns>_<mode>_mem): Likewise.
2909 (fctiw<u>z_<mode>_mem): Likewise.
2910 (fix<uns>_<mode>_mem): Likewise.
2911 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2912 the register allocator from doing a direct move to the GPRs to do
2913 a store, and instead use the ISA 3.0 store byte/half-word from
2914 vector register instruction. For IEEE 128-bit floating point,
2915 also optimize stores of 32-bit ints.
2916 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2918 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2920 * genextract.c (push_pathstr_operand): New function to support
2922 (walk_rtx): Call push_pathstr_operand.
2923 (print_path): Support [a-zA-Z].
2925 2018-02-07 Richard Biener <rguenther@suse.de>
2927 PR tree-optimization/84037
2928 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2929 (cse_and_gimplify_to_preheader): Declare.
2930 (vect_get_place_in_interleaving_chain): Likewise.
2931 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2933 (_loop_vec_info::~_loop_vec_info): Delete it.
2934 (cse_and_gimplify_to_preheader): New function.
2935 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2936 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2937 (vectorizable_load): Likewise. For grouped stores always base
2938 the IV on the first element.
2939 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2940 condition before gimplifying.
2942 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2944 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2945 *DIV_EXPR and *MOD_EXPR.
2947 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2950 * config/i386/i386.c (ix86_option_override_internal): Mask out
2951 the CF_SET bit when checking -fcf-protection.
2953 2018-02-07 Tom de Vries <tom@codesourcery.com>
2956 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2959 2018-02-07 Richard Biener <rguenther@suse.de>
2961 PR tree-optimization/84204
2962 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2965 PR tree-optimization/84205
2966 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2967 special-case isl_ast_op_zdiv_r.
2969 PR tree-optimization/84223
2970 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2971 Only add conditions from within the region.
2972 (gather_bbs::after_dom_children): Adjust.
2974 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2977 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2978 * config/avr/avr.md: Only post-reload split REG-REG moves if
2979 either register is GENERAL_REG_P.
2981 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2983 PR tree-optimization/84235
2984 * tree-ssa-scopedtables.c
2985 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2986 if the subtraction is performed in floating point type where NaNs are
2987 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2988 build 1. Formatting fix.
2990 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2993 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2994 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2995 and skip it regardless of bb boundaries. Use CALL_P macro,
2996 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2997 unnecessarily, formatting fix.
2999 2018-02-06 Michael Collison <michael.collison@arm.com>
3001 * config/arm/thumb2.md:
3002 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3003 (*thumb_mov_notscc): Ditto.
3005 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3008 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3009 unsigned_fix, not "s".
3011 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3013 * configure.ac (gcc_fn_eh_frame_ro): New function.
3014 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3015 correct .eh_frame permissions.
3016 * configure: Regenerate.
3018 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3020 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3023 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3025 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3026 Display warning message for -mno-speculate-indirect-jumps.
3028 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3030 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3032 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3034 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3036 PR tree-optimization/84225
3037 * tree-eh.c (find_trapping_overflow): Only call
3038 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3040 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3043 * config/i386/i386.c: Reimplement the check of possible options
3044 -mibt/-mshstk conbination. Change error messages.
3045 * doc/invoke.texi: Fix a typo: remove extra '='.
3047 2018-02-06 Marek Polacek <polacek@redhat.com>
3049 PR tree-optimization/84228
3050 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3052 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3055 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3056 emitted arch directives.
3057 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3058 __ARM_FEATURE_COPROC before changing architectures.
3060 2018-02-06 Richard Biener <rguenther@suse.de>
3062 * config/i386/i386.c (print_reg): Fix typo.
3063 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3065 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3067 * configure: Regenerate.
3069 2018-02-05 Martin Sebor <msebor@redhat.com>
3071 PR tree-optimization/83369
3072 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3075 2018-02-05 Martin Liska <mliska@suse.cz>
3077 * doc/invoke.texi: Cherry-pick upstream r323995.
3079 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3081 * ira.c (ira_init_register_move_cost): Adjust comment.
3083 2018-02-05 Martin Liska <mliska@suse.cz>
3085 PR gcov-profile/84137
3086 * doc/gcov.texi: Fix typo in documentation.
3088 2018-02-05 Martin Liska <mliska@suse.cz>
3090 PR gcov-profile/83879
3091 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3092 using dlopen functionality.
3094 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3096 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3097 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3098 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3099 _mm_maskz_range_round_ss): New intrinsics.
3100 (__builtin_ia32_rangesd128_round)
3101 (__builtin_ia32_rangess128_round): Remove.
3102 (__builtin_ia32_rangesd128_mask_round,
3103 __builtin_ia32_rangess128_mask_round): New builtins.
3104 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3105 __builtin_ia32_rangess128_round): Remove.
3106 (__builtin_ia32_rangesd128_mask_round,
3107 __builtin_ia32_rangess128_mask_round): New builtins.
3108 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3109 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3110 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3111 "<round_saeonly_constraint>")): Changed to ...
3112 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3113 "<round_saeonly_scalar_constraint>")): ... this.
3114 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3115 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3116 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3117 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3118 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3120 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3122 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3124 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3125 Remove all values except native, 8540 and 8548.
3127 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3129 * config/i386/i386.c (ix86_output_function_return): Pass
3130 INVALID_REGNUM, instead of -1, as invalid register number to
3131 indirect_thunk_name and output_indirect_thunk.
3133 2018-02-02 Julia Koval <julia.koval@intel.com>
3135 * config.gcc: Add -march=icelake.
3136 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3137 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3138 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3139 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3140 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3141 (processor_target_table): Add icelake.
3142 (ix86_option_override_internal): Handle new PTAs.
3143 (get_builtin_code_for_version): Handle icelake.
3144 (M_INTEL_COREI7_ICELAKE): New.
3145 (fold_builtin_cpu): Handle icelake.
3146 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3147 * doc/invoke.texi: Add -march=icelake.
3149 2018-02-02 Julia Koval <julia.koval@intel.com>
3151 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3152 to wide_int_bitmask.
3153 * wide-int-bitmask.h: New.
3155 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3158 * config/i386/i386.md: Replace Pmode with word_mode in
3159 builtin_setjmp_setup and builtin_longjmp to support x32.
3161 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3165 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3167 (rs6000_supported_cpu_names): New static variable.
3168 (linux_cpu_translation_table): Likewise.
3169 (elf_platform) <cpu>: Define new static variable and use it.
3170 Translate kernel AT_PLATFORM name to canonical name if needed.
3171 Error if platform name is unknown.
3173 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3176 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3178 2018-02-01 Jeff Law <law@redhat.com>
3181 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3182 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3183 the scratch if RELEASE_VIA_POP is false.
3184 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3185 If we have to save a temporary register, decrement SIZE appropriately.
3186 Pass new arguments to release_scratch_register_on_entry.
3187 (ix86_adjust_stack_and_probe): Likewise.
3188 (ix86_emit_probe_stack_range): Pass new arguments to
3189 release_scratch_register_on_entry.
3191 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3193 PR rtl-optimization/84157
3194 * combine.c (change_zero_ext): Use REG_P predicate in
3195 front of HARD_REGISTER_P predicate.
3197 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3199 * config/avr/avr.c (avr_option_override): Move disabling of
3200 -fdelete-null-pointer-checks to...
3201 * common/config/avr/avr-common.c (avr_option_optimization_table):
3204 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3206 PR tree-optimization/81635
3207 * tree-data-ref.c (split_constant_offset_1): For types that
3208 wrap on overflow, try to use range info to prove that wrapping
3211 2018-02-01 Renlin Li <renlin.li@arm.com>
3214 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3216 (aarch64_register_move_cost): Likewise.
3217 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3219 (REG_CLASS_NAMES): Likewise.
3220 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3221 TAILCALL_ADDR_REGS. Remove IP registers.
3222 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3224 2018-02-01 Richard Biener <rguenther@suse.de>
3226 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3227 for specifying RPO order and allow NULL for that.
3228 * domwalk.c (dom_walker::dom_walker): Likewise.
3229 (dom_walker::walk): Handle NULL RPO order.
3230 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3232 (rewrite_update_dom_walker): Likewise.
3233 (mark_def_dom_walker): Likewise.
3235 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3237 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3238 (aarch64_maybe_expand_sve_subreg_move): Declare.
3239 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3240 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3242 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3243 that are semantically a reverse operation.
3244 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3245 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3246 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3248 (aarch64_can_change_mode_class): For big-endian, forbid changes
3249 between two SVE modes if they have different element sizes.
3251 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3253 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3254 the TImode handling for big-endian targets.
3256 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3258 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3259 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3261 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3262 Remove BSWAP handing for big-endian targets and use the form of
3263 LD1RQ appropariate for the mode.
3265 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3267 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3268 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3271 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3274 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3275 check for operands that need to go through aarch64_sve_reload_be.
3277 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3279 PR tree-optimization/81661
3280 PR tree-optimization/84117
3281 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3282 * tree-eh.c: Include gimplify.h.
3283 (find_trapping_overflow, replace_trapping_overflow,
3284 rewrite_to_non_trapping_overflow): New functions.
3285 * tree-vect-loop.c: Include tree-eh.h.
3286 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3287 * tree-data-ref.c: Include tree-eh.h.
3288 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3290 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3292 PR rtl-optimization/84123
3293 * combine.c (change_zero_ext): Check if hard register satisfies
3294 can_change_dest_mode before calling gen_lowpart_SUBREG.
3296 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3299 * ira.c (ira_init_register_move_cost): Remove assert.
3301 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3303 PR rtl-optimization/84071
3304 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3305 * doc/tm.texi: Regenerate.
3307 2018-01-31 Richard Biener <rguenther@suse.de>
3309 PR tree-optimization/84132
3310 * tree-data-ref.c (analyze_miv_subscript): Properly
3311 check whether evolution_function_is_affine_multivariate_p
3312 before calling gcd_of_steps_may_divide_p.
3314 2018-01-31 Julia Koval <julia.koval@intel.com>
3317 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3318 * config/i386/i386.md (rdpid_rex64) New.
3319 (rdpid): Make 32bit only.
3321 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3324 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3325 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3327 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3330 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3332 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3334 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3336 PR rtl-optimization/84071
3337 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3338 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3340 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3342 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3343 (arc_attribute_table): Add 'aux' attribute.
3344 (arc_in_small_data_p): Consider aux like variables.
3345 (arc_is_aux_reg_p): New function.
3346 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3347 (arc_get_aux_arg): New function.
3348 (prepare_move_operands): Handle aux-register access.
3349 (arc_handle_aux_attribute): New function.
3350 * doc/extend.texi (ARC Variable attributes): Add subsection.
3352 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3354 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3355 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3356 (arc_attribute_table): Add 'uncached' attribute.
3357 (arc_print_operand): Print '.di' flag for uncached memory
3359 (arc_in_small_data_p): Do not consider for small data the uncached
3361 (arc_is_uncached_mem_p): New function.
3362 * config/arc/predicates.md (compact_store_memory_operand): Check
3363 for uncached memory accesses.
3364 (nonvol_nonimm_operand): Likewise.
3365 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3367 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3370 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3371 falign-loops=): Add Optimization flag.
3373 2018-01-30 Jeff Law <law@redhat.com>
3376 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3377 INT_REGISTERS_SAVED. Check it prior to calling
3378 get_scratch_register_on_entry.
3379 (ix86_adjust_stack_and_probe): Similarly.
3380 (ix86_emit_probe_stack_range): Similarly.
3381 (ix86_expand_prologue): Corresponding changes.
3383 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3386 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3387 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3389 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3392 * lra-constraints.c (curr_insn_transform): Process AND in the
3395 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3397 PR rtl-optimization/83986
3398 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3399 dependence against last_pending_memory_flush in addition to
3402 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3404 PR tree-optimization/81611
3405 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3408 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3411 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3414 2018-01-30 Richard Biener <rguenther@suse.de>
3415 Jakub Jelinek <jakub@redhat.com>
3417 PR tree-optimization/84111
3418 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3419 inner loops added during recursion, as they don't have up-to-date
3422 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3425 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3426 (can_inline_edge_by_limits_p): ... here.
3427 (can_early_inline_edge_p, check_callers,
3428 update_caller_keys, update_callee_keys, recursive_inlining,
3429 add_new_edges_to_heap, speculation_useful_p,
3430 inline_small_functions,
3431 inline_small_functions, flatten_function,
3432 inline_to_all_callers_1): Update.
3434 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3436 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3439 2018-01-30 Richard Biener <rguenther@suse.de>
3441 PR tree-optimization/83008
3442 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3443 invariant and constant vector uses in stmts when they need
3446 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3449 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3450 * configure: Regenerate.
3452 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3454 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3456 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3457 Use gen_rtx_REG rather than gen_lowpart.
3459 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3461 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3462 rather than 0 when creating partial subregs.
3464 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3466 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3469 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3472 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3473 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3474 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3475 flags. This restores the settings used before the 2017-07-24.
3476 Turning off pre increment/decrement/modify allows IVOPTS to
3477 optimize DF/SF loops where the index is an int.
3479 2018-01-29 Richard Biener <rguenther@suse.de>
3480 Kelvin Nilsen <kelvin@gcc.gnu.org>
3483 * tree-vect-stmts.c (vectorizable_call): Don't call
3484 targetm.vectorize_builtin_md_vectorized_function if callee is
3487 2018-01-22 Carl Love <cel@us.ibm.com>
3489 * doc/extend.tex: Fix typo in second arg in
3490 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3492 2018-01-29 Richard Biener <rguenther@suse.de>
3494 PR tree-optimization/84086
3495 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3496 (flush_ssaname_freelist): When SSA names were released reset
3497 the SCEV hash table.
3499 2018-01-29 Richard Biener <rguenther@suse.de>
3501 PR tree-optimization/84057
3502 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3503 removed paths when removing edges.
3505 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3507 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3508 -mfunction-return=@var{choice}.
3510 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3513 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3515 (layout::print_source_line): Likewise.
3516 (test_get_line_width_without_trailing_whitespace): Add test cases.
3518 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3521 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3524 2018-01-26 Jim Wilson <jimw@sifive.com>
3526 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3528 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3531 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3533 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3534 and CMP + SUB-immediate -> SUBS.
3536 2018-01-26 Martin Sebor <msebor@redhat.com>
3538 PR tree-optimization/83896
3539 * tree-ssa-strlen.c (get_string_len): Rename...
3540 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3541 Avoid assuming length is constant.
3542 (handle_char_store): Use HOST_WIDE_INT for string length.
3544 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3547 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3548 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3550 2018-01-26 Richard Biener <rguenther@suse.de>
3552 PR rtl-optimization/84003
3553 * dse.c (record_store): Only record redundant stores when
3554 the earlier store aliases at least all accesses the later one does.
3556 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3558 PR rtl-optimization/83985
3559 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3560 REG_CFA_RESTORE insns.
3561 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3562 REG_CFA_RESTORE insns here.
3565 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3566 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3568 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3570 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3571 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3572 (arc_init): Likewise.
3573 (arc_override_options): Likewise.
3574 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3576 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3578 * config/arc/arc.h (TARGET_DBNZ): Define.
3579 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3580 properly set the tune attribute.
3581 (dbnz): Use TARGET_DBNZ guard.
3582 * config/arc/arc.opt (mtune): Add core3 option.
3584 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3586 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3587 recognize new pic like addresses.
3588 (arc_delegitimize_address): Clean up.
3590 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3592 * config/arc/arc-arches.def: Option mrf16 valid for all
3594 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3595 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3596 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3597 * config/arc/arc-tables.opt: Regenerate.
3598 * config/arc/arc.c (arc_conditional_register_usage): Handle
3599 reduced register file case.
3600 (arc_file_start): Set must have build attributes.
3601 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3603 * config/arc/arc.opt (mrf16): Add new option.
3604 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3605 * config/arc/genmultilib.awk: Handle new mrf16 option.
3606 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3607 * config/arc/t-multilib: Regenerate.
3608 * doc/invoke.texi (ARC Options): Document mrf16 option.
3610 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3612 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3613 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3614 (arc_attribute_table): Add 'secure_call' attribute.
3615 (arc_print_operand): Print secure call operand.
3616 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3618 (arc_is_secure_call_p): New function. * config/arc/arc.md
3619 (call_i): Add support for sjli instruction.
3620 (call_value_i): Likewise.
3621 * config/arc/constraints.md (Csc): New constraint.
3623 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3624 John Eric Martin <John.Martin@emmicro-us.com>
3626 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3627 * config/arc/arc.c (_arc_jli_section): New struct.
3628 (arc_jli_section): New type.
3629 (rc_jli_sections): New static variable.
3630 (arc_handle_jli_attribute): New function.
3631 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3632 (arc_file_end): New function.
3633 (TARGET_ASM_FILE_END): Define.
3634 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3635 (arc_add_jli_section): New function.
3636 (jli_call_scan): Likewise.
3637 (arc_reorg): Call jli_call_scan.
3638 (arc_output_addsi): Remove 'S' from printing asm operand.
3639 (arc_is_jli_call_p): New function.
3640 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3642 (movhi_insn): Likewise.
3643 (movsi_insn): Likewise.
3644 (movsi_set_cc_insn): Likewise.
3645 (loadqi_update): Likewise.
3646 (load_zeroextendqisi_update): Likewise.
3647 (load_signextendqisi_update): Likewise.
3648 (loadhi_update): Likewise.
3649 (load_zeroextendhisi_update): Likewise.
3650 (load_signextendhisi_update): Likewise.
3651 (loadsi_update): Likewise.
3652 (loadsf_update): Likewise.
3653 (movsicc_insn): Likewise.
3654 (bset_insn): Likewise.
3655 (bxor_insn): Likewise.
3656 (bclr_insn): Likewise.
3657 (bmsk_insn): Likewise.
3658 (bicsi3_insn): Likewise.
3659 (cmpsi_cc_c_insn): Likewise.
3660 (movsi_ne): Likewise.
3661 (movsi_cond_exec): Likewise.
3662 (clrsbsi2): Likewise.
3673 (call_i): Remove 'S' asm letter, add jli instruction.
3674 (call_value_i): Likewise.
3675 * config/arc/arc.op (mjli-always): New option.
3676 * config/arc/constraints.md (Cji): New constraint.
3677 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3679 (subsf3_fpx): Likewise.
3680 (mulsf3_fpx): Likewise.
3681 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3683 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3685 * doc/invoke.texi (ARC): Document mjli-always option.
3687 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3689 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3690 avoid addition with 0 and use incw and decw where possible.
3692 2018-01-26 Richard Biener <rguenther@suse.de>
3694 PR tree-optimization/81082
3695 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3696 association if it requires casting to unsigned.
3697 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3698 from fold_plusminus_mult_expr to catch important cases late when
3699 range info is available.
3701 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3703 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3704 * configure.ac (hidden_linkonce): New test.
3705 * configure: Regenerate.
3706 * config.in: Regenerate.
3708 2018-01-26 Julia Koval <julia.koval@intel.com>
3710 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3711 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3712 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3713 _mm_mask_bitshuffle_epi64_mask): Fix type.
3714 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3715 USI_FTYPE_V4DI_V4DI_USI): Remove.
3716 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3717 __builtin_ia32_vpshufbitqmb256_mask,
3718 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3719 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3720 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3722 2018-01-26 Alan Modra <amodra@gmail.com>
3725 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3726 UNSPEC_VBPERMQ. Sort other unspecs.
3728 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3730 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3732 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3735 * predict.c (drop_profile): Do not push/pop cfun; update also
3737 (handle_missing_profiles): Fix logic looking for zero profiles.
3739 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3742 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3743 on functions with #pragma omp declare simd or functions with simd
3745 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3746 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3747 Remove trailing \n from warning_at calls.
3749 2018-01-25 Tom de Vries <tom@codesourcery.com>
3752 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3753 for neutered workers.
3755 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3758 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3759 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3761 2018-01-24 Jeff Law <law@redhat.com>
3764 * i386.c (get_probe_interval): Move to earlier point.
3765 (ix86_compute_frame_layout): If -fstack-clash-protection and
3766 the frame is larger than the probe interval, then use pushes
3767 to save registers rather than reg->mem moves.
3768 (ix86_expand_prologue): Remove conditional for int_registers_saved
3771 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3774 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3775 min/max for never referenced object.
3777 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3780 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3782 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3783 attributes from DECL_ATTRIBUTES (decl) without affecting
3784 DECL_ATTRIBUTES (current_function_decl).
3785 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3786 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3788 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3790 PR tree-optimization/83979
3791 * fold-const.c (fold_comparison): Use constant_boolean_node
3792 instead of boolean_{true,false}_node.
3794 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3796 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3799 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3801 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3802 Simplify the clause that sets the length attribute.
3803 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3804 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3805 clause that sets the length attribute.
3806 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3808 2018-01-24 Tom de Vries <tom@codesourcery.com>
3811 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3812 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3813 Add strict parameter.
3814 (prevent_branch_around_nothing): Insert dummy insn between branch to
3815 label and label with no ptx insn inbetween.
3816 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3818 2018-01-24 Tom de Vries <tom@codesourcery.com>
3821 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3822 for neutered threads in warp.
3823 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3825 2018-01-24 Richard Biener <rguenther@suse.de>
3827 PR tree-optimization/83176
3828 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3831 2018-01-24 Richard Biener <rguenther@suse.de>
3833 PR tree-optimization/82819
3834 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3835 code generating pluses that are no-ops in the target precision.
3837 2018-01-24 Richard Biener <rguenther@suse.de>
3840 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3842 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3844 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3845 to merge probabilities.
3846 * predict.c (probably_never_executed): Also mark as cold functions
3847 with global 0 profile and guessed local profile.
3848 * profile-count.c (profile_probability::combine_with_count): New
3850 * profile-count.h (profile_probability::operator*,
3851 profile_probability::operator*=, profile_probability::operator/,
3852 profile_probability::operator/=): Reduce precision to adjusted
3853 and set value to guessed on contradictory divisions.
