1 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/84946
4 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5 bitsize + bitsize in poly_uint64 rather than poly_int64.
8 * dwarf2asm.c: Include fold-const.c.
9 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10 of decl rather than decl itself.
12 PR rtl-optimization/84643
13 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
18 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
19 calling assemble_variable.
21 2018-03-19 Sudakshina Das <sudi.das@arm.com>
24 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
25 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
27 2018-03-19 Jim Wilson <jimw@sifive.com>
30 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
31 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
32 (riscv_first_stack_step): Likewise.
33 (riscv_option_override): Use STACK_BOUNDARY instead of
35 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
37 (BIGGEST_ALIGNMENT): Set to 128.
38 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
39 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
42 2018-03-19 Richard Biener <rguenther@suse.de>
44 PR tree-optimization/84933
45 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
46 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
48 2018-03-19 Richard Biener <rguenther@suse.de>
50 PR tree-optimization/84859
51 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
52 (cond_if_else_store_replacement): Perform sinking operation on
53 single-store BBs regardless of MAX_STORES_TO_SINK setting.
54 Generalize what a BB with a single eligible store is.
56 2018-03-19 Richard Biener <rguenther@suse.de>
58 PR tree-optimization/84929
59 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
60 chrec_is_positive against non-chrec arg.
62 2018-03-19 Tamar Christina <tamar.christina@arm.com>
65 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
67 2018-03-18 Martin Liska <mliska@suse.cz>
69 PR rtl-optimization/84635
70 * regrename.c (build_def_use): Use matches_mode only when
73 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
75 PR tree-optimization/84913
76 * tree-vect-loop.c (vectorizable_reduction): Don't try to
77 vectorize chains of COND_EXPRs.
79 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
81 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
83 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
85 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
87 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
89 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
91 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
92 Kito Cheng <kito.cheng@gmail.com>
94 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
95 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
96 (nds32_adjust_reg_alloc_order): New function.
97 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
99 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
101 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
102 nds32_print_operand, nds32_print_operand_address): Use
103 HOST_WIDE_INT_PRINT_DEC instead.
105 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
107 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
109 2018-03-17 Jakub Jelinek <jakub@redhat.com>
112 * config/i386/i386.c (initial_ix86_tune_features,
113 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
115 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
116 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
117 rather than 1u << ix86_tune. Formatting fix.
118 (ix86_option_override_internal): Change ix86_arch_mask from
119 unsigned int to unsigned HOST_WIDE_INT, initialize to
120 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
121 (ix86_function_specific_restore): Likewise.
123 2018-03-16 Jakub Jelinek <jakub@redhat.com>
126 * postreload.c (reload_combine_recognize_pattern): Perform
127 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
128 truncate_int_for_mode the result for the destination's mode.
131 * hsa-gen.c (mem_type_for_type): Fix comment typo.
132 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
134 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
137 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
140 * lra-assigns.c (lra_split_hard_reg_for): Don't use
141 regno_allocno_class_array and sorted_pseudos.
142 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
143 insns where regno is used.
145 2018-03-16 Martin Liska <mliska@suse.cz>
148 * multiple_target.c (create_dispatcher_calls): Redirect
149 reference in the symbol table.
151 2018-03-16 Martin Liska <mliska@suse.cz>
154 * multiple_target.c (create_dispatcher_calls): Redirect also
157 2018-03-16 Jakub Jelinek <jakub@redhat.com>
161 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
162 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
163 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
165 2018-03-16 Julia Koval <julia.koval@intel.com>
167 * doc/invoke.texi (Skylake Server): Add CLWB.
168 Cannonlake): Remove CLWB.
170 2018-03-16 Jakub Jelinek <jakub@redhat.com>
172 PR tree-optimization/84841
173 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
175 (FLOAT_ONE_CONST_TYPE): Define.
176 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
177 (sort_by_operand_rank): Put entries with higher constant_type last
178 rather than first to match comments.
180 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
182 * config/nios2/nios2.md (movsi_internal): Fix thinko in
185 2018-03-15 Jakub Jelinek <jakub@redhat.com>
188 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
189 check and use address of target always.
191 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
194 * config/i386/i386.c (indirect_thunk_needed): Update comments.
195 (indirect_thunk_bnd_needed): Likewise.
196 (indirect_thunks_used): Likewise.
197 (indirect_thunks_bnd_used): Likewise.
198 (indirect_return_needed): New.
199 (indirect_return_bnd_needed): Likewise.
200 (output_indirect_thunk_function): Add a bool argument for
202 (output_indirect_thunk_function): Don't generate alias for
203 function return thunk.
204 (ix86_code_end): Call output_indirect_thunk_function to generate
205 function return thunks.
206 (ix86_output_function_return): Set indirect_return_bnd_needed
207 and indirect_return_needed instead of indirect_thunk_bnd_needed
208 and indirect_thunk_needed.
210 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
212 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
213 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
214 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
216 2018-03-15 David Malcolm <dmalcolm@redhat.com>
217 Paul Hua <paul.hua.gm@gmail.com>
220 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
222 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
224 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
225 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
228 2018-03-15 Tamar Christina <tamar.christina@arm.com>
231 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
232 instead of GET_MODE_SIZE when comparing Units.
234 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
237 * varasm.c (hash_section): Return an unchangeble hash value
238 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
239 Return !aarch64_can_use_per_function_literal_pools_p ().
241 2018-03-15 Jakub Jelinek <jakub@redhat.com>
244 * optabs.c (emit_conditional_move): Pass address of cmode's copy
245 rather than address of cmode as last argument to prepare_cmp_insn.
247 2018-03-15 Julia Koval <julia.koval@intel.com>
249 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
250 F_AVX512VNNI, F_AVX512BITALG): New.
252 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
255 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
256 insn for floating-point loads and stores.
258 2018-03-14 Carl Love <cel@us.ibm.com>
260 * config/rs6000/rs6000-c.c: Add macro definitions for
261 ALTIVEC_BUILTIN_VEC_PERMXOR.
262 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
263 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
264 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
265 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
267 * config/doc/extend.texi: Add prototypes for vec_permxor.
269 2018-03-14 David Malcolm <dmalcolm@redhat.com>
272 * diagnostic-show-locus.c (class layout_point): Convert m_line
273 from int to linenum_type.
274 (line_span::comparator): Use linenum "compare" function when
275 comparing line numbers.
276 (test_line_span): New function.
277 (layout_range::contains_point): Convert param "row" from int to
279 (layout_range::intersects_line_p): Likewise.
280 (layout::will_show_line_p): Likewise.
281 (layout::print_source_line): Likewise.
282 (layout::should_print_annotation_line_p): Likewise.
283 (layout::print_annotation_line): Likewise.
284 (layout::print_leading_fixits): Likewise.
285 (layout::annotation_line_showed_range_p): Likewise.
286 (struct line_corrections): Likewise for field m_row.
287 (line_corrections::line_corrections): Likewise for param "row".
288 (layout::print_trailing_fixits): Likewise.
289 (layout::get_state_at_point): Likewise.
290 (layout::get_x_bound_for_row): Likewise.
291 (layout::print_line): Likewise.
292 (diagnostic_show_locus): Likewise for locals "last_line" and
294 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
295 * input.c (selftest::test_linenum_comparisons): New function.
296 (selftest::input_c_tests): Call it.
297 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
298 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
299 * selftest.h (ASSERT_GT): New macro.
300 (ASSERT_GT_AT): New macro.
301 (ASSERT_LT): New macro.
302 (ASSERT_LT_AT): New macro.
304 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
306 PR rtl-optimization/84780
307 * combine.c (distribute_links): Don't make a link based on pc_rtx.
309 2018-03-14 Martin Liska <mliska@suse.cz>
311 * tree.c (record_node_allocation_statistics): Use
313 (get_stats_node_kind): New function extracted from
314 record_node_allocation_statistics.
315 (free_node): Use get_stats_node_kind.
317 2018-03-14 Richard Biener <rguenther@suse.de>
319 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
320 that the value-set of ANTIC_IN doesn't grow.
323 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
325 (BB_VISITED_WITH_VISITED_SUCCS): New define.
326 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
328 2018-03-14 Julia Koval <julia.koval@intel.com>
330 * config.gcc (icelake-client, icelake-server): New.
332 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
333 (initial_ix86_arch_features): Ditto.
334 (PTA_SKYLAKE): Add SGX.
335 (PTA_ICELAKE): Remove.
336 (PTA_ICELAKE_CLIENT): New.
337 (PTA_ICELAKE_SERVER): New.
338 (ix86_option_override_internal): Split up icelake on icelake client and
340 (get_builtin_code_for_version): Ditto.
341 (fold_builtin_cpu): Ditto.
342 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
343 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
344 * config/i386/i386.h (processor_type): Ditto.
345 * doc/invoke.texi: Ditto.
347 2018-03-14 Jakub Jelinek <jakub@redhat.com>
350 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
351 INTEGER_CST offset, add it together with bitpos / 8 and
352 sign extend based on POINTER_SIZE.
356 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
359 * config/i386/constraints.md (Yc): New register constraint.
360 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
361 Use Yc constraint for alternative 2 of operand 0. Remove
362 preferred_for_speed attribute.
364 2018-03-14 Richard Biener <rguenther@suse.de>
366 PR tree-optimization/84830
367 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
368 with the old one to avoid oscillations.
370 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
373 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
375 (assign_by_spills): Return a flag of reload assignment failure.
376 Do not process the reload assignment failures. Do not spill other
377 reload pseudos if they has the same reg class. Update n if
379 (lra_assign): Add a return arg. Set up from the result of
380 assign_by_spills call.
381 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
382 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
383 usage_insns if it is not NULL.
384 (spill_hard_reg_in_range): New function.
385 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
386 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
388 (lra_assign): Change prototype.
389 * lra.c (lra): Add code to deal with fails by splitting hard reg
392 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
394 * config/riscv/riscv.opt (mrelax): New option.
395 * config/riscv/riscv.c (riscv_file_start): Emit ".option
396 "norelax" when riscv_mrelax is disabled.
397 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
399 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
402 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
403 reassociation for int modes.
405 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
407 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
408 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
410 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
411 * config/aarch64/aarch64-sve.md
412 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
413 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
414 (*extend<mode><Vwide>2): Rename to...
415 (aarch64_sve_extend<mode><Vwide>2): ...this.
416 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
417 renaming the old pattern to...
418 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
420 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
421 define_expand, renaming the old pattern to...
422 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
423 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
424 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
425 account when deciding which SVE instruction the optab should use.
426 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
428 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
430 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
431 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
432 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
433 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
434 (tlsdesc_small_<mode>): Turn a define_expand and use
435 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
436 (tlsdesc_small_advsimd_<mode>): ...this.
437 (tlsdesc_small_sve_<mode>): New pattern.
439 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
441 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
442 (UNSPEC_UMUL_HIGHPART): New constants.
443 (MUL_HIGHPART): New int iteraor.
444 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
445 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
447 (*<su>mul<mode>3_highpart): New define_insn.
449 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
452 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
455 2018-03-13 Martin Liska <mliska@suse.cz>
458 * (sem_item_optimizer::sem_item_optimizer): Initialize new
460 (sem_item_optimizer::~sem_item_optimizer): Release it.
461 (sem_item_optimizer::merge_classes): Register variable aliases.
462 (sem_item_optimizer::fixup_pt_set): New function.
463 (sem_item_optimizer::fixup_points_to_sets): Likewise.
464 * ipa-icf.h: Declare new variables and functions.
466 2018-03-13 Jakub Jelinek <jakub@redhat.com>
469 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
470 integer_pow2p@2 and test integer_pow2p in condition.
471 (A < 0 ? C : 0): Similarly for @1.
474 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
475 characters starting at p contain '\0' character, don't look beyond
479 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
480 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
483 * reg-stack.c (change_stack): Change update_end var from int to
484 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
485 also call set_block_for_insn on the newly added insns and rescan.
488 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
492 * tree.c (stabilize_reference_1): Return save_expr (e) for
493 STATEMENT_LIST even if it doesn't have side-effects.
495 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
497 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
499 2018-03-12 Renlin Li <renlin.li@arm.com>
501 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
502 aarch64_output_scalar_simd_mov_immediate.
504 2018-03-12 Martin Sebor <msebor@redhat.com>
506 PR tree-optimization/83456
507 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
508 for perfectly overlapping calls to memcpy.
509 (gimple_fold_builtin_memory_chk): Same.
510 (gimple_fold_builtin_strcpy): Handle no-warning.
511 (gimple_fold_builtin_stxcpy_chk): Same.
512 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
514 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
516 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
517 parameter. Use it for SFmode.
518 (rs6000_function_arg_advance_1): Adjust.
519 (rs6000_function_arg): Adjust.
520 (rs6000_gimplify_va_arg): Pass false for that new parameter.
522 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
524 PR rtl-optimization/84169
525 PR rtl-optimization/84780
526 * combine.c (can_combine_p): Check for a 2-insn combination whether
527 the destination register is used between the two insns, too.
529 2018-03-12 Richard Biener <rguenther@suse.de>
531 PR tree-optimization/84803
532 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
533 for refs DR analysis didn't process.
535 2018-03-12 Richard Biener <rguenther@suse.de>
537 PR tree-optimization/84777
538 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
539 force-vectorize loops ignore whether we are optimizing for size.
541 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
543 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
544 (TARGET_MD_ASM_ADJUST): Define.
546 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
547 Kito Cheng <kito.cheng@gmail.com>
548 Chung-Ju Wu <jasonwucj@gmail.com>
550 * config/nds32/nds32.c (nds32_compute_stack_frame,
551 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
552 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
553 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
554 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
555 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
556 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
557 * config/nds32/nds32.md (prologue, epilogue): Use macro
558 NDS32_V3PUSH_AVAILABLE_P to do checking.
560 2018-03-11 Jakub Jelinek <jakub@redhat.com>
563 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
564 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
565 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
566 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
567 addition of most attributes on !orig_type_die or the attribute not
568 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
570 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
571 Chung-Ju Wu <jasonwucj@gmail.com>
573 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
575 * config/nds32/nds32.opt (mvh): New option.
577 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
578 Chung-Ju Wu <jasonwucj@gmail.com>
580 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
582 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
583 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
586 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
587 Chung-Ju Wu <jasonwucj@gmail.com>
589 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
591 * config/nds32/nds32-multiple.md (strlensi): New pattern.
592 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
594 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
595 Kito Cheng <kito.cheng@gmail.com>
596 Chung-Ju Wu <jasonwucj@gmail.com>
598 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
599 UNSPEC_FFMISM and UNSPEC_FLMISM.
600 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
601 for ffb, ffmism and flmism.
602 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
603 (unspec_ffmism): Ditto.
604 (unspec_flmism): Ditto.
605 (nds32_expand_builtin_impl): Check if string extension is available.
606 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
607 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
609 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
612 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
615 * lra-assigns.c (assign_by_spills): Return a flag of reload
616 assignment failure. Do not process the reload assignment
617 failures. Do not spill other reload pseudos if they has the same
619 (lra_assign): Add a return arg. Set up from the result of
620 assign_by_spills call.
621 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
622 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
623 usage_insns if it is not NULL.
624 (spill_hard_reg_in_range): New function.
625 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
626 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
628 (lra_assign): Change prototype.
629 * lra.c (lra): Add code to deal with fails by splitting hard reg
632 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
635 * config/i386/i386.opt: Replace Enforcment with Enforcement.
637 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
640 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
641 (dw_val_node): Add val_symbolic_view.
642 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
643 (symview_upper_bound): New.
644 (new_line_info_table): Initialize symviews_since_reset.
645 (dwarf2out_source_line): Count symviews_since_reset and set
647 (dw_val_equal_p): Handle symview.
648 (add_AT_symview): New.
649 (print_dw_val): Handle symview.
650 (attr_checksum, attr_checksum_ordered): Likewise.
651 (same_dw_val_p, size_of_die): Likewise.
652 (value_format, output_die): Likewise.
653 (add_high_low_attributes): Use add_AT_symview for entry_view.
654 (dwarf2out_finish): Reset symview_upper_bound, clear
657 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
660 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
661 Add strict argument and use it.
662 (rs6000_split_multireg_move): Update for new strict argument.
663 (mem_operand_gpr): Disallow all non-offsettable addresses.
664 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
666 2018-03-09 Jakub Jelinek <jakub@redhat.com>
669 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
670 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
671 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
674 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
675 decl, use remap_type if we want to use the type.
677 2018-03-09 Martin Sebor <msebor@redhat.com>
679 PR tree-optimization/84526
680 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
682 (builtin_access::generic_overlap): Be prepared to handle non-array
685 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
687 PR rtl-optimization/84682
688 * lra-constraints.c (process_address_1): Check is_address flag
689 for address constraints.
690 (process_alt_operands): Likewise.
691 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
692 preprocess_constraints.
693 * recog.h (preprocess_constraints): Add oploc parameter.
695 * recog.c (preprocess_constraints): Test address_operand for
696 CT_ADDRESS constraints.
698 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
701 * lra-assigns.c (assign_by_spills): Return a flag of reload
702 assignment failure. Do not process the reload assignment
703 failures. Do not spill other reload pseudos if they has the same
705 (lra_assign): Add a return arg. Set up from the result of
706 assign_by_spills call.
707 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
708 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
709 usage_insns if it is not NULL.
710 (spill_hard_reg_in_range): New function.
711 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
712 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
714 (lra_assign): Change prototype.
715 * lra.c (lra): Add code to deal with fails by splitting hard reg
718 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
721 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
722 Accept complain bool parameter. Only emit errors if it is true.
723 (arm_parse_cpu_option_name): Likewise.
724 (arm_target_thumb_only): Adjust callers of the above.
725 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
726 prototype to take a default true bool parameter.
727 (arm_parse_arch_option_name): Likewise.
729 2018-03-09 David Malcolm <dmalcolm@redhat.com>
730 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
734 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
735 * configure: Regenerate.
736 * configure.ac ("linker --version-script option"): New.
737 ("linker soname option"): New.
739 2018-03-09 Richard Biener <rguenther@suse.de>
741 PR tree-optimization/84775
742 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
743 immediate uses of predicate stmts and mark them modified.
746 PR tree-optimization/84178
747 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
749 (version_loop_for_if_conversion): Delay update_ssa call.
750 (tree_if_conversion): Delay update_ssa until after predicate
753 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
756 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
757 when the function accesses prior frames.
759 2018-03-08 Jakub Jelinek <jakub@redhat.com>
762 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
763 gen_llsym, otherwise call maybe_gen_llsym.
766 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
767 has ',' character inside of it.
769 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
772 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
773 as clobbering CC_REGNUM.
775 2018-03-08 Richard Biener <rguenther@suse.de>
778 * tree-scalar-evolution.c: Include tree-into-ssa.h.
779 (follow_copies_to_constant): Do not follow SSA names registered
782 2018-03-08 Richard Biener <rguenther@suse.de>
784 PR tree-optimization/84178
785 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
787 (version_loop_for_if_conversion): Delay update_ssa call.
788 (tree_if_conversion): Delay update_ssa until after predicate
791 2018-03-08 David Malcolm <dmalcolm@redhat.com>
793 PR tree-optimization/84178
794 * tree-if-conv.c (release_bb_predicate): Remove the
795 the assertion that the stmts have NULL use_ops.
796 Discard the statements, asserting that they haven't
797 yet been added to a BB.
799 2018-03-08 Richard Biener <rguenther@suse.de>
801 PR tree-optimization/84746
802 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
803 (phi_translate): Pass in destination ANTIC_OUT set.
804 (phi_translate_1): Likewise. For a simplified result lookup
805 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
806 (phi_translate_set): Adjust.
807 (do_pre_regular_insertion): Likewise.
808 (do_pre_partial_partial_insertion): Likewise.
810 2018-03-08 Martin Liska <mliska@suse.cz>
812 PR gcov-profile/84735
813 * doc/gcov.texi: Document usage of profile files.
814 * gcov-io.h: Document changes in the format.
816 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
820 * dwarf2out.c (struct dw_line_info_table): Update comments for
822 (FORCE_RESET_NEXT_VIEW): New.
823 (FORCE_RESETTING_VIEW_P): New.
824 (RESETTING_VIEW_P): Check for -1 too.
825 (ZERO_VIEW_P): Likewise.
826 (new_line_info_table): Force-reset next view.
827 (dwarf2out_begin_function): Likewise.
828 (dwarf2out_source_line): Simplify zero_view_p initialization.
829 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
830 view directly. Omit view when omitting .loc at line 0.
832 2018-03-08 Jakub Jelinek <jakub@redhat.com>
834 PR tree-optimization/84740
835 * tree-switch-conversion.c (process_switch): Call build_constructors
836 only if info.phi_count is non-zero.
838 PR tree-optimization/84739
839 * tree-tailcall.c (find_tail_calls): Check call arguments against
840 DECL_ARGUMENTS (current_function_decl) rather than
841 DECL_ARGUMENTS (func) when checking for tail recursion.
843 2018-03-07 Jakub Jelinek <jakub@redhat.com>
845 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
846 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
847 Volker Reichelt's entry and add entries for people that perform
848 GCC fuzzy testing and report numerous bugs.
850 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
853 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
854 readonly data in sdata, if that is disabled.
855 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
856 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
857 -mreadonly-in-sdata option.
859 2018-03-07 Martin Sebor <msebor@redhat.com>
861 PR tree-optimization/84468
862 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
863 basic block when looking for nul assignment.
865 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
868 * except.h (output_function_exception_table): Adjust prototype.
869 * except.c (output_function_exception_table): Remove FNNAME parameter
870 and add SECTION parameter. Ouput one part of the table at a time.
871 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
872 the first part of the exception table and emit unwind directives.
873 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
874 (i386_pe_seh_cold_init): Likewise.
875 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
876 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
877 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
878 (ix86_output_call_insn): Emit a nop in one more case for SEH.
879 * config/i386/winnt.c: Include except.h.
880 (struct seh_frame_state): Add reg_offset, after_prologue and
881 in_cold_section fields.
882 (i386_pe_seh_end_prologue): Set seh->after_prologue.
883 (i386_pe_seh_cold_init): New function.
884 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
885 to seh->in_cold_section.
886 (seh_emit_push): Record the offset of the push.
887 (seh_emit_save): Record the offet of the save.
888 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
889 Test seh->after_prologue to disregard the epilogue.
890 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
891 (i386_pe_end_cold_function): New function.
893 2018-03-07 Jakub Jelinek <jakub@redhat.com>
896 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
897 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
900 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
902 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
903 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
906 * multiple_target.c: Include tree-inline.h and intl.h.
907 (expand_target_clones): Diagnose and fail if node->definition and
908 !tree_versionable_function_p (node->decl).
910 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
912 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
914 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
915 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
916 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
918 2018-03-06 Jakub Jelinek <jakub@redhat.com>
921 * combine.c (try_combine): Use reg_or_subregno instead of handling
922 just paradoxical SUBREGs and REGs.
924 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
926 * config/arc/arc.c (arc_finalize_pic): Remove function.
927 (arc_must_save_register): We use single base PIC register, remove
928 checks to save/restore the PIC register.
929 (arc_expand_prologue): Likewise.
930 * config/arc/arc-protos.h (arc_set_default_type_attributes):
932 (arc_verify_short): Likewise.
933 (arc_attr_type): Likewise.
934 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
935 (walk_stores): Likewise.
936 (arc_address_cost): Make it static.
937 (arc_verify_short): Likewise.
938 (branch_dest): Likewise.
939 (arc_attr_type): Likewise.
940 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
941 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
942 (arc_final_prescan_insn): Remove inserting the nops due to
943 hardware hazards. It is done in reorg step.