3854 (profile_probability::combine_with_freq): Remove.
3855 (profile_probability::combine_wiht_count): Declare.
3856 (profile_count::force_nonzero):: Set to adjusted.
3857 (profile_count::probability_in):: Set quality to adjusted.
3858 * tree-ssa-tail-merge.c (replace_block_by): Use
3861 2018-01-23 Andrew Waterman <andrew@sifive.com>
3862 Jim Wilson <jimw@sifive.com>
3864 * config/riscv/riscv.c (riscv_stack_boundary): New.
3865 (riscv_option_override): Set riscv_stack_boundary. Handle
3866 riscv_preferred_stack_boundary_arg.
3867 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3868 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3869 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3870 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3871 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3872 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3874 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3877 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3878 of struct ix86_frame.
3879 (ix86_expand_epilogue): Likewise. Add a local variable for
3880 the reg_save_offset field in struct ix86_frame.
3882 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3884 PR tree-optimization/82604
3885 * tree-loop-distribution.c (enum partition_kind): New enum item
3886 PKIND_PARTIAL_MEMSET.
3887 (partition_builtin_p): Support above new enum item.
3888 (generate_code_for_partition): Ditto.
3889 (compute_access_range): Differentiate cases that equality can be
3890 proven at all loops, the innermost loops or no loops.
3891 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3892 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3893 (finalize_partitions, distribute_loop): Don't fuse partition of
3894 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3895 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3898 2018-01-23 Martin Liska <mliska@suse.cz>
3900 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3901 order to ignore the predictor.
3902 (PRED_POLYMORPHIC_CALL): Likewise.
3903 (PRED_RECURSIVE_CALL): Likewise.
3905 2018-01-23 Martin Liska <mliska@suse.cz>
3907 * tree-profile.c (tree_profiling): Print function header to
3908 aware reader which function we are working on.
3909 * value-prof.c (gimple_find_values_to_profile): Do not print
3910 not interesting value histograms.
3912 2018-01-23 Martin Liska <mliska@suse.cz>
3914 * profile-count.h (enum profile_quality): Add
3915 profile_uninitialized as the first value. Do not number values
3916 as they are zero based.
3917 (profile_count::verify): Update sanity check.
3918 (profile_probability::verify): Likewise.
3920 2018-01-23 Nathan Sidwell <nathan@acm.org>
3922 * doc/invoke.texi (ffor-scope): Deprecate.
3924 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3926 PR tree-optimization/83510
3927 * domwalk.c (set_all_edges_as_executable): New function.
3928 (dom_walker::dom_walker): Convert bool param
3929 "skip_unreachable_blocks" to enum reachability. Move setup of
3930 edge flags to set_all_edges_as_executable and only do it when
3931 reachability is REACHABLE_BLOCKS.
3932 * domwalk.h (enum dom_walker::reachability): New enum.
3933 (dom_walker::dom_walker): Convert bool param
3934 "skip_unreachable_blocks" to enum reachability.
3935 (set_all_edges_as_executable): New decl.
3936 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3937 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3939 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3940 but converting true to REACHABLE_BLOCKS.
3941 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3943 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3944 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3945 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3947 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3948 if check_all_array_refs will be called.
3950 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3952 * tree.c (selftest::test_location_wrappers): Add more test
3955 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3957 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3958 (selftest::test_bit_in_range): Likewise.
3960 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3963 * doc/sourcebuild.texi (vect_float): Say that the selector
3964 only describes the situation when -funsafe-math-optimizations is on.
3965 (vect_float_strict): Document.
3967 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3969 PR tree-optimization/83965
3970 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3971 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3972 instead of checking only for a reduction.
3973 (vect_recog_widen_sum_pattern): Likewise.
3975 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3977 * predict.c (probably_never_executed): Only use precise profile info.
3978 (compute_function_frequency): Skip after inlining hack since we now
3979 have quality checking.
3981 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3983 * profile-count.h (profile_probability::very_unlikely,
3984 profile_probability::unlikely, profile_probability::even): Set
3985 precision to guessed.
3987 2018-01-23 Richard Biener <rguenther@suse.de>
3989 PR tree-optimization/83963
3990 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3991 Properly terminate dominator walk when crossing the exit edge not
3992 when visiting its source block.
3994 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3997 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3998 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4000 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4002 PR tree-optimization/83957
4003 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4004 semicolon after for body surrounded by braces.
4006 PR tree-optimization/83081
4007 * profile-count.h (profile_probability::split): New method.
4008 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4009 Use profile_probability::split.
4010 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4011 when splitting a single conditional jump into 2.
4013 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4015 PR tree-optimization/69452
4016 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4019 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4021 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4022 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4023 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4025 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4027 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4028 * config/rl78/rl78.md (movdi): New define_expand.
4029 * config/rl78/rl78.c (rl78_split_movdi): New function.
4031 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4034 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4036 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4037 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4038 128-bit to produce an UNSPEC move to get the double word with the
4039 signbit and then a shift directly to do signbit.
4040 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4041 implementation with a new version that just does either a direct
4042 move or a regular move. Move memory interface to separate insns.
4043 Move insns so they are next to the expander.
4044 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4045 with signbit move. Split big and little endian case.
4046 (signbit<mode>2_dm_mem_le): Likewise.
4047 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4048 (signbit<mode>2_dm2): Likewise.
4050 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4052 * config/rl78/rl78.md (anddi3): New define_expand.
4054 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4056 * config/rl78/rl78.md (umindi3): New define_expand.
4058 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4060 * config/rl78/rl78.md (smindi3): New define_expand.
4062 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4064 * config/rl78/rl78.md (smaxdi3): New define_expand.
4066 2018-01-22 Carl Love <cel@us.ibm.com>
4068 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4069 LVX_V1TI): Add macro expansion.
4070 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4071 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4072 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4073 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4074 Change check to determine if the instruction is a byte reversing
4075 entry. Fix typo in comment.
4076 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4077 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4078 Add def_builtin calls for new builtins.
4079 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4080 Add define_insn expansion.
4082 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4084 * config/rl78/rl78.md (umaxdi3): New define_expand.
4086 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4088 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4089 for non-QImode registers.
4091 2018-01-22 Richard Biener <rguenther@suse.de>
4093 PR tree-optimization/83963
4094 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4095 including the loop exit block.
4096 (scop_detection::merge_sese): Likewise.
4097 (scop_detection::add_scop): Do it here instead.
4099 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4101 * doc/sourcebuild.texi (arm_softfloat): Document.
4103 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4106 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4107 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4108 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4110 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4111 David Edelsohn <dje.gcc@gmail.com>
4114 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4115 Change "crset eq" to "crset 2".
4116 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4117 (*call_indirect_aix<mode>_nospec): Likewise.
4118 (*call_value_indirect_aix<mode>_nospec): Likewise.
4119 (*call_indirect_elfv2<mode>_nospec): Likewise.
4120 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4121 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4122 change assembly output from . to $.
4123 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4124 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4125 (*tablejump<mode>_internal1_nospec): Likewise.
4127 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4130 * config/sh/sh_optimize_sett_clrt.cc:
4131 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4133 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4135 PR tree-optimization/83940
4136 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4137 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4138 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4139 use it to pass back the definition type.
4140 (vect_check_store_rhs): Likewise rhs_dt_out.
4141 (vect_build_gather_load_calls): Add a mask_dt argument and use
4142 it instead of a call to vect_is_simple_use.
4143 (vectorizable_store): Update calls to vect_check_load_store_mask
4144 and vect_check_store_rhs. Use the dt returned by the latter instead
4145 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4146 instead of calls to vect_is_simple_use. Pass the scalar rather
4147 than the vector operand to vect_is_simple_use when handling
4148 second and subsequent copies of an rhs value.
4149 (vectorizable_load): Update calls to vect_check_load_store_mask
4150 and vect_build_gather_load_calls. Use the cached mask_dt and
4151 gs_info.offset_dt instead of calls to vect_is_simple_use.
4153 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4156 * tree-emutls.c: Include gimplify.h.
4157 (lower_emutls_2): New function.
4158 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4159 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4160 it before further processing.
4163 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4164 UINTVAL (trueop1) instead of INTVAL (op1).
4166 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4170 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4171 INCOMING_FRAME_SP_OFFSET if not defined.
4172 (scan_trace): Add ENTRY argument. If true and
4173 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4174 emit a note to adjust the CFA offset.
4175 (create_cfi_notes): Adjust scan_trace callers.
4176 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4177 INCOMING_FRAME_SP_OFFSET in the CIE.
4178 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4179 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4181 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4182 * doc/tm.texi: Regenerated.
4184 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4186 PR rtl-optimization/83147
4187 * lra-constraints.c (remove_inheritance_pseudos): Use
4188 lra_substitute_pseudo_within_insn.
4190 2018-01-19 Tom de Vries <tom@codesourcery.com>
4191 Cesar Philippidis <cesar@codesourcery.com>
4194 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4196 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4199 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4200 spaces for function labels.
4202 2018-01-19 Martin Liska <mliska@suse.cz>
4204 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4205 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4206 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4207 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4208 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4209 (PRED_CONST_RETURN): Change from 69 to 65.
4210 (PRED_NULL_RETURN): Change from 91 to 71.
4211 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4212 (PRED_LOOP_GUARD): Change from 66 to 73.
4214 2018-01-19 Martin Liska <mliska@suse.cz>
4216 * predict.c (predict_insn_def): Add new assert.
4217 (struct branch_predictor): Change type to signed integer.
4218 (test_prediction_value_range): Amend test to cover
4220 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4221 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4222 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4223 (PRED_LOOP_IV_COMPARE): Likewise.
4224 * predict.h (PROB_UNINITIALIZED): Define new constant.
4226 2018-01-19 Martin Liska <mliska@suse.cz>
4228 * predict.c (dump_prediction): Add new format for
4229 analyze_brprob.py script which is enabled with -details
4231 * profile-count.h (precise_p): New function.
4233 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4235 PR tree-optimization/83922
4236 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4237 there are no statements that need masking.
4238 (vect_active_double_reduction_p): New function.
4239 (vect_analyze_loop_operations): Use it when handling phis that
4240 are not in the loop header.
4242 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4244 PR tree-optimization/83914
4245 * tree-vect-loop.c (vectorizable_induction): Don't convert
4246 init_expr or apply the peeling adjustment for inductions
4247 that are nested within the vectorized loop.
4249 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4251 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4254 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4258 * function.h (gimplify_parameters): Add gimple_seq * argument.
4259 * function.c: Include gimple.h and options.h.
4260 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4261 for the added local temporaries if needed.
4262 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4263 if there are any parameter cleanups, wrap whole body into a
4264 try/finally with the cleanups.
4266 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4269 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4270 Use GET_MODE_CLASS for scalar floating point.
4272 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4276 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4277 Fix call of call_cgraph_insertion_hooks.
4279 2018-01-18 Martin Sebor <msebor@redhat.com>
4281 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4283 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4286 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4289 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4292 * common.opt: (-ffile-prefix-map): New option.
4293 * opts.c (common_handle_option): Defer it.
4294 * opts-global.c (handle_common_deferred_options): Handle it.
4295 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4296 * file-prefix-map.h: New file.
4297 (remap_debug_filename, add_debug_prefix_map): ...here.
4298 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4299 * final.c (debug_prefix_map, add_debug_prefix_map
4300 remap_debug_filename): Move to...
4301 * file-prefix-map.c: New file.
4302 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4303 generalize, get rid of alloca(), use strrchr() instead of strchr().
4304 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4305 Implement in terms of add_prefix_map().
4306 (remap_macro_filename, remap_debug_filename): Implement in term of
4308 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4309 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4310 * dbxout.c: Include file-prefix-map.h.
4311 * varasm.c: Likewise.
4312 * vmsdbgout.c: Likewise.
4313 * xcoffout.c: Likewise.
4314 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4315 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4316 * doc/invoke.texi (-ffile-prefix-map): Document.
4317 (-fdebug-prefix-map): Update description.
4319 2018-01-18 Martin Liska <mliska@suse.cz>
4321 * config/i386/i386.c (indirect_thunk_name): Document that also
4323 (output_indirect_thunk): Document why both instructions
4324 (pause and lfence) are generated.
4326 2018-01-18 Richard Biener <rguenther@suse.de>
4328 PR tree-optimization/83887
4329 * graphite-scop-detection.c
4330 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4331 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4332 (scop_detection::merge_sese): Re-implement with a flood-fill
4333 algorithm that properly finds a SESE region if it exists.
4335 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4338 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4339 pointer_diff optimizations use view_convert instead of convert.
4341 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4343 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4344 Generate different code for -mno-speculate-indirect-jumps.
4345 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4346 (*call_indirect_aix<mode>): Disable for
4347 -mno-speculate-indirect-jumps.
4348 (*call_indirect_aix<mode>_nospec): New define_insn.
4349 (*call_value_indirect_aix<mode>): Disable for
4350 -mno-speculate-indirect-jumps.
4351 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4352 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4353 -mno-speculate-indirect-jumps.
4354 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4356 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4358 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4359 long double type, set the flags for noting the default long double
4360 type, even if we don't pass or return a long double type.
4362 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4365 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4368 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4370 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4371 support for merge[hl].
4372 (fold_mergehl_helper): New helper function.
4373 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4374 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4375 (altivec_vmrglw_direct): Add xxmrglw insn.
4377 2018-01-17 Andrew Waterman <andrew@sifive.com>
4379 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4380 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4382 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4385 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4386 call the lto_location_cache before reading the
4387 DECL_SOURCE_LOCATION of the types.
4389 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4390 Richard Sandiford <richard.sandiford@linaro.org>
4392 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4393 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4394 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4395 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4396 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4398 * config/aarch64/constraints.md (aarch64_movti_operand):
4400 * config/aarch64/predicates.md (Uti): Add new constraint.
4402 2018-01-17 Carl Love <cel@us.ibm.com>
4404 * config/rs6000/vsx.md (define_expand xl_len_r,
4405 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4406 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4408 (define_expand, define_insn): Move the shift left from the
4409 define_insn to the define_expand for lxvl and stxvl instructions.
4410 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4411 and XL_LEN_R definitions to PURE.
4413 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4415 * config/i386/i386.c (indirect_thunk_name): Declare regno
4416 as unsigned int. Compare regno with INVALID_REGNUM.
4417 (output_indirect_thunk): Ditto.
4418 (output_indirect_thunk_function): Ditto.
4419 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4420 in the call to output_indirect_thunk_function.
4422 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4425 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4426 rather than the size of inner_type to determine the stack slot size
4427 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4429 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4432 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4435 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4437 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4438 endian Linux systems to optionally enable multilibs for selecting
4439 the long double type if the user configured an explicit type.
4440 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4441 have no long double multilibs if not defined.
4442 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4443 warn if the user used -mabi={ieee,ibm}longdouble and we built
4444 multilibs for long double.
4445 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4446 appropriate multilib option.
4447 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4449 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4450 for building long double multilibs.
4451 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4453 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4455 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4458 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4460 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4463 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4466 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4469 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4471 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4472 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4474 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4476 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4477 different rtl trees depending on TARGET_64BIT.
4478 (rs6000_gen_lvx): Likewise.
4480 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4482 * config/visium/visium.md (nop): Tweak comment.
4483 (hazard_nop): Likewise.
4485 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4487 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4488 -mspeculate-indirect-jumps.
4489 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4490 for -mno-speculate-indirect-jumps.
4491 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4492 (*call_value_indirect_elfv2<mode>): Disable for
4493 -mno-speculate-indirect-jumps.
4494 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4495 (indirect_jump): Emit different RTL for
4496 -mno-speculate-indirect-jumps.
4497 (*indirect_jump<mode>): Disable for
4498 -mno-speculate-indirect-jumps.
4499 (*indirect_jump<mode>_nospec): New define_insn.
4500 (tablejump): Emit different RTL for
4501 -mno-speculate-indirect-jumps.
4502 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4503 (tablejumpsi_nospec): New define_expand.
4504 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4505 (tablejumpdi_nospec): New define_expand.
4506 (*tablejump<mode>_internal1): Disable for
4507 -mno-speculate-indirect-jumps.
4508 (*tablejump<mode>_internal1_nospec): New define_insn.
4509 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4512 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4514 * caller-save.c (insert_save): Drop unnecessary parameter. All
4517 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4518 Richard Biener <rguenth@suse.de>
4521 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4522 return early, inline manually is_gimple_sizepos. Make sure if we
4523 call gimplify_expr we don't end up with a gimple constant.
4524 * tree.c (variably_modified_type_p): Don't return true for
4525 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4526 * gimplify.h (is_gimple_sizepos): Remove.
4528 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4530 PR tree-optimization/83857
4531 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4532 vectorizable_live_operation for pure SLP statements.
4533 (vectorizable_live_operation): Handle PHIs.
4535 2018-01-16 Richard Biener <rguenther@suse.de>
4537 PR tree-optimization/83867
4538 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4539 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4541 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4544 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4545 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4546 If off is not INTEGER_CST, issue a may not be aligned warning
4547 rather than isn't aligned. Use isn%'t rather than isn't.
4548 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4550 <case MULT_EXPR>: Improve the case when bottom and one of the
4551 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4552 operand, in that case check if the other operand is multiple of
4553 bottom divided by the INTEGER_CST operand.
4555 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4558 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4559 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4560 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4561 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4562 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4563 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4564 (pa_function_arg_size): New function.
4566 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4568 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4569 in a separate statement.
4571 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4573 PR tree-optimization/83847
4574 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4575 group gathers and scatters.
4577 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4579 PR rtl-optimization/86620
4580 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4582 PR rtl-optimization/83213
4583 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4584 to last if both are JUMP_INSNs.
4586 PR tree-optimization/83843
4587 * gimple-ssa-store-merging.c
4588 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4589 store_immediate_info for bswap/nop orig_stores.
4591 2018-01-15 Andrew Waterman <andrew@sifive.com>
4593 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4595 <UDIV>: Increase cost if !TARGET_DIV.
4597 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4599 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4600 (define_attr "cr_logical_3op"): New.
4601 (cceq_ior_compare): Adjust.
4602 (cceq_ior_compare_complement): Adjust.
4603 (*cceq_rev_compare): Adjust.
4604 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4605 (is_cracked_insn): Adjust.
4606 (insn_must_be_first_in_group): Adjust.
4607 * config/rs6000/40x.md: Adjust.
4608 * config/rs6000/440.md: Adjust.
4609 * config/rs6000/476.md: Adjust.
4610 * config/rs6000/601.md: Adjust.
4611 * config/rs6000/603.md: Adjust.
4612 * config/rs6000/6xx.md: Adjust.
4613 * config/rs6000/7450.md: Adjust.
4614 * config/rs6000/7xx.md: Adjust.
4615 * config/rs6000/8540.md: Adjust.
4616 * config/rs6000/cell.md: Adjust.
4617 * config/rs6000/e300c2c3.md: Adjust.
4618 * config/rs6000/e500mc.md: Adjust.
4619 * config/rs6000/e500mc64.md: Adjust.
4620 * config/rs6000/e5500.md: Adjust.
4621 * config/rs6000/e6500.md: Adjust.
4622 * config/rs6000/mpc.md: Adjust.
4623 * config/rs6000/power4.md: Adjust.
4624 * config/rs6000/power5.md: Adjust.
4625 * config/rs6000/power6.md: Adjust.
4626 * config/rs6000/power7.md: Adjust.
4627 * config/rs6000/power8.md: Adjust.
4628 * config/rs6000/power9.md: Adjust.
4629 * config/rs6000/rs64.md: Adjust.
4630 * config/rs6000/titan.md: Adjust.
4632 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4634 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4635 ix86_indirect_branch_register logic.
4637 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4639 * config/i386/constraints.md (Bs): Update
4640 ix86_indirect_branch_register check. Don't check
4641 ix86_indirect_branch_register with GOT_memory_operand.
4643 * config/i386/predicates.md (GOT_memory_operand): Don't check
4644 ix86_indirect_branch_register here.
4645 (GOT32_symbol_operand): Likewise.
4647 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4649 * config/i386/predicates.md (constant_call_address_operand):
4650 Rewrite ix86_indirect_branch_register logic.
4651 (sibcall_insn_operand): Likewise.
4653 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4655 * config/i386/constraints.md (Bs): Replace
4656 ix86_indirect_branch_thunk_register with
4657 ix86_indirect_branch_register.
4659 * config/i386/i386.md (indirect_jump): Likewise.
4660 (tablejump): Likewise.
4661 (*sibcall_memory): Likewise.
4662 (*sibcall_value_memory): Likewise.
4663 Peepholes of indirect call and jump via memory: Likewise.
4664 * config/i386/i386.opt: Likewise.
4665 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4666 (GOT_memory_operand): Likewise.
4667 (call_insn_operand): Likewise.
4668 (sibcall_insn_operand): Likewise.
4669 (GOT32_symbol_operand): Likewise.
4671 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4674 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4675 type rather than type addr's type points to.
4676 (expand_omp_atomic_mutex): Likewise.
4677 (expand_omp_atomic): Likewise.
4679 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4682 * config/i386/i386.c (output_indirect_thunk_function): Use
4683 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4684 for __x86_return_thunk.
4686 2018-01-15 Richard Biener <rguenther@suse.de>
4689 * expmed.c (extract_bit_field_1): Fix typo.
4691 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4694 * config/arm/iterators.md (VF): New mode iterator.
4695 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4696 Remove integer-related logic from pattern.
4697 (neon_vabd<mode>_3): Likewise.
4699 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4702 * common.opt (fstrict-overflow): No longer an alias.
4703 (fwrapv-pointer): New option.
4704 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4705 also for pointer types based on flag_wrapv_pointer.
4706 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4707 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4708 opts->x_flag_wrapv got set.
4709 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4710 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4711 POINTER_TYPE_OVERFLOW_UNDEFINED.
4712 * match.pd: Likewise in address comparison pattern.
4713 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4715 2018-01-15 Richard Biener <rguenther@suse.de>
4718 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4719 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4720 Reset type names to their identifier if their TYPE_DECL doesn't
4721 have linkage (and thus is used for ODR and devirt).
4722 (save_debug_info_for_decl): Remove.
4723 (save_debug_info_for_type): Likewise.
4724 (add_tree_to_fld_list): Adjust.
4725 * tree-pretty-print.c (dump_generic_node): Make dumping of
4726 type names more robust.
4728 2018-01-15 Richard Biener <rguenther@suse.de>
4730 * BASE-VER: Bump to 8.0.1.
4732 2018-01-14 Martin Sebor <msebor@redhat.com>
4735 * builtins.c (check_access): Avoid warning when the no-warning bit
4738 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4740 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4741 * ira-color (allocno_hard_regs_compare): Likewise.
4743 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4746 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4747 Use .pushsection/.popsection.
4749 2018-01-14 Martin Sebor <msebor@redhat.com>
4752 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4754 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4756 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4757 entry from extra_headers.
4758 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4759 extra_headers, make the list bitwise identical to the i?86-*-* one.
4761 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4763 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4764 -mcmodel=large with -mindirect-branch=thunk,
4765 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4766 -mfunction-return=thunk-extern.
4767 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4768 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4769 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4771 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4773 * config/i386/i386.c (print_reg): Print the name of the full
4774 integer register without '%'.
4775 (ix86_print_operand): Handle 'V'.
4776 * doc/extend.texi: Document 'V' modifier.
4778 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4780 * config/i386/constraints.md (Bs): Disallow memory operand for
4781 -mindirect-branch-register.
4783 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4784 (GOT_memory_operand): Likewise.
4785 (call_insn_operand): Likewise.
4786 (sibcall_insn_operand): Likewise.
4787 (GOT32_symbol_operand): Likewise.
4788 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4789 for -mindirect-branch-register.
4790 (tablejump): Likewise.
4791 (*sibcall_memory): Likewise.
4792 (*sibcall_value_memory): Likewise.
4793 Disallow peepholes of indirect call and jump via memory for
4794 -mindirect-branch-register.
4795 (*call_pop): Replace m with Bw.
4796 (*call_value_pop): Likewise.
4797 (*sibcall_pop_memory): Replace m with Bs.
4798 * config/i386/i386.opt (mindirect-branch-register): New option.
4799 * doc/invoke.texi: Document -mindirect-branch-register option.
4801 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4803 * config/i386/i386-protos.h (ix86_output_function_return): New.
4804 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4805 set function_return_type.
4806 (indirect_thunk_name): Add ret_p to indicate thunk for function
4808 (output_indirect_thunk_function): Pass false to
4809 indirect_thunk_name.
4810 (ix86_output_indirect_branch_via_reg): Likewise.
4811 (ix86_output_indirect_branch_via_push): Likewise.
4812 (output_indirect_thunk_function): Create alias for function
4813 return thunk if regno < 0.
4814 (ix86_output_function_return): New function.
4815 (ix86_handle_fndecl_attribute): Handle function_return.
4816 (ix86_attribute_table): Add function_return.
4817 * config/i386/i386.h (machine_function): Add
4818 function_return_type.
4819 * config/i386/i386.md (simple_return_internal): Use
4820 ix86_output_function_return.
4821 (simple_return_internal_long): Likewise.
4822 * config/i386/i386.opt (mfunction-return=): New option.
4823 (indirect_branch): Mention -mfunction-return=.
4824 * doc/extend.texi: Document function_return function attribute.
4825 * doc/invoke.texi: Document -mfunction-return= option.
4827 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4829 * config/i386/i386-opts.h (indirect_branch): New.
4830 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4831 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4832 with local indirect jump when converting indirect call and jump.
4833 (ix86_set_indirect_branch_type): New.
4834 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4835 (indirectlabelno): New.
4836 (indirect_thunk_needed): Likewise.
4837 (indirect_thunk_bnd_needed): Likewise.
4838 (indirect_thunks_used): Likewise.
4839 (indirect_thunks_bnd_used): Likewise.
4840 (INDIRECT_LABEL): Likewise.
4841 (indirect_thunk_name): Likewise.
4842 (output_indirect_thunk): Likewise.
4843 (output_indirect_thunk_function): Likewise.
4844 (ix86_output_indirect_branch_via_reg): Likewise.
4845 (ix86_output_indirect_branch_via_push): Likewise.
4846 (ix86_output_indirect_branch): Likewise.
4847 (ix86_output_indirect_jmp): Likewise.
4848 (ix86_code_end): Call output_indirect_thunk_function if needed.
4849 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4851 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4852 (ix86_attribute_table): Add indirect_branch.
4853 * config/i386/i386.h (machine_function): Add indirect_branch_type
4854 and has_local_indirect_jump.
4855 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4857 (tablejump): Likewise.
4858 (*indirect_jump): Use ix86_output_indirect_jmp.
4859 (*tablejump_1): Likewise.
4860 (simple_return_indirect_internal): Likewise.
4861 * config/i386/i386.opt (mindirect-branch=): New option.
4862 (indirect_branch): New.
4865 (thunk-inline): Likewise.
4866 (thunk-extern): Likewise.
4867 * doc/extend.texi: Document indirect_branch function attribute.
4868 * doc/invoke.texi: Document -mindirect-branch= option.
4870 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4873 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4875 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4877 * ipa-inline.c (want_inline_small_function_p): Return false if
4878 inlining has already failed with CIF_FINAL_ERROR.
4879 (update_caller_keys): Call want_inline_small_function_p before
4881 (update_callee_keys): Likewise.
4883 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4885 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4887 (rs6000_quadword_masked_address_p): Likewise.
4888 (quad_aligned_load_p): Likewise.
4889 (quad_aligned_store_p): Likewise.
4890 (const_load_sequence_p): Add comment to describe the outer-most loop.
4891 (mimic_memory_attributes_and_flags): New function.
4892 (rs6000_gen_stvx): Likewise.
4893 (replace_swapped_aligned_store): Likewise.
4894 (rs6000_gen_lvx): Likewise.
4895 (replace_swapped_aligned_load): Likewise.
4896 (replace_swapped_load_constant): Capitalize argument name in
4897 comment describing this function.
4898 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4899 and stores that access quad-word aligned addresses and replace
4900 with stvx or lvx instructions when appropriate.
4901 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4902 New function prototype.
4903 (rs6000_quadword_masked_address_p): Likewise.
4904 (rs6000_gen_lvx): Likewise.
4905 (rs6000_gen_stvx): Likewise.
4906 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4907 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4908 when memory address is aligned.
4909 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4910 this split to select lvx instruction when memory address is aligned.
4911 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4912 instruction when memory address is aligned.
4913 (*vsx_le_perm_load_v16qi): Likewise.
4914 (four unnamed splitters): Modify to select the stvx instruction
4915 when memory is aligned.
4917 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4919 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4920 which appears in the queue multiple times.
4922 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4923 Alan Hayward <alan.hayward@arm.com>
4924 David Sherwood <david.sherwood@arm.com>
4926 * tree-vectorizer.h (vec_lower_bound): New structure.
4927 (_loop_vec_info): Add check_nonzero and lower_bounds.
4928 (LOOP_VINFO_CHECK_NONZERO): New macro.
4929 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4930 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4931 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4932 fields. Make seg_len the distance travelled, not including the
4934 (dr_direction_indicator): Declare.
4935 (dr_zero_step_indicator): Likewise.
4936 (dr_known_forward_stride_p): Likewise.
4937 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4939 (runtime_alias_check_p): Allow runtime alias checks with
4941 (operator ==): Compare access_size and align.
4942 (prune_runtime_alias_test_list): Rework for new distinction between
4943 the access_size and seg_len.
4944 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4946 (get_segment_min_max): New function.
4947 (create_intersect_range_checks): Use it.
4948 (dr_step_indicator): New function.
4949 (dr_direction_indicator): Likewise.
4950 (dr_zero_step_indicator): Likewise.
4951 (dr_known_forward_stride_p): Likewise.
4952 * tree-loop-distribution.c (data_ref_segment_size): Return
4953 DR_STEP * (niters - 1).
4954 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4956 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4957 (vect_preserves_scalar_order_p): New function, split out from...
4958 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4959 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4960 (vect_vfa_access_size): New function.
4961 (vect_vfa_align): Likewise.
4962 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4963 (dump_lower_bound): New function.
4964 (vect_check_lower_bound): Likewise.
4965 (vect_small_gap_p): Likewise.
4966 (vectorizable_with_step_bound_p): Likewise.
4967 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4968 depencies if the vectorization factor is 1. Convert the checks
4969 for nonzero steps into checks on the bounds of DR_STEP. Try using
4970 a bunds check for variable steps if the minimum required step is
4971 relatively small. Update calls to the dr_with_seg_len
4972 constructor and to vect_compile_time_alias.
4973 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4975 (vect_loop_versioning): Call it.
4976 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4978 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4981 Alan Hayward <alan.hayward@arm.com>
4982 David Sherwood <david.sherwood@arm.com>
4984 * doc/sourcebuild.texi (vect_scatter_store): Document.
4985 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4987 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4989 * genopinit.c (main): Add supports_vec_scatter_store and
4990 supports_vec_scatter_store_cached to target_optabs.
4991 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4992 IFN_MASK_SCATTER_STORE.
4993 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4995 * internal-fn.h (internal_store_fn_p): Declare.
4996 (internal_fn_stored_value_index): Likewise.
4997 * internal-fn.c (scatter_store_direct): New macro.
4998 (expand_scatter_store_optab_fn): New function.
4999 (direct_scatter_store_optab_supported_p): New macro.
5000 (internal_store_fn_p): New function.
5001 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5002 IFN_MASK_SCATTER_STORE.
5003 (internal_fn_mask_index): Likewise.
5004 (internal_fn_stored_value_index): New function.
5005 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5007 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5008 * optabs-query.c (supports_vec_scatter_store_p): New function.
5009 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5010 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5011 true for scatter stores.
5012 (vect_gather_scatter_fn_p): Handle scatter stores too.
5013 (vect_check_gather_scatter): Consider using scatter stores if
5014 supports_vec_scatter_store_p.
5015 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5017 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5018 internal_fn_stored_value_index.
5019 (check_load_store_masking): Handle scatter stores too.
5020 (vect_get_store_rhs): Make public.
5021 (vectorizable_call): Use internal_store_fn_p.
5022 (vectorizable_store): Handle scatter store internal functions.
5023 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5024 when deciding whether the end of the group has been reached.
5025 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5026 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5027 (mask_scatter_store<mode>): New insns.
5029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5030 Alan Hayward <alan.hayward@arm.com>
5031 David Sherwood <david.sherwood@arm.com>
5033 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5034 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5035 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5037 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5038 Use vect_truncate_gather_scatter_offset if we can't treat the
5039 operation as a normal gather load or scatter store.
5040 (get_group_load_store_type): Take the gather_scatter_info
5041 as argument. Try using a gather load or scatter store for
5042 single-element groups.
5043 (get_load_store_type): Update calls to get_group_load_store_type
5044 and vect_use_strided_gather_scatters_p.
5046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5047 Alan Hayward <alan.hayward@arm.com>
5048 David Sherwood <david.sherwood@arm.com>
5050 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5051 optional tree argument.
5052 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5054 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5055 but continue to use the current value as a fallback.
5056 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5057 to compare the updates.
5058 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5059 (get_load_store_type): Use it when handling a strided access.
5060 (vect_get_strided_load_store_ops): New function.
5061 (vect_get_data_ptr_increment): Likewise.
5062 (vectorizable_load): Handle strided gather loads. Always pass
5063 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5066 Alan Hayward <alan.hayward@arm.com>
5067 David Sherwood <david.sherwood@arm.com>
5069 * doc/md.texi (gather_load@var{m}): Document.
5070 (mask_gather_load@var{m}): Likewise.
5071 * genopinit.c (main): Add supports_vec_gather_load and
5072 supports_vec_gather_load_cached to target_optabs.
5073 * optabs-tree.c (init_tree_optimization_optabs): Use
5074 ggc_cleared_alloc to allocate target_optabs.
5075 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5076 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5078 * internal-fn.h (internal_load_fn_p): Declare.
5079 (internal_gather_scatter_fn_p): Likewise.
5080 (internal_fn_mask_index): Likewise.
5081 (internal_gather_scatter_fn_supported_p): Likewise.
5082 * internal-fn.c (gather_load_direct): New macro.
5083 (expand_gather_load_optab_fn): New function.
5084 (direct_gather_load_optab_supported_p): New macro.
5085 (direct_internal_fn_optab): New function.
5086 (internal_load_fn_p): Likewise.
5087 (internal_gather_scatter_fn_p): Likewise.
5088 (internal_fn_mask_index): Likewise.
5089 (internal_gather_scatter_fn_supported_p): Likewise.
5090 * optabs-query.c (supports_at_least_one_mode_p): New function.
5091 (supports_vec_gather_load_p): Likewise.
5092 * optabs-query.h (supports_vec_gather_load_p): Declare.
5093 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5094 and memory_type field.
5095 (NUM_PATTERNS): Bump to 15.
5096 * tree-vect-data-refs.c: Include internal-fn.h.
5097 (vect_gather_scatter_fn_p): New function.
5098 (vect_describe_gather_scatter_call): Likewise.
5099 (vect_check_gather_scatter): Try using internal functions for
5100 gather loads. Recognize existing calls to a gather load function.
5101 (vect_analyze_data_refs): Consider using gather loads if
5102 supports_vec_gather_load_p.
5103 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5104 (vect_get_gather_scatter_offset_type): Likewise.
5105 (vect_convert_mask_for_vectype): Likewise.
5106 (vect_add_conversion_to_patterm): Likewise.
5107 (vect_try_gather_scatter_pattern): Likewise.
5108 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5109 (vect_vect_recog_func_ptrs): Add it.
5110 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5111 internal_fn_mask_index and internal_gather_scatter_fn_p.
5112 (check_load_store_masking): Take the gather_scatter_info as an
5113 argument and handle gather loads.
5114 (vect_get_gather_scatter_ops): New function.
5115 (vectorizable_call): Check internal_load_fn_p.
5116 (vectorizable_load): Likewise. Handle gather load internal
5118 (vectorizable_store): Update call to check_load_store_masking.
5119 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5120 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5121 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5122 (aarch64_gather_scale_operand_d): New predicates.
5123 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5124 (mask_gather_load<mode>): New insns.
5126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5127 Alan Hayward <alan.hayward@arm.com>
5128 David Sherwood <david.sherwood@arm.com>
5130 * optabs.def (fold_left_plus_optab): New optab.
5131 * doc/md.texi (fold_left_plus_@var{m}): Document.
5132 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5133 * internal-fn.c (fold_left_direct): Define.
5134 (expand_fold_left_optab_fn): Likewise.
5135 (direct_fold_left_optab_supported_p): Likewise.
5136 * fold-const-call.c (fold_const_fold_left): New function.
5137 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5138 * tree-parloops.c (valid_reduction_p): New function.
5139 (gather_scalar_reductions): Use it.
5140 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5141 (vect_finish_replace_stmt): Declare.
5142 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5143 (needs_fold_left_reduction_p): New function, split out from...
5144 (vect_is_simple_reduction): ...here. Accept reductions that
5145 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5146 (vect_force_simple_reduction): Also store the reduction type in
5147 the assignment's STMT_VINFO_REDUC_TYPE.
5148 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5149 (merge_with_identity): New function.
5150 (vect_expand_fold_left): Likewise.
5151 (vectorize_fold_left_reduction): Likewise.
5152 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5153 scalar phi in place for it. Check for target support and reject
5154 cases that would reassociate the operation. Defer the transform
5155 phase to vectorize_fold_left_reduction.
5156 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5157 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5158 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5162 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5163 call to ifc_temp_var.
5165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5166 Alan Hayward <alan.hayward@arm.com>
5167 David Sherwood <david.sherwood@arm.com>
5169 * target.def (legitimize_address_displacement): Take the original
5170 offset as a poly_int.
5171 * targhooks.h (default_legitimize_address_displacement): Update
5173 * targhooks.c (default_legitimize_address_displacement): Likewise.
5174 * doc/tm.texi: Regenerate.
5175 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5176 as an argument, moving assert of ad->disp == ad->disp_term to...
5177 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5178 Try calling targetm.legitimize_address_displacement before expanding
5179 the address rather than afterwards, and adjust for the new interface.
5180 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5181 Match the new hook interface. Handle SVE addresses.
5182 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5187 * Makefile.in (OBJS): Add early-remat.o.
5188 * target.def (select_early_remat_modes): New hook.
5189 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5190 * doc/tm.texi: Regenerate.
5191 * targhooks.h (default_select_early_remat_modes): Declare.
5192 * targhooks.c (default_select_early_remat_modes): New function.
5193 * timevar.def (TV_EARLY_REMAT): New timevar.
5194 * passes.def (pass_early_remat): New pass.
5195 * tree-pass.h (make_pass_early_remat): Declare.
5196 * early-remat.c: New file.
5197 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5199 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5202 Alan Hayward <alan.hayward@arm.com>
5203 David Sherwood <david.sherwood@arm.com>
5205 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5206 vfm1 with a bound_epilog parameter.
5207 (vect_do_peeling): Update calls accordingly, and move the prologue
5208 call earlier in the function. Treat the base bound_epilog as 0 for
5209 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5210 this base when peeling for gaps.
5211 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5212 with fully-masked loops.
5213 (vect_estimate_min_profitable_iters): Handle the single peeled
5214 iteration in that case.
5216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5217 Alan Hayward <alan.hayward@arm.com>
5218 David Sherwood <david.sherwood@arm.com>
5220 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5221 single-element interleaving even if the size is not a power of 2.
5222 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5223 accesses for single-element interleaving if the group size is
5226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5227 Alan Hayward <alan.hayward@arm.com>
5228 David Sherwood <david.sherwood@arm.com>
5230 * doc/md.texi (fold_extract_last_@var{m}): Document.
5231 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5232 * optabs.def (fold_extract_last_optab): New optab.
5233 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5234 * internal-fn.c (fold_extract_direct): New macro.
5235 (expand_fold_extract_optab_fn): Likewise.
5236 (direct_fold_extract_optab_supported_p): Likewise.
5237 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5238 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5239 EXTRACT_LAST_REDUCTION.
5240 (get_initial_def_for_reduction): Do not create an initial vector
5241 for EXTRACT_LAST_REDUCTION reductions.
5242 (vectorizable_reduction): Leave the scalar phi in place for
5243 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5244 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5245 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5246 transform phase to vectorizable_condition.
5247 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5249 (vect_finish_stmt_generation): ...here.
5250 (vect_finish_replace_stmt): New function.
5251 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5252 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5254 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5257 Alan Hayward <alan.hayward@arm.com>
5258 David Sherwood <david.sherwood@arm.com>
5260 * doc/md.texi (extract_last_@var{m}): Document.
5261 * optabs.def (extract_last_optab): New optab.
5262 * internal-fn.def (EXTRACT_LAST): New internal function.
5263 * internal-fn.c (cond_unary_direct): New macro.
5264 (expand_cond_unary_optab_fn): Likewise.
5265 (direct_cond_unary_optab_supported_p): Likewise.
5266 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5267 loops using EXTRACT_LAST.
5268 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5269 (extract_last_<mode>): ...this optab.
5270 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
5277 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5278 * doc/tm.texi: Regenerate.
5279 * targhooks.h (default_empty_mask_is_expensive): Declare.
5280 * targhooks.c (default_empty_mask_is_expensive): New function.
5281 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5282 if the target says that empty masks are expensive.
5283 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5285 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5288 Alan Hayward <alan.hayward@arm.com>
5289 David Sherwood <david.sherwood@arm.com>
5291 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5292 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5293 (vect_use_loop_mask_for_alignment_p): New function.
5294 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5295 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5296 niters_skip argument. Make sure that the first niters_skip elements
5297 of the first iteration are inactive.
5298 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5299 Update call to vect_set_loop_masks_directly.
5300 (get_misalign_in_elems): New function, split out from...
5301 (vect_gen_prolog_loop_niters): ...here.
5302 (vect_update_init_of_dr): Take a code argument that specifies whether
5303 the adjustment should be added or subtracted.
5304 (vect_update_init_of_drs): Likewise.
5305 (vect_prepare_for_masked_peels): New function.
5306 (vect_do_peeling): Skip prologue peeling if we're using a mask
5307 instead. Update call to vect_update_inits_of_drs.
5308 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5310 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5311 alignment. Do not include the number of peeled iterations in
5312 the minimum threshold in that case.
5313 (vectorizable_induction): Adjust the start value down by
5314 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5315 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5316 Take the number of skipped iterations into account when calculating
5318 * tree-vect-stmts.c (vect_gen_while_not): New function.
5320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5321 Alan Hayward <alan.hayward@arm.com>
5322 David Sherwood <david.sherwood@arm.com>
5324 * doc/sourcebuild.texi (vect_fully_masked): Document.
5325 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5327 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5329 (vect_analyze_loop_2): ...here. Don't check the vectorization
5330 factor against the number of loop iterations if the loop is
5333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5334 Alan Hayward <alan.hayward@arm.com>
5335 David Sherwood <david.sherwood@arm.com>
5337 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5338 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5339 (dump_groups): Update accordingly.
5340 (iv_use::mem_type): New member variable.
5341 (address_p): New function.
5342 (record_use): Add a mem_type argument and initialize the new
5344 (record_group_use): Add a mem_type argument. Use address_p.
5345 Remove obsolete null checks of base_object. Update call to record_use.
5346 (find_interesting_uses_op): Update call to record_group_use.
5347 (find_interesting_uses_cond): Likewise.
5348 (find_interesting_uses_address): Likewise.
5349 (get_mem_type_for_internal_fn): New function.