944 (insn_length_variant_t): Remove.
945 (insn_length_parameters_t): Likewise.
946 (arc_insn_length_parameters): Likewise.
947 (arc_get_insn_variants): Likewise.
948 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
950 2018-03-06 Jakub Jelinek <jakub@redhat.com>
953 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
956 PR tree-optimization/84687
957 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
959 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
961 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
963 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
964 Rename to ppc_speculation_barrier.
965 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
966 __builtin_ppc_speculation_barrier.
968 2018-03-05 Jakub Jelinek <jakub@redhat.com>
971 * combine.c (combine_simplify_rtx): Don't try to simplify if
972 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
975 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
977 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
978 to 32 bytes when compiling for POWER9.
980 2018-03-05 Jakub Jelinek <jakub@redhat.com>
983 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
984 regparm >= 3 with no arg reg available also for calls with
985 flag_force_indirect_call. Pass decl to ix86_function_regparm.
988 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
990 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
992 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
995 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
997 2018-03-05 Richard Biener <rguenther@suse.de>
999 PR tree-optimization/84486
1000 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1001 When inserting a __builtin_assume_aligned call set the LHS
1002 SSA name alignment info accordingly.
1004 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1006 PR tree-optimization/84114
1007 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1008 Avoid reassociation of FLOAT_MODE addition.
1010 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1012 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1013 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1014 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1015 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1016 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1017 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1018 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1020 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1021 __builtin_ia32_wbinvd): New builtins.
1022 (SPECIAL_ARGS2): New.
1023 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1024 (SPECIAL_ARGS2): New.
1025 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1026 (ix86_valid_target_attribute_inner_p): Ditto.
1027 (ix86_init_mmx_sse_builtins): Add special_args2.
1028 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1029 TARGET_WBNOINVD_P): New.
1030 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1031 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1032 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1033 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1034 * config/i386/pconfigintrin.h: New file.
1035 * config/i386/wbnoinvdintrin.h: Ditto.
1036 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1037 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1039 2018-03-05 Richard Biener <rguenther@suse.de>
1041 PR tree-optimization/84670
1042 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1044 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1045 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1046 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1047 doesn't grow if all successors (recursively) were visited at least
1050 2018-03-05 Richard Biener <rguenther@suse.de>
1052 PR tree-optimization/84650
1053 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1054 if executed in the loop pipeline.
1056 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1058 * doc/configfiles.texi (Configuration Files): Move info about
1059 conditionalizing $target-protos.h to...
1060 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1061 differs from $target-protos.h.
1063 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1064 Chung-Ju Wu <jasonwucj@gmail.com>
1066 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1067 * config/nds32/nds32-multiple.md (setmemsi): Define.
1068 * config/nds32/nds32-memory-manipulation.c
1069 (nds32_gen_dup_4_byte_to_word_value): New.
1070 (emit_setmem_word_loop): New.
1071 (emit_setmem_byte_loop): New.
1072 (nds32_expand_setmem_loop): New.
1073 (nds32_expand_setmem_loop_v3m): New.
1074 (nds32_expand_setmem_unroll): New.
1075 (nds32_expand_setmem): New.
1077 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1078 Chung-Ju Wu <jasonwucj@gmail.com>
1080 * config/nds32/nds32-memory-manipulation.c
1081 (nds32_emit_load_store): New.
1082 (nds32_emit_post_inc_load_store): New.
1083 (nds32_emit_mem_move): New.
1084 (nds32_emit_mem_move_block): New.
1085 (nds32_expand_movmemsi_loop_unknown_size): New.
1086 (nds32_expand_movmemsi_loop_known_size): New.
1087 (nds32_expand_movmemsi_loop): New.
1088 (nds32_expand_movmemsi_unroll): New.
1089 (nds32_expand_movmemqi): Rename ...
1090 (nds32_expand_movmemsi): ... to this.
1091 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1092 (movmemsi): ... to this.
1093 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1094 (nds32_expand_movmemsi): ... to this.
1096 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1097 Monk Chiang <sh.chiang04@gmail.com>
1098 Chung-Ju Wu <jasonwucj@gmail.com>
1100 * config/nds32/nds32-protos.h
1101 (nds32_expand_load_multiple): New arguments.
1102 (nds32_expand_store_multiple): Ditto.
1103 (nds32_valid_multiple_load_store): Rename ...
1104 (nds32_valid_multiple_load_store_p): ... to this.
1105 * config/nds32/nds32-memory-manipulation.c
1106 (nds32_expand_load_multiple): Refine implementation.
1107 (nds32_expand_store_multiple): Ditto.
1108 * config/nds32/nds32-multiple.md
1109 (load_multiple): Update nds32_expand_load_multiple interface.
1110 (store_multiple): Update nds32_expand_store_multiple interface.
1111 * config/nds32/nds32-predicates.c
1112 (nds32_valid_multiple_load_store): Rename ...
1113 (nds32_valid_multiple_load_store_p): ... to this and refine
1115 * config/nds32/predicates.md
1116 (nds32_load_multiple_and_update_address_operation): New predicate.
1117 (nds32_store_multiple_and_update_address_operation): New predicate.
1119 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1120 Chung-Ju Wu <jasonwucj@gmail.com>
1122 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1123 (combo): New attribute.
1124 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1126 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1128 * config/nds32/nds32.opt: Change -mcmodel= default value.
1130 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1131 Monk Chiang <sh.chiang04@gmail.com>
1132 Chung-Ju Wu <jasonwucj@gmail.com>
1134 * config/nds32/constants.md (unspec_element): New enum.
1135 * config/nds32/constraints.md (Umw): New constraint.
1136 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1137 * config/nds32/nds32-intrinsic.md: Likewise.
1138 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1139 (nds32_valid_smw_lwm_base_p): New.
1140 (nds32_output_smw_single_word): New.
1141 (nds32_output_lmw_single_word): New.
1142 (nds32_expand_unaligned_load): New.
1143 (nds32_expand_unaligned_store): New.
1144 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1145 (nds32_output_smw_single_word): Declare.
1146 (nds32_output_lmw_single_word): Declare.
1147 (nds32_expand_unaligned_load): Declare.
1148 (nds32_expand_unaligned_store): Declare.
1149 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1150 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1151 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1152 NDS32_BUILTIN_UASTORE_DW.
1153 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1156 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1157 Kito Cheng <kito.cheng@gmail.com>
1158 Chung-Ju Wu <jasonwucj@gmail.com>
1160 * config/nds32/nds32-intrinsic.c
1161 (nds32_expand_builtin_null_ftype_reg): Delete.
1162 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1163 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1164 (nds32_read_argument): New.
1165 (nds32_legitimize_target): Ditto.
1166 (nds32_legitimize_argument): Ditto.
1167 (nds32_check_constant_argument): Ditto.
1168 (nds32_expand_unop_builtin): Ditto.
1169 (nds32_expand_unopimm_builtin): Ditto.
1170 (nds32_expand_binop_builtin): Ditto.
1171 (nds32_builtin_decl_impl): Ditto.
1172 (builtin_description): Ditto.
1173 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1174 (nds32_init_builtins_impl): Ditto.
1175 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1176 (nds32_builtin_decl): New.
1177 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1178 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1180 2018-03-02 Jeff Law <law@redhat.com>
1182 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1183 (redundant_insn, fill_simple_delay_slots): Likewise.
1184 (fill_slots_from_thread): Likewise.
1185 * resource.c (mark_referenced_resources): Likewise.
1186 (mark_set_resources, find_dead_or_set_registers): Likewise.
1188 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1190 * substring-locations.h (format_warning_va): Formatting fix for
1192 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1194 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1195 * substring-locations.c: Include intl.h.
1196 (format_warning_va): Turned into small wrapper around
1197 format_warning_n_va, renamed to ...
1198 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1199 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1201 (format_warning_at_substring_n): New function.
1202 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1203 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1204 format_warning_at_substring with just a shorter name instead of
1205 const function pointer.
1206 (fmtwarn_n): New function.
1207 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1208 appropriate, get rid of all the fmtstr temporaries, move conditionals
1209 with G_() wrapped string literals directly into fmtwarn arguments,
1210 cast dir.len to (int), formatting fixes.
1212 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1214 * doc/invoke.texi: Remove "Cilk Plus" references.
1216 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1217 Richard Biener <rguenther@suse.de>
1220 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1221 for error or warning attributes if CALL_FROM_THUNK_P is set.
1224 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1227 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1228 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1231 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1232 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1234 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1235 instead of -1U in last predictors element's probability member.
1237 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1240 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1241 arguments if they are comparable.
1243 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1245 PR tree-optimization/84634
1246 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1247 masks and masked_loop_p with a single loop_masks, making sure it's
1248 null for bb vectorization.
1250 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1252 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1253 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1254 loop->force_vectorize to check whether there is no alias.
1256 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1259 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1261 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1263 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1264 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1265 instead of a loop around prev_real_insn.
1266 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1270 * config/i386/i386.c (ix86_print_operand): Use conditional
1271 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1274 2018-03-02 Richard Biener <rguenther@suse.de>
1276 PR tree-optimization/84427
1277 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1278 (bitmap_set_subtract_values): Rewrite to handle multiple
1281 (prune_clobbered_mems): Likewise.
1282 (phi_translate): Take edge instead of pred/phiblock.
1283 (phi_translate_1): Likewise.
1284 (phi_translate_set): Likewise. Insert all translated
1285 exprs for a value into the set, keeping possibly multiple
1286 expressions per value.
1287 (compute_antic_aux): Adjust for phi_translate changes.
1288 When intersecting union the expressions and prune those
1289 not in the final value set, keeping possibly multiple
1290 expressions per value. Do not use value-insertion
1291 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1292 all expressions. Add verification that the value-sets
1293 only shrink during iteration.
1294 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1295 (do_pre_regular_insertion): Likewise.
1296 (do_pre_partial_partial_insertion): Likewise.
1298 2018-03-02 Richard Biener <rguenther@suse.de>
1301 * config/darwin.c (saved_debug_info_level): New static global.
1302 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1303 (darwin_asm_lto_end): Restore debug info generation settings.
1305 2018-03-01 Martin Liska <mliska@suse.cz>
1308 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1311 2018-03-01 Richard Biener <rguenther@suse.de>
1314 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1315 type annotation with LTO when debug was disabled at compile-time.
1317 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1319 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1321 (mips_final_postscan_insn): Likewise.
1323 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1325 PR rtl-optimization/84528
1326 * alias.c (init_alias_target): Add commentary.
1327 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1328 a unique base value if the frame pointer is not eliminated
1329 to the stack pointer.
1331 2018-03-01 Tom de Vries <tom@codesourcery.com>
1333 PR rtl-optimization/83327
1334 * lra-int.h (hard_regs_spilled_into): Declare.
1335 * lra.c (hard_regs_spilled_into): Define.
1336 (init_reg_info): Init hard_regs_spilled_into.
1337 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1338 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1339 (process_bb_lives): Handle hard_regs_spilled_into.
1340 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1341 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1343 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1345 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1346 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1347 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1348 * config/rs6000/aix72.h: New file.
1350 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1352 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1353 instead of warning_at with conditional singular and plural messages
1357 * stor-layout.c (update_alignment_for_field): For
1358 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1359 && !DECL_PACKED (field), do the alignment update, just use
1360 only desired_align instead of MAX (type_align, desired_align)
1362 (place_field): Don't do known_align < desired_align handling
1363 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1364 is non-NULL, instead do it after rli->prev_field handling and
1365 only if not within a bitfield word. For DECL_PACKED (field)
1366 use type_align of BITS_PER_UNIT.
1368 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1370 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1371 superfluous parentheses and trailing spaces.
1373 2018-02-28 Richard Biener <rguenther@suse.de>
1375 PR tree-optimization/84584
1376 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1377 SCoPs with fake exit edge.
1379 2018-02-28 Martin Liska <mliska@suse.cz>
1382 * timevar.c (timer::print): Fix format to properly print 100%
1385 2018-02-28 Richard Biener <rguenther@suse.de>
1388 * genmatch.c (capture_info::walk_match): Do not mark
1389 captured expressions without operands as expr_p given
1390 they act more like predicates and should be subject to
1391 "lost tail" side-effect preserving.
1393 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1395 PR rtl-optimization/81611
1396 * auto-inc-dec.c (attempt_change): Move dead note from
1397 mem_insn if it's the next use of regno
1398 (find_address): Take address use of reg holding
1399 non-incremented value. Add parm to limit search to the named
1401 (merge_in_block): Attempt to use a mem insn that is the next
1402 use of the original regno.
1404 2018-02-27 Martin Sebor <msebor@redhat.com>
1407 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1408 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1410 2018-02-27 Martin Sebor <msebor@redhat.com>
1412 PR translation/84207
1413 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1414 n argument to unsigned HOST_WIDE_INT.
1415 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1416 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1417 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1418 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1420 2018-02-27 Richard Biener <rguenther@suse.de>
1422 PR tree-optimization/84512
1423 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1424 Do not use the estimate returned from record_stmt_cost for
1425 the scalar iteration cost but sum properly using add_stmt_cost.
1427 2018-02-27 Richard Biener <rguenther@suse.de>
1429 PR tree-optimization/84466
1430 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1431 Adjust last change to less strictly validate use operands.
1433 2018-02-27 Martin Liska <mliska@suse.cz>
1435 PR gcov-profile/84548
1436 * gcov.c (process_file): Allow partial overlap and consider it
1437 also as group functions.
1438 (output_lines): Properly calculate range of lines for a group.
1440 2018-02-27 Martin Liska <mliska@suse.cz>
1442 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1443 'ggc' suffixes. Change first column width.
1444 (timer::print): Fix formatting of the column.
1446 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1448 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1449 preserve inline entry blocks for the sake of debug inline
1450 entry point markers alone.
1451 (remove_unused_locals): Suggest in comments a better place to
1452 force the preservation of inline entry blocks that are
1453 otherwise unused, but do not preserve them.
1455 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1457 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1459 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1462 * config/i386/constraints.md (Bs): Replace
1463 ix86_indirect_branch_register with
1464 TARGET_INDIRECT_BRANCH_REGISTER.
1466 * config/i386/i386.md (indirect_jump): Likewise.
1467 (tablejump): Likewise.
1468 (*sibcall_memory): Likewise.
1469 (*sibcall_value_memory): Likewise.
1470 Peepholes of indirect call and jump via memory: Likewise.
1471 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1472 (*sibcall_value_GOT_32): Likewise.
1473 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1474 (GOT_memory_operand): Likewise.
1475 (call_insn_operand): Likewise.
1476 (sibcall_insn_operand): Likewise.
1477 (GOT32_symbol_operand): Likewise.
1478 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1480 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1482 PR rtl-optimization/83496
1483 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1484 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1485 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1486 redundant insn, if any.
1487 (relax_delay_slots): Likewise.
1488 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1490 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1492 PR tree-optimization/83965
1493 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1494 that grouped statements are part of a reduction chain. Return
1495 true if the statement is not marked as a reduction itself but
1497 (vect_recog_dot_prod_pattern): Don't check whether the statement
1498 is part of a group here.
1499 (vect_recog_sad_pattern): Likewise.
1500 (vect_recog_widen_sum_pattern): Likewise.
1502 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1505 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1507 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1510 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1512 (ix86_output_indirect_function_return): New prototype.
1513 (ix86_split_simple_return_pop_internal): Likewise.
1514 * config/i386/i386.c (indirect_return_via_cx): New.
1515 (indirect_return_via_cx_bnd): Likewise.
1516 (indirect_thunk_name): Handle return va CX_REG.
1517 (output_indirect_thunk_function): Create alias for
1518 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1519 (ix86_output_indirect_jmp): Remove the bool argument.
1520 (ix86_output_indirect_function_return): New function.
1521 (ix86_split_simple_return_pop_internal): Likewise.
1522 * config/i386/i386.md (*indirect_jump): Don't pass false
1523 to ix86_output_indirect_jmp.
1524 (*tablejump_1): Likewise.
1525 (simple_return_pop_internal): Change it to define_insn_and_split.
1526 Call ix86_split_simple_return_pop_internal to split it for
1528 (simple_return_indirect_internal): Call
1529 ix86_output_indirect_function_return instead of
1530 ix86_output_indirect_jmp.
1532 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1535 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1536 memset and value initialization afterwards.
1538 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1540 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1542 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1545 * common/config/aarch64/aarch64-common.c
1546 (aarch_option_optimization_table[]): Switch
1547 off fomit-frame-pointer
1549 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1550 Chung-Ju Wu <jasonwucj@gmail.com>
1552 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1554 (store_multiple): Ditto.
1556 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1558 * config.gcc: Add --with-cpu support for nds32 target.
1559 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1560 * config/nds32/nds32.opt: Add -mcpu= option.
1562 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1564 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1565 isel=yes): Warn for these deprecated options.
1567 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1569 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1570 ISA_2_5_MASKS_EMBEDDED.
1572 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1574 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1575 p->max as pointers rather than using iterative_hash_expr.
1577 2018-02-23 Carl Love <cel@us.ibm.com>
1579 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1580 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1582 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1583 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1584 P8V_BUILTIN_VEC_VUNSIGNED2.
1586 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1589 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1590 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1592 * lra-constraints.c (curr_insn_transform): Set up
1593 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1595 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1596 * lra-eliminations.c (spill_pseudos): Ditto.
1597 (process_insn_for_elimination): Ditto.
1598 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1599 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1600 LRA_NON_CLOBBERED_ALT.
1602 2018-02-22 Martin Sebor <msebor@redhat.com>
1604 PR tree-optimization/84480
1605 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1606 to maybe_diag_stxncpy_trunc. Call it.
1607 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1608 from gimple_fold_builtin_strcpy. Print inlining stack.
1609 (handle_builtin_stxncpy): Print inlining stack.
1610 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1612 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1615 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1616 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1617 and -fcheck-pointer-bounds are used together.
1618 (indirect_thunk_prefix): New enum.
1619 (indirect_thunk_need_prefix): New function.
1620 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1621 "_nt" instead of "_bnd" for NOTRACK prefix.
1622 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1623 (output_indirect_thunk_function): Likewise.
1625 (ix86_code_end): Update output_indirect_thunk_function calls.
1626 (ix86_output_indirect_branch_via_reg): Replace
1627 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1628 (ix86_output_indirect_branch_via_push): Likewise.
1629 (ix86_output_function_return): Likewise.
1630 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1631 incompatible with -fcf-protection=branch and
1632 -fcheck-pointer-bounds.
1634 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1637 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1638 Change gcc_assert call to output_operand_lossage.
1640 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1642 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1644 2018-02-22 DJ Delorie <dj@redhat.com>
1645 Sebastian Perta <sebastian.perta@renesas.com>
1646 Oleg Endo <olegendo@gcc.gnu.org>
1648 * config/rx/rx.c (rx_rtx_costs): New function.
1649 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1651 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1653 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1655 2018-02-22 Martin Liska <mliska@suse.cz>
1658 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1659 Add "native" as a possible value.
1661 2018-02-22 Martin Liska <mliska@suse.cz>
1664 * config/i386/i386.c (ix86_option_override_internal):
1665 Add "native" as a possible value for -march and -mtune.
1667 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1670 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1671 to all type variants.
1673 PR tree-optimization/84503
1674 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1675 width as info->bitpos + info->bitsize - start.
1676 (merged_store_group::merge_overlapping): Simplify width computation.
1677 (check_no_overlap): New function.
1678 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1679 start + width and last_order of the group, fail if check_no_overlap
1681 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1682 to group if check_no_overlap fails.
1684 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1686 * config/rs6000/altivec.md: Delete contraint arguments to
1687 define_expand, define_split, and define_peephole2, and in
1688 define_insn_and_split if always unused.
1689 * config/rs6000/darwin.md: Ditto.
1690 * config/rs6000/dfp.md: Ditto.
1691 * config/rs6000/rs6000.md: Ditto.
1692 * config/rs6000/sync.md: Ditto.
1693 * config/rs6000/vector.md: Ditto.
1694 * config/rs6000/vsx.md: Ditto.
1696 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1698 * config/rs6000/altivec.md: Write output control strings as braced
1699 blocks instead of double-quoted strings.
1700 * config/rs6000/darwin.md: Ditto.
1701 * config/rs6000/rs6000.md: Ditto.
1702 * config/rs6000/vector.md: Ditto.
1703 * config/rs6000/vsx.md: Ditto.
1705 2018-02-21 Jason Merrill <jason@redhat.com>
1707 PR c++/84314 - ICE with templates and fastcall attribute.
1708 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1710 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1712 * ipa-cp.c (determine_versionability): Fix comment typos.
1714 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1717 * ipa-cp.c (determine_versionability): Do not version functions caling
1720 2018-02-21 Martin Liska <mliska@suse.cz>
1723 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1724 Add "native" as a possible value.
1725 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1726 the macro when native cpu detection is available.
1728 2018-02-21 Martin Liska <mliska@suse.cz>
1731 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1732 Add "native" as a possible value.
1733 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1734 when native cpu detection is available.
1736 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1737 Martin Sebor <msebor@redhat.com>
1739 PR tree-optimization/84478
1740 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1742 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1743 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1744 support which is conservatively correct, for 2 only stay conservative
1745 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1746 argument to the 2 argument get_range_strlen, adjust 6 arg
1747 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1749 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1750 (gimple_fold_builtin_strlen): Pass true as last argument to
1753 2018-02-20 Martin Sebor <msebor@redhat.com>
1756 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1757 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1758 (builtin_memref::builtin_memref): Factor out parts into
1759 set_base_and_offset and call it.
1761 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1764 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1765 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1766 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1767 search at the associated MODE_INT.
1769 2018-02-20 Jeff Law <law@redhat.com>
1772 PR tree-optimization/81592
1774 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1775 for range data rather than using global data.
1776 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1777 range data rather than using global data.
1778 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1779 pass it to children as needed.
1780 (struct directive::fmtresult): Similarly.
1781 (struct directive::set_width): Similarly.
1782 (struct directive::set_precision): Similarly.
1783 (format_integer, format_directive, parse_directive): Similarly.
1784 (format_none): Accept unnamed vr_values parameter.
1785 (format_percent, format_floating, format_character): Similarly.
1786 (format_string, format_plain): Similarly.
1787 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1788 the EVRP range analyzer for range data rather than using global data.
1789 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1790 gimple-ssa-evrp-analyze.h
1791 (class sprintf_dom_walker): Add after_dom_children member function.
1792 Add evrp_range_analyzer member.
1793 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1794 range analyzer as needed.
1795 (sprintf_dom_walker::after_dom_children): New member function.
1796 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1798 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1799 (evrp_range_analyzer::pop_to_marker): Likewise.
1801 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1803 PR tree-optimization/84419
1804 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1805 with the required type if its current type is compatible but
1808 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1811 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1812 after vectorization.
1814 2018-02-20 Martin Liska <mliska@suse.cz>
1817 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1818 possible values if we don't have a hint.
1820 2018-02-20 Martin Liska <mliska@suse.cz>
1824 * final.c (shorten_branches): Build align_tab array with one
1826 * opts.c (finish_options): Add alignment option limit check.
1827 (MAX_CODE_ALIGN): Likewise.
1828 (MAX_CODE_ALIGN_VALUE): Likewise.
1829 * doc/invoke.texi: Document maximum allowed option value for
1830 all -falign-* options.
1832 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1835 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1836 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1837 * var-tracking.c (emit_note_insn_var_location): Remove all references
1838 to NOTE_INSN_CALL_ARG_LOCATION.
1839 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1840 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1841 Use copy_rtx_if_shared.