5350 (find_address_like_use): Likewise.
5351 (find_interesting_uses_stmt): Try find_address_like_use before
5352 calling find_interesting_uses_op.
5353 (addr_offset_valid_p): Use the iv mem_type field as the type
5354 of the addressed memory.
5355 (add_autoinc_candidates): Likewise.
5356 (get_address_cost): Likewise.
5357 (split_small_address_groups_p): Use address_p.
5358 (split_address_groups): Likewise.
5359 (add_iv_candidate_for_use): Likewise.
5360 (autoinc_possible_for_pair): Likewise.
5361 (rewrite_groups): Likewise.
5362 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5363 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5364 (get_alias_ptr_type_for_ptr_address): New function.
5365 (rewrite_use_address): Rewrite address uses in calls that were
5366 identified by find_address_like_use.
5368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5369 Alan Hayward <alan.hayward@arm.com>
5370 David Sherwood <david.sherwood@arm.com>
5372 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5374 * gimple-expr.h (is_gimple_addressable: Likewise.
5375 * gimple-expr.c (is_gimple_address): Likewise.
5376 * internal-fn.c (expand_call_mem_ref): New function.
5377 (expand_mask_load_optab_fn): Use it.
5378 (expand_mask_store_optab_fn): Likewise.
5380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5381 Alan Hayward <alan.hayward@arm.com>
5382 David Sherwood <david.sherwood@arm.com>
5384 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5385 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5386 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5387 (cond_umax@var{mode}): Document.
5388 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5389 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5390 (cond_umin_optab, cond_umax_optab): New optabs.
5391 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5392 (COND_IOR, COND_XOR): New internal functions.
5393 * internal-fn.h (get_conditional_internal_fn): Declare.
5394 * internal-fn.c (cond_binary_direct): New macro.
5395 (expand_cond_binary_optab_fn): Likewise.
5396 (direct_cond_binary_optab_supported_p): Likewise.
5397 (get_conditional_internal_fn): New function.
5398 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5399 Cope with reduction statements that are vectorized as calls rather
5401 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5402 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5403 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5404 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5405 (UNSPEC_COND_EOR): New unspecs.
5406 (optab): Add mappings for them.
5407 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5408 (sve_int_op, sve_fp_op): New int attributes.
5410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5411 Alan Hayward <alan.hayward@arm.com>
5412 David Sherwood <david.sherwood@arm.com>
5414 * optabs.def (while_ult_optab): New optab.
5415 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5416 * internal-fn.def (WHILE_ULT): New internal function.
5417 * internal-fn.h (direct_internal_fn_supported_p): New override
5418 that takes two types as argument.
5419 * internal-fn.c (while_direct): New macro.
5420 (expand_while_optab_fn): New function.
5421 (convert_optab_supported_p): Likewise.
5422 (direct_while_optab_supported_p): New macro.
5423 * wide-int.h (wi::udiv_ceil): New function.
5424 * tree-vectorizer.h (rgroup_masks): New structure.
5425 (vec_loop_masks): New typedef.
5426 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5428 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5429 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5430 (vect_max_vf): New function.
5431 (slpeel_make_loop_iterate_ntimes): Delete.
5432 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5433 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5434 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5435 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5436 internal-fn.h, stor-layout.h and optabs-query.h.
5437 (vect_set_loop_mask): New function.
5438 (add_preheader_seq): Likewise.
5439 (add_header_seq): Likewise.
5440 (interleave_supported_p): Likewise.
5441 (vect_maybe_permute_loop_masks): Likewise.
5442 (vect_set_loop_masks_directly): Likewise.
5443 (vect_set_loop_condition_masked): Likewise.
5444 (vect_set_loop_condition_unmasked): New function, split out from
5445 slpeel_make_loop_iterate_ntimes.
5446 (slpeel_make_loop_iterate_ntimes): Rename to..
5447 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5448 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5449 (vect_do_peeling): Update call accordingly.
5450 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5452 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5453 mask_compare_type, can_fully_mask_p and fully_masked_p.
5454 (release_vec_loop_masks): New function.
5455 (_loop_vec_info): Use it to free the loop masks.
5456 (can_produce_all_loop_masks_p): New function.
5457 (vect_get_max_nscalars_per_iter): Likewise.
5458 (vect_verify_full_masking): Likewise.
5459 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5460 retries, and free the mask rgroups before retrying. Check loop-wide
5461 reasons for disallowing fully-masked loops. Make the final decision
5462 about whether use a fully-masked loop or not.
5463 (vect_estimate_min_profitable_iters): Do not assume that peeling
5464 for the number of iterations will be needed for fully-masked loops.
5465 (vectorizable_reduction): Disable fully-masked loops.
5466 (vectorizable_live_operation): Likewise.
5467 (vect_halve_mask_nunits): New function.
5468 (vect_double_mask_nunits): Likewise.
5469 (vect_record_loop_mask): Likewise.
5470 (vect_get_loop_mask): Likewise.
5471 (vect_transform_loop): Handle the case in which the final loop
5472 iteration might handle a partial vector. Call vect_set_loop_condition
5473 instead of slpeel_make_loop_iterate_ntimes.
5474 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5475 (check_load_store_masking): New function.
5476 (prepare_load_store_mask): Likewise.
5477 (vectorizable_store): Handle fully-masked loops.
5478 (vectorizable_load): Likewise.
5479 (supportable_widening_operation): Use vect_halve_mask_nunits for
5481 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5482 (vect_gen_while): New function.
5483 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5484 (aarch64_uqdec<mode>): New insn.
5486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5487 Alan Hayward <alan.hayward@arm.com>
5488 David Sherwood <david.sherwood@arm.com>
5490 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5491 (reduc_xor_scal_optab): New optabs.
5492 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5493 (reduc_xor_scal_@var{m}): Document.
5494 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5495 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5497 * fold-const-call.c (fold_const_call): Handle them.
5498 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5499 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5500 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5501 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5502 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5503 (UNSPEC_XORV): New unspecs.
5504 (optab): Add entries for them.
5505 (BITWISEV): New int iterator.
5506 (bit_reduc_op): New int attributes.
5508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5509 Alan Hayward <alan.hayward@arm.com>
5510 David Sherwood <david.sherwood@arm.com>
5512 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5513 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5514 * optabs.def (vec_shl_insert_optab): New optab.
5515 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5516 (duplicate_and_interleave): Likewise.
5517 * tree-vect-loop.c: Include internal-fn.h.
5518 (neutral_op_for_slp_reduction): New function, split out from
5519 get_initial_defs_for_reduction.
5520 (get_initial_def_for_reduction): Handle option 2 for variable-length
5521 vectors by loading the neutral value into a vector and then shifting
5522 the initial value into element 0.
5523 (get_initial_defs_for_reduction): Replace the code argument with
5524 the neutral value calculated by neutral_op_for_slp_reduction.
5525 Use gimple_build_vector for constant-length vectors.
5526 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5527 but the first group_size elements have a neutral value.
5528 Use duplicate_and_interleave otherwise.
5529 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5530 Update call to get_initial_defs_for_reduction. Handle SLP
5531 reductions for variable-length vectors by creating one vector
5532 result for each scalar result, with the elements associated
5533 with other scalar results stubbed out with the neutral value.
5534 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5535 Require IFN_VEC_SHL_INSERT for double reductions on
5536 variable-length vectors, or SLP reductions that have
5537 a neutral value. Require can_duplicate_and_interleave_p
5538 support for variable-length unchained SLP reductions if there
5539 is no neutral value, such as for MIN/MAX reductions. Also require
5540 the number of vector elements to be a multiple of the number of
5541 SLP statements when doing variable-length unchained SLP reductions.
5542 Update call to vect_create_epilog_for_reduction.
5543 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5544 and remove initial values.
5545 (duplicate_and_interleave): Make public.
5546 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5547 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5550 Alan Hayward <alan.hayward@arm.com>
5551 David Sherwood <david.sherwood@arm.com>
5553 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5554 (can_duplicate_and_interleave_p): New function.
5555 (vect_get_and_check_slp_defs): Take the vector of statements
5556 rather than just the current one. Remove excess parentheses.
5557 Restriction rejectinon of vect_constant_def and vect_external_def
5558 for variable-length vectors to boolean types, or types for which
5559 can_duplicate_and_interleave_p is false.
5560 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5561 (duplicate_and_interleave): New function.
5562 (vect_get_constant_vectors): Use gimple_build_vector for
5563 constant-length vectors and suitable variable-length constant
5564 vectors. Use duplicate_and_interleave for other variable-length
5565 vectors. Don't defer the update when inserting new statements.
5567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5568 Alan Hayward <alan.hayward@arm.com>
5569 David Sherwood <david.sherwood@arm.com>
5571 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5572 min_profitable_iters doesn't go negative.
5574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5575 Alan Hayward <alan.hayward@arm.com>
5576 David Sherwood <david.sherwood@arm.com>
5578 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5579 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5580 * optabs.def (vec_mask_load_lanes_optab): New optab.
5581 (vec_mask_store_lanes_optab): Likewise.
5582 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5583 (MASK_STORE_LANES): Likewise.
5584 * internal-fn.c (mask_load_lanes_direct): New macro.
5585 (mask_store_lanes_direct): Likewise.
5586 (expand_mask_load_optab_fn): Handle masked operations.
5587 (expand_mask_load_lanes_optab_fn): New macro.
5588 (expand_mask_store_optab_fn): Handle masked operations.
5589 (expand_mask_store_lanes_optab_fn): New macro.
5590 (direct_mask_load_lanes_optab_supported_p): Likewise.
5591 (direct_mask_store_lanes_optab_supported_p): Likewise.
5592 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5594 (vect_load_lanes_supported): Likewise.
5595 * tree-vect-data-refs.c (strip_conversion): New function.
5596 (can_group_stmts_p): Likewise.
5597 (vect_analyze_data_ref_accesses): Use it instead of checking
5598 for a pair of assignments.
5599 (vect_store_lanes_supported): Take a masked_p parameter.
5600 (vect_load_lanes_supported): Likewise.
5601 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5602 vect_store_lanes_supported and vect_load_lanes_supported.
5603 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5604 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5605 parameter. Don't allow gaps for masked accesses.
5606 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5607 and vect_load_lanes_supported.
5608 (get_load_store_type): Take a masked_p parameter and update
5609 call to get_group_load_store_type.
5610 (vectorizable_store): Update call to get_load_store_type.
5611 Handle IFN_MASK_STORE_LANES.
5612 (vectorizable_load): Update call to get_load_store_type.
5613 Handle IFN_MASK_LOAD_LANES.
5615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5616 Alan Hayward <alan.hayward@arm.com>
5617 David Sherwood <david.sherwood@arm.com>
5619 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5621 * config/aarch64/aarch64-protos.h
5622 (aarch64_sve_struct_memory_operand_p): Declare.
5623 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5624 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5625 (VPRED, vpred): Handle SVE structure modes.
5626 * config/aarch64/constraints.md (Utx): New constraint.
5627 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5628 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5629 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5630 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5631 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5632 structure modes. Split into pieces after RA.
5633 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5634 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5636 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5637 SVE structure modes.
5638 (aarch64_classify_address): Likewise.
5639 (sizetochar): Move earlier in file.
5640 (aarch64_print_operand): Handle SVE register lists.
5641 (aarch64_array_mode): New function.
5642 (aarch64_sve_struct_memory_operand_p): Likewise.
5643 (TARGET_ARRAY_MODE): Redefine.
5645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5646 Alan Hayward <alan.hayward@arm.com>
5647 David Sherwood <david.sherwood@arm.com>
5649 * target.def (array_mode): New target hook.
5650 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5651 * doc/tm.texi: Regenerate.
5652 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5653 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5654 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5656 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5660 Alan Hayward <alan.hayward@arm.com>
5661 David Sherwood <david.sherwood@arm.com>
5663 * fold-const.c (fold_binary_loc): Check the argument types
5664 rather than the result type when testing for a vector operation.
5666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5668 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5669 * doc/tm.texi: Regenerate.
5671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5672 Alan Hayward <alan.hayward@arm.com>
5673 David Sherwood <david.sherwood@arm.com>
5675 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5676 (sve): Document new AArch64 extension.
5677 * doc/md.texi (w): Extend the description of the AArch64
5678 constraint to include SVE vectors.
5679 (Upl, Upa): Document new AArch64 predicate constraints.
5680 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5682 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5683 (msve-vector-bits=): New option.
5684 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5685 SVE when these are disabled.
5686 (sve): New extension.
5687 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5688 modes. Adjust their number of units based on aarch64_sve_vg.
5689 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5690 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5691 aarch64_addr_query_type.
5692 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5693 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5694 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5695 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5696 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5697 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5698 (aarch64_simd_imm_zero_p): Delete.
5699 (aarch64_check_zero_based_sve_index_immediate): Declare.
5700 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5701 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5702 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5703 (aarch64_sve_float_mul_immediate_p): Likewise.
5704 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5706 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5707 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5708 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5709 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5710 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5711 (aarch64_regmode_natural_size): Likewise.
5712 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5713 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5715 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5716 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5717 for VG and the SVE predicate registers.
5718 (V_ALIASES): Add a "z"-prefixed alias.
5719 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5720 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5721 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5722 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5723 (REG_CLASS_NAMES): Add entries for them.
5724 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5725 and the predicate registers.
5726 (aarch64_sve_vg): Declare.
5727 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5728 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5729 (REGMODE_NATURAL_SIZE): Define.
5730 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5732 * config/aarch64/aarch64.c: Include cfgrtl.h.
5733 (simd_immediate_info): Add a constructor for series vectors,
5734 and an associated step field.
5735 (aarch64_sve_vg): New variable.
5736 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5737 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5738 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5739 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5740 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5741 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5742 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5743 (aarch64_get_mask_mode): New functions.
5744 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5745 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5746 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5747 predicate modes and predicate registers. Explicitly restrict
5748 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5749 to store a vector mode if it is recognized by
5750 aarch64_classify_vector_mode.
5751 (aarch64_regmode_natural_size): New function.
5752 (aarch64_hard_regno_caller_save_mode): Return the original mode
5754 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5755 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5756 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5757 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5759 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5760 does not overlap dest if the function is frame-related. Handle
5762 (aarch64_split_add_offset): New function.
5763 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5764 them aarch64_add_offset.
5765 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5766 and update call to aarch64_sub_sp.
5767 (aarch64_add_cfa_expression): New function.
5768 (aarch64_expand_prologue): Pass extra temporary registers to the
5769 functions above. Handle the case in which we need to emit new
5770 DW_CFA_expressions for registers that were originally saved
5771 relative to the stack pointer, but now have to be expressed
5772 relative to the frame pointer.
5773 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5775 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5776 IP0 and IP1 values for SVE frames.
5777 (aarch64_expand_vec_series): New function.
5778 (aarch64_expand_sve_widened_duplicate): Likewise.
5779 (aarch64_expand_sve_const_vector): Likewise.
5780 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5781 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5782 into the register, rather than emitting a SET directly.
5783 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5784 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5785 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5786 (offset_9bit_signed_scaled_p): New functions.
5787 (aarch64_replicate_bitmask_imm): New function.
5788 (aarch64_bitmask_imm): Use it.
5789 (aarch64_cannot_force_const_mem): Reject expressions involving
5790 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5791 (aarch64_classify_index): Handle SVE indices, by requiring
5792 a plain register index with a scale that matches the element size.
5793 (aarch64_classify_address): Handle SVE addresses. Assert that
5794 the mode of the address is VOIDmode or an integer mode.
5795 Update call to aarch64_classify_symbol.
5796 (aarch64_classify_symbolic_expression): Update call to
5797 aarch64_classify_symbol.
5798 (aarch64_const_vec_all_in_range_p): New function.
5799 (aarch64_print_vector_float_operand): Likewise.
5800 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5801 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5802 and the FP immediates 1.0 and 0.5.
5803 (aarch64_print_address_internal): Handle SVE addresses.
5804 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5805 (aarch64_regno_regclass): Handle predicate registers.
5806 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5808 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5809 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5810 (aarch64_convert_sve_vector_bits): New function.
5811 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5812 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5814 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5815 Handle SVE vector and predicate modes. Accept VL-based constants
5816 that need only one temporary register, and VL offsets that require
5817 no temporary registers.
5818 (aarch64_conditional_register_usage): Mark the predicate registers
5819 as fixed if SVE isn't available.
5820 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5821 Return true for SVE vector and predicate modes.
5822 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5823 rather than an unsigned int. Handle SVE modes.
5824 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5826 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5828 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5829 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5830 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5831 (aarch64_sve_float_mul_immediate_p): New functions.
5832 (aarch64_sve_valid_immediate): New function.
5833 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5834 Explicitly reject structure modes. Check for INDEX constants.
5835 Handle PTRUE and PFALSE constants.
5836 (aarch64_check_zero_based_sve_index_immediate): New function.
5837 (aarch64_simd_imm_zero_p): Delete.
5838 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5839 vector modes. Accept constants in the range of CNT[BHWD].
5840 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5841 ask for an Advanced SIMD mode.
5842 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5843 (aarch64_simd_vector_alignment): Handle SVE predicates.
5844 (aarch64_vectorize_preferred_vector_alignment): New function.
5845 (aarch64_simd_vector_alignment_reachable): Use it instead of
5847 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5848 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5850 (MAX_VECT_LEN): Delete.
5851 (expand_vec_perm_d): Add a vec_flags field.
5852 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5853 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5854 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5856 (aarch64_evpc_rev): Rename to...
5857 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5858 (aarch64_evpc_rev_global): New function.
5859 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5860 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5862 (aarch64_evpc_sve_tbl): New function.
5863 (aarch64_expand_vec_perm_const_1): Update after rename of
5864 aarch64_evpc_rev. Handle SVE permutes too, trying
5865 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5866 than aarch64_evpc_tbl.
5867 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5868 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5869 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5870 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5871 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5872 (aarch64_expand_sve_vcond): New functions.
5873 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5874 of aarch64_vector_mode_p.
5875 (aarch64_dwarf_poly_indeterminate_value): New function.
5876 (aarch64_compute_pressure_classes): Likewise.
5877 (aarch64_can_change_mode_class): Likewise.
5878 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5879 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5880 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5881 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5882 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5883 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5884 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5885 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5887 (Dn, Dl, Dr): Accept const as well as const_vector.
5888 (Dz): Likewise. Compare against CONST0_RTX.
5889 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5890 of "vector" where appropriate.
5891 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5892 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5893 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5894 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5895 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5896 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5897 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5898 (v_int_equiv): Extend to SVE modes.
5899 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5901 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5902 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5903 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5904 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5905 (SVE_COND_FP_CMP): New int iterators.
5906 (perm_hilo): Handle the new unpack unspecs.
5907 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5909 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5910 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5911 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5912 (aarch64_equality_operator, aarch64_constant_vector_operand)
5913 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5914 (aarch64_sve_nonimmediate_operand): Likewise.
5915 (aarch64_sve_general_operand): Likewise.
5916 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5917 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5918 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5919 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5920 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5921 (aarch64_sve_float_arith_immediate): Likewise.
5922 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5923 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5924 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5925 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5926 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5927 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5928 (aarch64_sve_float_arith_operand): Likewise.
5929 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5930 (aarch64_sve_float_mul_operand): Likewise.
5931 (aarch64_sve_vec_perm_operand): Likewise.
5932 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5933 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5934 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5935 as well as const_vector.
5936 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5937 in file. Use CONST0_RTX and CONSTM1_RTX.
5938 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5939 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5940 Use aarch64_simd_imm_zero.
5941 * config/aarch64/aarch64-sve.md: New file.
5942 * config/aarch64/aarch64.md: Include it.
5943 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5944 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5945 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5946 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5947 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5948 (sve): New attribute.
5949 (enabled): Disable instructions with the sve attribute unless
5951 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5952 aarch64_expand_mov_immediate.
5953 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5954 CNT[BHSD] immediates.
5955 (movti): Split CONST_POLY_INT moves into two halves.
5956 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5957 Split additions that need a temporary here if the destination
5958 is the stack pointer.
5959 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5960 (*add<mode>3_poly_1): New instruction.
5961 (set_clobber_cc): New expander.
5963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5965 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5966 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5967 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5968 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5969 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5970 Change innermode from fixed_mode_size to machine_mode.
5971 (simplify_subreg): Update call accordingly. Handle a constant-sized
5972 subreg of a variable-length CONST_VECTOR.
5974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5975 Alan Hayward <alan.hayward@arm.com>
5976 David Sherwood <david.sherwood@arm.com>
5978 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5979 (add_offset_to_base): New function, split out from...
5980 (create_mem_ref): ...here. When handling a scale other than 1,
5981 check first whether the address is valid without the offset.
5982 Add it into the base if so, leaving the index and scale as-is.
5984 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5987 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5988 fold_for_warn before checking if arg2 is INTEGER_CST.
5990 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5992 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5993 (store_multiple_operation): Delete.
5994 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5995 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5996 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5997 guarded by TARGET_STRING.
5998 (rs6000_output_load_multiple): Delete.
5999 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6000 OPTION_MASK_STRING / TARGET_STRING handling.
6001 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6002 (const rs6000_opt_masks) <"string">: Change mask to 0.
6003 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6004 (MASK_STRING): Delete.
6005 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6007 (load_multiple): Delete.
6014 (store_multiple): Delete.
6021 (movmemsi_8reg): Delete.
6022 (corresponding unnamed define_insn): Delete.
6023 (movmemsi_6reg): Delete.
6024 (corresponding unnamed define_insn): Delete.
6025 (movmemsi_4reg): Delete.
6026 (corresponding unnamed define_insn): Delete.
6027 (movmemsi_2reg): Delete.
6028 (corresponding unnamed define_insn): Delete.
6029 (movmemsi_1reg): Delete.
6030 (corresponding unnamed define_insn): Delete.
6031 * config/rs6000/rs6000.opt (mno-string): New.
6032 (mstring): Replace by deprecation warning stub.
6033 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6035 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6037 * regrename.c (regrename_do_replace): If replacing the same
6038 reg multiple times, try to reuse last created gen_raw_REG.
6041 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6042 main to workaround a bug in GDB.
6044 2018-01-12 Tom de Vries <tom@codesourcery.com>
6047 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6049 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6051 PR rtl-optimization/80481
6052 * ira-color.c (get_cap_member): New function.
6053 (allocnos_conflict_by_live_ranges_p): Use it.
6054 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6055 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6057 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6060 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6061 (*saddl_se_1): Ditto.
6063 (*ssubl_se_1): Ditto.
6065 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6067 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6068 rather than wi::to_widest for DR_INITs.
6069 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6070 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6071 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6073 (vect_analyze_group_access_1): Note that here.
6075 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6077 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6078 polynomial type sizes.
6080 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6082 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6083 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6084 (gimple_add_tmp_var): Likewise.
6086 2018-01-12 Martin Liska <mliska@suse.cz>
6088 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6089 (gimple_alloc_sizes): Likewise.
6090 (dump_gimple_statistics): Use PRIu64 in printf format.
6091 * gimple.h: Change uint64_t to int.
6093 2018-01-12 Martin Liska <mliska@suse.cz>
6095 * tree-core.h: Use uint64_t instead of int.
6096 * tree.c (tree_node_counts): Likewise.
6097 (tree_node_sizes): Likewise.
6098 (dump_tree_statistics): Use PRIu64 in printf format.
6100 2018-01-12 Martin Liska <mliska@suse.cz>
6102 * Makefile.in: As qsort_chk is implemented in vec.c, add
6103 vec.o to linkage of gencfn-macros.
6104 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6105 passing the info to record_node_allocation_statistics.
6106 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6108 * ggc-common.c (struct ggc_usage): Add operator== and use
6109 it in operator< and compare function.
6110 * mem-stats.h (struct mem_usage): Likewise.
6111 * vec.c (struct vec_usage): Remove operator< and compare
6112 function. Can be simply inherited.
6114 2018-01-12 Martin Jambor <mjambor@suse.cz>
6117 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6118 * tree-ssa-math-opts.c: Include domwalk.h.
6119 (convert_mult_to_fma_1): New function.
6120 (fma_transformation_info): New type.
6121 (fma_deferring_state): Likewise.
6122 (cancel_fma_deferring): New function.
6123 (result_of_phi): Likewise.
6124 (last_fma_candidate_feeds_initial_phi): Likewise.
6125 (convert_mult_to_fma): Added deferring logic, split actual
6126 transformation to convert_mult_to_fma_1.
6127 (math_opts_dom_walker): New type.
6128 (math_opts_dom_walker::after_dom_children): New method, body moved
6129 here from pass_optimize_widening_mul::execute, added deferring logic
6131 (pass_optimize_widening_mul::execute): Moved most of code to
6132 math_opts_dom_walker::after_dom_children.
6133 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6134 * config/i386/i386.c (ix86_option_override_internal): Added
6135 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6137 2018-01-12 Richard Biener <rguenther@suse.de>
6140 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6141 inline instance vars.
6143 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6146 * config/rx/rx.c (rx_is_restricted_memory_address):
6149 2018-01-12 Richard Biener <rguenther@suse.de>
6151 PR tree-optimization/80846
6152 * target.def (split_reduction): New target hook.
6153 * targhooks.c (default_split_reduction): New function.
6154 * targhooks.h (default_split_reduction): Declare.
6155 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6156 target requests first reduce vectors by combining low and high
6158 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6159 (get_vectype_for_scalar_type_and_size): Export.
6160 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6161 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6162 * doc/tm.texi: Regenerate.
6163 * config/i386/i386.c (ix86_split_reduction): Implement
6164 TARGET_VECTORIZE_SPLIT_REDUCTION.
6166 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6169 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6170 in PIC mode except for TARGET_VXWORKS_RTP.
6171 * config/sparc/sparc.c: Include cfgrtl.h.
6172 (TARGET_INIT_PIC_REG): Define.
6173 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6174 (sparc_pic_register_p): New predicate.
6175 (sparc_legitimate_address_p): Use it.
6176 (sparc_legitimize_pic_address): Likewise.
6177 (sparc_delegitimize_address): Likewise.
6178 (sparc_mode_dependent_address_p): Likewise.
6179 (gen_load_pcrel_sym): Remove 4th parameter.
6180 (load_got_register): Adjust call to above. Remove obsolete stuff.
6181 (sparc_expand_prologue): Do not call load_got_register here.
6182 (sparc_flat_expand_prologue): Likewise.
6183 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6184 (sparc_use_pseudo_pic_reg): New function.
6185 (sparc_init_pic_reg): Likewise.
6186 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6187 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6189 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6191 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6192 Add item for branch_cost.
6194 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6196 PR rtl-optimization/83565
6197 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6198 not extend the result to a larger mode for rotate operations.
6199 (num_sign_bit_copies1): Likewise.
6201 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6204 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6206 Use values-Xc.o for -pedantic.
6207 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6209 2018-01-12 Martin Liska <mliska@suse.cz>
6212 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6214 (possible_polymorphic_call_targets): Use it.
6215 (ipa_devirt): Likewise.
6217 2018-01-12 Martin Liska <mliska@suse.cz>
6219 * profile-count.h (enum profile_quality): Use 0 as invalid
6220 enum value of profile_quality.
6222 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6224 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6225 -mext-string options.
6227 2018-01-12 Richard Biener <rguenther@suse.de>
6229 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6230 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6231 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6233 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6235 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6237 * configure.ac (--with-long-double-format): Add support for the
6238 configuration option to change the default long double format on
6240 * config.gcc (powerpc*-linux*-*): Likewise.
6241 * configure: Regenerate.
6242 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6243 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6244 used without modification.
6246 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6248 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6249 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6250 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6251 MISC_BUILTIN_SPEC_BARRIER.
6252 (rs6000_init_builtins): Likewise.
6253 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6255 (speculation_barrier): New define_insn.
6256 * doc/extend.texi: Document __builtin_speculation_barrier.
6258 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6261 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6262 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6263 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6265 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6266 integral modes instead of "ss" and "sd".
6267 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6268 vectors with 32-bit and 64-bit elements.
6269 (vecdupssescalarmodesuffix): New mode attribute.
6270 (vec_dup<mode>): Use it.
6272 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6275 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6276 frame if argument is passed on stack.
6278 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6281 * ree.c (combine_reaching_defs): Optimize also
6282 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6283 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6285 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6288 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6290 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6293 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6294 after they are computed.
6296 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6298 PR tree-optimization/83695
6299 * gimple-loop-linterchange.cc
6300 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6301 reset cached scev information after interchange.
6302 (pass_linterchange::execute): Remove call to scev_reset_htab.
6304 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6306 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6307 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6308 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6309 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6310 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6311 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6312 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6313 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6314 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6315 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6316 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6317 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6318 (V_lane_reg): Likewise.
6319 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6321 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6322 (vfmal_lane_low<mode>_intrinsic,
6323 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6324 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6325 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6326 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6327 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6328 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6330 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6332 * config/arm/arm-cpus.in (fp16fml): New feature.
6333 (ALL_SIMD): Add fp16fml.
6334 (armv8.2-a): Add fp16fml as an option.
6335 (armv8.3-a): Likewise.
6336 (armv8.4-a): Add fp16fml as part of fp16.
6337 * config/arm/arm.h (TARGET_FP16FML): Define.
6338 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6340 * config/arm/arm-modes.def (V2HF): Define.
6341 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6342 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6343 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6344 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6345 vfmsl_low, vfmsl_high): New set of builtins.
6346 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6347 (vfml_op): New code attribute.
6348 (VFMLHALVES): New int iterator.
6349 (VFML, VFMLSEL): New mode attributes.
6350 (V_reg): Define mapping for V2HF.
6351 (V_hi, V_lo): New mode attributes.
6352 (VF_constraint): Likewise.
6353 (vfml_half, vfml_half_selector): New int attributes.
6354 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6356 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6357 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6359 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6360 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6361 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6362 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6364 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6365 Document new effective target and option set.
6367 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6369 * config/arm/arm-cpus.in (armv8_4): New feature.
6370 (ARMv8_4a): New fgroup.
6371 (armv8.4-a): New arch.
6372 * config/arm/arm-tables.opt: Regenerate.
6373 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6374 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6375 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6376 Add matching rules for -march=armv8.4-a and extensions.
6377 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6379 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6382 * config/rx/rx.md (BW): New mode attribute.
6383 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6385 2018-01-11 Richard Biener <rguenther@suse.de>
6387 PR tree-optimization/83435
6388 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6389 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6390 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6392 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6393 Alan Hayward <alan.hayward@arm.com>
6394 David Sherwood <david.sherwood@arm.com>
6396 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6398 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6399 (aarch64_print_address_internal): Use it to check for a zero offset.
6401 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6402 Alan Hayward <alan.hayward@arm.com>
6403 David Sherwood <david.sherwood@arm.com>
6405 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6406 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6407 Return a poly_int64 rather than a HOST_WIDE_INT.
6408 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6409 rather than a HOST_WIDE_INT.
6410 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6411 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6412 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6413 final_offset from HOST_WIDE_INT to poly_int64.
6414 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6415 to_constant when getting the number of units in an Advanced SIMD
6417 (aarch64_builtin_vectorized_function): Check for a constant number
6419 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6421 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6422 attribute instead of GET_MODE_NUNITS.
6423 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6424 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6425 GET_MODE_SIZE for fixed-size registers.
6426 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6427 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6428 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6429 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6430 (aarch64_print_operand, aarch64_print_address_internal)
6431 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6432 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6433 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6434 Handle polynomial GET_MODE_SIZE.
6435 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6436 wider than SImode without modification.
6437 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6438 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6439 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6440 passing and returning SVE modes.
6441 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6442 rather than GEN_INT.
6443 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6444 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6445 (aarch64_allocate_and_probe_stack_space): Likewise.
6446 (aarch64_layout_frame): Cope with polynomial offsets.
6447 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6448 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6450 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6451 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6452 poly_int64 rather than a HOST_WIDE_INT.
6453 (aarch64_get_separate_components, aarch64_process_components)
6454 (aarch64_expand_prologue, aarch64_expand_epilogue)
6455 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6456 (aarch64_anchor_offset): New function, split out from...
6457 (aarch64_legitimize_address): ...here.
6458 (aarch64_builtin_vectorization_cost): Handle polynomial
6459 TYPE_VECTOR_SUBPARTS.
6460 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6462 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6463 number of elements from the PARALLEL rather than the mode.
6464 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6465 rather than GET_MODE_BITSIZE.
6466 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6467 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6468 (aarch64_expand_vec_perm_const_1): Handle polynomial
6469 d->perm.length () and d->perm elements.
6470 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6471 Apply to_constant to d->perm elements.
6472 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6473 polynomial CONST_VECTOR_NUNITS.
6474 (aarch64_move_pointer): Take amount as a poly_int64 rather
6476 (aarch64_progress_pointer): Avoid temporary variable.
6477 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6478 the mode attribute instead of GET_MODE.
6480 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6481 Alan Hayward <alan.hayward@arm.com>
6482 David Sherwood <david.sherwood@arm.com>
6484 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6485 x exists before using it.
6486 (aarch64_add_constant_internal): Rename to...
6487 (aarch64_add_offset_1): ...this. Replace regnum with separate
6488 src and dest rtxes. Handle the case in which they're different,
6489 including when the offset is zero. Replace scratchreg with an rtx.
6490 Use 2 additions if there is no spare register into which we can
6491 move a 16-bit constant.
6492 (aarch64_add_constant): Delete.
6493 (aarch64_add_offset): Replace reg with separate src and dest
6494 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6495 Use aarch64_add_offset_1.
6496 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6497 an rtx rather than an int. Take the delta as a poly_int64
6498 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6499 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6500 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6501 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6502 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6504 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6505 aarch64_add_constant.
6507 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6509 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6510 Use scalar_float_mode.
6512 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6514 * config/aarch64/aarch64-simd.md
6515 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6516 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6517 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6518 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6519 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6520 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6521 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6522 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6523 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6524 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6526 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6529 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6530 targ_options->x_arm_arch_string is non NULL.
6532 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6534 * config/aarch64/aarch64.h
6535 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6537 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6540 * expmed.c (emit_store_flag_force): Swap if const op0
6541 and change VOIDmode to mode of op0.
6543 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6545 PR rtl-optimization/83761
6546 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6547 than bytes to mode_for_size.
6549 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6552 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6553 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6556 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6559 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6560 when in layout mode.
6561 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6562 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6565 2018-01-10 Michael Collison <michael.collison@arm.com>
6567 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6568 * config/aarch64/aarch64-option-extension.def: Add
6569 AARCH64_OPT_EXTENSION of 'fp16fml'.
6570 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6571 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6572 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6573 * config/aarch64/constraints.md (Ui7): New constraint.
6574 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6575 (VFMLA_SEL_W): Ditto.
6578 (VFMLA16_LOW): New int iterator.
6579 (VFMLA16_HIGH): Ditto.
6580 (UNSPEC_FMLAL): New unspec.
6581 (UNSPEC_FMLSL): Ditto.
6582 (UNSPEC_FMLAL2): Ditto.
6583 (UNSPEC_FMLSL2): Ditto.
6584 (f16mac): New code attribute.
6585 * config/aarch64/aarch64-simd-builtins.def
6586 (aarch64_fmlal_lowv2sf): Ditto.
6587 (aarch64_fmlsl_lowv2sf): Ditto.
6588 (aarch64_fmlalq_lowv4sf): Ditto.
6589 (aarch64_fmlslq_lowv4sf): Ditto.
6590 (aarch64_fmlal_highv2sf): Ditto.
6591 (aarch64_fmlsl_highv2sf): Ditto.
6592 (aarch64_fmlalq_highv4sf): Ditto.
6593 (aarch64_fmlslq_highv4sf): Ditto.
6594 (aarch64_fmlal_lane_lowv2sf): Ditto.
6595 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6596 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6597 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6598 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6599 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6600 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6601 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6602 (aarch64_fmlal_lane_highv2sf): Ditto.
6603 (aarch64_fmlsl_lane_highv2sf): Ditto.
6604 (aarch64_fmlal_laneq_highv2sf): Ditto.
6605 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6606 (aarch64_fmlalq_lane_highv4sf): Ditto.
6607 (aarch64_fmlsl_lane_highv4sf): Ditto.
6608 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6609 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6610 * config/aarch64/aarch64-simd.md:
6611 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6612 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6613 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6614 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6615 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6616 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6617 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6618 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6619 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6620 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6621 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6622 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6623 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6624 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6625 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6626 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6627 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6628 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6629 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6630 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6631 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6632 (vfmlsl_low_u32): Ditto.
6633 (vfmlalq_low_u32): Ditto.
6634 (vfmlslq_low_u32): Ditto.
6635 (vfmlal_high_u32): Ditto.
6636 (vfmlsl_high_u32): Ditto.
6637 (vfmlalq_high_u32): Ditto.
6638 (vfmlslq_high_u32): Ditto.
6639 (vfmlal_lane_low_u32): Ditto.
6640 (vfmlsl_lane_low_u32): Ditto.
6641 (vfmlal_laneq_low_u32): Ditto.
6642 (vfmlsl_laneq_low_u32): Ditto.
6643 (vfmlalq_lane_low_u32): Ditto.
6644 (vfmlslq_lane_low_u32): Ditto.
6645 (vfmlalq_laneq_low_u32): Ditto.
6646 (vfmlslq_laneq_low_u32): Ditto.
6647 (vfmlal_lane_high_u32): Ditto.
6648 (vfmlsl_lane_high_u32): Ditto.
6649 (vfmlal_laneq_high_u32): Ditto.
6650 (vfmlsl_laneq_high_u32): Ditto.
6651 (vfmlalq_lane_high_u32): Ditto.
6652 (vfmlslq_lane_high_u32): Ditto.
6653 (vfmlalq_laneq_high_u32): Ditto.
6654 (vfmlslq_laneq_high_u32): Ditto.
6655 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6656 (AARCH64_FL_FOR_ARCH8_4): New.
6657 (AARCH64_ISA_F16FML): New ISA flag.
6658 (TARGET_F16FML): New feature flag for fp16fml.
6659 (doc/invoke.texi): Document new fp16fml option.
6661 2018-01-10 Michael Collison <michael.collison@arm.com>
6663 * config/aarch64/aarch64-builtins.c:
6664 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6665 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6666 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6667 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6668 (AARCH64_ISA_SHA3): New ISA flag.
6669 (TARGET_SHA3): New feature flag for sha3.
6670 * config/aarch64/iterators.md (sha512_op): New int attribute.
6671 (CRYPTO_SHA512): New int iterator.
6672 (UNSPEC_SHA512H): New unspec.
6673 (UNSPEC_SHA512H2): Ditto.
6674 (UNSPEC_SHA512SU0): Ditto.
6675 (UNSPEC_SHA512SU1): Ditto.
6676 * config/aarch64/aarch64-simd-builtins.def
6677 (aarch64_crypto_sha512hqv2di): New builtin.
6678 (aarch64_crypto_sha512h2qv2di): Ditto.
6679 (aarch64_crypto_sha512su0qv2di): Ditto.
6680 (aarch64_crypto_sha512su1qv2di): Ditto.
6681 (aarch64_eor3qv8hi): Ditto.
6682 (aarch64_rax1qv2di): Ditto.
6683 (aarch64_xarqv2di): Ditto.
6684 (aarch64_bcaxqv8hi): Ditto.
6685 * config/aarch64/aarch64-simd.md:
6686 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6687 (aarch64_crypto_sha512su0qv2di): Ditto.
6688 (aarch64_crypto_sha512su1qv2di): Ditto.
6689 (aarch64_eor3qv8hi): Ditto.
6690 (aarch64_rax1qv2di): Ditto.
6691 (aarch64_xarqv2di): Ditto.
6692 (aarch64_bcaxqv8hi): Ditto.
6693 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6694 (vsha512h2q_u64): Ditto.
6695 (vsha512su0q_u64): Ditto.
6696 (vsha512su1q_u64): Ditto.
6697 (veor3q_u16): Ditto.
6698 (vrax1q_u64): Ditto.
6700 (vbcaxq_u16): Ditto.
6701 * config/arm/types.md (crypto_sha512): New type attribute.
6702 (crypto_sha3): Ditto.
6703 (doc/invoke.texi): Document new sha3 option.
6705 2018-01-10 Michael Collison <michael.collison@arm.com>
6707 * config/aarch64/aarch64-builtins.c:
6708 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6709 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6710 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6711 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6712 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6713 (AARCH64_ISA_SM4): New ISA flag.
6714 (TARGET_SM4): New feature flag for sm4.
6715 * config/aarch64/aarch64-simd-builtins.def
6716 (aarch64_sm3ss1qv4si): Ditto.
6717 (aarch64_sm3tt1aq4si): Ditto.
6718 (aarch64_sm3tt1bq4si): Ditto.
6719 (aarch64_sm3tt2aq4si): Ditto.
6720 (aarch64_sm3tt2bq4si): Ditto.
6721 (aarch64_sm3partw1qv4si): Ditto.
6722 (aarch64_sm3partw2qv4si): Ditto.
6723 (aarch64_sm4eqv4si): Ditto.
6724 (aarch64_sm4ekeyqv4si): Ditto.
6725 * config/aarch64/aarch64-simd.md:
6726 (aarch64_sm3ss1qv4si): Ditto.
6727 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6728 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6729 (aarch64_sm4eqv4si): Ditto.
6730 (aarch64_sm4ekeyqv4si): Ditto.
6731 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6732 (sm3part_op): Ditto.
6733 (CRYPTO_SM3TT): Ditto.
6734 (CRYPTO_SM3PART): Ditto.
6735 (UNSPEC_SM3SS1): New unspec.
6736 (UNSPEC_SM3TT1A): Ditto.
6737 (UNSPEC_SM3TT1B): Ditto.
6738 (UNSPEC_SM3TT2A): Ditto.
6739 (UNSPEC_SM3TT2B): Ditto.
6740 (UNSPEC_SM3PARTW1): Ditto.
6741 (UNSPEC_SM3PARTW2): Ditto.
6742 (UNSPEC_SM4E): Ditto.
6743 (UNSPEC_SM4EKEY): Ditto.
6744 * config/aarch64/constraints.md (Ui2): New constraint.
6745 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6746 * config/arm/types.md (crypto_sm3): New type attribute.
6747 (crypto_sm4): Ditto.
6748 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6749 (vsm3tt1aq_u32): Ditto.
6750 (vsm3tt1bq_u32): Ditto.
6751 (vsm3tt2aq_u32): Ditto.
6752 (vsm3tt2bq_u32): Ditto.
6753 (vsm3partw1q_u32): Ditto.
6754 (vsm3partw2q_u32): Ditto.
6755 (vsm4eq_u32): Ditto.
6756 (vsm4ekeyq_u32): Ditto.
6757 (doc/invoke.texi): Document new sm4 option.
6759 2018-01-10 Michael Collison <michael.collison@arm.com>
6761 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6762 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6763 (AARCH64_FL_FOR_ARCH8_4): New.
6764 (AARCH64_FL_V8_4): New flag.
6765 (doc/invoke.texi): Document new armv8.4-a option.
6767 2018-01-10 Michael Collison <michael.collison@arm.com>
6769 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6770 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6771 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6772 * config/aarch64/aarch64-option-extension.def: Add
6773 AARCH64_OPT_EXTENSION of 'sha2'.
6774 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6775 (crypto): Disable sha2 and aes if crypto disabled.
6776 (crypto): Enable aes and sha2 if enabled.
6777 (simd): Disable sha2 and aes if simd disabled.
6778 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6780 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6781 (TARGET_SHA2): New feature flag for sha2.