1842 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1843 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1844 (dwarf2out_var_location): Remove handling of
1845 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1847 * final.c (final_scan_insn): Remove all references to
1848 NOTE_INSN_CALL_ARG_LOCATION.
1849 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1850 before dumping final insns.
1851 * except.c (emit_note_eh_region_end): Remove all references to
1852 NOTE_INSN_CALL_ARG_LOCATION.
1853 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1854 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1855 * config/arc/arc.c (hwloop_optimize): Likewise.
1856 * config/arm/arm.c (create_fix_barrier): Likewise.
1857 * config/s390/s390.c (s390_chunkify_start): Likewise.
1858 * config/sh/sh.c (find_barrier): Likewise.
1859 * config/i386/i386.c (rest_of_insert_endbranch,
1860 ix86_seh_fixup_eh_fallthru): Likewise.
1861 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1862 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1863 * config/frv/frv.c (frv_function_prologue): Likewise.
1864 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1866 (note_outside_basic_block_p): Remove all references to
1867 NOTE_INSN_CALL_ARG_LOCATION.
1868 * gengtype.c (adjust_field_rtx_def): Likewise.
1869 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1871 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1872 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1875 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1878 PR tree-optimization/84452
1879 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1880 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1883 2018-02-19 Martin Liska <mliska@suse.cz>
1886 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1888 2018-02-19 Martin Liska <mliska@suse.cz>
1889 Richard Sandiford <richard.sandiford@linaro.org>
1891 PR tree-optimization/82491
1892 * gimple-fold.c (get_base_constructor): Make earlier bail out
1895 2018-02-19 Carl Love <cel@us.ibm.com>
1897 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1898 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1900 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1901 P8V_BUILTIN_VEC_NEG.
1903 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1905 * config/rl78/rl78.md (movdf): New define expand.
1907 2018-02-19 Martin Liska <mliska@suse.cz>
1910 * doc/invoke.texi: Fix typo.
1911 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1913 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1915 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1916 handle rs6000_single_float and rs6000_double_float specially for
1919 2018-02-16 Jeff Law <law@redhat.com>
1921 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1924 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1927 * ipa-inline.c (inline_small_functions): Fix a typo.
1929 2018-02-16 Nathan Sidwell <nathan@acm.org>
1931 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1933 2018-02-16 Carl Love <cel@us.ibm.com>
1935 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1936 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1937 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1938 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1939 expansion to P8V_BUILTIN_VEC_FLOAT2.
1941 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1943 PR rtl-optimization/70023
1944 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1945 src_regno into account.
1947 2018-02-16 Carl Love <cel@us.ibm.com>
1949 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1950 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1951 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1952 * config/rs6000/rs6000.c: Remove case statements for
1953 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1954 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1955 and P9V_BUILTIN_VEC_VINSERT4B.
1956 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1957 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1958 * config/rs6000/vsx.md:
1959 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1962 2018-02-16 Carl Love <cel@us.ibm.com>
1964 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1966 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1968 * config/rs6000/rs6000-c.c: Add the definitions for
1969 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1970 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1971 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1972 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1973 definition for insert4b and define insn *insert3b_internal.
1974 * doc/extend.texi: Add documentation for vec_extract4b.
1976 2018-02-16 Nathan Sidwell <nathan@acm.org>
1978 * doc/extend.texi (Backwards Compatibility): Mention friend
1979 injection. Note for-scope is deprecated.
1980 * doc/invoke.texi (-ffriend-injection): Deprecate.
1982 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1984 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1985 that moved to I2, also allow destinations that are a paradoxical
1986 subreg (instead of a normal reg).
1988 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1991 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1994 2018-02-16 Richard Biener <rguenther@suse.de>
1996 PR tree-optimization/84037
1997 PR tree-optimization/84016
1999 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2000 Adjust vec_construct for the fact we need additional higher latency
2001 128bit inserts for AVX256 and AVX512 vector builds.
2002 (ix86_add_stmt_cost): Scale vector construction cost for
2005 2018-02-16 Richard Biener <rguenther@suse.de>
2007 PR tree-optimization/84417
2008 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2009 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2010 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2012 2018-02-16 Martin Liska <mliska@suse.cz>
2015 * internal-fn.def (ASAN_CHECK): Set proper flags.
2016 (ASAN_MARK): Likewise.
2018 2018-02-16 Julia Koval <julia.koval@intel.com>
2020 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2021 from PTA_CANNONLAKE.
2023 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2026 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2027 Use ++iter rather than iter++ for std::list iterators.
2028 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2029 defer deleting them until all nodes in the forest are processed. Do
2030 free even leaf nodes. Change to_process into auto_vec.
2033 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2034 * vec.h (vec_default_construct): Use memset instead of placement new
2035 if BROKEN_VALUE_INITIALIZATION is defined.
2036 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2037 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2040 PR rtl-optimization/83723
2041 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2042 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2043 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2045 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2047 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2049 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2051 PR rtl-optimization/81443
2052 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2053 from inner REGs to paradoxical SUBREGs.
2055 2018-02-16 Richard Biener <rguenther@suse.de>
2057 PR tree-optimization/84399
2058 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2059 For operands we can analyze at their definition make sure we can
2060 analyze them at each use as well.
2062 2018-02-16 Richard Biener <rguenther@suse.de>
2064 PR tree-optimization/84190
2065 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2066 volatile accesses if the decl isn't volatile.
2068 2018-02-15 Jason Merrill <jason@redhat.com>
2070 PR c++/84314 - ICE with templates and fastcall attribute.
2071 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2072 TYPE_CANONICAL on an existing type.
2074 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2076 PR tree-optimization/84383
2077 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2078 dstoff nor call operand_equal_p if dstbase is NULL.
2080 PR tree-optimization/84334
2081 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2082 also a CONSTANT_CLASS_P, punt.
2084 2018-02-14 Jim Wilson <jimw@sifive.com>
2086 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2087 first SMALL_OPERAND check. New local min_second_step. Move assert
2088 to where locals are set. Add TARGET_RVC support.
2089 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2091 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2093 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2095 2018-02-14 Martin Sebor <msebor@redhat.com>
2097 PR tree-optimization/83698
2098 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2099 arrays constrain the offset range to their bounds.
2100 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2101 (builtin_access::overlap): Avoid setting the size of overlap if it's
2103 (maybe_diag_overlap): Also consider arrays when deciding what values
2104 of offsets to include in diagnostics.
2106 2018-02-14 Martin Sebor <msebor@redhat.com>
2109 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2110 that correspond to the kind of a declaration.
2112 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2115 * config/pa/pa.md: Load address of PIC label using the linkage table
2116 if the label is nonlocal.
2118 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2120 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2121 warning message if user requests -maltivec=be.
2122 * doc/invoke.texi: Document deprecation of -maltivec=be.
2124 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2127 * config/rs6000/rs6000-c.c: Update definitions for
2128 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2129 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2131 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2134 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2135 add _get_ssp intrinsics. Remove argument from
2136 __builtin_ia32_rdssp[d|q].
2137 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2138 * config/i386/i386-builtin.def: Remove argument from
2139 __builtin_ia32_rdssp[d|q].
2140 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2141 ix86_expand_special_args_builtin for _rdssp[d|q].
2142 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2143 Clear register before usage.
2144 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2145 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2147 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2149 PR tree-optimization/84357
2150 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2151 operand 1 of an ARRAY_REF too.
2153 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2156 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2157 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2159 (set_of_reg): New struct.
2160 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2161 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2162 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2164 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2165 Split into bitclr, bitset, bitinvert patterns if appropriate.
2166 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2167 use rx_fuse_in_memory_bitop.
2168 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2169 to named insn, correct maximum insn length.
2171 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2174 * machmode.def: Define a complex mode for PARTIAL_INT.
2175 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2177 * doc/rtl.texi: Document CSPImode.
2178 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2180 (msp430_hard_regno_nregs_with_padding): Likewise.
2182 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2185 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2187 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2189 PR rtl-optimization/84169
2190 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2191 we generated a parallel as new i3 and we split that to new i2 and i3
2192 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2193 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2194 those to i2, not i1. Partially rewrite this scan code.
2196 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2199 * stor-layout.c (place_field): For variable length fields, adjust
2200 offset_align afterwards not just based on the field's alignment,
2201 but also on the size.
2204 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2205 of exps and logs in the use_exp2 case.
2207 2018-02-13 Jeff Law <law@redhat.com>
2209 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2212 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2215 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2219 * common.opt (gas-loc-support, gas-locview-support): New.
2220 (ginline-points, ginternal-reset-location-views): New.
2221 * doc/invoke.texi: Document them. Use @itemx where intended.
2222 (gvariable-location-views): Adjust.
2223 * target.def (reset_location_view): New.
2224 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2225 (TARGET_RESET_LOCATION_VIEW): New.
2226 * doc/tm.texi: Rebuilt.
2227 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2228 (dwarf2out_default_as_locview_support): New.
2229 (output_asm_line_debug_info): Use option variables.
2230 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2231 (output_loc_list): Likewise.
2232 (add_high_low_attributes): Check option variables.
2233 Don't output entry view attribute in strict mode.
2234 (gen_inlined_subroutine_die): Check option variables.
2235 (dwarf2out_inline_entry): Likewise.
2236 (init_sections_and_labels): Likewise.
2237 (dwarf2out_early_finish): Likewise.
2238 (maybe_reset_location_view): New, from...
2239 (dwarf2out_var_location): ... here. Call it.
2240 * debug.h (dwarf2out_default_as_loc_support): Declare.
2241 (dwarf2out_default_as_locview_support): Declare.
2242 * hooks.c (hook_int_rtx_insn_0): New.
2243 * hooks.h (hook_int_rtx_insn_0): Declare.
2244 * toplev.c (process_options): Take -gas-loc-support and
2245 -gas-locview-support from dwarf2out. Enable
2246 -gvariable-location-views by default only with locview
2247 assembler support. Enable -ginternal-reset-location-views by
2248 default only if the target defines the corresponding hook.
2249 Enable -ginline-points by default if location views are
2250 enabled; force it disabled if statement frontiers are
2252 * tree-inline.c (expand_call_inline): Check option variables.
2253 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2255 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2257 PR tree-optimization/84321
2258 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2259 handling. Also check whether the anti-range contains any values
2260 that satisfy the mask; switch to a VR_RANGE if not.
2262 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2265 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2267 2018-02-13 Martin Jambor <mjambor@suse.cz>
2270 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2271 of call statements, also set location of a load to a temporary.
2273 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2275 * config/rl78/rl78.c (add_vector_labels): New function.
2276 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2277 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2278 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2279 which checks that no arguments are passed.
2280 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2281 * doc/extend.texi: Documentation for the new attribute.
2283 2018-02-13 Andreas Schwab <schwab@suse.de>
2285 * config/riscv/linux.h (CPP_SPEC): Define.
2287 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2290 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2291 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2292 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2293 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2294 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2295 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2296 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2297 temporarily for AES and PCLMUL builtins.
2299 PR tree-optimization/84339
2300 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2301 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2305 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2306 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2307 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2309 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2311 * omp-simd-clone.h: New file.
2312 * omp-simd-clone.c: Include omp-simd-clone.h.
2313 (expand_simd_clones): No longer static.
2314 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2315 cgraph.h and omp-simd-clone.h.
2316 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2317 (vect_recog_widen_shift_pattern): Formatting fix.
2318 (vect_pattern_recog_1): Don't check optab for calls.
2321 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2322 operands[2] into a REG before using gen_lowpart on it.
2324 2018-02-12 Jeff Law <law@redhat.com>
2327 * config/sh/sh.c (find_barrier): Consider a sibling call
2330 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2331 successfully back substituting a reg.
2333 2018-02-12 Richard Biener <rguenther@suse.de>
2335 PR tree-optimization/84037
2336 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2337 parameter, move visited init to caller.
2338 (vect_slp_analyze_operations): Separate cost from validity
2339 check, initialize visited once for all instances.
2340 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2342 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2344 (vect_model_promotion_demotion_cost): Likewise.
2345 (vectorizable_bswap): Guard cost modeling with !slp_node
2346 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2348 (vectorizable_call): Likewise.
2349 (vectorizable_conversion): Likewise.
2350 (vectorizable_assignment): Likewise.
2351 (vectorizable_shift): Likewise.
2352 (vectorizable_operation): Likewise.
2353 (vectorizable_store): Likewise.
2354 (vectorizable_load): Likewise.
2355 (vectorizable_condition): Likewise.
2356 (vectorizable_comparison): Likewise.
2358 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2361 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2362 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2363 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2365 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2368 * params.def (inline-min-speedup): Increase from 8 to 15.
2369 (max-inline-insns-auto): Decrease from 40 to 30.
2370 * ipa-split.c (consider_split): Add some buffer for function to
2371 be considered inlining candidate.
2372 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2375 2018-02-12 Richard Biener <rguenther@suse.de>
2377 PR tree-optimization/84037
2378 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2379 matched stmts if we cannot swap the non-matched ones.
2381 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2383 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2384 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2385 _mm_maskz_scalef_round_ss): New intrinsics.
2386 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2387 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2388 __builtin_ia32_scalefss_round): Remove.
2389 (__builtin_ia32_scalefsd_mask_round,
2390 __builtin_ia32_scalefss_mask_round): New intrinsics.
2391 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2392 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2393 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2394 "<round_constraint>")): Changed to ...
2395 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2396 "<round_scalar_constraint>")): ... this.
2397 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2398 %0, %1, %2<round_op3>}"): Changed to ...
2399 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2400 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2401 %2<round_scalar_mask_op3>}"): ... this.
2402 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2404 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2406 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2407 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2408 (_mm_maskz_sqrt_round_ss): New intrinsics.
2409 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2410 (__builtin_ia32_sqrtsd_mask_round)
2411 (__builtin_ia32_sqrtss_mask_round): New builtins.
2412 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2413 (__builtin_ia32_sqrtss_round): Remove.
2414 (__builtin_ia32_sqrtsd_mask_round)
2415 (__builtin_ia32_sqrtss_mask_round): New builtins.
2416 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2417 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2418 ((match_operand:VF_128 1 "vector_operand"
2419 "xBm,<round_constraint>")): Changed to ...
2420 ((match_operand:VF_128 1 "vector_operand"
2421 "xBm,<round_scalar_constraint>")): ... this.
2422 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2423 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2424 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2425 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2426 %<iptr>1<round_scalar_mask_op3>}): ... this.
2427 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2428 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2430 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2433 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2434 Cast vec_cmpeq result to correct type.
2435 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2436 Cast vec_cmpgt result to correct type.
2438 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2440 * final.c (final_scan_insn_1): Renamed from...
2441 (final_scan_insn): ... this. New wrapper, to recover
2442 seen from the outermost call in recursive ones.
2443 * config/sparc/sparc.c (output_return): Drop seen from call.
2444 (output_sibcall): Likewise.
2445 * config/visium/visium.c (output_branch): Likewise.
2447 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2449 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2452 2018-02-10 Alan Modra <amodra@gmail.com>
2455 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2456 Specify LR as an input.
2458 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2461 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2462 remove_member_access_dummy_vars): New functions.
2463 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2464 lower_omp_1, execute_lower_omp): Use them.
2466 PR rtl-optimization/84308
2467 * shrink-wrap.c (spread_components): Release todo vector.
2469 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2471 PR rtl-optimization/57193
2472 * ira-color.c (struct allocno_color_data): Add member
2473 conflict_allocno_hard_prefs.
2474 (update_conflict_allocno_hard_prefs): New.
2475 (bucket_allocno_compare_func): Add a preference based on
2476 conflict_allocno_hard_prefs.
2477 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2478 (color_allocnos): Remove a dead code. Initiate
2479 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2481 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2484 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2485 constraint from =wa to wa. Avoid a subreg on the output operand,
2486 instead use a pseudo and subreg it in a move.
2487 (p9_xxbrd_<mode>): Changed to ...
2488 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2489 (p9_xxbrd_v2df): New expander.
2490 (p9_xxbrw_<mode>): Changed to ...
2491 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2492 (p9_xxbrw_v4sf): New expander.
2494 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2496 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2498 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2501 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2502 multiply in 32-bit mode.
2503 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2504 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2507 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2509 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2510 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2511 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2512 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2514 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2517 * dwarf2out.c (is_trivial_indirect_ref): New function.
2518 (dwarf2out_late_global_decl): Do not generate a location
2519 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2520 and that are not defined in the current unit.
2522 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2524 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2525 instead of a libcall for UNORDERED.
2527 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2530 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2531 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2533 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2536 * config/s390/s390.c (s390_set_current_function): Invoke
2537 s390_indirect_branch_settings also if fndecl didn't change.
2539 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2541 * config/rs6000/rs6000.md (blockage): Set length to zero.
2543 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2545 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2547 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2550 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2551 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2555 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2556 PARALLEL incoming that failed vt_get_decl_and_offset check.
2559 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2560 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2562 (get_variable_section): For decls in named .bss* sections pass true as
2563 second argument to bss_initializer_p.
2565 2018-02-09 Marek Polacek <polacek@redhat.com>
2566 Jakub Jelinek <jakub@redhat.com>
2569 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2570 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2571 Sync some changes from cxx_fold_indirect_ref.
2573 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2575 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2577 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2578 (BLOCK_INLINE_ENTRY_LABEL): New.
2579 (dwarf2out_var_location): Disregard inline entry markers.
2580 (inline_entry_data): New struct.
2581 (inline_entry_data_hasher): New hashtable type.
2582 (inline_entry_data_hasher::hash): New.
2583 (inline_entry_data_hasher::equal): New.
2584 (inline_entry_data_table): New variable.
2585 (add_high_low_attributes): Add DW_AT_entry_pc and
2586 DW_AT_GNU_entry_view attributes if a pending entry is found
2587 in inline_entry_data_table. Add old entry_pc attribute only
2588 if debug nonbinding markers are disabled.
2589 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2590 markers are enabled.
2591 (block_within_block_p, dwarf2out_inline_entry): New.
2592 (dwarf2out_finish): Check that no entries remained in
2593 inline_entry_data_table.
2594 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2595 (final_scan_insn, notice_source_line): Likewise.
2596 (rest_of_clean_state): Skip inline entry markers.
2597 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2599 * gimple.c (gimple_build_debug_inline_entry): New.
2600 * gimple.h (enum gimple_debug_subcode): Add
2601 GIMPLE_DEBUG_INLINE_ENTRY.
2602 (gimple_build_debug_inline_entry): Declare.
2603 (gimple_debug_inline_entry_p): New.
2604 (gimple_debug_nonbind_marker_p): Adjust.
2605 * insn-notes.def (INLINE_ENTRY): New.
2606 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2607 inline entry marker notes.
2608 (print_insn): Likewise.
2609 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2610 (INSN_DEBUG_MARKER_KIND): Likewise.
2611 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2612 * tree-inline.c (expand_call_inline): Build and insert
2613 debug_inline_entry stmt.
2614 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2615 inline entry blocks early, if nonbind markers are enabled.
2616 (dump_scope_block): Dump fragment info.
2617 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2618 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2619 (gimple_build_debug_inline_entry): New.
2620 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2621 Enable/disable inline entry points too.
2622 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2623 (DEBUG_INSN): Describe inline entry markers.
2625 * common.opt (gvariable-location-views): New.
2626 (gvariable-location-views=incompat5): New.
2627 * config.in: Rebuilt.
2628 * configure: Rebuilt.
2629 * configure.ac: Test assembler for view support.
2630 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2631 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2632 * dwarf2out.c (var_loc_view): New typedef.
2633 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2634 (dwarf2out_locviews_in_attribute): New.
2635 (dwarf2out_locviews_in_loclist): New.
2636 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2637 (enum dw_line_info_opcode): Add LI_adv_address.
2638 (struct dw_line_info_table): Add view.
2639 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2640 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2641 (zero_view_p): New variable.
2642 (ZERO_VIEW_P): New macro.
2643 (output_asm_line_debug_info): New.
2644 (struct var_loc_node): Add view.
2645 (add_AT_view_list, AT_loc_list): New.
2646 (add_var_loc_to_decl): Add view param. Test it against last.
2647 (new_loc_list): Add view params. Record them.
2648 (AT_loc_list_ptr): Handle loc and view lists.
2649 (view_list_to_loc_list_val_node): New.
2650 (print_dw_val): Handle dw_val_class_view_list.
2651 (size_of_die): Likewise.
2652 (value_format): Likewise.
2653 (loc_list_has_views): New.
2654 (gen_llsym): Set vl_symbol too.
2655 (maybe_gen_llsym, skip_loc_list_entry): New.
2656 (dwarf2out_maybe_output_loclist_view_pair): New.
2657 (output_loc_list): Output view list or entries too.
2658 (output_view_list_offset): New.
2659 (output_die): Handle dw_val_class_view_list.
2660 (output_dwarf_version): New.
2661 (output_compilation_unit_header): Use it.
2662 (output_skeleton_debug_sections): Likewise.
2663 (output_rnglists, output_line_info): Likewise.
2664 (output_pubnames, output_aranges): Update version comments.
2665 (output_one_line_info_table): Output view numbers in asm comments.
2666 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2667 Call maybe_gen_llsym.
2668 (loc_list_from_tree_1): Adjust.
2669 (add_AT_location_description): Create view list attribute if
2670 needed, check it's absent otherwise.
2671 (convert_cfa_to_fb_loc_list): Adjust.
2672 (maybe_emit_file): Call output_asm_line_debug_info for test.
2673 (dwarf2out_var_location): Reset views as needed. Precompute
2674 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2675 attribute. Set view.
2676 (new_line_info_table): Reset next view.
2677 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2678 (dwarf2out_source_line): Likewise. Output view resets and labels to
2679 the assembler, or select appropriate line info opcodes.
2680 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2681 (optimize_string_length): Catch it. Adjust.
2682 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2683 dw_val_class_view_list, and remove it if no longer needed.
2684 (hash_loc_list): Hash view numbers.
2685 (loc_list_hasher::equal): Compare them.
2686 (optimize_location_lists): Check whether a view list symbol is
2687 needed, and whether the locview attribute is present, and
2688 whether they match. Remove the locview attribute if no longer
2690 (index_location_lists): Call skip_loc_list_entry for test.
2691 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2692 Use output_dwarf_version.
2693 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2694 (struct dw_val_node): Add val_view_list.
2695 * final.c (SEEN_NEXT_VIEW): New.
2696 (set_next_view_needed): New.
2697 (clear_next_view_needed): New.
2698 (maybe_output_next_view): New.
2699 (final_start_function): Rename to...
2700 (final_start_function_1): ... this. Take pointer to FIRST,
2701 add SEEN parameter. Emit param bindings in the initial view.
2702 (final_start_function): Reintroduce SEEN-less interface.
2703 (final): Rename to...
2704 (final_1): ... this. Take SEEN parameter. Output final pending
2705 next view at the end.
2706 (final): Reintroduce seen-less interface.
2707 (final_scan_insn): Output pending next view before switching
2708 sections or ending a block. Mark the next view as needed when
2709 outputting variable locations. Notify debug backend of section
2710 changes, and of location view changes.
2711 (rest_of_handle_final): Adjust.
2712 * toplev.c (process_options): Autodetect value for debug variable
2713 location views option. Warn on incompat5 without -gdwarf-5.
2714 * doc/invoke.texi (gvariable-location-views): New.
2715 (gvariable-location-views=incompat5): New.
2716 (gno-variable-location-views): New.
2718 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2720 PR tree-optimization/84136
2721 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2722 that the result of find_edge is non-NULL.
2724 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2727 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2728 storing integer register in SImode. Fix cost of 256 and 512
2729 byte aligned SSE register store.
2731 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2733 * config/i386/i386.c (ix86_multiplication_cost): Fix
2734 multiplication cost for TARGET_AVX512DQ.