6782 (TARGET_AES): New feature flag for aes.
6783 * config/aarch64/aarch64-simd.md:
6784 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6785 conditional on TARGET_AES.
6786 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6787 (aarch64_crypto_sha1hsi): Make pattern conditional
6789 (aarch64_crypto_sha1hv4si): Ditto.
6790 (aarch64_be_crypto_sha1hv4si): Ditto.
6791 (aarch64_crypto_sha1su1v4si): Ditto.
6792 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6793 (aarch64_crypto_sha1su0v4si): Ditto.
6794 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6795 (aarch64_crypto_sha256su0v4si): Ditto.
6796 (aarch64_crypto_sha256su1v4si): Ditto.
6797 (doc/invoke.texi): Document new aes and sha2 options.
6799 2018-01-10 Martin Sebor <msebor@redhat.com>
6801 PR tree-optimization/83781
6802 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6805 2018-01-11 Martin Sebor <msebor@gmail.com>
6806 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6808 PR tree-optimization/83501
6809 PR tree-optimization/81703
6811 * tree-ssa-strlen.c (get_string_cst): Rename...
6812 (get_string_len): ...to this. Handle global constants.
6813 (handle_char_store): Adjust.
6815 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6816 Jim Wilson <jimw@sifive.com>
6818 * config/riscv/riscv-protos.h (riscv_output_return): New.
6819 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6820 (riscv_attribute_table, riscv_output_return),
6821 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6822 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6823 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6825 (riscv_expand_prologue): Add early return for naked function.
6826 (riscv_expand_epilogue): Likewise.
6827 (riscv_function_ok_for_sibcall): Return false for naked function.
6828 (riscv_set_current_function): New.
6829 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6830 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6831 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6832 * doc/extend.texi (RISC-V Function Attributes): New.
6834 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6836 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6837 check for 128-bit long double before checking TCmode.
6838 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6839 128-bit long doubles before checking TFmode or TCmode.
6840 (FLOAT128_IBM_P): Likewise.
6842 2018-01-10 Martin Sebor <msebor@redhat.com>
6844 PR tree-optimization/83671
6845 * builtins.c (c_strlen): Unconditionally return zero for the empty
6847 Use -Warray-bounds for warnings.
6848 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6849 for non-constant array indices with COMPONENT_REF, arrays of
6850 arrays, and pointers to arrays.
6851 (gimple_fold_builtin_strlen): Determine and set length range for
6852 non-constant character arrays.
6854 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6857 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6860 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6862 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6864 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6867 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6868 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6869 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6870 indexed_or_indirect_operand predicate.
6871 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6872 (*vsx_le_perm_load_v8hi): Likewise.
6873 (*vsx_le_perm_load_v16qi): Likewise.
6874 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6875 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6876 (*vsx_le_perm_store_v8hi): Likewise.
6877 (*vsx_le_perm_store_v16qi): Likewise.
6878 (eight unnamed splitters): Likewise.
6880 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6882 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6883 * config/rs6000/emmintrin.h: Likewise.
6884 * config/rs6000/mmintrin.h: Likewise.
6885 * config/rs6000/xmmintrin.h: Likewise.
6887 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6890 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6892 * tree.c (tree_nop_conversion): Return true for location wrapper
6894 (maybe_wrap_with_location): New function.
6895 (selftest::check_strip_nops): New function.
6896 (selftest::test_location_wrappers): New function.
6897 (selftest::tree_c_tests): Call it.
6898 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6899 (maybe_wrap_with_location): New decl.
6900 (EXPR_LOCATION_WRAPPER_P): New macro.
6901 (location_wrapper_p): New inline function.
6902 (tree_strip_any_location_wrapper): New inline function.
6904 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6907 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6908 stack_realign_offset for the largest alignment of stack slot
6910 (ix86_find_max_used_stack_alignment): New function.
6911 (ix86_finalize_stack_frame_flags): Use it. Set
6912 max_used_stack_alignment if we don't realign stack.
6913 * config/i386/i386.h (machine_function): Add
6914 max_used_stack_alignment.
6916 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6918 * config/arm/arm.opt (-mbranch-cost): New option.
6919 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6922 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6925 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6926 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6928 2018-01-10 Richard Biener <rguenther@suse.de>
6931 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6932 early out so it also covers the case where we have a non-NULL
6935 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6937 PR tree-optimization/83753
6938 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6939 for non-strided grouped accesses if the number of elements is 1.
6941 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6944 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6945 * i386.h (TARGET_USE_GATHER): Define.
6946 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6948 2018-01-10 Martin Liska <mliska@suse.cz>
6951 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6952 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6954 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6955 CLEANUP_NO_PARTITIONING is not set.
6957 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6959 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6960 for vectors, as a partial revert of r254296.
6961 * rtl.h (const_vec_p): Delete.
6962 (const_vec_duplicate_p): Don't test for vector CONSTs.
6963 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6964 * expmed.c (make_tree): Likewise.
6967 * common.md (E, F): Use CONSTANT_P instead of checking for
6969 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6970 checking for CONST_VECTOR.
6972 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6975 * predict.c (force_edge_cold): Handle in more sane way edges
6978 2018-01-09 Carl Love <cel@us.ibm.com>
6980 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6982 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6983 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6984 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6985 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6986 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6987 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6988 * config/rs6000/rs6000-protos.h: Add extern defition for
6989 rs6000_generate_float2_double_code.
6990 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6992 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6993 (float2_v2df): Add define_expand.
6995 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6998 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6999 op_mode in the force_to_mode call.
7001 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7003 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7004 instead of checking each element individually.
7005 (aarch64_evpc_uzp): Likewise.
7006 (aarch64_evpc_zip): Likewise.
7007 (aarch64_evpc_ext): Likewise.
7008 (aarch64_evpc_rev): Likewise.
7009 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7010 instead of checking each element individually. Return true without
7012 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7013 whether all selected elements come from the same input, instead of
7014 checking each element individually. Remove calls to gen_rtx_REG,
7015 start_sequence and end_sequence and instead assert that no rtl is
7018 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7020 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7021 order of HIGH and CONST checks.
7023 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7025 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7026 if the destination isn't an SSA_NAME.
7028 2018-01-09 Richard Biener <rguenther@suse.de>
7030 PR tree-optimization/83668
7031 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7033 (canonicalize_loop_form): ... here, renamed from ...
7034 (canonicalize_loop_closed_ssa_form): ... this and amended to
7035 swap successor edges for loop exit blocks to make us use
7036 the RPO order we need for initial schedule generation.
7038 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7040 PR tree-optimization/64811
7041 * match.pd: When optimizing comparisons with Inf, avoid
7042 introducing or losing exceptions from comparisons with NaN.
7044 2018-01-09 Martin Liska <mliska@suse.cz>
7047 * asan.c (shadow_mem_size): Add gcc_assert.
7049 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7051 Don't save registers in main().
7054 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7055 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7056 * config/avr/avr.c (avr_set_current_function): Don't error if
7057 naked, OS_task or OS_main are specified at the same time.
7058 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7060 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7062 * common/config/avr/avr-common.c (avr_option_optimization_table):
7063 Switch on -mmain-is-OS_task for optimizing compilations.
7065 2018-01-09 Richard Biener <rguenther@suse.de>
7067 PR tree-optimization/83572
7068 * graphite.c: Include cfganal.h.
7069 (graphite_transform_loops): Connect infinite loops to exit
7070 and remove fake edges at the end.
7072 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7074 * ipa-inline.c (edge_badness): Revert accidental checkin.
7076 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7079 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7080 symbols; not inline clones.
7082 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7085 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7086 hard registers. Formatting fixes.
7088 PR preprocessor/83722
7089 * gcc.c (try_generate_repro): Pass
7090 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7091 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7094 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7095 Kito Cheng <kito.cheng@gmail.com>
7097 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7098 (riscv_leaf_function_p): Delete.
7099 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7101 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7103 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7105 (do_ifelse): New function.
7106 (do_isel): New function.
7107 (do_sub3): New function.
7108 (do_add3): New function.
7109 (do_load_mask_compare): New function.
7110 (do_overlap_load_compare): New function.
7111 (expand_compare_loop): New function.
7112 (expand_block_compare): Call expand_compare_loop() when appropriate.
7113 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7115 (-mblock-compare-inline-loop-limit): New option.
7117 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7120 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7121 Reverse order of second and third operands in first alternative.
7122 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7123 of first and second elements in UNSPEC_VPERMR vector.
7124 (altivec_expand_vec_perm_le): Likewise.
7126 2018-01-08 Jeff Law <law@redhat.com>
7128 PR rtl-optimizatin/81308
7129 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7130 (process_switch): If group_case_labels makes a change, then set
7132 (pass_convert_switch::execute): If a switch is converted, then
7133 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7135 PR rtl-optimization/81308
7136 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7139 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7141 PR target/83663 - Revert r255946
7142 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7143 generation for cases where splatting a value is not useful.
7144 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7145 across a vec_duplicate and a paradoxical subreg forming a vector
7146 mode to a vec_concat.
7148 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7150 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7151 -march=armv8.3-a variants.
7152 * config/arm/t-multilib: Likewise.
7153 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7155 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7157 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7159 (cceq_ior_compare_complement): Give it a name so I can use it, and
7160 change boolean_or_operator predicate to boolean_operator so it can
7161 be used to generate a crand.
7162 (eqne): New code iterator.
7163 (bd/bd_neg): New code_attrs.
7164 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7165 a single define_insn.
7166 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7167 decrement (bdnzt/bdnzf/bdzt/bdzf).
7168 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7169 with the new names of the branch decrement patterns, and added the
7170 names of the branch decrement conditional patterns.
7172 2018-01-08 Richard Biener <rguenther@suse.de>
7174 PR tree-optimization/83563
7175 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7178 2018-01-08 Richard Biener <rguenther@suse.de>
7181 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7183 2018-01-08 Richard Biener <rguenther@suse.de>
7185 PR tree-optimization/83685
7186 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7187 references to abnormals.
7189 2018-01-08 Richard Biener <rguenther@suse.de>
7192 * dwarf2out.c (output_indirect_strings): Handle empty
7193 skeleton_debug_str_hash.
7194 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7196 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7198 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7199 (emit_store_direct): Likewise.
7200 (arc_trampoline_adjust_address): Likewise.
7201 (arc_asm_trampoline_template): New function.
7202 (arc_initialize_trampoline): Use asm_trampoline_template.
7203 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7204 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7205 * config/arc/arc.md (flush_icache): Delete pattern.
7207 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7209 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7210 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7213 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7216 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7217 by not USED_FOR_TARGET.
7218 (make_pass_resolve_sw_modes): Likewise.
7220 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7222 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7225 2018-01-08 Richard Biener <rguenther@suse.de>
7228 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7230 2018-01-08 Richard Biener <rguenther@suse.de>
7233 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7235 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7238 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7239 basic blocks with a small number of successors.
7240 (convert_control_dep_chain_into_preds): Improve handling of
7242 (dump_predicates): Split apart into...
7243 (dump_pred_chain): ...here...
7244 (dump_pred_info): ...and here.
7245 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7246 (can_chain_union_be_invalidated_p): Improve check for invalidation
7248 (uninit_uses_cannot_happen): Avoid unnecessary if
7249 convert_control_dep_chain_into_preds yielded nothing.
7251 2018-01-06 Martin Sebor <msebor@redhat.com>
7253 PR tree-optimization/83640
7254 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7255 subtracting negative offset from size.
7256 (builtin_access::overlap): Adjust offset bounds of the access to fall
7257 within the size of the object if possible.
7259 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7261 PR rtl-optimization/83699
7262 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7263 extract_bit_field_as_subreg to cases in which the extracted
7264 value is also a vector.
7266 * lra-constraints.c (process_alt_operands): Test for the equivalence
7267 substitutions when detecting a possible reload cycle.
7269 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7272 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7273 by default if flag_selective_schedling{,2}. Formatting fixes.
7275 PR rtl-optimization/83682
7276 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7277 if it has non-VECTOR_MODE element mode.
7278 (vec_duplicate_p): Likewise.
7281 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7282 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7284 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7287 * config/i386/i386-builtin.def
7288 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7289 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7290 Require also OPTION_MASK_ISA_AVX512F in addition to
7291 OPTION_MASK_ISA_GFNI.
7292 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7293 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7294 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7295 to OPTION_MASK_ISA_GFNI.
7296 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7297 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7298 OPTION_MASK_ISA_AVX512BW.
7299 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7300 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7301 addition to OPTION_MASK_ISA_GFNI.
7302 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7303 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7304 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7305 to OPTION_MASK_ISA_GFNI.
7306 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7307 a requirement for all ISAs rather than any of them with a few
7309 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7311 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7312 bitmasks to be enabled with 3 exceptions, instead of requiring any
7313 enabled ISA with lots of exceptions.
7314 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7315 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7316 Change avx512bw in isa attribute to avx512f.
7317 * config/i386/sgxintrin.h: Add license boilerplate.
7318 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7319 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7320 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7321 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7323 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7324 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7325 temporarily sse2 rather than sse if not enabled already.
7328 * config/i386/sse.md (VI248_VLBW): Rename to ...
7329 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7330 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7331 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7332 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7333 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7334 mode iterator instead of VI248_VLBW.
7336 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7338 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7339 (record_modified): Skip clobbers; add debug output.
7340 (param_change_prob): Use sreal frequencies.
7342 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7344 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7345 punt for user-aligned variables.
7347 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7349 * tree-chrec.c (chrec_contains_symbols): Return true for
7352 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7355 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7356 of (x|y) == x for BICS pattern.
7358 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7360 PR tree-optimization/83605
7361 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7362 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7365 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7367 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7368 * config/epiphany/rtems.h: New file.
7370 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7371 Uros Bizjak <ubizjak@gmail.com>
7374 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7375 QIreg_operand instead of register_operand predicate.
7376 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7377 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7378 comments instead of -fmitigate[-_]rop.
7380 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7383 * cgraphunit.c (symbol_table::compile): Switch to text_section
7384 before calling assembly_start debug hook.
7385 * run-rtl-passes.c (run_rtl_passes): Likewise.
7388 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7390 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7391 range_int_cst_p rather than !symbolic_range_p before calling
7392 extract_range_from_multiplicative_op_1.
7394 2018-01-04 Jeff Law <law@redhat.com>
7396 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7397 redundant test in assertion.
7399 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7401 * doc/rtl.texi: Document machine_mode wrapper classes.
7403 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7405 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7408 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7410 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7411 the VEC_PERM_EXPR fold to fail.
7413 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7416 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7417 to switched_sections.
7419 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7422 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7425 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7428 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7429 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7431 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7434 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7435 is BLKmode and bitpos not zero or mode change is needed.
7437 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7440 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7443 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7446 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7447 instead of MULT rtx. Update all corresponding splitters.
7449 (*ssub<modesuffix>): Ditto.
7451 (*cmp_sadd_di): Update split patterns.
7452 (*cmp_sadd_si): Ditto.
7453 (*cmp_sadd_sidi): Ditto.
7454 (*cmp_ssub_di): Ditto.
7455 (*cmp_ssub_si): Ditto.
7456 (*cmp_ssub_sidi): Ditto.
7457 * config/alpha/predicates.md (const23_operand): New predicate.
7458 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7459 Look for ASHIFT, not MULT inner operand.
7460 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7462 2018-01-04 Martin Liska <mliska@suse.cz>
7464 PR gcov-profile/83669
7465 * gcov.c (output_intermediate_file): Add version to intermediate
7467 * doc/gcov.texi: Document new field 'version' in intermediate
7468 file format. Fix location of '-k' option of gcov command.
7470 2018-01-04 Martin Liska <mliska@suse.cz>
7473 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7475 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7477 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7479 2018-01-03 Martin Sebor <msebor@redhat.com>
7481 PR tree-optimization/83655
7482 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7483 checking calls with invalid arguments.
7485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7487 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7488 (vectorizable_mask_load_store): Delete.
7489 (vectorizable_call): Return false for masked loads and stores.
7490 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7491 instead of gimple_assign_rhs1.
7492 (vectorizable_load): Handle IFN_MASK_LOAD.
7493 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7497 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7499 (vectorizable_mask_load_store): ...here.
7500 (vectorizable_load): ...and here.
7502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7504 * tree-vect-stmts.c (vect_build_all_ones_mask)
7505 (vect_build_zero_merge_argument): New functions, split out from...
7506 (vectorizable_load): ...here.
7508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7510 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7512 (vectorizable_mask_load_store): ...here.
7513 (vectorizable_store): ...and here.
7515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7517 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7519 (vectorizable_mask_load_store): ...here.
7521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7523 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7524 (vect_model_store_cost): Take a vec_load_store_type instead of a
7526 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7527 (vect_model_store_cost): Take a vec_load_store_type instead of a
7529 (vectorizable_mask_load_store): Update accordingly.
7530 (vectorizable_store): Likewise.
7531 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7535 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7536 IFN_MASK_LOAD calls here rather than...
7537 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7540 Alan Hayward <alan.hayward@arm.com>
7541 David Sherwood <david.sherwood@arm.com>
7543 * expmed.c (extract_bit_field_1): For vector extracts,
7544 fall back to extract_bit_field_as_subreg if vec_extract
7547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7548 Alan Hayward <alan.hayward@arm.com>
7549 David Sherwood <david.sherwood@arm.com>
7551 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7552 they are variable or constant sized.
7553 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7554 slots for constant-sized data.
7556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7557 Alan Hayward <alan.hayward@arm.com>
7558 David Sherwood <david.sherwood@arm.com>
7560 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7561 handling COND_EXPRs with boolean comparisons, try to find a better
7562 basis for the mask type than the boolean itself.
7564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7566 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7567 is calculated and how it can be overridden.
7568 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7569 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7571 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7575 Alan Hayward <alan.hayward@arm.com>
7576 David Sherwood <david.sherwood@arm.com>
7578 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7579 Remove the mode argument.
7580 (aarch64_simd_valid_immediate): Remove the mode and inverse
7582 * config/aarch64/iterators.md (bitsize): New iterator.
7583 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7584 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7585 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7586 aarch64_simd_valid_immediate.
7587 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7588 (aarch64_reg_or_bic_imm): Likewise.
7589 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7590 with an insn_type enum and msl with a modifier_type enum.
7591 Replace element_width with a scalar_mode. Change the shift
7592 to unsigned int. Add constructors for scalar_float_mode and
7593 scalar_int_mode elements.
7594 (aarch64_vect_float_const_representable_p): Delete.
7595 (aarch64_can_const_movi_rtx_p)
7596 (aarch64_simd_scalar_immediate_valid_for_move)
7597 (aarch64_simd_make_constant): Update call to
7598 aarch64_simd_valid_immediate.
7599 (aarch64_advsimd_valid_immediate_hs): New function.
7600 (aarch64_advsimd_valid_immediate): Likewise.
7601 (aarch64_simd_valid_immediate): Remove mode and inverse
7602 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7603 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7604 and aarch64_float_const_representable_p on the result.
7605 (aarch64_output_simd_mov_immediate): Remove mode argument.
7606 Update call to aarch64_simd_valid_immediate and use of
7607 simd_immediate_info.
7608 (aarch64_output_scalar_simd_mov_immediate): Update call
7611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7612 Alan Hayward <alan.hayward@arm.com>
7613 David Sherwood <david.sherwood@arm.com>
7615 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7616 (mode_nunits): Likewise CONST_MODE_NUNITS.
7617 * machmode.def (ADJUST_NUNITS): Document.
7618 * genmodes.c (mode_data::need_nunits_adj): New field.
7619 (blank_mode): Update accordingly.
7620 (adj_nunits): New variable.
7621 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7623 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7624 listed in adj_nunits.
7625 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7626 listed in adj_nunits. Don't emit case statements for such modes.
7627 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7628 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7629 nothing if adj_nunits is nonnull.
7630 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7631 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7632 (emit_mode_fbit): Update use of print_maybe_const_decl.
7633 (emit_move_size): Likewise. Treat the array as non-const
7635 (emit_mode_adjustments): Handle adj_nunits.
7637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7639 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7640 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7641 (VECTOR_MODES): Use it.
7642 (make_vector_modes): Take the prefix as an argument.
7644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7645 Alan Hayward <alan.hayward@arm.com>
7646 David Sherwood <david.sherwood@arm.com>
7648 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7649 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7650 for MODE_VECTOR_BOOL.
7651 * machmode.def (VECTOR_BOOL_MODE): Document.
7652 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7653 (make_vector_bool_mode): New function.
7654 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7656 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7657 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7659 * stor-layout.c (int_mode_for_mode): Likewise.
7660 * tree.c (build_vector_type_for_mode): Likewise.
7661 * varasm.c (output_constant_pool_2): Likewise.
7662 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7663 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7664 for MODE_VECTOR_BOOL.
7665 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7666 of mode class checks.
7667 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7668 instead of a list of mode class checks.
7669 (expand_vector_scalar_condition): Likewise.
7670 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7673 Alan Hayward <alan.hayward@arm.com>
7674 David Sherwood <david.sherwood@arm.com>
7676 * machmode.h (mode_size): Change from unsigned short to
7678 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7679 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7680 or if measurement_type is not polynomial.
7681 (fixed_size_mode::includes_p): Check for constant-sized modes.
7682 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7683 return a poly_uint16 rather than an unsigned short.
7684 (emit_mode_size): Change the type of mode_size from unsigned short
7685 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7686 (emit_mode_adjustments): Cope with polynomial vector sizes.
7687 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7689 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7691 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7692 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7693 * caller-save.c (setup_save_areas): Likewise.
7694 (replace_reg_with_saved_mem): Likewise.
7695 * calls.c (emit_library_call_value_1): Likewise.
7696 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7697 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7698 (gen_lowpart_for_combine): Likewise.
7699 * convert.c (convert_to_integer_1): Likewise.
7700 * cse.c (equiv_constant, cse_insn): Likewise.
7701 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7702 (cselib_subst_to_values): Likewise.
7703 * dce.c (word_dce_process_block): Likewise.
7704 * df-problems.c (df_word_lr_mark_ref): Likewise.
7705 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7706 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7707 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7708 (rtl_for_decl_location): Likewise.
7709 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7710 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7711 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7712 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7713 (expand_expr_real_1): Likewise.
7714 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7715 (pad_below): Likewise.
7716 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7717 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7718 * ira.c (get_subreg_tracking_sizes): Likewise.
7719 * ira-build.c (ira_create_allocno_objects): Likewise.
7720 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7721 (ira_sort_regnos_for_alter_reg): Likewise.
7722 * ira-costs.c (record_operand_costs): Likewise.
7723 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7724 (resolve_simple_move): Likewise.
7725 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7726 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7727 (lra_constraints): Likewise.
7728 (CONST_POOL_OK_P): Reject variable-sized modes.
7729 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7730 (add_pseudo_to_slot, lra_spill): Likewise.
7731 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7732 * optabs-query.c (get_best_extraction_insn): Likewise.
7733 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7734 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7735 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7736 * recog.c (offsettable_address_addr_space_p): Likewise.
7737 * regcprop.c (maybe_mode_change): Likewise.
7738 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7739 * regrename.c (build_def_use): Likewise.
7740 * regstat.c (dump_reg_info): Likewise.
7741 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7742 (find_reloads, find_reloads_subreg_address): Likewise.
7743 * reload1.c (eliminate_regs_1): Likewise.
7744 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7745 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7746 (simplify_binary_operation_1, simplify_subreg): Likewise.
7747 * targhooks.c (default_function_arg_padding): Likewise.
7748 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7749 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7750 (verify_gimple_assign_ternary): Likewise.
7751 * tree-inline.c (estimate_move_cost): Likewise.
7752 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7753 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7754 (get_address_cost_ainc): Likewise.
7755 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7756 (vect_supportable_dr_alignment): Likewise.
7757 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7758 (vectorizable_reduction): Likewise.
7759 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7760 (vectorizable_operation, vectorizable_load): Likewise.
7761 * tree.c (build_same_sized_truth_vector_type): Likewise.
7762 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7763 * var-tracking.c (emit_note_insn_var_location): Likewise.
7764 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7765 (ADDR_VEC_ALIGN): Likewise.
7767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7768 Alan Hayward <alan.hayward@arm.com>
7769 David Sherwood <david.sherwood@arm.com>
7771 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7773 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7774 or if measurement_type is polynomial.
7775 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7776 * combine.c (make_extraction): Likewise.
7777 * dse.c (find_shift_sequence): Likewise.
7778 * dwarf2out.c (mem_loc_descriptor): Likewise.
7779 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7780 (extract_bit_field, extract_low_bits): Likewise.
7781 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7782 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7783 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7784 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7785 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7786 * reload.c (find_reloads): Likewise.
7787 * reload1.c (alter_reg): Likewise.
7788 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7789 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7790 * tree-if-conv.c (predicate_mem_writes): Likewise.
7791 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7792 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7793 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7794 * valtrack.c (dead_debug_insert_temp): Likewise.
7795 * varasm.c (mergeable_constant_section): Likewise.
7796 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7799 Alan Hayward <alan.hayward@arm.com>
7800 David Sherwood <david.sherwood@arm.com>
7802 * expr.c (expand_assignment): Cope with polynomial mode sizes
7803 when assigning to a CONCAT.
7805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7806 Alan Hayward <alan.hayward@arm.com>
7807 David Sherwood <david.sherwood@arm.com>
7809 * machmode.h (mode_precision): Change from unsigned short to
7811 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7813 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7814 or if measurement_type is not polynomial.
7815 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7816 in which the mode is already known to be a scalar_int_mode.
7817 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7818 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7820 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7821 for GET_MODE_PRECISION.
7822 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7823 for GET_MODE_PRECISION.
7824 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7826 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7827 (expand_field_assignment, make_extraction): Likewise.
7828 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7829 (get_last_value): Likewise.
7830 * convert.c (convert_to_integer_1): Likewise.
7831 * cse.c (cse_insn): Likewise.
7832 * expr.c (expand_expr_real_1): Likewise.
7833 * lra-constraints.c (simplify_operand_subreg): Likewise.
7834 * optabs-query.c (can_atomic_load_p): Likewise.
7835 * optabs.c (expand_atomic_load): Likewise.
7836 (expand_atomic_store): Likewise.
7837 * ree.c (combine_reaching_defs): Likewise.
7838 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7839 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7840 * tree.h (type_has_mode_precision_p): Likewise.
7841 * ubsan.c (instrument_si_overflow): Likewise.
7843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7844 Alan Hayward <alan.hayward@arm.com>
7845 David Sherwood <david.sherwood@arm.com>
7847 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7848 polynomial numbers of units.
7849 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7850 (valid_vector_subparts_p): New function.
7851 (build_vector_type): Remove temporary shim and take the number
7852 of units as a poly_uint64 rather than an int.
7853 (build_opaque_vector_type): Take the number of units as a
7854 poly_uint64 rather than an int.
7855 * tree.c (build_vector_from_ctor): Handle polynomial
7856 TYPE_VECTOR_SUBPARTS.
7857 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7858 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7859 (build_vector_from_val): If the number of units is variable,
7860 use build_vec_duplicate_cst for constant operands and
7861 VEC_DUPLICATE_EXPR otherwise.
7862 (make_vector_type): Remove temporary is_constant ().
7863 (build_vector_type, build_opaque_vector_type): Take the number of
7864 units as a poly_uint64 rather than an int.
7865 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7867 * cfgexpand.c (expand_debug_expr): Likewise.
7868 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7869 (store_constructor, expand_expr_real_1): Likewise.
7870 (const_scalar_mask_from_tree): Likewise.
7871 * fold-const-call.c (fold_const_reduction): Likewise.
7872 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7873 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7874 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7875 (fold_relational_const): Likewise.
7876 (native_interpret_vector): Likewise. Change the size from an
7877 int to an unsigned int.
7878 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7879 TYPE_VECTOR_SUBPARTS.
7880 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7881 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7882 duplicating a non-constant operand into a variable-length vector.
7883 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7884 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7885 * ipa-icf.c (sem_variable::equals): Likewise.
7886 * match.pd: Likewise.
7887 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7888 * print-tree.c (print_node): Likewise.
7889 * stor-layout.c (layout_type): Likewise.
7890 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7891 * tree-cfg.c (verify_gimple_comparison): Likewise.
7892 (verify_gimple_assign_binary): Likewise.
7893 (verify_gimple_assign_ternary): Likewise.
7894 (verify_gimple_assign_single): Likewise.
7895 * tree-pretty-print.c (dump_generic_node): Likewise.
7896 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7897 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7898 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7899 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7900 (vect_shift_permute_load_chain): Likewise.
7901 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7902 (expand_vector_condition, optimize_vector_constructor): Likewise.
7903 (lower_vec_perm, get_compute_type): Likewise.
7904 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7905 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7906 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7907 (vect_recog_mask_conversion_pattern): Likewise.
7908 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7909 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7910 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7911 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7912 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7913 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7914 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7915 (supportable_widening_operation): Likewise.
7916 (supportable_narrowing_operation): Likewise.
7917 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7919 * varasm.c (output_constant): Likewise.
7921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7922 Alan Hayward <alan.hayward@arm.com>
7923 David Sherwood <david.sherwood@arm.com>
7925 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7926 so that both the length == 3 and length != 3 cases set up their
7927 own permute vectors. Add comments explaining why we know the
7928 number of elements is constant.
7929 (vect_permute_load_chain): Likewise.
7931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7932 Alan Hayward <alan.hayward@arm.com>
7933 David Sherwood <david.sherwood@arm.com>
7935 * machmode.h (mode_nunits): Change from unsigned char to
7937 (ONLY_FIXED_SIZE_MODES): New macro.
7938 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7939 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7940 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7942 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7943 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7944 or if measurement_type is not polynomial.
7945 * genmodes.c (ZERO_COEFFS): New macro.
7946 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7948 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7949 Use ZERO_COEFFS when emitting initializers.
7950 * data-streamer.h (bp_pack_poly_value): New function.
7951 (bp_unpack_poly_value): Likewise.
7952 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7953 for GET_MODE_NUNITS.
7954 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7955 for GET_MODE_NUNITS.
7956 * tree.c (make_vector_type): Remove temporary shim and make
7957 the real function take the number of units as a poly_uint64
7959 (build_vector_type_for_mode): Handle polynomial nunits.
7960 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7961 * emit-rtl.c (const_vec_series_p_1): Likewise.
7962 (gen_rtx_CONST_VECTOR): Likewise.
7963 * fold-const.c (test_vec_duplicate_folding): Likewise.
7964 * genrecog.c (validate_pattern): Likewise.
7965 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7966 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7967 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7968 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7969 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7970 * rtlanal.c (subreg_get_info): Likewise.
7971 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7972 (vect_grouped_load_supported): Likewise.
7973 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7974 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7975 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7976 (simplify_const_unary_operation, simplify_binary_operation_1)
7977 (simplify_const_binary_operation, simplify_ternary_operation)
7978 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7979 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7980 instead of CONST_VECTOR_NUNITS.
7981 * varasm.c (output_constant_pool_2): Likewise.
7982 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7983 explicit-encoded elements in the XVEC for variable-length vectors.
7985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7990 Alan Hayward <alan.hayward@arm.com>
7991 David Sherwood <david.sherwood@arm.com>
7993 * coretypes.h (fixed_size_mode): Declare.
7994 (fixed_size_mode_pod): New typedef.
7995 * builtins.h (target_builtins::x_apply_args_mode)
7996 (target_builtins::x_apply_result_mode): Change type to
7997 fixed_size_mode_pod.
7998 * builtins.c (apply_args_size, apply_result_size, result_vector)
7999 (expand_builtin_apply_args_1, expand_builtin_apply)
8000 (expand_builtin_return): Update accordingly.
8002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8004 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8005 * cselib.c (cselib_hash_rtx): Likewise.
8006 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8007 CONST_VECTOR encoding.
8009 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8010 Jeff Law <law@redhat.com>
8013 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8014 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8015 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8016 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8019 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8020 explicitly probe *sp in a noreturn function if there were any callee
8021 register saves or frame pointer is needed.
8023 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8026 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8027 BLKmode for ternary, binary or unary expressions.
8030 * var-tracking.c (delete_vta_debug_insn): New inline function.
8031 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8032 insns from get_insns () to NULL instead of each bb separately.
8033 Use delete_vta_debug_insn. No longer static.
8034 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8035 delete_vta_debug_insns callers.
8036 * rtl.h (delete_vta_debug_insns): Declare.
8037 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8038 instead of variable_tracking_main.
8040 2018-01-03 Martin Sebor <msebor@redhat.com>
8042 PR tree-optimization/83603
8043 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8044 arguments past the endof the argument list in functions declared
8045 without a prototype.
8046 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8047 Avoid checking when arguments are null.
8049 2018-01-03 Martin Sebor <msebor@redhat.com>
8052 * doc/extend.texi (attribute const): Fix a typo.
8053 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8054 issuing -Wsuggest-attribute for void functions.
8056 2018-01-03 Martin Sebor <msebor@redhat.com>
8058 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8059 offset_int::from instead of wide_int::to_shwi.
8060 (maybe_diag_overlap): Remove assertion.
8061 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8062 * gimple-ssa-sprintf.c (format_directive): Same.
8063 (parse_directive): Same.
8064 (sprintf_dom_walker::compute_format_length): Same.
8065 (try_substitute_return_value): Same.
8067 2018-01-03 Jeff Law <law@redhat.com>
8070 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8071 non-constant residual for zero at runtime and avoid probing in
8072 that case. Reorganize code for trailing problem to mirror handling
8075 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8077 PR tree-optimization/83501
8078 * tree-ssa-strlen.c (get_string_cst): New.
8079 (handle_char_store): Call get_string_cst.
8081 2018-01-03 Martin Liska <mliska@suse.cz>
8083 PR tree-optimization/83593
8084 * tree-ssa-strlen.c: Include tree-cfg.h.
8085 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8086 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8087 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8089 (strlen_dom_walker::before_dom_children): Call
8090 gimple_purge_dead_eh_edges. Dump tranformation with details
8092 (strlen_dom_walker::before_dom_children): Update call by adding
8093 new argument cleanup_eh.
8094 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8096 2018-01-03 Martin Liska <mliska@suse.cz>
8099 * cif-code.def (VARIADIC_THUNK): New enum value.
8100 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8103 2018-01-03 Jan Beulich <jbeulich@suse.com>
8105 * sse.md (mov<mode>_internal): Tighten condition for when to use
8106 vmovdqu<ssescalarsize> for TI and OI modes.
8108 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8110 Update copyright years.
8112 2018-01-03 Martin Liska <mliska@suse.cz>
8115 * ipa-visibility.c (function_and_variable_visibility): Skip
8116 functions with noipa attribure.
8118 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8120 * gcc.c (process_command): Update copyright notice dates.
8121 * gcov-dump.c (print_version): Ditto.
8122 * gcov.c (print_version): Ditto.
8123 * gcov-tool.c (print_version): Ditto.
8124 * gengtype.c (create_file): Ditto.
8125 * doc/cpp.texi: Bump @copying's copyright year.
8126 * doc/cppinternals.texi: Ditto.
8127 * doc/gcc.texi: Ditto.
8128 * doc/gccint.texi: Ditto.
8129 * doc/gcov.texi: Ditto.
8130 * doc/install.texi: Ditto.
8131 * doc/invoke.texi: Ditto.
8133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8135 * vector-builder.h (vector_builder::m_full_nelts): Change from
8136 unsigned int to poly_uint64.
8137 (vector_builder::full_nelts): Update prototype accordingly.
8138 (vector_builder::new_vector): Likewise.
8139 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8140 (vector_builder::operator ==): Likewise.
8141 (vector_builder::finalize): Likewise.
8142 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8143 Take the number of elements as a poly_uint64 rather than an
8145 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8146 from unsigned int to poly_uint64.
8147 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8148 (vec_perm_indices::new_vector): Likewise.
8149 (vec_perm_indices::length): Likewise.
8150 (vec_perm_indices::nelts_per_input): Likewise.
8151 (vec_perm_indices::input_nelts): Likewise.
8152 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8153 number of elements per input as a poly_uint64 rather than an
8154 unsigned int. Use the original encoding for variable-length
8155 vectors, rather than clamping each individual element.
8156 For the second and subsequent elements in each pattern,
8157 clamp the step and base before clamping their sum.
8158 (vec_perm_indices::series_p): Handle polynomial element counts.
8159 (vec_perm_indices::all_in_range_p): Likewise.
8160 (vec_perm_indices_to_tree): Likewise.
8161 (vec_perm_indices_to_rtx): Likewise.
8162 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8163 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8164 (tree_vector_builder::new_binary_operation): Handle polynomial
8165 element counts. Return false if we need to know the number
8166 of elements at compile time.
8167 * fold-const.c (fold_vec_perm): Punt if the number of elements
8168 isn't known at compile time.
8170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8172 * vec-perm-indices.h (vec_perm_builder): Change element type
8173 from HOST_WIDE_INT to poly_int64.
8174 (vec_perm_indices::element_type): Update accordingly.
8175 (vec_perm_indices::clamp): Handle polynomial element_types.
8176 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8177 (vec_perm_indices::all_in_range_p): Likewise.
8178 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8180 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8181 polynomial vec_perm_indices element types.
8182 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8183 * fold-const.c (fold_vec_perm): Likewise.
8184 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8185 * tree-vect-generic.c (lower_vec_perm): Likewise.
8186 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8187 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8188 element type to HOST_WIDE_INT.
8190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8191 Alan Hayward <alan.hayward@arm.com>
8192 David Sherwood <david.sherwood@arm.com>
8194 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8195 rather than an int. Use plus_constant.
8196 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8197 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8200 Alan Hayward <alan.hayward@arm.com>
8201 David Sherwood <david.sherwood@arm.com>
8203 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8204 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
8211 mode sizes. Require a constant size for BLKmode parameters
8212 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8213 forces a parameter to be padded at the lsb end in order to
8214 fill a complete number of words, require the parameter size
8215 to be ordered wrt UNITS_PER_WORD.
8217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8218 Alan Hayward <alan.hayward@arm.com>
8219 David Sherwood <david.sherwood@arm.com>
8221 * reload1.c (spill_stack_slot_width): Change element type
8222 from unsigned int to poly_uint64_pod.
8223 (alter_reg): Treat mode sizes as polynomial.
8225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8226 Alan Hayward <alan.hayward@arm.com>
8227 David Sherwood <david.sherwood@arm.com>
8229 * reload.c (complex_word_subreg_p): New function.
8230 (reload_inner_reg_of_subreg, push_reload): Use it.
8232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8233 Alan Hayward <alan.hayward@arm.com>
8234 David Sherwood <david.sherwood@arm.com>
8236 * lra-constraints.c (process_alt_operands): Reject matched
8237 operands whose sizes aren't ordered.
8238 (match_reload): Refer to this check here.
8240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8241 Alan Hayward <alan.hayward@arm.com>
8242 David Sherwood <david.sherwood@arm.com>
8244 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8245 that the mode size is in the set {1, 2, 4, 8, 16}.
8247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8248 Alan Hayward <alan.hayward@arm.com>
8249 David Sherwood <david.sherwood@arm.com>
8251 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8252 Use plus_constant instead of gen_rtx_PLUS.
8254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8255 Alan Hayward <alan.hayward@arm.com>
8256 David Sherwood <david.sherwood@arm.com>
8258 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8259 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8260 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8261 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8262 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8263 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8264 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8265 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8266 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8267 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8269 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8270 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8271 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8272 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8273 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8274 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8275 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8276 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8277 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8278 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8280 * expr.c (emit_move_resolve_push): Treat the input and result
8281 of PUSH_ROUNDING as a poly_int64.
8282 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8283 (emit_push_insn): Likewise.
8284 * lra-eliminations.c (mark_not_eliminable): Likewise.
8285 * recog.c (push_operand): Likewise.
8286 * reload1.c (elimination_effects): Likewise.
8287 * rtlanal.c (nonzero_bits1): Likewise.
8288 * calls.c (store_one_arg): Likewise. Require the padding to be
8289 known at compile time.
8291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8292 Alan Hayward <alan.hayward@arm.com>
8293 David Sherwood <david.sherwood@arm.com>
8295 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8296 Use plus_constant instead of gen_rtx_PLUS.
8298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8299 Alan Hayward <alan.hayward@arm.com>
8300 David Sherwood <david.sherwood@arm.com>
8302 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8306 Alan Hayward <alan.hayward@arm.com>
8307 David Sherwood <david.sherwood@arm.com>
8309 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8310 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8311 via stack temporaries. Treat the mode size as polynomial too.
8313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8314 Alan Hayward <alan.hayward@arm.com>
8315 David Sherwood <david.sherwood@arm.com>
8317 * expr.c (expand_expr_real_2): When handling conversions involving
8318 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8319 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8320 as a poly_uint64 too.
8322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8323 Alan Hayward <alan.hayward@arm.com>
8324 David Sherwood <david.sherwood@arm.com>
8326 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8329 Alan Hayward <alan.hayward@arm.com>
8330 David Sherwood <david.sherwood@arm.com>
8332 * combine.c (can_change_dest_mode): Handle polynomial
8333 REGMODE_NATURAL_SIZE.
8334 * expmed.c (store_bit_field_1): Likewise.
8335 * expr.c (store_constructor): Likewise.
8336 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8337 and polynomial REGMODE_NATURAL_SIZE.
8338 (gen_lowpart_common): Likewise.
8339 * reginfo.c (record_subregs_of_mode): Likewise.
8340 * rtlanal.c (read_modify_subreg_p): Likewise.
8342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8343 Alan Hayward <alan.hayward@arm.com>
8344 David Sherwood <david.sherwood@arm.com>
8346 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8347 numbers of elements.
8349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8350 Alan Hayward <alan.hayward@arm.com>
8351 David Sherwood <david.sherwood@arm.com>
8353 * match.pd: Cope with polynomial numbers of vector elements.
8355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8356 Alan Hayward <alan.hayward@arm.com>
8357 David Sherwood <david.sherwood@arm.com>
8359 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8360 in a POINTER_PLUS_EXPR.
8362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8363 Alan Hayward <alan.hayward@arm.com>
8364 David Sherwood <david.sherwood@arm.com>
8366 * omp-simd-clone.c (simd_clone_subparts): New function.
8367 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8368 (ipa_simd_modify_function_body): Likewise.
8370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8371 Alan Hayward <alan.hayward@arm.com>
8372 David Sherwood <david.sherwood@arm.com>
8374 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8375 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8376 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8377 (expand_vector_condition, vector_element): Likewise.
8378 (subparts_gt): New function.
8379 (get_compute_type): Use subparts_gt.
8380 (count_type_subparts): Delete.
8381 (expand_vector_operations_1): Use subparts_gt instead of
8382 count_type_subparts.
8384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8385 Alan Hayward <alan.hayward@arm.com>
8386 David Sherwood <david.sherwood@arm.com>
8388 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8389 (vect_compile_time_alias): ...this new function. Do the calculation
8390 on poly_ints rather than trees.
8391 (vect_prune_runtime_alias_test_list): Update call accordingly.
8393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8394 Alan Hayward <alan.hayward@arm.com>
8395 David Sherwood <david.sherwood@arm.com>
8397 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8399 (vect_schedule_slp_instance): Likewise.
8401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8402 Alan Hayward <alan.hayward@arm.com>
8403 David Sherwood <david.sherwood@arm.com>
8405 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8406 constant and extern definitions for variable-length vectors.
8407 (vect_get_constant_vectors): Note that the number of units
8408 is known to be constant.