2736 2018-02-08 Marek Polacek <polacek@redhat.com>
2738 PR tree-optimization/84238
2739 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2742 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2744 PR tree-optimization/84265
2745 * tree-vect-stmts.c (vectorizable_store): Don't treat
2746 VMAT_CONTIGUOUS accesses as grouped.
2747 (vectorizable_load): Likewise.
2749 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2751 PR tree-optimization/81635
2752 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2753 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2754 (test_round_for_mask): New functions.
2755 (wide_int_cc_tests): Call test_round_for_mask.
2756 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2757 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2758 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2759 range returned by get_range_info.
2761 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2764 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2765 * symtab.c: Include builtins.h
2766 (symtab_node::output_to_lto_symbol_table_p): Move here
2767 from lto-streamer-out.c:output_symbol_p.
2768 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2769 (output_symbol_p): Move all logic to symtab.c
2770 (produce_symtab): Update.
2772 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2774 * config/s390/s390-opts.h (enum indirect_branch): Define.
2775 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2776 (s390_indirect_branch_via_thunk)
2777 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2778 (enum s390_indirect_branch_type): Define.
2779 * config/s390/s390.c (struct s390_frame_layout, struct
2780 machine_function): Remove.
2781 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2782 (indirect_branch_table_label_no, indirect_branch_table_name):
2784 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2785 (enum s390_indirect_branch_option): Define.
2786 (s390_return_addr_from_memory): New function.
2787 (s390_handle_string_attribute): New function.
2788 (s390_attribute_table): Add new attribute handler.
2789 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2790 (s390_indirect_branch_via_thunk): New function.
2791 (s390_indirect_branch_via_inline_thunk): New function.
2792 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2793 sibling call optimization for non z10 compiles.
2794 (s390_emit_call): Force indirect branch target to be a single
2795 register. Add r1 clobber for non-z10 compiles.
2796 (s390_emit_epilogue): Emit return jump via return_use expander.
2797 (s390_reorg): Handle JUMP_INSNs as execute targets.
2798 (s390_option_override_internal): Perform validity checks for the
2799 new command line options.
2800 (s390_indirect_branch_attrvalue): New function.
2801 (s390_indirect_branch_settings): New function.
2802 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2803 (s390_output_indirect_thunk_function): New function.
2804 (s390_code_end): Implement target hook.
2805 (s390_case_values_threshold): Implement target hook.
2806 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2808 * config/s390/s390.h (struct s390_frame_layout)
2809 (struct machine_function): Move here from s390.c.
2810 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2811 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2812 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2813 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2814 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2815 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2816 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2817 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2818 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2819 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2820 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2821 (mnemonic attribute): Add values which aren't recognized
2823 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2824 pattern for branch conversion. Fix mnemonic attribute.
2825 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2826 indirect branch via thunk if requested.
2827 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2828 ("*indirect_jump"): Disable for branch conversion using out of
2830 ("indirect_jump_via_thunk<mode>_z10")
2831 ("indirect_jump_via_thunk<mode>")
2832 ("indirect_jump_via_inlinethunk<mode>_z10")
2833 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2834 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2835 ("casesi_jump_via_inlinethunk<mode>_z10")
2836 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2837 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2838 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2839 ("*indirect2_jump"): Disable for branch conversion.
2840 ("casesi_jump"): Turn into expander and expand patterns for branch
2842 ("return_use"): New expander.
2843 ("*return"): Emit return via thunk and rename it to ...
2844 ("*return<mode>"): ... this one.
2845 * config/s390/s390.opt: Add new options and and enum for the
2848 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2850 * lra-constraints.c (match_reload): Unconditionally use
2851 gen_lowpart_SUBREG, rather than selecting between that
2852 and equivalent gen_rtx_SUBREG code.
2854 2018-02-08 Richard Biener <rguenther@suse.de>
2856 PR tree-optimization/84233
2857 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2858 changed flag instead of boguously re-using phi_inserted.
2860 2018-02-08 Martin Jambor <mjambor@suse.cz>
2862 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2863 static local variables.
2865 2018-02-08 Richard Biener <rguenther@suse.de>
2867 PR tree-optimization/84278
2868 * tree-vect-stmts.c (vectorizable_store): When looking for
2869 smaller vector types to perform grouped strided loads/stores
2870 make sure the mode is supported by the target.
2871 (vectorizable_load): Likewise.
2873 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2875 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2876 Increase LDP/STP opportunities by adding adjacent callee-saves.
2878 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2880 PR rtl-optimization/84068
2881 PR rtl-optimization/83459
2882 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2884 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2886 PR tree-optimization/84224
2887 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2888 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2891 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2894 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2895 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2896 count, remove one USE.
2898 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2900 * doc/install.texi (Configuration): Document the
2901 --with-long-double-format={ibm,ieee} PowerPC configuration
2905 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2906 Convert from define_expand to be define_insn_and_split. Rework
2907 float/double/_Float128 conversions to QI/HI/SImode to work with
2908 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2909 conversions to QI/HImode types did a store and then a load to
2910 truncate the value. For conversions to VSX registers, don't split
2911 the insn, instead emit the code directly. Use the code iterator
2912 any_fix to combine signed and unsigned conversions.
2913 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2914 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2915 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2916 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2917 (fix_<mode>di2_hw): Likewise.
2918 (fixuns_<mode>di2_hw): Likewise.
2919 (fix_<mode>si2_hw): Likewise.
2920 (fixuns_<mode>si2_hw): Likewise.
2921 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2922 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2923 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2924 fix<uns>_trunc<SFDF:mode>si2_p8.
2925 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2927 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2928 (fix<uns>_<mode>_mem): Likewise.
2929 (fctiw<u>z_<mode>_mem): Likewise.
2930 (fix<uns>_<mode>_mem): Likewise.
2931 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2932 the register allocator from doing a direct move to the GPRs to do
2933 a store, and instead use the ISA 3.0 store byte/half-word from
2934 vector register instruction. For IEEE 128-bit floating point,
2935 also optimize stores of 32-bit ints.
2936 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2938 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2940 * genextract.c (push_pathstr_operand): New function to support
2942 (walk_rtx): Call push_pathstr_operand.
2943 (print_path): Support [a-zA-Z].
2945 2018-02-07 Richard Biener <rguenther@suse.de>
2947 PR tree-optimization/84037
2948 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2949 (cse_and_gimplify_to_preheader): Declare.
2950 (vect_get_place_in_interleaving_chain): Likewise.
2951 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2953 (_loop_vec_info::~_loop_vec_info): Delete it.
2954 (cse_and_gimplify_to_preheader): New function.
2955 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2956 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2957 (vectorizable_load): Likewise. For grouped stores always base
2958 the IV on the first element.
2959 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2960 condition before gimplifying.
2962 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2964 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2965 *DIV_EXPR and *MOD_EXPR.
2967 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2970 * config/i386/i386.c (ix86_option_override_internal): Mask out
2971 the CF_SET bit when checking -fcf-protection.
2973 2018-02-07 Tom de Vries <tom@codesourcery.com>
2976 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2979 2018-02-07 Richard Biener <rguenther@suse.de>
2981 PR tree-optimization/84204
2982 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2985 PR tree-optimization/84205
2986 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2987 special-case isl_ast_op_zdiv_r.
2989 PR tree-optimization/84223
2990 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2991 Only add conditions from within the region.
2992 (gather_bbs::after_dom_children): Adjust.
2994 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2997 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2998 * config/avr/avr.md: Only post-reload split REG-REG moves if
2999 either register is GENERAL_REG_P.
3001 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3003 PR tree-optimization/84235
3004 * tree-ssa-scopedtables.c
3005 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3006 if the subtraction is performed in floating point type where NaNs are
3007 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3008 build 1. Formatting fix.
3010 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3013 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3014 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3015 and skip it regardless of bb boundaries. Use CALL_P macro,
3016 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3017 unnecessarily, formatting fix.
3019 2018-02-06 Michael Collison <michael.collison@arm.com>
3021 * config/arm/thumb2.md:
3022 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3023 (*thumb_mov_notscc): Ditto.
3025 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3028 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3029 unsigned_fix, not "s".
3031 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3033 * configure.ac (gcc_fn_eh_frame_ro): New function.
3034 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3035 correct .eh_frame permissions.
3036 * configure: Regenerate.
3038 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3040 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3043 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3045 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3046 Display warning message for -mno-speculate-indirect-jumps.
3048 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3050 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3052 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3054 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3056 PR tree-optimization/84225
3057 * tree-eh.c (find_trapping_overflow): Only call
3058 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3060 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3063 * config/i386/i386.c: Reimplement the check of possible options
3064 -mibt/-mshstk conbination. Change error messages.
3065 * doc/invoke.texi: Fix a typo: remove extra '='.
3067 2018-02-06 Marek Polacek <polacek@redhat.com>
3069 PR tree-optimization/84228
3070 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3072 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3075 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3076 emitted arch directives.
3077 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3078 __ARM_FEATURE_COPROC before changing architectures.
3080 2018-02-06 Richard Biener <rguenther@suse.de>
3082 * config/i386/i386.c (print_reg): Fix typo.
3083 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3085 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3087 * configure: Regenerate.
3089 2018-02-05 Martin Sebor <msebor@redhat.com>
3091 PR tree-optimization/83369
3092 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3095 2018-02-05 Martin Liska <mliska@suse.cz>
3097 * doc/invoke.texi: Cherry-pick upstream r323995.
3099 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3101 * ira.c (ira_init_register_move_cost): Adjust comment.
3103 2018-02-05 Martin Liska <mliska@suse.cz>
3105 PR gcov-profile/84137
3106 * doc/gcov.texi: Fix typo in documentation.
3108 2018-02-05 Martin Liska <mliska@suse.cz>
3110 PR gcov-profile/83879
3111 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3112 using dlopen functionality.
3114 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3116 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3117 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3118 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3119 _mm_maskz_range_round_ss): New intrinsics.
3120 (__builtin_ia32_rangesd128_round)
3121 (__builtin_ia32_rangess128_round): Remove.
3122 (__builtin_ia32_rangesd128_mask_round,
3123 __builtin_ia32_rangess128_mask_round): New builtins.
3124 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3125 __builtin_ia32_rangess128_round): Remove.
3126 (__builtin_ia32_rangesd128_mask_round,
3127 __builtin_ia32_rangess128_mask_round): New builtins.
3128 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3129 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3130 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3131 "<round_saeonly_constraint>")): Changed to ...
3132 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3133 "<round_saeonly_scalar_constraint>")): ... this.
3134 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3135 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3136 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3137 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3138 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3140 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3142 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3144 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3145 Remove all values except native, 8540 and 8548.
3147 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3149 * config/i386/i386.c (ix86_output_function_return): Pass
3150 INVALID_REGNUM, instead of -1, as invalid register number to
3151 indirect_thunk_name and output_indirect_thunk.
3153 2018-02-02 Julia Koval <julia.koval@intel.com>
3155 * config.gcc: Add -march=icelake.
3156 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3157 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3158 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3159 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3160 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3161 (processor_target_table): Add icelake.
3162 (ix86_option_override_internal): Handle new PTAs.
3163 (get_builtin_code_for_version): Handle icelake.
3164 (M_INTEL_COREI7_ICELAKE): New.
3165 (fold_builtin_cpu): Handle icelake.
3166 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3167 * doc/invoke.texi: Add -march=icelake.
3169 2018-02-02 Julia Koval <julia.koval@intel.com>
3171 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3172 to wide_int_bitmask.
3173 * wide-int-bitmask.h: New.
3175 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3178 * config/i386/i386.md: Replace Pmode with word_mode in
3179 builtin_setjmp_setup and builtin_longjmp to support x32.
3181 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3185 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3187 (rs6000_supported_cpu_names): New static variable.
3188 (linux_cpu_translation_table): Likewise.
3189 (elf_platform) <cpu>: Define new static variable and use it.
3190 Translate kernel AT_PLATFORM name to canonical name if needed.
3191 Error if platform name is unknown.
3193 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3196 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3198 2018-02-01 Jeff Law <law@redhat.com>
3201 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3202 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3203 the scratch if RELEASE_VIA_POP is false.
3204 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3205 If we have to save a temporary register, decrement SIZE appropriately.
3206 Pass new arguments to release_scratch_register_on_entry.
3207 (ix86_adjust_stack_and_probe): Likewise.
3208 (ix86_emit_probe_stack_range): Pass new arguments to
3209 release_scratch_register_on_entry.
3211 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3213 PR rtl-optimization/84157
3214 * combine.c (change_zero_ext): Use REG_P predicate in
3215 front of HARD_REGISTER_P predicate.
3217 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3219 * config/avr/avr.c (avr_option_override): Move disabling of
3220 -fdelete-null-pointer-checks to...
3221 * common/config/avr/avr-common.c (avr_option_optimization_table):
3224 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3226 PR tree-optimization/81635
3227 * tree-data-ref.c (split_constant_offset_1): For types that
3228 wrap on overflow, try to use range info to prove that wrapping
3231 2018-02-01 Renlin Li <renlin.li@arm.com>
3234 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3236 (aarch64_register_move_cost): Likewise.
3237 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3239 (REG_CLASS_NAMES): Likewise.
3240 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3241 TAILCALL_ADDR_REGS. Remove IP registers.
3242 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3244 2018-02-01 Richard Biener <rguenther@suse.de>
3246 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3247 for specifying RPO order and allow NULL for that.
3248 * domwalk.c (dom_walker::dom_walker): Likewise.
3249 (dom_walker::walk): Handle NULL RPO order.
3250 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3252 (rewrite_update_dom_walker): Likewise.
3253 (mark_def_dom_walker): Likewise.
3255 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3257 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3258 (aarch64_maybe_expand_sve_subreg_move): Declare.
3259 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3260 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3262 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3263 that are semantically a reverse operation.
3264 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3265 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3266 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3268 (aarch64_can_change_mode_class): For big-endian, forbid changes
3269 between two SVE modes if they have different element sizes.
3271 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3273 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3274 the TImode handling for big-endian targets.
3276 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3278 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3279 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3281 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3282 Remove BSWAP handing for big-endian targets and use the form of
3283 LD1RQ appropariate for the mode.
3285 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3287 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3288 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3291 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3294 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3295 check for operands that need to go through aarch64_sve_reload_be.
3297 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3299 PR tree-optimization/81661
3300 PR tree-optimization/84117
3301 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3302 * tree-eh.c: Include gimplify.h.
3303 (find_trapping_overflow, replace_trapping_overflow,
3304 rewrite_to_non_trapping_overflow): New functions.
3305 * tree-vect-loop.c: Include tree-eh.h.
3306 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3307 * tree-data-ref.c: Include tree-eh.h.
3308 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3310 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3312 PR rtl-optimization/84123
3313 * combine.c (change_zero_ext): Check if hard register satisfies
3314 can_change_dest_mode before calling gen_lowpart_SUBREG.
3316 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3319 * ira.c (ira_init_register_move_cost): Remove assert.
3321 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3323 PR rtl-optimization/84071
3324 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3325 * doc/tm.texi: Regenerate.
3327 2018-01-31 Richard Biener <rguenther@suse.de>
3329 PR tree-optimization/84132
3330 * tree-data-ref.c (analyze_miv_subscript): Properly
3331 check whether evolution_function_is_affine_multivariate_p
3332 before calling gcd_of_steps_may_divide_p.
3334 2018-01-31 Julia Koval <julia.koval@intel.com>
3337 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3338 * config/i386/i386.md (rdpid_rex64) New.
3339 (rdpid): Make 32bit only.
3341 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3344 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3345 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3347 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3350 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3352 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3354 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3356 PR rtl-optimization/84071
3357 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3358 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3360 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3362 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3363 (arc_attribute_table): Add 'aux' attribute.
3364 (arc_in_small_data_p): Consider aux like variables.
3365 (arc_is_aux_reg_p): New function.
3366 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3367 (arc_get_aux_arg): New function.
3368 (prepare_move_operands): Handle aux-register access.
3369 (arc_handle_aux_attribute): New function.
3370 * doc/extend.texi (ARC Variable attributes): Add subsection.
3372 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3374 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3375 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3376 (arc_attribute_table): Add 'uncached' attribute.
3377 (arc_print_operand): Print '.di' flag for uncached memory
3379 (arc_in_small_data_p): Do not consider for small data the uncached
3381 (arc_is_uncached_mem_p): New function.
3382 * config/arc/predicates.md (compact_store_memory_operand): Check
3383 for uncached memory accesses.
3384 (nonvol_nonimm_operand): Likewise.
3385 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3387 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3390 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3391 falign-loops=): Add Optimization flag.
3393 2018-01-30 Jeff Law <law@redhat.com>
3396 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3397 INT_REGISTERS_SAVED. Check it prior to calling
3398 get_scratch_register_on_entry.
3399 (ix86_adjust_stack_and_probe): Similarly.
3400 (ix86_emit_probe_stack_range): Similarly.
3401 (ix86_expand_prologue): Corresponding changes.
3403 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3406 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3407 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3409 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3412 * lra-constraints.c (curr_insn_transform): Process AND in the
3415 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3417 PR rtl-optimization/83986
3418 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3419 dependence against last_pending_memory_flush in addition to
3422 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3424 PR tree-optimization/81611
3425 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3428 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3431 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3434 2018-01-30 Richard Biener <rguenther@suse.de>
3435 Jakub Jelinek <jakub@redhat.com>
3437 PR tree-optimization/84111
3438 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3439 inner loops added during recursion, as they don't have up-to-date
3442 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3445 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3446 (can_inline_edge_by_limits_p): ... here.
3447 (can_early_inline_edge_p, check_callers,
3448 update_caller_keys, update_callee_keys, recursive_inlining,
3449 add_new_edges_to_heap, speculation_useful_p,
3450 inline_small_functions,
3451 inline_small_functions, flatten_function,
3452 inline_to_all_callers_1): Update.
3454 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3456 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3459 2018-01-30 Richard Biener <rguenther@suse.de>
3461 PR tree-optimization/83008
3462 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3463 invariant and constant vector uses in stmts when they need
3466 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3469 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3470 * configure: Regenerate.
3472 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3474 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3476 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3477 Use gen_rtx_REG rather than gen_lowpart.
3479 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3481 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3482 rather than 0 when creating partial subregs.
3484 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3486 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3489 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3492 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3493 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3494 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3495 flags. This restores the settings used before the 2017-07-24.
3496 Turning off pre increment/decrement/modify allows IVOPTS to
3497 optimize DF/SF loops where the index is an int.
3499 2018-01-29 Richard Biener <rguenther@suse.de>
3500 Kelvin Nilsen <kelvin@gcc.gnu.org>
3503 * tree-vect-stmts.c (vectorizable_call): Don't call
3504 targetm.vectorize_builtin_md_vectorized_function if callee is
3507 2018-01-22 Carl Love <cel@us.ibm.com>
3509 * doc/extend.tex: Fix typo in second arg in
3510 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3512 2018-01-29 Richard Biener <rguenther@suse.de>
3514 PR tree-optimization/84086
3515 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3516 (flush_ssaname_freelist): When SSA names were released reset
3517 the SCEV hash table.
3519 2018-01-29 Richard Biener <rguenther@suse.de>
3521 PR tree-optimization/84057
3522 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3523 removed paths when removing edges.
3525 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3527 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3528 -mfunction-return=@var{choice}.
3530 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3533 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3535 (layout::print_source_line): Likewise.
3536 (test_get_line_width_without_trailing_whitespace): Add test cases.
3538 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3541 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3544 2018-01-26 Jim Wilson <jimw@sifive.com>
3546 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3548 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3551 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3553 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3554 and CMP + SUB-immediate -> SUBS.
3556 2018-01-26 Martin Sebor <msebor@redhat.com>
3558 PR tree-optimization/83896
3559 * tree-ssa-strlen.c (get_string_len): Rename...
3560 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3561 Avoid assuming length is constant.
3562 (handle_char_store): Use HOST_WIDE_INT for string length.
3564 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3567 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3568 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3570 2018-01-26 Richard Biener <rguenther@suse.de>
3572 PR rtl-optimization/84003
3573 * dse.c (record_store): Only record redundant stores when
3574 the earlier store aliases at least all accesses the later one does.
3576 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3578 PR rtl-optimization/83985
3579 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3580 REG_CFA_RESTORE insns.
3581 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3582 REG_CFA_RESTORE insns here.
3585 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3586 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3588 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3590 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3591 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3592 (arc_init): Likewise.
3593 (arc_override_options): Likewise.
3594 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3596 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3598 * config/arc/arc.h (TARGET_DBNZ): Define.
3599 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3600 properly set the tune attribute.
3601 (dbnz): Use TARGET_DBNZ guard.
3602 * config/arc/arc.opt (mtune): Add core3 option.
3604 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3606 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3607 recognize new pic like addresses.
3608 (arc_delegitimize_address): Clean up.
3610 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3612 * config/arc/arc-arches.def: Option mrf16 valid for all
3614 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3615 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3616 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3617 * config/arc/arc-tables.opt: Regenerate.
3618 * config/arc/arc.c (arc_conditional_register_usage): Handle
3619 reduced register file case.
3620 (arc_file_start): Set must have build attributes.
3621 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3623 * config/arc/arc.opt (mrf16): Add new option.
3624 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3625 * config/arc/genmultilib.awk: Handle new mrf16 option.
3626 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3627 * config/arc/t-multilib: Regenerate.
3628 * doc/invoke.texi (ARC Options): Document mrf16 option.
3630 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3632 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3633 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3634 (arc_attribute_table): Add 'secure_call' attribute.
3635 (arc_print_operand): Print secure call operand.
3636 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3638 (arc_is_secure_call_p): New function. * config/arc/arc.md
3639 (call_i): Add support for sjli instruction.
3640 (call_value_i): Likewise.
3641 * config/arc/constraints.md (Csc): New constraint.
3643 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3644 John Eric Martin <John.Martin@emmicro-us.com>
3646 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3647 * config/arc/arc.c (_arc_jli_section): New struct.
3648 (arc_jli_section): New type.
3649 (rc_jli_sections): New static variable.
3650 (arc_handle_jli_attribute): New function.
3651 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3652 (arc_file_end): New function.
3653 (TARGET_ASM_FILE_END): Define.
3654 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3655 (arc_add_jli_section): New function.
3656 (jli_call_scan): Likewise.
3657 (arc_reorg): Call jli_call_scan.
3658 (arc_output_addsi): Remove 'S' from printing asm operand.
3659 (arc_is_jli_call_p): New function.
3660 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3662 (movhi_insn): Likewise.
3663 (movsi_insn): Likewise.
3664 (movsi_set_cc_insn): Likewise.
3665 (loadqi_update): Likewise.
3666 (load_zeroextendqisi_update): Likewise.
3667 (load_signextendqisi_update): Likewise.
3668 (loadhi_update): Likewise.
3669 (load_zeroextendhisi_update): Likewise.
3670 (load_signextendhisi_update): Likewise.
3671 (loadsi_update): Likewise.
3672 (loadsf_update): Likewise.
3673 (movsicc_insn): Likewise.
3674 (bset_insn): Likewise.
3675 (bxor_insn): Likewise.
3676 (bclr_insn): Likewise.
3677 (bmsk_insn): Likewise.
3678 (bicsi3_insn): Likewise.
3679 (cmpsi_cc_c_insn): Likewise.
3680 (movsi_ne): Likewise.
3681 (movsi_cond_exec): Likewise.
3682 (clrsbsi2): Likewise.
3693 (call_i): Remove 'S' asm letter, add jli instruction.
3694 (call_value_i): Likewise.
3695 * config/arc/arc.op (mjli-always): New option.
3696 * config/arc/constraints.md (Cji): New constraint.
3697 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3699 (subsf3_fpx): Likewise.
3700 (mulsf3_fpx): Likewise.
3701 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3703 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3705 * doc/invoke.texi (ARC): Document mjli-always option.
3707 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3709 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3710 avoid addition with 0 and use incw and decw where possible.
3712 2018-01-26 Richard Biener <rguenther@suse.de>
3714 PR tree-optimization/81082
3715 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3716 association if it requires casting to unsigned.
3717 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3718 from fold_plusminus_mult_expr to catch important cases late when
3719 range info is available.
3721 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3723 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3724 * configure.ac (hidden_linkonce): New test.
3725 * configure: Regenerate.
3726 * config.in: Regenerate.
3728 2018-01-26 Julia Koval <julia.koval@intel.com>
3730 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3731 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3732 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3733 _mm_mask_bitshuffle_epi64_mask): Fix type.
3734 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3735 USI_FTYPE_V4DI_V4DI_USI): Remove.
3736 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3737 __builtin_ia32_vpshufbitqmb256_mask,
3738 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3739 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3740 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3742 2018-01-26 Alan Modra <amodra@gmail.com>
3745 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3746 UNSPEC_VBPERMQ. Sort other unspecs.
3748 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3750 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3752 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3755 * predict.c (drop_profile): Do not push/pop cfun; update also
3757 (handle_missing_profiles): Fix logic looking for zero profiles.
3759 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3762 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3763 on functions with #pragma omp declare simd or functions with simd
3765 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3766 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3767 Remove trailing \n from warning_at calls.
3769 2018-01-25 Tom de Vries <tom@codesourcery.com>
3772 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3773 for neutered workers.
3775 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3778 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3779 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3781 2018-01-24 Jeff Law <law@redhat.com>
3784 * i386.c (get_probe_interval): Move to earlier point.
3785 (ix86_compute_frame_layout): If -fstack-clash-protection and
3786 the frame is larger than the probe interval, then use pushes
3787 to save registers rather than reg->mem moves.
3788 (ix86_expand_prologue): Remove conditional for int_registers_saved
3791 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3794 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3795 min/max for never referenced object.
3797 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3800 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3802 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3803 attributes from DECL_ATTRIBUTES (decl) without affecting
3804 DECL_ATTRIBUTES (current_function_decl).
3805 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3806 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3808 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3810 PR tree-optimization/83979
3811 * fold-const.c (fold_comparison): Use constant_boolean_node
3812 instead of boolean_{true,false}_node.
3814 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3816 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3819 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3821 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3822 Simplify the clause that sets the length attribute.
3823 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3824 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3825 clause that sets the length attribute.
3826 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3828 2018-01-24 Tom de Vries <tom@codesourcery.com>
3831 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3832 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3833 Add strict parameter.
3834 (prevent_branch_around_nothing): Insert dummy insn between branch to
3835 label and label with no ptx insn inbetween.
3836 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3838 2018-01-24 Tom de Vries <tom@codesourcery.com>
3841 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3842 for neutered threads in warp.
3843 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3845 2018-01-24 Richard Biener <rguenther@suse.de>
3847 PR tree-optimization/83176
3848 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3851 2018-01-24 Richard Biener <rguenther@suse.de>
3853 PR tree-optimization/82819
3854 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3855 code generating pluses that are no-ops in the target precision.
3857 2018-01-24 Richard Biener <rguenther@suse.de>
3860 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3862 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3864 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3865 to merge probabilities.
3866 * predict.c (probably_never_executed): Also mark as cold functions
3867 with global 0 profile and guessed local profile.
3868 * profile-count.c (profile_probability::combine_with_count): New
3870 * profile-count.h (profile_probability::operator*,
3871 profile_probability::operator*=, profile_probability::operator/,
3872 profile_probability::operator/=): Reduce precision to adjusted
3873 and set value to guessed on contradictory divisions.
3874 (profile_probability::combine_with_freq): Remove.
3875 (profile_probability::combine_wiht_count): Declare.
3876 (profile_count::force_nonzero):: Set to adjusted.
3877 (profile_count::probability_in):: Set quality to adjusted.
3878 * tree-ssa-tail-merge.c (replace_block_by): Use
3881 2018-01-23 Andrew Waterman <andrew@sifive.com>
3882 Jim Wilson <jimw@sifive.com>
3884 * config/riscv/riscv.c (riscv_stack_boundary): New.
3885 (riscv_option_override): Set riscv_stack_boundary. Handle
3886 riscv_preferred_stack_boundary_arg.
3887 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3888 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3889 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3890 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3891 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3892 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3894 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3897 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3898 of struct ix86_frame.
3899 (ix86_expand_epilogue): Likewise. Add a local variable for
3900 the reg_save_offset field in struct ix86_frame.
3902 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3904 PR tree-optimization/82604
3905 * tree-loop-distribution.c (enum partition_kind): New enum item
3906 PKIND_PARTIAL_MEMSET.
3907 (partition_builtin_p): Support above new enum item.
3908 (generate_code_for_partition): Ditto.
3909 (compute_access_range): Differentiate cases that equality can be
3910 proven at all loops, the innermost loops or no loops.
3911 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3912 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3913 (finalize_partitions, distribute_loop): Don't fuse partition of
3914 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3915 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3918 2018-01-23 Martin Liska <mliska@suse.cz>
3920 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3921 order to ignore the predictor.
3922 (PRED_POLYMORPHIC_CALL): Likewise.
3923 (PRED_RECURSIVE_CALL): Likewise.
3925 2018-01-23 Martin Liska <mliska@suse.cz>
3927 * tree-profile.c (tree_profiling): Print function header to
3928 aware reader which function we are working on.
3929 * value-prof.c (gimple_find_values_to_profile): Do not print
3930 not interesting value histograms.
3932 2018-01-23 Martin Liska <mliska@suse.cz>
3934 * profile-count.h (enum profile_quality): Add
3935 profile_uninitialized as the first value. Do not number values
3936 as they are zero based.
3937 (profile_count::verify): Update sanity check.
3938 (profile_probability::verify): Likewise.
3940 2018-01-23 Nathan Sidwell <nathan@acm.org>
3942 * doc/invoke.texi (ffor-scope): Deprecate.
3944 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3946 PR tree-optimization/83510
3947 * domwalk.c (set_all_edges_as_executable): New function.
3948 (dom_walker::dom_walker): Convert bool param
3949 "skip_unreachable_blocks" to enum reachability. Move setup of
3950 edge flags to set_all_edges_as_executable and only do it when
3951 reachability is REACHABLE_BLOCKS.
3952 * domwalk.h (enum dom_walker::reachability): New enum.
3953 (dom_walker::dom_walker): Convert bool param
3954 "skip_unreachable_blocks" to enum reachability.
3955 (set_all_edges_as_executable): New decl.
3956 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3957 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3959 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3960 but converting true to REACHABLE_BLOCKS.
3961 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3963 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3964 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3965 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3967 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3968 if check_all_array_refs will be called.
3970 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3972 * tree.c (selftest::test_location_wrappers): Add more test
3975 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3977 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3978 (selftest::test_bit_in_range): Likewise.
3980 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3983 * doc/sourcebuild.texi (vect_float): Say that the selector
3984 only describes the situation when -funsafe-math-optimizations is on.
3985 (vect_float_strict): Document.
3987 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3989 PR tree-optimization/83965
3990 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3991 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3992 instead of checking only for a reduction.
3993 (vect_recog_widen_sum_pattern): Likewise.
3995 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3997 * predict.c (probably_never_executed): Only use precise profile info.
3998 (compute_function_frequency): Skip after inlining hack since we now
3999 have quality checking.
4001 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4003 * profile-count.h (profile_probability::very_unlikely,
4004 profile_probability::unlikely, profile_probability::even): Set
4005 precision to guessed.
4007 2018-01-23 Richard Biener <rguenther@suse.de>
4009 PR tree-optimization/83963
4010 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4011 Properly terminate dominator walk when crossing the exit edge not
4012 when visiting its source block.
4014 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4017 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4018 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4020 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4022 PR tree-optimization/83957
4023 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4024 semicolon after for body surrounded by braces.
4026 PR tree-optimization/83081
4027 * profile-count.h (profile_probability::split): New method.
4028 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4029 Use profile_probability::split.
4030 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4031 when splitting a single conditional jump into 2.
4033 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4035 PR tree-optimization/69452
4036 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4039 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4041 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4042 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4043 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4045 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4047 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4048 * config/rl78/rl78.md (movdi): New define_expand.
4049 * config/rl78/rl78.c (rl78_split_movdi): New function.
4051 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4054 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4056 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4057 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4058 128-bit to produce an UNSPEC move to get the double word with the
4059 signbit and then a shift directly to do signbit.
4060 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4061 implementation with a new version that just does either a direct
4062 move or a regular move. Move memory interface to separate insns.
4063 Move insns so they are next to the expander.
4064 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4065 with signbit move. Split big and little endian case.
4066 (signbit<mode>2_dm_mem_le): Likewise.
4067 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4068 (signbit<mode>2_dm2): Likewise.
4070 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4072 * config/rl78/rl78.md (anddi3): New define_expand.
4074 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4076 * config/rl78/rl78.md (umindi3): New define_expand.
4078 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4080 * config/rl78/rl78.md (smindi3): New define_expand.
4082 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4084 * config/rl78/rl78.md (smaxdi3): New define_expand.
4086 2018-01-22 Carl Love <cel@us.ibm.com>
4088 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4089 LVX_V1TI): Add macro expansion.
4090 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4091 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4092 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4093 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4094 Change check to determine if the instruction is a byte reversing
4095 entry. Fix typo in comment.
4096 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4097 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4098 Add def_builtin calls for new builtins.
4099 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4100 Add define_insn expansion.
4102 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4104 * config/rl78/rl78.md (umaxdi3): New define_expand.
4106 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4108 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4109 for non-QImode registers.
4111 2018-01-22 Richard Biener <rguenther@suse.de>
4113 PR tree-optimization/83963
4114 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4115 including the loop exit block.
4116 (scop_detection::merge_sese): Likewise.
4117 (scop_detection::add_scop): Do it here instead.
4119 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4121 * doc/sourcebuild.texi (arm_softfloat): Document.
4123 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4126 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4127 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4128 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4130 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4131 David Edelsohn <dje.gcc@gmail.com>
4134 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4135 Change "crset eq" to "crset 2".
4136 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4137 (*call_indirect_aix<mode>_nospec): Likewise.
4138 (*call_value_indirect_aix<mode>_nospec): Likewise.
4139 (*call_indirect_elfv2<mode>_nospec): Likewise.
4140 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4141 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4142 change assembly output from . to $.
4143 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4144 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4145 (*tablejump<mode>_internal1_nospec): Likewise.
4147 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4150 * config/sh/sh_optimize_sett_clrt.cc:
4151 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4153 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4155 PR tree-optimization/83940
4156 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4157 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4158 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4159 use it to pass back the definition type.
4160 (vect_check_store_rhs): Likewise rhs_dt_out.
4161 (vect_build_gather_load_calls): Add a mask_dt argument and use
4162 it instead of a call to vect_is_simple_use.
4163 (vectorizable_store): Update calls to vect_check_load_store_mask
4164 and vect_check_store_rhs. Use the dt returned by the latter instead
4165 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4166 instead of calls to vect_is_simple_use. Pass the scalar rather
4167 than the vector operand to vect_is_simple_use when handling
4168 second and subsequent copies of an rhs value.
4169 (vectorizable_load): Update calls to vect_check_load_store_mask
4170 and vect_build_gather_load_calls. Use the cached mask_dt and
4171 gs_info.offset_dt instead of calls to vect_is_simple_use.
4173 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4176 * tree-emutls.c: Include gimplify.h.
4177 (lower_emutls_2): New function.
4178 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4179 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4180 it before further processing.
4183 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4184 UINTVAL (trueop1) instead of INTVAL (op1).
4186 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4190 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4191 INCOMING_FRAME_SP_OFFSET if not defined.
4192 (scan_trace): Add ENTRY argument. If true and
4193 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4194 emit a note to adjust the CFA offset.
4195 (create_cfi_notes): Adjust scan_trace callers.
4196 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4197 INCOMING_FRAME_SP_OFFSET in the CIE.
4198 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4199 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4201 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4202 * doc/tm.texi: Regenerated.
4204 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4206 PR rtl-optimization/83147
4207 * lra-constraints.c (remove_inheritance_pseudos): Use
4208 lra_substitute_pseudo_within_insn.
4210 2018-01-19 Tom de Vries <tom@codesourcery.com>
4211 Cesar Philippidis <cesar@codesourcery.com>
4214 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4216 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4219 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4220 spaces for function labels.
4222 2018-01-19 Martin Liska <mliska@suse.cz>
4224 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4225 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4226 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4227 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4228 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4229 (PRED_CONST_RETURN): Change from 69 to 65.
4230 (PRED_NULL_RETURN): Change from 91 to 71.
4231 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4232 (PRED_LOOP_GUARD): Change from 66 to 73.
4234 2018-01-19 Martin Liska <mliska@suse.cz>
4236 * predict.c (predict_insn_def): Add new assert.
4237 (struct branch_predictor): Change type to signed integer.
4238 (test_prediction_value_range): Amend test to cover
4240 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4241 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4242 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4243 (PRED_LOOP_IV_COMPARE): Likewise.
4244 * predict.h (PROB_UNINITIALIZED): Define new constant.
4246 2018-01-19 Martin Liska <mliska@suse.cz>
4248 * predict.c (dump_prediction): Add new format for
4249 analyze_brprob.py script which is enabled with -details
4251 * profile-count.h (precise_p): New function.
4253 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4255 PR tree-optimization/83922
4256 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4257 there are no statements that need masking.
4258 (vect_active_double_reduction_p): New function.
4259 (vect_analyze_loop_operations): Use it when handling phis that
4260 are not in the loop header.
4262 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4264 PR tree-optimization/83914
4265 * tree-vect-loop.c (vectorizable_induction): Don't convert
4266 init_expr or apply the peeling adjustment for inductions
4267 that are nested within the vectorized loop.
4269 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4271 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4274 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4278 * function.h (gimplify_parameters): Add gimple_seq * argument.
4279 * function.c: Include gimple.h and options.h.
4280 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4281 for the added local temporaries if needed.
4282 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4283 if there are any parameter cleanups, wrap whole body into a
4284 try/finally with the cleanups.
4286 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4289 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4290 Use GET_MODE_CLASS for scalar floating point.
4292 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4296 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4297 Fix call of call_cgraph_insertion_hooks.
4299 2018-01-18 Martin Sebor <msebor@redhat.com>
4301 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4303 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4306 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4309 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4312 * common.opt: (-ffile-prefix-map): New option.
4313 * opts.c (common_handle_option): Defer it.
4314 * opts-global.c (handle_common_deferred_options): Handle it.
4315 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4316 * file-prefix-map.h: New file.
4317 (remap_debug_filename, add_debug_prefix_map): ...here.
4318 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4319 * final.c (debug_prefix_map, add_debug_prefix_map
4320 remap_debug_filename): Move to...
4321 * file-prefix-map.c: New file.
4322 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4323 generalize, get rid of alloca(), use strrchr() instead of strchr().
4324 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4325 Implement in terms of add_prefix_map().
4326 (remap_macro_filename, remap_debug_filename): Implement in term of
4328 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4329 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4330 * dbxout.c: Include file-prefix-map.h.
4331 * varasm.c: Likewise.
4332 * vmsdbgout.c: Likewise.
4333 * xcoffout.c: Likewise.
4334 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4335 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4336 * doc/invoke.texi (-ffile-prefix-map): Document.
4337 (-fdebug-prefix-map): Update description.
4339 2018-01-18 Martin Liska <mliska@suse.cz>
4341 * config/i386/i386.c (indirect_thunk_name): Document that also
4343 (output_indirect_thunk): Document why both instructions
4344 (pause and lfence) are generated.
4346 2018-01-18 Richard Biener <rguenther@suse.de>
4348 PR tree-optimization/83887
4349 * graphite-scop-detection.c
4350 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4351 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4352 (scop_detection::merge_sese): Re-implement with a flood-fill
4353 algorithm that properly finds a SESE region if it exists.
4355 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4358 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4359 pointer_diff optimizations use view_convert instead of convert.
4361 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4363 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4364 Generate different code for -mno-speculate-indirect-jumps.
4365 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4366 (*call_indirect_aix<mode>): Disable for
4367 -mno-speculate-indirect-jumps.
4368 (*call_indirect_aix<mode>_nospec): New define_insn.
4369 (*call_value_indirect_aix<mode>): Disable for
4370 -mno-speculate-indirect-jumps.
4371 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4372 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4373 -mno-speculate-indirect-jumps.
4374 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4376 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4378 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4379 long double type, set the flags for noting the default long double
4380 type, even if we don't pass or return a long double type.
4382 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4385 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4388 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4390 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4391 support for merge[hl].
4392 (fold_mergehl_helper): New helper function.
4393 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4394 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4395 (altivec_vmrglw_direct): Add xxmrglw insn.
4397 2018-01-17 Andrew Waterman <andrew@sifive.com>
4399 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4400 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4402 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4405 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4406 call the lto_location_cache before reading the
4407 DECL_SOURCE_LOCATION of the types.
4409 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4410 Richard Sandiford <richard.sandiford@linaro.org>
4412 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4413 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4414 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4415 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4416 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4418 * config/aarch64/constraints.md (aarch64_movti_operand):
4420 * config/aarch64/predicates.md (Uti): Add new constraint.
4422 2018-01-17 Carl Love <cel@us.ibm.com>
4424 * config/rs6000/vsx.md (define_expand xl_len_r,
4425 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4426 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4428 (define_expand, define_insn): Move the shift left from the
4429 define_insn to the define_expand for lxvl and stxvl instructions.
4430 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4431 and XL_LEN_R definitions to PURE.
4433 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4435 * config/i386/i386.c (indirect_thunk_name): Declare regno
4436 as unsigned int. Compare regno with INVALID_REGNUM.
4437 (output_indirect_thunk): Ditto.
4438 (output_indirect_thunk_function): Ditto.
4439 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4440 in the call to output_indirect_thunk_function.
4442 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4445 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4446 rather than the size of inner_type to determine the stack slot size
4447 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4449 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4452 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4455 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4457 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4458 endian Linux systems to optionally enable multilibs for selecting
4459 the long double type if the user configured an explicit type.
4460 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4461 have no long double multilibs if not defined.
4462 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4463 warn if the user used -mabi={ieee,ibm}longdouble and we built
4464 multilibs for long double.
4465 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4466 appropriate multilib option.
4467 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4469 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4470 for building long double multilibs.
4471 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4473 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4475 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4478 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4480 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4483 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4486 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4489 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4491 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4492 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4494 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4496 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4497 different rtl trees depending on TARGET_64BIT.
4498 (rs6000_gen_lvx): Likewise.
4500 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4502 * config/visium/visium.md (nop): Tweak comment.
4503 (hazard_nop): Likewise.
4505 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4507 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4508 -mspeculate-indirect-jumps.
4509 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4510 for -mno-speculate-indirect-jumps.
4511 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4512 (*call_value_indirect_elfv2<mode>): Disable for
4513 -mno-speculate-indirect-jumps.
4514 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4515 (indirect_jump): Emit different RTL for
4516 -mno-speculate-indirect-jumps.
4517 (*indirect_jump<mode>): Disable for
4518 -mno-speculate-indirect-jumps.
4519 (*indirect_jump<mode>_nospec): New define_insn.
4520 (tablejump): Emit different RTL for
4521 -mno-speculate-indirect-jumps.
4522 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4523 (tablejumpsi_nospec): New define_expand.
4524 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4525 (tablejumpdi_nospec): New define_expand.
4526 (*tablejump<mode>_internal1): Disable for
4527 -mno-speculate-indirect-jumps.
4528 (*tablejump<mode>_internal1_nospec): New define_insn.
4529 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4532 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4534 * caller-save.c (insert_save): Drop unnecessary parameter. All
4537 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4538 Richard Biener <rguenth@suse.de>
4541 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4542 return early, inline manually is_gimple_sizepos. Make sure if we
4543 call gimplify_expr we don't end up with a gimple constant.
4544 * tree.c (variably_modified_type_p): Don't return true for
4545 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4546 * gimplify.h (is_gimple_sizepos): Remove.
4548 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4550 PR tree-optimization/83857
4551 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4552 vectorizable_live_operation for pure SLP statements.
4553 (vectorizable_live_operation): Handle PHIs.
4555 2018-01-16 Richard Biener <rguenther@suse.de>
4557 PR tree-optimization/83867
4558 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4559 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4561 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4564 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4565 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4566 If off is not INTEGER_CST, issue a may not be aligned warning
4567 rather than isn't aligned. Use isn%'t rather than isn't.
4568 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4570 <case MULT_EXPR>: Improve the case when bottom and one of the
4571 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4572 operand, in that case check if the other operand is multiple of
4573 bottom divided by the INTEGER_CST operand.
4575 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4578 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4579 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4580 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4581 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4582 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4583 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4584 (pa_function_arg_size): New function.
4586 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4588 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4589 in a separate statement.
4591 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4593 PR tree-optimization/83847
4594 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4595 group gathers and scatters.
4597 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4599 PR rtl-optimization/86620
4600 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4602 PR rtl-optimization/83213
4603 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4604 to last if both are JUMP_INSNs.
4606 PR tree-optimization/83843
4607 * gimple-ssa-store-merging.c
4608 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4609 store_immediate_info for bswap/nop orig_stores.
4611 2018-01-15 Andrew Waterman <andrew@sifive.com>
4613 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4615 <UDIV>: Increase cost if !TARGET_DIV.
4617 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4619 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4620 (define_attr "cr_logical_3op"): New.
4621 (cceq_ior_compare): Adjust.
4622 (cceq_ior_compare_complement): Adjust.
4623 (*cceq_rev_compare): Adjust.
4624 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4625 (is_cracked_insn): Adjust.
4626 (insn_must_be_first_in_group): Adjust.
4627 * config/rs6000/40x.md: Adjust.
4628 * config/rs6000/440.md: Adjust.
4629 * config/rs6000/476.md: Adjust.
4630 * config/rs6000/601.md: Adjust.
4631 * config/rs6000/603.md: Adjust.
4632 * config/rs6000/6xx.md: Adjust.
4633 * config/rs6000/7450.md: Adjust.
4634 * config/rs6000/7xx.md: Adjust.
4635 * config/rs6000/8540.md: Adjust.
4636 * config/rs6000/cell.md: Adjust.
4637 * config/rs6000/e300c2c3.md: Adjust.
4638 * config/rs6000/e500mc.md: Adjust.
4639 * config/rs6000/e500mc64.md: Adjust.
4640 * config/rs6000/e5500.md: Adjust.
4641 * config/rs6000/e6500.md: Adjust.
4642 * config/rs6000/mpc.md: Adjust.
4643 * config/rs6000/power4.md: Adjust.
4644 * config/rs6000/power5.md: Adjust.
4645 * config/rs6000/power6.md: Adjust.
4646 * config/rs6000/power7.md: Adjust.
4647 * config/rs6000/power8.md: Adjust.
4648 * config/rs6000/power9.md: Adjust.
4649 * config/rs6000/rs64.md: Adjust.
4650 * config/rs6000/titan.md: Adjust.
4652 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4654 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4655 ix86_indirect_branch_register logic.
4657 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4659 * config/i386/constraints.md (Bs): Update
4660 ix86_indirect_branch_register check. Don't check
4661 ix86_indirect_branch_register with GOT_memory_operand.
4663 * config/i386/predicates.md (GOT_memory_operand): Don't check
4664 ix86_indirect_branch_register here.