8410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8411 Alan Hayward <alan.hayward@arm.com>
8412 David Sherwood <david.sherwood@arm.com>
8414 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8415 of units as polynomial. Choose between WIDE and NARROW based
8418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8419 Alan Hayward <alan.hayward@arm.com>
8420 David Sherwood <david.sherwood@arm.com>
8422 * tree-vect-stmts.c (simd_clone_subparts): New function.
8423 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8426 Alan Hayward <alan.hayward@arm.com>
8427 David Sherwood <david.sherwood@arm.com>
8429 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8430 vectors as polynomial. Use build_index_vector for
8433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8434 Alan Hayward <alan.hayward@arm.com>
8435 David Sherwood <david.sherwood@arm.com>
8437 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8438 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8439 for variable-length vectors.
8440 (vectorizable_mask_load_store): Treat the number of units as
8441 polynomial, asserting that it is constant if the condition has
8442 already been enforced.
8443 (vectorizable_store, vectorizable_load): Likewise.
8445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8446 Alan Hayward <alan.hayward@arm.com>
8447 David Sherwood <david.sherwood@arm.com>
8449 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8450 of units as polynomial. Punt if we can't tell at compile time
8451 which vector contains the final result.
8453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8454 Alan Hayward <alan.hayward@arm.com>
8455 David Sherwood <david.sherwood@arm.com>
8457 * tree-vect-loop.c (vectorizable_induction): Treat the number
8458 of units as polynomial. Punt on SLP inductions. Use an integer
8459 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8460 cast of such a series for variable-length floating-point
8463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8464 Alan Hayward <alan.hayward@arm.com>
8465 David Sherwood <david.sherwood@arm.com>
8467 * tree.h (build_index_vector): Declare.
8468 * tree.c (build_index_vector): New function.
8469 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8470 of units as polynomial, forcibly converting it to a constant if
8471 vectorizable_reduction has already enforced the condition.
8472 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8473 to create a {1,2,3,...} vector.
8474 (vectorizable_reduction): Treat the number of units as polynomial.
8475 Choose vectype_in based on the largest scalar element size rather
8476 than the smallest number of units. Enforce the restrictions
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8484 number of units as polynomial.
8486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8487 Alan Hayward <alan.hayward@arm.com>
8488 David Sherwood <david.sherwood@arm.com>
8490 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8491 * target.def (autovectorize_vector_sizes): Return the vector sizes
8492 by pointer, using vector_sizes rather than a bitmask.
8493 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8494 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8495 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8497 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8498 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8499 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8500 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8501 * omp-general.c (omp_max_vf): Likewise.
8502 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8503 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8504 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8505 * tree-vect-slp.c (vect_slp_bb): Likewise.
8506 * doc/tm.texi: Regenerate.
8507 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8509 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8510 the vector size as a poly_uint64 rather than an unsigned int.
8511 (current_vector_size): Change from an unsigned int to a poly_uint64.
8512 (get_vectype_for_scalar_type): Update accordingly.
8513 * tree.h (build_truth_vector_type): Take the size and number of
8514 units as a poly_uint64 rather than an unsigned int.
8515 (build_vector_type): Add a temporary overload that takes
8516 the number of units as a poly_uint64 rather than an unsigned int.
8517 * tree.c (make_vector_type): Likewise.
8518 (build_truth_vector_type): Take the number of units as a poly_uint64
8519 rather than an unsigned int.
8521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8522 Alan Hayward <alan.hayward@arm.com>
8523 David Sherwood <david.sherwood@arm.com>
8525 * target.def (get_mask_mode): Take the number of units and length
8526 as poly_uint64s rather than unsigned ints.
8527 * targhooks.h (default_get_mask_mode): Update accordingly.
8528 * targhooks.c (default_get_mask_mode): Likewise.
8529 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8530 * doc/tm.texi: Regenerate.
8532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8533 Alan Hayward <alan.hayward@arm.com>
8534 David Sherwood <david.sherwood@arm.com>
8536 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8537 * omp-general.c (omp_max_vf): Likewise.
8538 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8539 (expand_omp_simd): Handle polynomial safelen.
8540 * omp-low.c (omplow_simd_context): Add a default constructor.
8541 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8542 (lower_rec_simd_input_clauses): Update accordingly.
8543 (lower_rec_input_clauses): Likewise.
8545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8546 Alan Hayward <alan.hayward@arm.com>
8547 David Sherwood <david.sherwood@arm.com>
8549 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8550 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8551 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8552 (vect_analyze_slp_cost): Likewise.
8553 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8554 (vect_model_load_cost): Likewise.
8556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8557 Alan Hayward <alan.hayward@arm.com>
8558 David Sherwood <david.sherwood@arm.com>
8560 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8561 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8562 from an unsigned int * to a poly_uint64_pod *.
8563 (calculate_unrolling_factor): New function.
8564 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8567 Alan Hayward <alan.hayward@arm.com>
8568 David Sherwood <david.sherwood@arm.com>
8570 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8571 from an unsigned int to a poly_uint64.
8572 (_loop_vec_info::slp_unrolling_factor): Likewise.
8573 (_loop_vec_info::vectorization_factor): Change from an int
8575 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8576 (vect_get_num_vectors): New function.
8577 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8578 (vect_get_num_copies): Use vect_get_num_vectors.
8579 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8580 to an unsigned int *.
8581 (vect_analyze_data_refs): Change min_vf from an int * to a
8583 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8584 than an unsigned HOST_WIDE_INT.
8585 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8586 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8587 to an unsigned int *.
8588 (vect_analyze_data_ref_dependences): Likewise.
8589 (vect_compute_data_ref_alignment): Handle polynomial vf.
8590 (vect_enhance_data_refs_alignment): Likewise.
8591 (vect_prune_runtime_alias_test_list): Likewise.
8592 (vect_shift_permute_load_chain): Likewise.
8593 (vect_supportable_dr_alignment): Likewise.
8594 (dependence_distance_ge_vf): Take the vectorization factor as a
8595 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8596 (vect_analyze_data_refs): Change min_vf from an int * to a
8598 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8599 vfm1 as a poly_uint64 rather than an int. Make the same change
8600 for the returned bound_scalar.
8601 (vect_gen_vector_loop_niters): Handle polynomial vf.
8602 (vect_do_peeling): Likewise. Update call to
8603 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8604 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8606 * tree-vect-loop.c (vect_determine_vectorization_factor)
8607 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8608 (vect_get_known_peeling_cost): Likewise.
8609 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8610 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8611 (vect_transform_loop): Likewise. Use the lowest possible VF when
8612 updating the upper bounds of the loop.
8613 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8615 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8616 polynomial unroll factors.
8617 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8618 (vect_make_slp_decision): Likewise.
8619 (vect_supported_load_permutation_p): Likewise, and polynomial
8621 (vect_analyze_slp_cost): Handle polynomial vf.
8622 (vect_slp_analyze_node_operations): Likewise.
8623 (vect_slp_analyze_bb_1): Likewise.
8624 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8625 than an unsigned HOST_WIDE_INT.
8626 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8627 (vectorizable_load): Handle polynomial vf.
8628 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8630 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8633 Alan Hayward <alan.hayward@arm.com>
8634 David Sherwood <david.sherwood@arm.com>
8636 * match.pd: Handle bit operations involving three constants
8637 and try to fold one pair.
8639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8641 * tree-vect-loop-manip.c: Include gimple-fold.h.
8642 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8643 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8644 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8645 Add a path that uses a step of VF instead of 1, but disable it
8647 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8648 and niters_no_overflow parameters. Update calls to
8649 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8650 Create a new SSA name if the latter choses to use a ste other
8651 than zero, and return it via niters_vector_mult_vf_var.
8652 * tree-vect-loop.c (vect_transform_loop): Update calls to
8653 vect_do_peeling, vect_gen_vector_loop_niters and
8654 slpeel_make_loop_iterate_ntimes.
8655 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8656 (vect_gen_vector_loop_niters): Update declarations after above changes.
8658 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8660 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8661 128-bit round to integer instructions.
8662 (ceil<mode>2): Likewise.
8663 (btrunc<mode>2): Likewise.
8664 (round<mode>2): Likewise.
8666 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8668 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8669 unaligned VSX load/store on P8/P9.
8670 (expand_block_clear): Allow the use of unaligned VSX
8671 load/store on P8/P9.
8673 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8675 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8677 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8678 swap associated with both a load and a store.
8680 2018-01-02 Andrew Waterman <andrew@sifive.com>
8682 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8683 * config/riscv/riscv.md (clear_cache): Use it.
8685 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8687 * web.c: Remove out-of-date comment.
8689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8691 * expr.c (fixup_args_size_notes): Check that any existing
8692 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8693 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8694 (emit_single_push_insn): ...here.
8696 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8698 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8699 (const_vector_encoded_nelts): New function.
8700 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8701 (const_vector_int_elt, const_vector_elt): Declare.
8702 * emit-rtl.c (const_vector_int_elt_1): New function.
8703 (const_vector_elt): Likewise.
8704 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8705 of CONST_VECTOR_ELT.
8707 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8709 * expr.c: Include rtx-vector-builder.h.
8710 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8711 directly on the tree encoding.
8712 (const_vector_from_tree): Likewise.
8713 * optabs.c: Include rtx-vector-builder.h.
8714 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8715 sequence of "u" values.
8716 * vec-perm-indices.c: Include rtx-vector-builder.h.
8717 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8718 directly on the vec_perm_indices encoding.
8720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8722 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8723 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8724 * rtx-vector-builder.h: New file.
8725 * rtx-vector-builder.c: Likewise.
8726 * rtl.h (rtx_def::u2): Add a const_vector field.
8727 (CONST_VECTOR_NPATTERNS): New macro.
8728 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8729 (CONST_VECTOR_DUPLICATE_P): Likewise.
8730 (CONST_VECTOR_STEPPED_P): Likewise.
8731 (CONST_VECTOR_ENCODED_ELT): Likewise.
8732 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8733 (unwrap_const_vec_duplicate): Likewise.
8734 (const_vec_series_p): Check for a non-duplicated vector encoding.
8735 Say that the function only returns true for integer vectors.
8736 * emit-rtl.c: Include rtx-vector-builder.h.
8737 (gen_const_vec_duplicate_1): Delete.
8738 (gen_const_vector): Call gen_const_vec_duplicate instead of
8739 gen_const_vec_duplicate_1.
8740 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8741 (gen_const_vec_duplicate): Use rtx_vector_builder.
8742 (gen_const_vec_series): Likewise.
8743 (gen_rtx_CONST_VECTOR): Likewise.
8744 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8745 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8746 Build a new vector rather than modifying a CONST_VECTOR in-place.
8747 (handle_special_swappables): Update call accordingly.
8748 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8749 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8750 Build a new vector rather than modifying a CONST_VECTOR in-place.
8751 (handle_special_swappables): Update call accordingly.
8753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8755 * simplify-rtx.c (simplify_const_binary_operation): Use
8756 CONST_VECTOR_ELT instead of XVECEXP.
8758 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8760 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8761 the selector elements to be different from the data elements
8762 if the selector is a VECTOR_CST.
8763 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8764 ssizetype for the selector.
8766 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8768 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8769 before testing each element individually.
8770 * tree-vect-generic.c (lower_vec_perm): Likewise.
8772 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8774 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8775 * selftest-run-tests.c (selftest::run_tests): Call it.
8776 * vector-builder.h (vector_builder::operator ==): New function.
8777 (vector_builder::operator !=): Likewise.
8778 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8779 (vec_perm_indices::all_from_input_p): New function.
8780 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8781 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8782 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8783 instead of reading the VECTOR_CST directly. Detect whether both
8784 vector inputs are the same before constructing the vec_perm_indices,
8785 and update the number of inputs argument accordingly. Use the
8786 utility functions added above. Only construct sel2 if we need to.
8788 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8790 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8791 the broadcast of the low byte.
8792 (expand_mult_highpart): Use an explicit encoding for the permutes.
8793 * optabs-query.c (can_mult_highpart_p): Likewise.
8794 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8795 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8796 (vectorizable_bswap): Likewise.
8797 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8798 explicit encoding for the power-of-2 permutes.
8799 (vect_permute_store_chain): Likewise.
8800 (vect_grouped_load_supported): Likewise.
8801 (vect_permute_load_chain): Likewise.
8803 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8805 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8806 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8807 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8808 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8809 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8810 (vect_gen_perm_mask_any): Likewise.
8812 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8814 * int-vector-builder.h: New file.
8815 * vec-perm-indices.h: Include int-vector-builder.h.
8816 (vec_perm_indices): Redefine as an int_vector_builder.
8817 (auto_vec_perm_indices): Delete.
8818 (vec_perm_builder): Redefine as a stand-alone class.
8819 (vec_perm_indices::vec_perm_indices): New function.
8820 (vec_perm_indices::clamp): Likewise.
8821 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8822 (vec_perm_indices::new_vector): New function.
8823 (vec_perm_indices::new_expanded_vector): Update for new
8824 vec_perm_indices class.
8825 (vec_perm_indices::rotate_inputs): New function.
8826 (vec_perm_indices::all_in_range_p): Operate directly on the
8827 encoded form, without computing elided elements.
8828 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8829 encoding. Update for new vec_perm_indices class.
8830 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8831 the given vec_perm_builder.
8832 (expand_vec_perm_var): Update vec_perm_builder constructor.
8833 (expand_mult_highpart): Use vec_perm_builder instead of
8834 auto_vec_perm_indices.
8835 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8836 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8837 or double series encoding as appropriate.
8838 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8839 vec_perm_indices instead of auto_vec_perm_indices.
8840 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8841 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8842 (vect_permute_store_chain): Likewise.
8843 (vect_grouped_load_supported): Likewise.
8844 (vect_permute_load_chain): Likewise.
8845 (vect_shift_permute_load_chain): Likewise.
8846 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8847 (vect_transform_slp_perm_load): Likewise.
8848 (vect_schedule_slp_instance): Likewise.
8849 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8850 (vectorizable_mask_load_store): Likewise.
8851 (vectorizable_bswap): Likewise.
8852 (vectorizable_store): Likewise.
8853 (vectorizable_load): Likewise.
8854 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8855 vec_perm_indices instead of auto_vec_perm_indices. Use
8856 tree_to_vec_perm_builder to read the vector from a tree.
8857 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8858 vec_perm_builder instead of a vec_perm_indices.
8859 (have_whole_vector_shift): Use vec_perm_builder and
8860 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8861 truncation to calc_vec_perm_mask_for_shift.
8862 (vect_create_epilog_for_reduction): Likewise.
8863 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8864 from auto_vec_perm_indices to vec_perm_indices.
8865 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8866 instead of changing individual elements.
8867 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8868 the vector in d.perm.
8869 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8870 from auto_vec_perm_indices to vec_perm_indices.
8871 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8872 instead of changing individual elements.
8873 (arm_vectorize_vec_perm_const): Use new_vector to install
8874 the vector in d.perm.
8875 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8876 Update vec_perm_builder constructor.
8877 (rs6000_expand_interleave): Likewise.
8878 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8879 (rs6000_expand_interleave): Likewise.
8881 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8883 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8884 to qimode could truncate the indices.
8885 * optabs.c (expand_vec_perm_var): Likewise.
8887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8889 * Makefile.in (OBJS): Add vec-perm-indices.o.
8890 * vec-perm-indices.h: New file.
8891 * vec-perm-indices.c: Likewise.
8892 * target.h (vec_perm_indices): Replace with a forward class
8894 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8895 * optabs.h: Include vec-perm-indices.h.
8896 (expand_vec_perm): Delete.
8897 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8898 (expand_vec_perm_const): Declare.
8899 * target.def (vec_perm_const_ok): Replace with...
8900 (vec_perm_const): ...this new hook.
8901 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8902 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8903 * doc/tm.texi: Regenerate.
8904 * optabs.def (vec_perm_const): Delete.
8905 * doc/md.texi (vec_perm_const): Likewise.
8906 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8907 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8908 expand_vec_perm for constant permutation vectors. Assert that
8909 the mode of variable permutation vectors is the integer equivalent
8910 of the mode that is being permuted.
8911 * optabs-query.h (selector_fits_mode_p): Declare.
8912 * optabs-query.c: Include vec-perm-indices.h.
8913 (selector_fits_mode_p): New function.
8914 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8915 is defined, instead of checking whether the vec_perm_const_optab
8916 exists. Use targetm.vectorize.vec_perm_const instead of
8917 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8918 fit in the vector mode before using a variable permute.
8919 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8920 vec_perm_indices instead of an rtx.
8921 (expand_vec_perm): Replace with...
8922 (expand_vec_perm_const): ...this new function. Take the selector
8923 as a vec_perm_indices rather than an rtx. Also take the mode of
8924 the selector. Update call to shift_amt_for_vec_perm_mask.
8925 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8926 Use vec_perm_indices::new_expanded_vector to expand the original
8927 selector into bytes. Check whether the indices fit in the vector
8928 mode before using a variable permute.
8929 (expand_vec_perm_var): Make global.
8930 (expand_mult_highpart): Use expand_vec_perm_const.
8931 * fold-const.c: Includes vec-perm-indices.h.
8932 * tree-ssa-forwprop.c: Likewise.
8933 * tree-vect-data-refs.c: Likewise.
8934 * tree-vect-generic.c: Likewise.
8935 * tree-vect-loop.c: Likewise.
8936 * tree-vect-slp.c: Likewise.
8937 * tree-vect-stmts.c: Likewise.
8938 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8940 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8941 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8942 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8943 (aarch64_vectorize_vec_perm_const): ...this new function.
8944 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8945 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8946 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8947 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8948 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8949 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8950 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8952 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8953 check for NEON modes.
8954 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8955 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8956 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8957 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8959 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8960 the old VEC_PERM_CONST conditions.
8961 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8962 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8963 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8964 (ia64_vectorize_vec_perm_const_ok): Merge into...
8965 (ia64_vectorize_vec_perm_const): ...this new function.
8966 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8967 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8968 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8969 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8970 * config/mips/mips.c (mips_expand_vec_perm_const)
8971 (mips_vectorize_vec_perm_const_ok): Merge into...
8972 (mips_vectorize_vec_perm_const): ...this new function.
8973 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8974 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8975 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8976 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8977 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8978 (rs6000_expand_vec_perm_const): Delete.
8979 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8981 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8982 (altivec_expand_vec_perm_const_le): Take each operand individually.
8983 Operate on constant selectors rather than rtxes.
8984 (altivec_expand_vec_perm_const): Likewise. Update call to
8985 altivec_expand_vec_perm_const_le.
8986 (rs6000_expand_vec_perm_const): Delete.
8987 (rs6000_vectorize_vec_perm_const_ok): Delete.
8988 (rs6000_vectorize_vec_perm_const): New function.
8989 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8990 an element count and rtx array.
8991 (rs6000_expand_extract_even): Update call accordingly.
8992 (rs6000_expand_interleave): Likewise.
8993 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8994 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8995 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8996 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8997 (rs6000_expand_vec_perm_const): Delete.
8998 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8999 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9000 (altivec_expand_vec_perm_const_le): Take each operand individually.
9001 Operate on constant selectors rather than rtxes.
9002 (altivec_expand_vec_perm_const): Likewise. Update call to
9003 altivec_expand_vec_perm_const_le.
9004 (rs6000_expand_vec_perm_const): Delete.
9005 (rs6000_vectorize_vec_perm_const_ok): Delete.
9006 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9007 reference to the SPE evmerge intructions.
9008 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9009 an element count and rtx array.
9010 (rs6000_expand_extract_even): Update call accordingly.
9011 (rs6000_expand_interleave): Likewise.
9012 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9013 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9015 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9017 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9019 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9020 vector mode and that that mode matches the mode of the data
9022 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9023 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9024 directly using expand_vec_perm_1 when forcing selectors into
9026 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9028 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9030 * optabs-query.h (can_vec_perm_p): Delete.
9031 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9032 * optabs-query.c (can_vec_perm_p): Split into...
9033 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9034 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9035 particular selector is valid.
9036 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9037 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9038 (vect_grouped_load_supported): Likewise.
9039 (vect_shift_permute_load_chain): Likewise.
9040 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9041 (vect_transform_slp_perm_load): Likewise.
9042 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9043 (vectorizable_bswap): Likewise.
9044 (vect_gen_perm_mask_checked): Likewise.
9045 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9046 implementations of variable permutation vectors into account
9047 when deciding which selector to use.
9048 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9049 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9050 with a false third argument.
9051 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9052 to test whether the constant selector is valid and can_vec_perm_var_p
9053 to test whether a variable selector is valid.
9055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9057 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9058 * optabs-query.c (can_vec_perm_p): Likewise.
9059 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9060 instead of vec_perm_indices.
9061 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9062 (vect_gen_perm_mask_checked): Likewise,
9063 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9064 (vect_gen_perm_mask_checked): Likewise,
9066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9068 * optabs-query.h (qimode_for_vec_perm): Declare.
9069 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9070 (qimode_for_vec_perm): ...this new function.
9071 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9073 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9075 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9076 does not have a conditional at the top.
9078 2018-01-02 Richard Biener <rguenther@suse.de>
9080 * ipa-inline.c (big_speedup_p): Fix expression.
9082 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9085 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9088 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9092 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9093 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9094 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9095 cond_taken_branch_cost 3->4.
9097 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9099 PR tree-optimization/83581
9100 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9101 TODO_cleanup_cfg if any changes have been made.
9104 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9105 convert_modes if target mode has the right side, but different mode
9109 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9110 last argument when extracting from CONCAT. If either from_real or
9111 from_imag is NULL, use expansion through memory. If result is not
9112 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9113 the parts directly to inner mode, if even that fails, use expansion
9117 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9118 check for bswap in mode rather than HImode and use that in expand_unop
9121 Copyright (C) 2018 Free Software Foundation, Inc.
9123 Copying and distribution of this file, with or without modification,
9124 are permitted in any medium without royalty provided the copyright
9125 notice and this notice are preserved.