4665 (GOT32_symbol_operand): Likewise.
4667 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4669 * config/i386/predicates.md (constant_call_address_operand):
4670 Rewrite ix86_indirect_branch_register logic.
4671 (sibcall_insn_operand): Likewise.
4673 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4675 * config/i386/constraints.md (Bs): Replace
4676 ix86_indirect_branch_thunk_register with
4677 ix86_indirect_branch_register.
4679 * config/i386/i386.md (indirect_jump): Likewise.
4680 (tablejump): Likewise.
4681 (*sibcall_memory): Likewise.
4682 (*sibcall_value_memory): Likewise.
4683 Peepholes of indirect call and jump via memory: Likewise.
4684 * config/i386/i386.opt: Likewise.
4685 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4686 (GOT_memory_operand): Likewise.
4687 (call_insn_operand): Likewise.
4688 (sibcall_insn_operand): Likewise.
4689 (GOT32_symbol_operand): Likewise.
4691 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4694 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4695 type rather than type addr's type points to.
4696 (expand_omp_atomic_mutex): Likewise.
4697 (expand_omp_atomic): Likewise.
4699 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4702 * config/i386/i386.c (output_indirect_thunk_function): Use
4703 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4704 for __x86_return_thunk.
4706 2018-01-15 Richard Biener <rguenther@suse.de>
4709 * expmed.c (extract_bit_field_1): Fix typo.
4711 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4714 * config/arm/iterators.md (VF): New mode iterator.
4715 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4716 Remove integer-related logic from pattern.
4717 (neon_vabd<mode>_3): Likewise.
4719 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4722 * common.opt (fstrict-overflow): No longer an alias.
4723 (fwrapv-pointer): New option.
4724 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4725 also for pointer types based on flag_wrapv_pointer.
4726 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4727 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4728 opts->x_flag_wrapv got set.
4729 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4730 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4731 POINTER_TYPE_OVERFLOW_UNDEFINED.
4732 * match.pd: Likewise in address comparison pattern.
4733 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4735 2018-01-15 Richard Biener <rguenther@suse.de>
4738 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4739 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4740 Reset type names to their identifier if their TYPE_DECL doesn't
4741 have linkage (and thus is used for ODR and devirt).
4742 (save_debug_info_for_decl): Remove.
4743 (save_debug_info_for_type): Likewise.
4744 (add_tree_to_fld_list): Adjust.
4745 * tree-pretty-print.c (dump_generic_node): Make dumping of
4746 type names more robust.
4748 2018-01-15 Richard Biener <rguenther@suse.de>
4750 * BASE-VER: Bump to 8.0.1.
4752 2018-01-14 Martin Sebor <msebor@redhat.com>
4755 * builtins.c (check_access): Avoid warning when the no-warning bit
4758 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4760 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4761 * ira-color (allocno_hard_regs_compare): Likewise.
4763 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4766 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4767 Use .pushsection/.popsection.
4769 2018-01-14 Martin Sebor <msebor@redhat.com>
4772 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4774 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4776 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4777 entry from extra_headers.
4778 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4779 extra_headers, make the list bitwise identical to the i?86-*-* one.
4781 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4783 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4784 -mcmodel=large with -mindirect-branch=thunk,
4785 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4786 -mfunction-return=thunk-extern.
4787 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4788 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4789 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4791 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4793 * config/i386/i386.c (print_reg): Print the name of the full
4794 integer register without '%'.
4795 (ix86_print_operand): Handle 'V'.
4796 * doc/extend.texi: Document 'V' modifier.
4798 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4800 * config/i386/constraints.md (Bs): Disallow memory operand for
4801 -mindirect-branch-register.
4803 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4804 (GOT_memory_operand): Likewise.
4805 (call_insn_operand): Likewise.
4806 (sibcall_insn_operand): Likewise.
4807 (GOT32_symbol_operand): Likewise.
4808 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4809 for -mindirect-branch-register.
4810 (tablejump): Likewise.
4811 (*sibcall_memory): Likewise.
4812 (*sibcall_value_memory): Likewise.
4813 Disallow peepholes of indirect call and jump via memory for
4814 -mindirect-branch-register.
4815 (*call_pop): Replace m with Bw.
4816 (*call_value_pop): Likewise.
4817 (*sibcall_pop_memory): Replace m with Bs.
4818 * config/i386/i386.opt (mindirect-branch-register): New option.
4819 * doc/invoke.texi: Document -mindirect-branch-register option.
4821 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4823 * config/i386/i386-protos.h (ix86_output_function_return): New.
4824 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4825 set function_return_type.
4826 (indirect_thunk_name): Add ret_p to indicate thunk for function
4828 (output_indirect_thunk_function): Pass false to
4829 indirect_thunk_name.
4830 (ix86_output_indirect_branch_via_reg): Likewise.
4831 (ix86_output_indirect_branch_via_push): Likewise.
4832 (output_indirect_thunk_function): Create alias for function
4833 return thunk if regno < 0.
4834 (ix86_output_function_return): New function.
4835 (ix86_handle_fndecl_attribute): Handle function_return.
4836 (ix86_attribute_table): Add function_return.
4837 * config/i386/i386.h (machine_function): Add
4838 function_return_type.
4839 * config/i386/i386.md (simple_return_internal): Use
4840 ix86_output_function_return.
4841 (simple_return_internal_long): Likewise.
4842 * config/i386/i386.opt (mfunction-return=): New option.
4843 (indirect_branch): Mention -mfunction-return=.
4844 * doc/extend.texi: Document function_return function attribute.
4845 * doc/invoke.texi: Document -mfunction-return= option.
4847 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4849 * config/i386/i386-opts.h (indirect_branch): New.
4850 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4851 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4852 with local indirect jump when converting indirect call and jump.
4853 (ix86_set_indirect_branch_type): New.
4854 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4855 (indirectlabelno): New.
4856 (indirect_thunk_needed): Likewise.
4857 (indirect_thunk_bnd_needed): Likewise.
4858 (indirect_thunks_used): Likewise.
4859 (indirect_thunks_bnd_used): Likewise.
4860 (INDIRECT_LABEL): Likewise.
4861 (indirect_thunk_name): Likewise.
4862 (output_indirect_thunk): Likewise.
4863 (output_indirect_thunk_function): Likewise.
4864 (ix86_output_indirect_branch_via_reg): Likewise.
4865 (ix86_output_indirect_branch_via_push): Likewise.
4866 (ix86_output_indirect_branch): Likewise.
4867 (ix86_output_indirect_jmp): Likewise.
4868 (ix86_code_end): Call output_indirect_thunk_function if needed.
4869 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4871 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4872 (ix86_attribute_table): Add indirect_branch.
4873 * config/i386/i386.h (machine_function): Add indirect_branch_type
4874 and has_local_indirect_jump.
4875 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4877 (tablejump): Likewise.
4878 (*indirect_jump): Use ix86_output_indirect_jmp.
4879 (*tablejump_1): Likewise.
4880 (simple_return_indirect_internal): Likewise.
4881 * config/i386/i386.opt (mindirect-branch=): New option.
4882 (indirect_branch): New.
4885 (thunk-inline): Likewise.
4886 (thunk-extern): Likewise.
4887 * doc/extend.texi: Document indirect_branch function attribute.
4888 * doc/invoke.texi: Document -mindirect-branch= option.
4890 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4893 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4895 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4897 * ipa-inline.c (want_inline_small_function_p): Return false if
4898 inlining has already failed with CIF_FINAL_ERROR.
4899 (update_caller_keys): Call want_inline_small_function_p before
4901 (update_callee_keys): Likewise.
4903 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4905 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4907 (rs6000_quadword_masked_address_p): Likewise.
4908 (quad_aligned_load_p): Likewise.
4909 (quad_aligned_store_p): Likewise.
4910 (const_load_sequence_p): Add comment to describe the outer-most loop.
4911 (mimic_memory_attributes_and_flags): New function.
4912 (rs6000_gen_stvx): Likewise.
4913 (replace_swapped_aligned_store): Likewise.
4914 (rs6000_gen_lvx): Likewise.
4915 (replace_swapped_aligned_load): Likewise.
4916 (replace_swapped_load_constant): Capitalize argument name in
4917 comment describing this function.
4918 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4919 and stores that access quad-word aligned addresses and replace
4920 with stvx or lvx instructions when appropriate.
4921 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4922 New function prototype.
4923 (rs6000_quadword_masked_address_p): Likewise.
4924 (rs6000_gen_lvx): Likewise.
4925 (rs6000_gen_stvx): Likewise.
4926 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4927 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4928 when memory address is aligned.
4929 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4930 this split to select lvx instruction when memory address is aligned.
4931 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4932 instruction when memory address is aligned.
4933 (*vsx_le_perm_load_v16qi): Likewise.
4934 (four unnamed splitters): Modify to select the stvx instruction
4935 when memory is aligned.
4937 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4939 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4940 which appears in the queue multiple times.
4942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4943 Alan Hayward <alan.hayward@arm.com>
4944 David Sherwood <david.sherwood@arm.com>
4946 * tree-vectorizer.h (vec_lower_bound): New structure.
4947 (_loop_vec_info): Add check_nonzero and lower_bounds.
4948 (LOOP_VINFO_CHECK_NONZERO): New macro.
4949 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4950 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4951 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4952 fields. Make seg_len the distance travelled, not including the
4954 (dr_direction_indicator): Declare.
4955 (dr_zero_step_indicator): Likewise.
4956 (dr_known_forward_stride_p): Likewise.
4957 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4959 (runtime_alias_check_p): Allow runtime alias checks with
4961 (operator ==): Compare access_size and align.
4962 (prune_runtime_alias_test_list): Rework for new distinction between
4963 the access_size and seg_len.
4964 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4966 (get_segment_min_max): New function.
4967 (create_intersect_range_checks): Use it.
4968 (dr_step_indicator): New function.
4969 (dr_direction_indicator): Likewise.
4970 (dr_zero_step_indicator): Likewise.
4971 (dr_known_forward_stride_p): Likewise.
4972 * tree-loop-distribution.c (data_ref_segment_size): Return
4973 DR_STEP * (niters - 1).
4974 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4976 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4977 (vect_preserves_scalar_order_p): New function, split out from...
4978 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4979 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4980 (vect_vfa_access_size): New function.
4981 (vect_vfa_align): Likewise.
4982 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4983 (dump_lower_bound): New function.
4984 (vect_check_lower_bound): Likewise.
4985 (vect_small_gap_p): Likewise.
4986 (vectorizable_with_step_bound_p): Likewise.
4987 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4988 depencies if the vectorization factor is 1. Convert the checks
4989 for nonzero steps into checks on the bounds of DR_STEP. Try using
4990 a bunds check for variable steps if the minimum required step is
4991 relatively small. Update calls to the dr_with_seg_len
4992 constructor and to vect_compile_time_alias.
4993 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4995 (vect_loop_versioning): Call it.
4996 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4998 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5001 Alan Hayward <alan.hayward@arm.com>
5002 David Sherwood <david.sherwood@arm.com>
5004 * doc/sourcebuild.texi (vect_scatter_store): Document.
5005 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5007 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5009 * genopinit.c (main): Add supports_vec_scatter_store and
5010 supports_vec_scatter_store_cached to target_optabs.
5011 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5012 IFN_MASK_SCATTER_STORE.
5013 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5015 * internal-fn.h (internal_store_fn_p): Declare.
5016 (internal_fn_stored_value_index): Likewise.
5017 * internal-fn.c (scatter_store_direct): New macro.
5018 (expand_scatter_store_optab_fn): New function.
5019 (direct_scatter_store_optab_supported_p): New macro.
5020 (internal_store_fn_p): New function.
5021 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5022 IFN_MASK_SCATTER_STORE.
5023 (internal_fn_mask_index): Likewise.
5024 (internal_fn_stored_value_index): New function.
5025 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5027 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5028 * optabs-query.c (supports_vec_scatter_store_p): New function.
5029 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5030 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5031 true for scatter stores.
5032 (vect_gather_scatter_fn_p): Handle scatter stores too.
5033 (vect_check_gather_scatter): Consider using scatter stores if
5034 supports_vec_scatter_store_p.
5035 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5037 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5038 internal_fn_stored_value_index.
5039 (check_load_store_masking): Handle scatter stores too.
5040 (vect_get_store_rhs): Make public.
5041 (vectorizable_call): Use internal_store_fn_p.
5042 (vectorizable_store): Handle scatter store internal functions.
5043 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5044 when deciding whether the end of the group has been reached.
5045 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5046 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5047 (mask_scatter_store<mode>): New insns.
5049 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5050 Alan Hayward <alan.hayward@arm.com>
5051 David Sherwood <david.sherwood@arm.com>
5053 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5054 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5055 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5057 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5058 Use vect_truncate_gather_scatter_offset if we can't treat the
5059 operation as a normal gather load or scatter store.
5060 (get_group_load_store_type): Take the gather_scatter_info
5061 as argument. Try using a gather load or scatter store for
5062 single-element groups.
5063 (get_load_store_type): Update calls to get_group_load_store_type
5064 and vect_use_strided_gather_scatters_p.
5066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5067 Alan Hayward <alan.hayward@arm.com>
5068 David Sherwood <david.sherwood@arm.com>
5070 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5071 optional tree argument.
5072 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5074 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5075 but continue to use the current value as a fallback.
5076 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5077 to compare the updates.
5078 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5079 (get_load_store_type): Use it when handling a strided access.
5080 (vect_get_strided_load_store_ops): New function.
5081 (vect_get_data_ptr_increment): Likewise.
5082 (vectorizable_load): Handle strided gather loads. Always pass
5083 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5086 Alan Hayward <alan.hayward@arm.com>
5087 David Sherwood <david.sherwood@arm.com>
5089 * doc/md.texi (gather_load@var{m}): Document.
5090 (mask_gather_load@var{m}): Likewise.
5091 * genopinit.c (main): Add supports_vec_gather_load and
5092 supports_vec_gather_load_cached to target_optabs.
5093 * optabs-tree.c (init_tree_optimization_optabs): Use
5094 ggc_cleared_alloc to allocate target_optabs.
5095 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5096 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5098 * internal-fn.h (internal_load_fn_p): Declare.
5099 (internal_gather_scatter_fn_p): Likewise.
5100 (internal_fn_mask_index): Likewise.
5101 (internal_gather_scatter_fn_supported_p): Likewise.
5102 * internal-fn.c (gather_load_direct): New macro.
5103 (expand_gather_load_optab_fn): New function.
5104 (direct_gather_load_optab_supported_p): New macro.
5105 (direct_internal_fn_optab): New function.
5106 (internal_load_fn_p): Likewise.
5107 (internal_gather_scatter_fn_p): Likewise.
5108 (internal_fn_mask_index): Likewise.
5109 (internal_gather_scatter_fn_supported_p): Likewise.
5110 * optabs-query.c (supports_at_least_one_mode_p): New function.
5111 (supports_vec_gather_load_p): Likewise.
5112 * optabs-query.h (supports_vec_gather_load_p): Declare.
5113 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5114 and memory_type field.
5115 (NUM_PATTERNS): Bump to 15.
5116 * tree-vect-data-refs.c: Include internal-fn.h.
5117 (vect_gather_scatter_fn_p): New function.
5118 (vect_describe_gather_scatter_call): Likewise.
5119 (vect_check_gather_scatter): Try using internal functions for
5120 gather loads. Recognize existing calls to a gather load function.
5121 (vect_analyze_data_refs): Consider using gather loads if
5122 supports_vec_gather_load_p.
5123 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5124 (vect_get_gather_scatter_offset_type): Likewise.
5125 (vect_convert_mask_for_vectype): Likewise.
5126 (vect_add_conversion_to_patterm): Likewise.
5127 (vect_try_gather_scatter_pattern): Likewise.
5128 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5129 (vect_vect_recog_func_ptrs): Add it.
5130 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5131 internal_fn_mask_index and internal_gather_scatter_fn_p.
5132 (check_load_store_masking): Take the gather_scatter_info as an
5133 argument and handle gather loads.
5134 (vect_get_gather_scatter_ops): New function.
5135 (vectorizable_call): Check internal_load_fn_p.
5136 (vectorizable_load): Likewise. Handle gather load internal
5138 (vectorizable_store): Update call to check_load_store_masking.
5139 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5140 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5141 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5142 (aarch64_gather_scale_operand_d): New predicates.
5143 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5144 (mask_gather_load<mode>): New insns.
5146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5147 Alan Hayward <alan.hayward@arm.com>
5148 David Sherwood <david.sherwood@arm.com>
5150 * optabs.def (fold_left_plus_optab): New optab.
5151 * doc/md.texi (fold_left_plus_@var{m}): Document.
5152 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5153 * internal-fn.c (fold_left_direct): Define.
5154 (expand_fold_left_optab_fn): Likewise.
5155 (direct_fold_left_optab_supported_p): Likewise.
5156 * fold-const-call.c (fold_const_fold_left): New function.
5157 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5158 * tree-parloops.c (valid_reduction_p): New function.
5159 (gather_scalar_reductions): Use it.
5160 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5161 (vect_finish_replace_stmt): Declare.
5162 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5163 (needs_fold_left_reduction_p): New function, split out from...
5164 (vect_is_simple_reduction): ...here. Accept reductions that
5165 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5166 (vect_force_simple_reduction): Also store the reduction type in
5167 the assignment's STMT_VINFO_REDUC_TYPE.
5168 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5169 (merge_with_identity): New function.
5170 (vect_expand_fold_left): Likewise.
5171 (vectorize_fold_left_reduction): Likewise.
5172 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5173 scalar phi in place for it. Check for target support and reject
5174 cases that would reassociate the operation. Defer the transform
5175 phase to vectorize_fold_left_reduction.
5176 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5177 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5178 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5182 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5183 call to ifc_temp_var.
5185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5186 Alan Hayward <alan.hayward@arm.com>
5187 David Sherwood <david.sherwood@arm.com>
5189 * target.def (legitimize_address_displacement): Take the original
5190 offset as a poly_int.
5191 * targhooks.h (default_legitimize_address_displacement): Update
5193 * targhooks.c (default_legitimize_address_displacement): Likewise.
5194 * doc/tm.texi: Regenerate.
5195 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5196 as an argument, moving assert of ad->disp == ad->disp_term to...
5197 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5198 Try calling targetm.legitimize_address_displacement before expanding
5199 the address rather than afterwards, and adjust for the new interface.
5200 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5201 Match the new hook interface. Handle SVE addresses.
5202 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5207 * Makefile.in (OBJS): Add early-remat.o.
5208 * target.def (select_early_remat_modes): New hook.
5209 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5210 * doc/tm.texi: Regenerate.
5211 * targhooks.h (default_select_early_remat_modes): Declare.
5212 * targhooks.c (default_select_early_remat_modes): New function.
5213 * timevar.def (TV_EARLY_REMAT): New timevar.
5214 * passes.def (pass_early_remat): New pass.
5215 * tree-pass.h (make_pass_early_remat): Declare.
5216 * early-remat.c: New file.
5217 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5219 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5222 Alan Hayward <alan.hayward@arm.com>
5223 David Sherwood <david.sherwood@arm.com>
5225 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5226 vfm1 with a bound_epilog parameter.
5227 (vect_do_peeling): Update calls accordingly, and move the prologue
5228 call earlier in the function. Treat the base bound_epilog as 0 for
5229 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5230 this base when peeling for gaps.
5231 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5232 with fully-masked loops.
5233 (vect_estimate_min_profitable_iters): Handle the single peeled
5234 iteration in that case.
5236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5237 Alan Hayward <alan.hayward@arm.com>
5238 David Sherwood <david.sherwood@arm.com>
5240 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5241 single-element interleaving even if the size is not a power of 2.
5242 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5243 accesses for single-element interleaving if the group size is
5246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5247 Alan Hayward <alan.hayward@arm.com>
5248 David Sherwood <david.sherwood@arm.com>
5250 * doc/md.texi (fold_extract_last_@var{m}): Document.
5251 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5252 * optabs.def (fold_extract_last_optab): New optab.
5253 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5254 * internal-fn.c (fold_extract_direct): New macro.
5255 (expand_fold_extract_optab_fn): Likewise.
5256 (direct_fold_extract_optab_supported_p): Likewise.
5257 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5258 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5259 EXTRACT_LAST_REDUCTION.
5260 (get_initial_def_for_reduction): Do not create an initial vector
5261 for EXTRACT_LAST_REDUCTION reductions.
5262 (vectorizable_reduction): Leave the scalar phi in place for
5263 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5264 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5265 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5266 transform phase to vectorizable_condition.
5267 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5269 (vect_finish_stmt_generation): ...here.
5270 (vect_finish_replace_stmt): New function.
5271 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5272 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5274 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5277 Alan Hayward <alan.hayward@arm.com>
5278 David Sherwood <david.sherwood@arm.com>
5280 * doc/md.texi (extract_last_@var{m}): Document.
5281 * optabs.def (extract_last_optab): New optab.
5282 * internal-fn.def (EXTRACT_LAST): New internal function.
5283 * internal-fn.c (cond_unary_direct): New macro.
5284 (expand_cond_unary_optab_fn): Likewise.
5285 (direct_cond_unary_optab_supported_p): Likewise.
5286 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5287 loops using EXTRACT_LAST.
5288 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5289 (extract_last_<mode>): ...this optab.
5290 (vec_extract<mode><Vel>): Update accordingly.
5292 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5293 Alan Hayward <alan.hayward@arm.com>
5294 David Sherwood <david.sherwood@arm.com>
5296 * target.def (empty_mask_is_expensive): New hook.
5297 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5298 * doc/tm.texi: Regenerate.
5299 * targhooks.h (default_empty_mask_is_expensive): Declare.
5300 * targhooks.c (default_empty_mask_is_expensive): New function.
5301 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5302 if the target says that empty masks are expensive.
5303 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5305 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5307 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5308 Alan Hayward <alan.hayward@arm.com>
5309 David Sherwood <david.sherwood@arm.com>
5311 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5312 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5313 (vect_use_loop_mask_for_alignment_p): New function.
5314 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5315 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5316 niters_skip argument. Make sure that the first niters_skip elements
5317 of the first iteration are inactive.
5318 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5319 Update call to vect_set_loop_masks_directly.
5320 (get_misalign_in_elems): New function, split out from...
5321 (vect_gen_prolog_loop_niters): ...here.
5322 (vect_update_init_of_dr): Take a code argument that specifies whether
5323 the adjustment should be added or subtracted.
5324 (vect_update_init_of_drs): Likewise.
5325 (vect_prepare_for_masked_peels): New function.
5326 (vect_do_peeling): Skip prologue peeling if we're using a mask
5327 instead. Update call to vect_update_inits_of_drs.
5328 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5330 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5331 alignment. Do not include the number of peeled iterations in
5332 the minimum threshold in that case.
5333 (vectorizable_induction): Adjust the start value down by
5334 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5335 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5336 Take the number of skipped iterations into account when calculating
5338 * tree-vect-stmts.c (vect_gen_while_not): New function.
5340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5341 Alan Hayward <alan.hayward@arm.com>
5342 David Sherwood <david.sherwood@arm.com>
5344 * doc/sourcebuild.texi (vect_fully_masked): Document.
5345 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5347 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5349 (vect_analyze_loop_2): ...here. Don't check the vectorization
5350 factor against the number of loop iterations if the loop is
5353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5354 Alan Hayward <alan.hayward@arm.com>
5355 David Sherwood <david.sherwood@arm.com>
5357 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5358 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5359 (dump_groups): Update accordingly.
5360 (iv_use::mem_type): New member variable.
5361 (address_p): New function.
5362 (record_use): Add a mem_type argument and initialize the new
5364 (record_group_use): Add a mem_type argument. Use address_p.
5365 Remove obsolete null checks of base_object. Update call to record_use.
5366 (find_interesting_uses_op): Update call to record_group_use.
5367 (find_interesting_uses_cond): Likewise.
5368 (find_interesting_uses_address): Likewise.
5369 (get_mem_type_for_internal_fn): New function.
5370 (find_address_like_use): Likewise.
5371 (find_interesting_uses_stmt): Try find_address_like_use before
5372 calling find_interesting_uses_op.
5373 (addr_offset_valid_p): Use the iv mem_type field as the type
5374 of the addressed memory.
5375 (add_autoinc_candidates): Likewise.
5376 (get_address_cost): Likewise.
5377 (split_small_address_groups_p): Use address_p.
5378 (split_address_groups): Likewise.
5379 (add_iv_candidate_for_use): Likewise.
5380 (autoinc_possible_for_pair): Likewise.
5381 (rewrite_groups): Likewise.
5382 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5383 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5384 (get_alias_ptr_type_for_ptr_address): New function.
5385 (rewrite_use_address): Rewrite address uses in calls that were
5386 identified by find_address_like_use.
5388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5389 Alan Hayward <alan.hayward@arm.com>
5390 David Sherwood <david.sherwood@arm.com>
5392 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5394 * gimple-expr.h (is_gimple_addressable: Likewise.
5395 * gimple-expr.c (is_gimple_address): Likewise.
5396 * internal-fn.c (expand_call_mem_ref): New function.
5397 (expand_mask_load_optab_fn): Use it.
5398 (expand_mask_store_optab_fn): Likewise.
5400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5401 Alan Hayward <alan.hayward@arm.com>
5402 David Sherwood <david.sherwood@arm.com>
5404 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5405 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5406 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5407 (cond_umax@var{mode}): Document.
5408 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5409 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5410 (cond_umin_optab, cond_umax_optab): New optabs.
5411 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5412 (COND_IOR, COND_XOR): New internal functions.
5413 * internal-fn.h (get_conditional_internal_fn): Declare.
5414 * internal-fn.c (cond_binary_direct): New macro.
5415 (expand_cond_binary_optab_fn): Likewise.
5416 (direct_cond_binary_optab_supported_p): Likewise.
5417 (get_conditional_internal_fn): New function.
5418 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5419 Cope with reduction statements that are vectorized as calls rather
5421 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5422 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5423 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5424 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5425 (UNSPEC_COND_EOR): New unspecs.
5426 (optab): Add mappings for them.
5427 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5428 (sve_int_op, sve_fp_op): New int attributes.
5430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5431 Alan Hayward <alan.hayward@arm.com>
5432 David Sherwood <david.sherwood@arm.com>
5434 * optabs.def (while_ult_optab): New optab.
5435 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5436 * internal-fn.def (WHILE_ULT): New internal function.
5437 * internal-fn.h (direct_internal_fn_supported_p): New override
5438 that takes two types as argument.
5439 * internal-fn.c (while_direct): New macro.
5440 (expand_while_optab_fn): New function.
5441 (convert_optab_supported_p): Likewise.
5442 (direct_while_optab_supported_p): New macro.
5443 * wide-int.h (wi::udiv_ceil): New function.
5444 * tree-vectorizer.h (rgroup_masks): New structure.
5445 (vec_loop_masks): New typedef.
5446 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5448 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5449 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5450 (vect_max_vf): New function.
5451 (slpeel_make_loop_iterate_ntimes): Delete.
5452 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5453 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5454 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5455 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5456 internal-fn.h, stor-layout.h and optabs-query.h.
5457 (vect_set_loop_mask): New function.
5458 (add_preheader_seq): Likewise.
5459 (add_header_seq): Likewise.
5460 (interleave_supported_p): Likewise.
5461 (vect_maybe_permute_loop_masks): Likewise.
5462 (vect_set_loop_masks_directly): Likewise.
5463 (vect_set_loop_condition_masked): Likewise.
5464 (vect_set_loop_condition_unmasked): New function, split out from
5465 slpeel_make_loop_iterate_ntimes.
5466 (slpeel_make_loop_iterate_ntimes): Rename to..
5467 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5468 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5469 (vect_do_peeling): Update call accordingly.
5470 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5472 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5473 mask_compare_type, can_fully_mask_p and fully_masked_p.
5474 (release_vec_loop_masks): New function.
5475 (_loop_vec_info): Use it to free the loop masks.
5476 (can_produce_all_loop_masks_p): New function.
5477 (vect_get_max_nscalars_per_iter): Likewise.
5478 (vect_verify_full_masking): Likewise.
5479 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5480 retries, and free the mask rgroups before retrying. Check loop-wide
5481 reasons for disallowing fully-masked loops. Make the final decision
5482 about whether use a fully-masked loop or not.
5483 (vect_estimate_min_profitable_iters): Do not assume that peeling
5484 for the number of iterations will be needed for fully-masked loops.
5485 (vectorizable_reduction): Disable fully-masked loops.
5486 (vectorizable_live_operation): Likewise.
5487 (vect_halve_mask_nunits): New function.
5488 (vect_double_mask_nunits): Likewise.
5489 (vect_record_loop_mask): Likewise.
5490 (vect_get_loop_mask): Likewise.
5491 (vect_transform_loop): Handle the case in which the final loop
5492 iteration might handle a partial vector. Call vect_set_loop_condition
5493 instead of slpeel_make_loop_iterate_ntimes.
5494 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5495 (check_load_store_masking): New function.
5496 (prepare_load_store_mask): Likewise.
5497 (vectorizable_store): Handle fully-masked loops.
5498 (vectorizable_load): Likewise.
5499 (supportable_widening_operation): Use vect_halve_mask_nunits for
5501 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5502 (vect_gen_while): New function.
5503 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5504 (aarch64_uqdec<mode>): New insn.
5506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5507 Alan Hayward <alan.hayward@arm.com>
5508 David Sherwood <david.sherwood@arm.com>
5510 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5511 (reduc_xor_scal_optab): New optabs.
5512 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5513 (reduc_xor_scal_@var{m}): Document.
5514 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5515 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5517 * fold-const-call.c (fold_const_call): Handle them.
5518 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5519 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5520 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5521 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5522 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5523 (UNSPEC_XORV): New unspecs.
5524 (optab): Add entries for them.
5525 (BITWISEV): New int iterator.
5526 (bit_reduc_op): New int attributes.
5528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5529 Alan Hayward <alan.hayward@arm.com>
5530 David Sherwood <david.sherwood@arm.com>
5532 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5533 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5534 * optabs.def (vec_shl_insert_optab): New optab.
5535 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5536 (duplicate_and_interleave): Likewise.
5537 * tree-vect-loop.c: Include internal-fn.h.
5538 (neutral_op_for_slp_reduction): New function, split out from
5539 get_initial_defs_for_reduction.
5540 (get_initial_def_for_reduction): Handle option 2 for variable-length
5541 vectors by loading the neutral value into a vector and then shifting
5542 the initial value into element 0.
5543 (get_initial_defs_for_reduction): Replace the code argument with
5544 the neutral value calculated by neutral_op_for_slp_reduction.
5545 Use gimple_build_vector for constant-length vectors.
5546 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5547 but the first group_size elements have a neutral value.
5548 Use duplicate_and_interleave otherwise.
5549 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5550 Update call to get_initial_defs_for_reduction. Handle SLP
5551 reductions for variable-length vectors by creating one vector
5552 result for each scalar result, with the elements associated
5553 with other scalar results stubbed out with the neutral value.
5554 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5555 Require IFN_VEC_SHL_INSERT for double reductions on
5556 variable-length vectors, or SLP reductions that have
5557 a neutral value. Require can_duplicate_and_interleave_p
5558 support for variable-length unchained SLP reductions if there
5559 is no neutral value, such as for MIN/MAX reductions. Also require
5560 the number of vector elements to be a multiple of the number of
5561 SLP statements when doing variable-length unchained SLP reductions.
5562 Update call to vect_create_epilog_for_reduction.
5563 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5564 and remove initial values.
5565 (duplicate_and_interleave): Make public.
5566 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5567 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5570 Alan Hayward <alan.hayward@arm.com>
5571 David Sherwood <david.sherwood@arm.com>
5573 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5574 (can_duplicate_and_interleave_p): New function.
5575 (vect_get_and_check_slp_defs): Take the vector of statements
5576 rather than just the current one. Remove excess parentheses.
5577 Restriction rejectinon of vect_constant_def and vect_external_def
5578 for variable-length vectors to boolean types, or types for which
5579 can_duplicate_and_interleave_p is false.
5580 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5581 (duplicate_and_interleave): New function.
5582 (vect_get_constant_vectors): Use gimple_build_vector for
5583 constant-length vectors and suitable variable-length constant
5584 vectors. Use duplicate_and_interleave for other variable-length
5585 vectors. Don't defer the update when inserting new statements.
5587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5588 Alan Hayward <alan.hayward@arm.com>
5589 David Sherwood <david.sherwood@arm.com>
5591 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5592 min_profitable_iters doesn't go negative.
5594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5595 Alan Hayward <alan.hayward@arm.com>
5596 David Sherwood <david.sherwood@arm.com>
5598 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5599 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5600 * optabs.def (vec_mask_load_lanes_optab): New optab.
5601 (vec_mask_store_lanes_optab): Likewise.
5602 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5603 (MASK_STORE_LANES): Likewise.
5604 * internal-fn.c (mask_load_lanes_direct): New macro.
5605 (mask_store_lanes_direct): Likewise.
5606 (expand_mask_load_optab_fn): Handle masked operations.
5607 (expand_mask_load_lanes_optab_fn): New macro.
5608 (expand_mask_store_optab_fn): Handle masked operations.
5609 (expand_mask_store_lanes_optab_fn): New macro.
5610 (direct_mask_load_lanes_optab_supported_p): Likewise.
5611 (direct_mask_store_lanes_optab_supported_p): Likewise.
5612 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5614 (vect_load_lanes_supported): Likewise.
5615 * tree-vect-data-refs.c (strip_conversion): New function.
5616 (can_group_stmts_p): Likewise.
5617 (vect_analyze_data_ref_accesses): Use it instead of checking
5618 for a pair of assignments.
5619 (vect_store_lanes_supported): Take a masked_p parameter.
5620 (vect_load_lanes_supported): Likewise.
5621 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5622 vect_store_lanes_supported and vect_load_lanes_supported.
5623 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5624 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5625 parameter. Don't allow gaps for masked accesses.
5626 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5627 and vect_load_lanes_supported.
5628 (get_load_store_type): Take a masked_p parameter and update
5629 call to get_group_load_store_type.
5630 (vectorizable_store): Update call to get_load_store_type.
5631 Handle IFN_MASK_STORE_LANES.
5632 (vectorizable_load): Update call to get_load_store_type.
5633 Handle IFN_MASK_LOAD_LANES.
5635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5636 Alan Hayward <alan.hayward@arm.com>
5637 David Sherwood <david.sherwood@arm.com>
5639 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5641 * config/aarch64/aarch64-protos.h
5642 (aarch64_sve_struct_memory_operand_p): Declare.
5643 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5644 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5645 (VPRED, vpred): Handle SVE structure modes.
5646 * config/aarch64/constraints.md (Utx): New constraint.
5647 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5648 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5649 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5650 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5651 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5652 structure modes. Split into pieces after RA.
5653 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5654 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5656 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5657 SVE structure modes.
5658 (aarch64_classify_address): Likewise.
5659 (sizetochar): Move earlier in file.
5660 (aarch64_print_operand): Handle SVE register lists.
5661 (aarch64_array_mode): New function.
5662 (aarch64_sve_struct_memory_operand_p): Likewise.
5663 (TARGET_ARRAY_MODE): Redefine.
5665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5666 Alan Hayward <alan.hayward@arm.com>
5667 David Sherwood <david.sherwood@arm.com>
5669 * target.def (array_mode): New target hook.
5670 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5671 * doc/tm.texi: Regenerate.
5672 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5673 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5674 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5676 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5680 Alan Hayward <alan.hayward@arm.com>
5681 David Sherwood <david.sherwood@arm.com>
5683 * fold-const.c (fold_binary_loc): Check the argument types
5684 rather than the result type when testing for a vector operation.
5686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5688 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5689 * doc/tm.texi: Regenerate.
5691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5692 Alan Hayward <alan.hayward@arm.com>
5693 David Sherwood <david.sherwood@arm.com>
5695 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5696 (sve): Document new AArch64 extension.
5697 * doc/md.texi (w): Extend the description of the AArch64
5698 constraint to include SVE vectors.
5699 (Upl, Upa): Document new AArch64 predicate constraints.
5700 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5702 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5703 (msve-vector-bits=): New option.
5704 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5705 SVE when these are disabled.
5706 (sve): New extension.
5707 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5708 modes. Adjust their number of units based on aarch64_sve_vg.
5709 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5710 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5711 aarch64_addr_query_type.
5712 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5713 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5714 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5715 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5716 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5717 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5718 (aarch64_simd_imm_zero_p): Delete.
5719 (aarch64_check_zero_based_sve_index_immediate): Declare.
5720 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5721 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5722 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5723 (aarch64_sve_float_mul_immediate_p): Likewise.
5724 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5726 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5727 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5728 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5729 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5730 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5731 (aarch64_regmode_natural_size): Likewise.
5732 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5733 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5735 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5736 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5737 for VG and the SVE predicate registers.
5738 (V_ALIASES): Add a "z"-prefixed alias.
5739 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5740 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5741 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5742 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5743 (REG_CLASS_NAMES): Add entries for them.
5744 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5745 and the predicate registers.
5746 (aarch64_sve_vg): Declare.
5747 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5748 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5749 (REGMODE_NATURAL_SIZE): Define.
5750 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5752 * config/aarch64/aarch64.c: Include cfgrtl.h.
5753 (simd_immediate_info): Add a constructor for series vectors,
5754 and an associated step field.
5755 (aarch64_sve_vg): New variable.
5756 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5757 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5758 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5759 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5760 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5761 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5762 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5763 (aarch64_get_mask_mode): New functions.
5764 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5765 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5766 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5767 predicate modes and predicate registers. Explicitly restrict
5768 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5769 to store a vector mode if it is recognized by
5770 aarch64_classify_vector_mode.
5771 (aarch64_regmode_natural_size): New function.
5772 (aarch64_hard_regno_caller_save_mode): Return the original mode
5774 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5775 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5776 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5777 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5779 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5780 does not overlap dest if the function is frame-related. Handle
5782 (aarch64_split_add_offset): New function.
5783 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5784 them aarch64_add_offset.
5785 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5786 and update call to aarch64_sub_sp.
5787 (aarch64_add_cfa_expression): New function.
5788 (aarch64_expand_prologue): Pass extra temporary registers to the
5789 functions above. Handle the case in which we need to emit new
5790 DW_CFA_expressions for registers that were originally saved
5791 relative to the stack pointer, but now have to be expressed
5792 relative to the frame pointer.
5793 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5795 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5796 IP0 and IP1 values for SVE frames.
5797 (aarch64_expand_vec_series): New function.
5798 (aarch64_expand_sve_widened_duplicate): Likewise.
5799 (aarch64_expand_sve_const_vector): Likewise.
5800 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5801 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5802 into the register, rather than emitting a SET directly.
5803 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5804 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5805 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5806 (offset_9bit_signed_scaled_p): New functions.
5807 (aarch64_replicate_bitmask_imm): New function.
5808 (aarch64_bitmask_imm): Use it.
5809 (aarch64_cannot_force_const_mem): Reject expressions involving
5810 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5811 (aarch64_classify_index): Handle SVE indices, by requiring
5812 a plain register index with a scale that matches the element size.
5813 (aarch64_classify_address): Handle SVE addresses. Assert that
5814 the mode of the address is VOIDmode or an integer mode.
5815 Update call to aarch64_classify_symbol.
5816 (aarch64_classify_symbolic_expression): Update call to
5817 aarch64_classify_symbol.
5818 (aarch64_const_vec_all_in_range_p): New function.
5819 (aarch64_print_vector_float_operand): Likewise.
5820 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5821 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5822 and the FP immediates 1.0 and 0.5.
5823 (aarch64_print_address_internal): Handle SVE addresses.
5824 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5825 (aarch64_regno_regclass): Handle predicate registers.
5826 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5828 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5829 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5830 (aarch64_convert_sve_vector_bits): New function.
5831 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5832 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5834 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5835 Handle SVE vector and predicate modes. Accept VL-based constants
5836 that need only one temporary register, and VL offsets that require
5837 no temporary registers.
5838 (aarch64_conditional_register_usage): Mark the predicate registers
5839 as fixed if SVE isn't available.
5840 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5841 Return true for SVE vector and predicate modes.
5842 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5843 rather than an unsigned int. Handle SVE modes.
5844 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5846 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5848 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5849 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5850 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5851 (aarch64_sve_float_mul_immediate_p): New functions.
5852 (aarch64_sve_valid_immediate): New function.
5853 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5854 Explicitly reject structure modes. Check for INDEX constants.
5855 Handle PTRUE and PFALSE constants.
5856 (aarch64_check_zero_based_sve_index_immediate): New function.
5857 (aarch64_simd_imm_zero_p): Delete.
5858 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5859 vector modes. Accept constants in the range of CNT[BHWD].
5860 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5861 ask for an Advanced SIMD mode.
5862 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5863 (aarch64_simd_vector_alignment): Handle SVE predicates.
5864 (aarch64_vectorize_preferred_vector_alignment): New function.
5865 (aarch64_simd_vector_alignment_reachable): Use it instead of
5867 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5868 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5870 (MAX_VECT_LEN): Delete.
5871 (expand_vec_perm_d): Add a vec_flags field.
5872 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5873 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5874 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5876 (aarch64_evpc_rev): Rename to...
5877 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5878 (aarch64_evpc_rev_global): New function.
5879 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5880 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5882 (aarch64_evpc_sve_tbl): New function.
5883 (aarch64_expand_vec_perm_const_1): Update after rename of
5884 aarch64_evpc_rev. Handle SVE permutes too, trying
5885 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5886 than aarch64_evpc_tbl.
5887 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5888 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5889 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5890 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5891 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5892 (aarch64_expand_sve_vcond): New functions.
5893 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5894 of aarch64_vector_mode_p.
5895 (aarch64_dwarf_poly_indeterminate_value): New function.
5896 (aarch64_compute_pressure_classes): Likewise.
5897 (aarch64_can_change_mode_class): Likewise.
5898 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5899 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5900 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5901 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5902 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5903 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5904 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5905 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5907 (Dn, Dl, Dr): Accept const as well as const_vector.
5908 (Dz): Likewise. Compare against CONST0_RTX.
5909 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5910 of "vector" where appropriate.
5911 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5912 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5913 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5914 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5915 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5916 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5917 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5918 (v_int_equiv): Extend to SVE modes.
5919 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5921 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5922 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5923 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5924 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5925 (SVE_COND_FP_CMP): New int iterators.
5926 (perm_hilo): Handle the new unpack unspecs.
5927 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5929 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5930 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5931 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5932 (aarch64_equality_operator, aarch64_constant_vector_operand)
5933 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5934 (aarch64_sve_nonimmediate_operand): Likewise.
5935 (aarch64_sve_general_operand): Likewise.
5936 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5937 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5938 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5939 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5940 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5941 (aarch64_sve_float_arith_immediate): Likewise.
5942 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5943 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5944 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5945 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5946 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5947 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5948 (aarch64_sve_float_arith_operand): Likewise.
5949 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5950 (aarch64_sve_float_mul_operand): Likewise.
5951 (aarch64_sve_vec_perm_operand): Likewise.
5952 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5953 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5954 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5955 as well as const_vector.
5956 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5957 in file. Use CONST0_RTX and CONSTM1_RTX.
5958 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5959 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5960 Use aarch64_simd_imm_zero.
5961 * config/aarch64/aarch64-sve.md: New file.
5962 * config/aarch64/aarch64.md: Include it.
5963 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5964 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5965 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5966 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5967 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5968 (sve): New attribute.
5969 (enabled): Disable instructions with the sve attribute unless
5971 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5972 aarch64_expand_mov_immediate.
5973 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5974 CNT[BHSD] immediates.
5975 (movti): Split CONST_POLY_INT moves into two halves.
5976 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5977 Split additions that need a temporary here if the destination
5978 is the stack pointer.
5979 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5980 (*add<mode>3_poly_1): New instruction.
5981 (set_clobber_cc): New expander.
5983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5985 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5986 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5987 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5988 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5989 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5990 Change innermode from fixed_mode_size to machine_mode.
5991 (simplify_subreg): Update call accordingly. Handle a constant-sized
5992 subreg of a variable-length CONST_VECTOR.
5994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5995 Alan Hayward <alan.hayward@arm.com>
5996 David Sherwood <david.sherwood@arm.com>
5998 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5999 (add_offset_to_base): New function, split out from...
6000 (create_mem_ref): ...here. When handling a scale other than 1,
6001 check first whether the address is valid without the offset.
6002 Add it into the base if so, leaving the index and scale as-is.
6004 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6007 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6008 fold_for_warn before checking if arg2 is INTEGER_CST.
6010 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6012 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6013 (store_multiple_operation): Delete.
6014 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6015 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6016 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6017 guarded by TARGET_STRING.
6018 (rs6000_output_load_multiple): Delete.
6019 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6020 OPTION_MASK_STRING / TARGET_STRING handling.
6021 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6022 (const rs6000_opt_masks) <"string">: Change mask to 0.
6023 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6024 (MASK_STRING): Delete.
6025 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6027 (load_multiple): Delete.
6034 (store_multiple): Delete.
6041 (movmemsi_8reg): Delete.
6042 (corresponding unnamed define_insn): Delete.
6043 (movmemsi_6reg): Delete.
6044 (corresponding unnamed define_insn): Delete.
6045 (movmemsi_4reg): Delete.
6046 (corresponding unnamed define_insn): Delete.
6047 (movmemsi_2reg): Delete.
6048 (corresponding unnamed define_insn): Delete.
6049 (movmemsi_1reg): Delete.
6050 (corresponding unnamed define_insn): Delete.
6051 * config/rs6000/rs6000.opt (mno-string): New.
6052 (mstring): Replace by deprecation warning stub.
6053 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6055 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6057 * regrename.c (regrename_do_replace): If replacing the same
6058 reg multiple times, try to reuse last created gen_raw_REG.
6061 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6062 main to workaround a bug in GDB.
6064 2018-01-12 Tom de Vries <tom@codesourcery.com>
6067 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6069 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6071 PR rtl-optimization/80481
6072 * ira-color.c (get_cap_member): New function.
6073 (allocnos_conflict_by_live_ranges_p): Use it.
6074 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6075 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6077 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6080 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6081 (*saddl_se_1): Ditto.
6083 (*ssubl_se_1): Ditto.
6085 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6087 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6088 rather than wi::to_widest for DR_INITs.
6089 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6090 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6091 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6093 (vect_analyze_group_access_1): Note that here.
6095 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6097 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6098 polynomial type sizes.
6100 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6102 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6103 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6104 (gimple_add_tmp_var): Likewise.
6106 2018-01-12 Martin Liska <mliska@suse.cz>
6108 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6109 (gimple_alloc_sizes): Likewise.
6110 (dump_gimple_statistics): Use PRIu64 in printf format.
6111 * gimple.h: Change uint64_t to int.
6113 2018-01-12 Martin Liska <mliska@suse.cz>
6115 * tree-core.h: Use uint64_t instead of int.
6116 * tree.c (tree_node_counts): Likewise.
6117 (tree_node_sizes): Likewise.
6118 (dump_tree_statistics): Use PRIu64 in printf format.
6120 2018-01-12 Martin Liska <mliska@suse.cz>
6122 * Makefile.in: As qsort_chk is implemented in vec.c, add
6123 vec.o to linkage of gencfn-macros.
6124 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6125 passing the info to record_node_allocation_statistics.
6126 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6128 * ggc-common.c (struct ggc_usage): Add operator== and use
6129 it in operator< and compare function.
6130 * mem-stats.h (struct mem_usage): Likewise.
6131 * vec.c (struct vec_usage): Remove operator< and compare
6132 function. Can be simply inherited.
6134 2018-01-12 Martin Jambor <mjambor@suse.cz>
6137 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6138 * tree-ssa-math-opts.c: Include domwalk.h.
6139 (convert_mult_to_fma_1): New function.
6140 (fma_transformation_info): New type.
6141 (fma_deferring_state): Likewise.
6142 (cancel_fma_deferring): New function.
6143 (result_of_phi): Likewise.
6144 (last_fma_candidate_feeds_initial_phi): Likewise.
6145 (convert_mult_to_fma): Added deferring logic, split actual
6146 transformation to convert_mult_to_fma_1.
6147 (math_opts_dom_walker): New type.
6148 (math_opts_dom_walker::after_dom_children): New method, body moved
6149 here from pass_optimize_widening_mul::execute, added deferring logic
6151 (pass_optimize_widening_mul::execute): Moved most of code to
6152 math_opts_dom_walker::after_dom_children.
6153 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6154 * config/i386/i386.c (ix86_option_override_internal): Added
6155 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6157 2018-01-12 Richard Biener <rguenther@suse.de>
6160 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6161 inline instance vars.
6163 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6166 * config/rx/rx.c (rx_is_restricted_memory_address):
6169 2018-01-12 Richard Biener <rguenther@suse.de>
6171 PR tree-optimization/80846
6172 * target.def (split_reduction): New target hook.
6173 * targhooks.c (default_split_reduction): New function.
6174 * targhooks.h (default_split_reduction): Declare.
6175 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6176 target requests first reduce vectors by combining low and high
6178 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6179 (get_vectype_for_scalar_type_and_size): Export.
6180 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6181 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6182 * doc/tm.texi: Regenerate.
6183 * config/i386/i386.c (ix86_split_reduction): Implement
6184 TARGET_VECTORIZE_SPLIT_REDUCTION.
6186 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6189 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6190 in PIC mode except for TARGET_VXWORKS_RTP.
6191 * config/sparc/sparc.c: Include cfgrtl.h.
6192 (TARGET_INIT_PIC_REG): Define.
6193 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6194 (sparc_pic_register_p): New predicate.
6195 (sparc_legitimate_address_p): Use it.
6196 (sparc_legitimize_pic_address): Likewise.
6197 (sparc_delegitimize_address): Likewise.
6198 (sparc_mode_dependent_address_p): Likewise.
6199 (gen_load_pcrel_sym): Remove 4th parameter.
6200 (load_got_register): Adjust call to above. Remove obsolete stuff.
6201 (sparc_expand_prologue): Do not call load_got_register here.
6202 (sparc_flat_expand_prologue): Likewise.
6203 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6204 (sparc_use_pseudo_pic_reg): New function.
6205 (sparc_init_pic_reg): Likewise.
6206 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6207 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6209 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6211 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6212 Add item for branch_cost.
6214 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6216 PR rtl-optimization/83565
6217 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6218 not extend the result to a larger mode for rotate operations.
6219 (num_sign_bit_copies1): Likewise.
6221 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6224 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6226 Use values-Xc.o for -pedantic.
6227 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6229 2018-01-12 Martin Liska <mliska@suse.cz>
6232 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6234 (possible_polymorphic_call_targets): Use it.
6235 (ipa_devirt): Likewise.
6237 2018-01-12 Martin Liska <mliska@suse.cz>
6239 * profile-count.h (enum profile_quality): Use 0 as invalid
6240 enum value of profile_quality.
6242 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6244 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6245 -mext-string options.
6247 2018-01-12 Richard Biener <rguenther@suse.de>
6249 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6250 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6251 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6253 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6255 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6257 * configure.ac (--with-long-double-format): Add support for the
6258 configuration option to change the default long double format on
6260 * config.gcc (powerpc*-linux*-*): Likewise.
6261 * configure: Regenerate.
6262 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6263 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6264 used without modification.
6266 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6268 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6269 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6270 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6271 MISC_BUILTIN_SPEC_BARRIER.
6272 (rs6000_init_builtins): Likewise.
6273 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6275 (speculation_barrier): New define_insn.
6276 * doc/extend.texi: Document __builtin_speculation_barrier.
6278 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6281 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6282 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6283 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6285 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6286 integral modes instead of "ss" and "sd".
6287 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6288 vectors with 32-bit and 64-bit elements.
6289 (vecdupssescalarmodesuffix): New mode attribute.
6290 (vec_dup<mode>): Use it.
6292 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6295 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6296 frame if argument is passed on stack.
6298 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6301 * ree.c (combine_reaching_defs): Optimize also
6302 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6303 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6305 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6308 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6310 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6313 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6314 after they are computed.
6316 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6318 PR tree-optimization/83695
6319 * gimple-loop-linterchange.cc
6320 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6321 reset cached scev information after interchange.
6322 (pass_linterchange::execute): Remove call to scev_reset_htab.
6324 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6326 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6327 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6328 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6329 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6330 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6331 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6332 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6333 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6334 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6335 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6336 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6337 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6338 (V_lane_reg): Likewise.
6339 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6341 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6342 (vfmal_lane_low<mode>_intrinsic,
6343 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6344 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6345 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6346 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6347 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6348 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6350 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6352 * config/arm/arm-cpus.in (fp16fml): New feature.
6353 (ALL_SIMD): Add fp16fml.
6354 (armv8.2-a): Add fp16fml as an option.
6355 (armv8.3-a): Likewise.
6356 (armv8.4-a): Add fp16fml as part of fp16.
6357 * config/arm/arm.h (TARGET_FP16FML): Define.
6358 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6360 * config/arm/arm-modes.def (V2HF): Define.
6361 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6362 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6363 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6364 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6365 vfmsl_low, vfmsl_high): New set of builtins.
6366 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6367 (vfml_op): New code attribute.
6368 (VFMLHALVES): New int iterator.
6369 (VFML, VFMLSEL): New mode attributes.
6370 (V_reg): Define mapping for V2HF.
6371 (V_hi, V_lo): New mode attributes.
6372 (VF_constraint): Likewise.
6373 (vfml_half, vfml_half_selector): New int attributes.
6374 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6376 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6377 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6379 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6380 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6381 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6382 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6384 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6385 Document new effective target and option set.
6387 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6389 * config/arm/arm-cpus.in (armv8_4): New feature.
6390 (ARMv8_4a): New fgroup.
6391 (armv8.4-a): New arch.
6392 * config/arm/arm-tables.opt: Regenerate.
6393 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6394 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6395 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6396 Add matching rules for -march=armv8.4-a and extensions.
6397 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6399 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6402 * config/rx/rx.md (BW): New mode attribute.
6403 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6405 2018-01-11 Richard Biener <rguenther@suse.de>
6407 PR tree-optimization/83435
6408 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6409 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6410 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6412 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6413 Alan Hayward <alan.hayward@arm.com>
6414 David Sherwood <david.sherwood@arm.com>
6416 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6418 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6419 (aarch64_print_address_internal): Use it to check for a zero offset.
6421 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6422 Alan Hayward <alan.hayward@arm.com>
6423 David Sherwood <david.sherwood@arm.com>
6425 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6426 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6427 Return a poly_int64 rather than a HOST_WIDE_INT.
6428 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6429 rather than a HOST_WIDE_INT.
6430 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6431 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6432 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6433 final_offset from HOST_WIDE_INT to poly_int64.
6434 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6435 to_constant when getting the number of units in an Advanced SIMD
6437 (aarch64_builtin_vectorized_function): Check for a constant number
6439 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6441 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6442 attribute instead of GET_MODE_NUNITS.
6443 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6444 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6445 GET_MODE_SIZE for fixed-size registers.
6446 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6447 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6448 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6449 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6450 (aarch64_print_operand, aarch64_print_address_internal)
6451 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6452 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6453 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6454 Handle polynomial GET_MODE_SIZE.
6455 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6456 wider than SImode without modification.
6457 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6458 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6459 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6460 passing and returning SVE modes.
6461 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6462 rather than GEN_INT.
6463 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6464 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6465 (aarch64_allocate_and_probe_stack_space): Likewise.
6466 (aarch64_layout_frame): Cope with polynomial offsets.
6467 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6468 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6470 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6471 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6472 poly_int64 rather than a HOST_WIDE_INT.
6473 (aarch64_get_separate_components, aarch64_process_components)
6474 (aarch64_expand_prologue, aarch64_expand_epilogue)
6475 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6476 (aarch64_anchor_offset): New function, split out from...
6477 (aarch64_legitimize_address): ...here.
6478 (aarch64_builtin_vectorization_cost): Handle polynomial
6479 TYPE_VECTOR_SUBPARTS.
6480 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6482 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6483 number of elements from the PARALLEL rather than the mode.
6484 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6485 rather than GET_MODE_BITSIZE.
6486 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6487 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6488 (aarch64_expand_vec_perm_const_1): Handle polynomial
6489 d->perm.length () and d->perm elements.
6490 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6491 Apply to_constant to d->perm elements.
6492 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6493 polynomial CONST_VECTOR_NUNITS.
6494 (aarch64_move_pointer): Take amount as a poly_int64 rather
6496 (aarch64_progress_pointer): Avoid temporary variable.
6497 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6498 the mode attribute instead of GET_MODE.
6500 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6501 Alan Hayward <alan.hayward@arm.com>
6502 David Sherwood <david.sherwood@arm.com>
6504 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6505 x exists before using it.
6506 (aarch64_add_constant_internal): Rename to...
6507 (aarch64_add_offset_1): ...this. Replace regnum with separate
6508 src and dest rtxes. Handle the case in which they're different,
6509 including when the offset is zero. Replace scratchreg with an rtx.
6510 Use 2 additions if there is no spare register into which we can
6511 move a 16-bit constant.
6512 (aarch64_add_constant): Delete.
6513 (aarch64_add_offset): Replace reg with separate src and dest
6514 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6515 Use aarch64_add_offset_1.
6516 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6517 an rtx rather than an int. Take the delta as a poly_int64
6518 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6519 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6520 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6521 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6522 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6524 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6525 aarch64_add_constant.
6527 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6529 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6530 Use scalar_float_mode.
6532 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6534 * config/aarch64/aarch64-simd.md
6535 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6536 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6537 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6538 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6539 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6540 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6541 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6542 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6543 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6544 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6546 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6549 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6550 targ_options->x_arm_arch_string is non NULL.
6552 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6554 * config/aarch64/aarch64.h
6555 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6557 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6560 * expmed.c (emit_store_flag_force): Swap if const op0
6561 and change VOIDmode to mode of op0.
6563 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6565 PR rtl-optimization/83761
6566 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6567 than bytes to mode_for_size.
6569 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6572 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6573 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6576 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6579 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6580 when in layout mode.
6581 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6582 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6585 2018-01-10 Michael Collison <michael.collison@arm.com>
6587 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6588 * config/aarch64/aarch64-option-extension.def: Add
6589 AARCH64_OPT_EXTENSION of 'fp16fml'.
6590 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6591 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6592 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6593 * config/aarch64/constraints.md (Ui7): New constraint.
6594 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6595 (VFMLA_SEL_W): Ditto.
6598 (VFMLA16_LOW): New int iterator.
6599 (VFMLA16_HIGH): Ditto.
6600 (UNSPEC_FMLAL): New unspec.
6601 (UNSPEC_FMLSL): Ditto.
6602 (UNSPEC_FMLAL2): Ditto.
6603 (UNSPEC_FMLSL2): Ditto.
6604 (f16mac): New code attribute.
6605 * config/aarch64/aarch64-simd-builtins.def
6606 (aarch64_fmlal_lowv2sf): Ditto.
6607 (aarch64_fmlsl_lowv2sf): Ditto.
6608 (aarch64_fmlalq_lowv4sf): Ditto.
6609 (aarch64_fmlslq_lowv4sf): Ditto.
6610 (aarch64_fmlal_highv2sf): Ditto.
6611 (aarch64_fmlsl_highv2sf): Ditto.
6612 (aarch64_fmlalq_highv4sf): Ditto.
6613 (aarch64_fmlslq_highv4sf): Ditto.
6614 (aarch64_fmlal_lane_lowv2sf): Ditto.
6615 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6616 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6617 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6618 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6619 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6620 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6621 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6622 (aarch64_fmlal_lane_highv2sf): Ditto.
6623 (aarch64_fmlsl_lane_highv2sf): Ditto.
6624 (aarch64_fmlal_laneq_highv2sf): Ditto.
6625 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6626 (aarch64_fmlalq_lane_highv4sf): Ditto.
6627 (aarch64_fmlsl_lane_highv4sf): Ditto.
6628 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6629 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6630 * config/aarch64/aarch64-simd.md:
6631 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6632 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6633 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6634 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6635 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6636 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6637 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6638 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6639 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6640 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6641 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6642 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6643 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6644 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6645 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6646 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6647 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6648 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6649 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6650 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6651 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6652 (vfmlsl_low_u32): Ditto.
6653 (vfmlalq_low_u32): Ditto.
6654 (vfmlslq_low_u32): Ditto.
6655 (vfmlal_high_u32): Ditto.
6656 (vfmlsl_high_u32): Ditto.
6657 (vfmlalq_high_u32): Ditto.
6658 (vfmlslq_high_u32): Ditto.
6659 (vfmlal_lane_low_u32): Ditto.
6660 (vfmlsl_lane_low_u32): Ditto.
6661 (vfmlal_laneq_low_u32): Ditto.
6662 (vfmlsl_laneq_low_u32): Ditto.
6663 (vfmlalq_lane_low_u32): Ditto.
6664 (vfmlslq_lane_low_u32): Ditto.
6665 (vfmlalq_laneq_low_u32): Ditto.
6666 (vfmlslq_laneq_low_u32): Ditto.
6667 (vfmlal_lane_high_u32): Ditto.
6668 (vfmlsl_lane_high_u32): Ditto.
6669 (vfmlal_laneq_high_u32): Ditto.
6670 (vfmlsl_laneq_high_u32): Ditto.
6671 (vfmlalq_lane_high_u32): Ditto.
6672 (vfmlslq_lane_high_u32): Ditto.
6673 (vfmlalq_laneq_high_u32): Ditto.
6674 (vfmlslq_laneq_high_u32): Ditto.
6675 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6676 (AARCH64_FL_FOR_ARCH8_4): New.
6677 (AARCH64_ISA_F16FML): New ISA flag.
6678 (TARGET_F16FML): New feature flag for fp16fml.
6679 (doc/invoke.texi): Document new fp16fml option.
6681 2018-01-10 Michael Collison <michael.collison@arm.com>
6683 * config/aarch64/aarch64-builtins.c:
6684 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6685 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6686 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6687 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6688 (AARCH64_ISA_SHA3): New ISA flag.
6689 (TARGET_SHA3): New feature flag for sha3.
6690 * config/aarch64/iterators.md (sha512_op): New int attribute.
6691 (CRYPTO_SHA512): New int iterator.
6692 (UNSPEC_SHA512H): New unspec.
6693 (UNSPEC_SHA512H2): Ditto.
6694 (UNSPEC_SHA512SU0): Ditto.
6695 (UNSPEC_SHA512SU1): Ditto.
6696 * config/aarch64/aarch64-simd-builtins.def
6697 (aarch64_crypto_sha512hqv2di): New builtin.
6698 (aarch64_crypto_sha512h2qv2di): Ditto.
6699 (aarch64_crypto_sha512su0qv2di): Ditto.
6700 (aarch64_crypto_sha512su1qv2di): Ditto.
6701 (aarch64_eor3qv8hi): Ditto.
6702 (aarch64_rax1qv2di): Ditto.
6703 (aarch64_xarqv2di): Ditto.
6704 (aarch64_bcaxqv8hi): Ditto.
6705 * config/aarch64/aarch64-simd.md:
6706 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6707 (aarch64_crypto_sha512su0qv2di): Ditto.
6708 (aarch64_crypto_sha512su1qv2di): Ditto.
6709 (aarch64_eor3qv8hi): Ditto.
6710 (aarch64_rax1qv2di): Ditto.
6711 (aarch64_xarqv2di): Ditto.
6712 (aarch64_bcaxqv8hi): Ditto.
6713 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6714 (vsha512h2q_u64): Ditto.
6715 (vsha512su0q_u64): Ditto.
6716 (vsha512su1q_u64): Ditto.
6717 (veor3q_u16): Ditto.
6718 (vrax1q_u64): Ditto.
6720 (vbcaxq_u16): Ditto.
6721 * config/arm/types.md (crypto_sha512): New type attribute.
6722 (crypto_sha3): Ditto.
6723 (doc/invoke.texi): Document new sha3 option.
6725 2018-01-10 Michael Collison <michael.collison@arm.com>
6727 * config/aarch64/aarch64-builtins.c:
6728 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6729 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6730 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6731 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6732 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6733 (AARCH64_ISA_SM4): New ISA flag.
6734 (TARGET_SM4): New feature flag for sm4.
6735 * config/aarch64/aarch64-simd-builtins.def
6736 (aarch64_sm3ss1qv4si): Ditto.
6737 (aarch64_sm3tt1aq4si): Ditto.
6738 (aarch64_sm3tt1bq4si): Ditto.
6739 (aarch64_sm3tt2aq4si): Ditto.
6740 (aarch64_sm3tt2bq4si): Ditto.
6741 (aarch64_sm3partw1qv4si): Ditto.
6742 (aarch64_sm3partw2qv4si): Ditto.
6743 (aarch64_sm4eqv4si): Ditto.
6744 (aarch64_sm4ekeyqv4si): Ditto.
6745 * config/aarch64/aarch64-simd.md:
6746 (aarch64_sm3ss1qv4si): Ditto.
6747 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6748 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6749 (aarch64_sm4eqv4si): Ditto.
6750 (aarch64_sm4ekeyqv4si): Ditto.
6751 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6752 (sm3part_op): Ditto.
6753 (CRYPTO_SM3TT): Ditto.
6754 (CRYPTO_SM3PART): Ditto.
6755 (UNSPEC_SM3SS1): New unspec.
6756 (UNSPEC_SM3TT1A): Ditto.
6757 (UNSPEC_SM3TT1B): Ditto.
6758 (UNSPEC_SM3TT2A): Ditto.
6759 (UNSPEC_SM3TT2B): Ditto.
6760 (UNSPEC_SM3PARTW1): Ditto.
6761 (UNSPEC_SM3PARTW2): Ditto.
6762 (UNSPEC_SM4E): Ditto.
6763 (UNSPEC_SM4EKEY): Ditto.
6764 * config/aarch64/constraints.md (Ui2): New constraint.
6765 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6766 * config/arm/types.md (crypto_sm3): New type attribute.
6767 (crypto_sm4): Ditto.
6768 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6769 (vsm3tt1aq_u32): Ditto.
6770 (vsm3tt1bq_u32): Ditto.
6771 (vsm3tt2aq_u32): Ditto.
6772 (vsm3tt2bq_u32): Ditto.
6773 (vsm3partw1q_u32): Ditto.
6774 (vsm3partw2q_u32): Ditto.
6775 (vsm4eq_u32): Ditto.
6776 (vsm4ekeyq_u32): Ditto.
6777 (doc/invoke.texi): Document new sm4 option.
6779 2018-01-10 Michael Collison <michael.collison@arm.com>
6781 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6782 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6783 (AARCH64_FL_FOR_ARCH8_4): New.
6784 (AARCH64_FL_V8_4): New flag.
6785 (doc/invoke.texi): Document new armv8.4-a option.
6787 2018-01-10 Michael Collison <michael.collison@arm.com>
6789 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6790 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6791 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6792 * config/aarch64/aarch64-option-extension.def: Add
6793 AARCH64_OPT_EXTENSION of 'sha2'.
6794 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6795 (crypto): Disable sha2 and aes if crypto disabled.
6796 (crypto): Enable aes and sha2 if enabled.
6797 (simd): Disable sha2 and aes if simd disabled.
6798 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6800 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6801 (TARGET_SHA2): New feature flag for sha2.
6802 (TARGET_AES): New feature flag for aes.
6803 * config/aarch64/aarch64-simd.md:
6804 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6805 conditional on TARGET_AES.
6806 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6807 (aarch64_crypto_sha1hsi): Make pattern conditional
6809 (aarch64_crypto_sha1hv4si): Ditto.
6810 (aarch64_be_crypto_sha1hv4si): Ditto.
6811 (aarch64_crypto_sha1su1v4si): Ditto.
6812 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6813 (aarch64_crypto_sha1su0v4si): Ditto.
6814 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6815 (aarch64_crypto_sha256su0v4si): Ditto.
6816 (aarch64_crypto_sha256su1v4si): Ditto.
6817 (doc/invoke.texi): Document new aes and sha2 options.
6819 2018-01-10 Martin Sebor <msebor@redhat.com>
6821 PR tree-optimization/83781
6822 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6825 2018-01-11 Martin Sebor <msebor@gmail.com>
6826 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6828 PR tree-optimization/83501
6829 PR tree-optimization/81703
6831 * tree-ssa-strlen.c (get_string_cst): Rename...
6832 (get_string_len): ...to this. Handle global constants.
6833 (handle_char_store): Adjust.
6835 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6836 Jim Wilson <jimw@sifive.com>
6838 * config/riscv/riscv-protos.h (riscv_output_return): New.
6839 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6840 (riscv_attribute_table, riscv_output_return),
6841 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6842 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6843 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6845 (riscv_expand_prologue): Add early return for naked function.
6846 (riscv_expand_epilogue): Likewise.
6847 (riscv_function_ok_for_sibcall): Return false for naked function.
6848 (riscv_set_current_function): New.
6849 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6850 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6851 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6852 * doc/extend.texi (RISC-V Function Attributes): New.
6854 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6856 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6857 check for 128-bit long double before checking TCmode.
6858 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6859 128-bit long doubles before checking TFmode or TCmode.
6860 (FLOAT128_IBM_P): Likewise.
6862 2018-01-10 Martin Sebor <msebor@redhat.com>
6864 PR tree-optimization/83671
6865 * builtins.c (c_strlen): Unconditionally return zero for the empty
6867 Use -Warray-bounds for warnings.
6868 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6869 for non-constant array indices with COMPONENT_REF, arrays of
6870 arrays, and pointers to arrays.
6871 (gimple_fold_builtin_strlen): Determine and set length range for
6872 non-constant character arrays.
6874 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6877 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6880 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6882 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6884 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6887 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6888 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6889 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6890 indexed_or_indirect_operand predicate.
6891 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6892 (*vsx_le_perm_load_v8hi): Likewise.
6893 (*vsx_le_perm_load_v16qi): Likewise.
6894 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6895 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6896 (*vsx_le_perm_store_v8hi): Likewise.
6897 (*vsx_le_perm_store_v16qi): Likewise.
6898 (eight unnamed splitters): Likewise.
6900 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6902 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6903 * config/rs6000/emmintrin.h: Likewise.
6904 * config/rs6000/mmintrin.h: Likewise.
6905 * config/rs6000/xmmintrin.h: Likewise.
6907 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6910 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6912 * tree.c (tree_nop_conversion): Return true for location wrapper
6914 (maybe_wrap_with_location): New function.
6915 (selftest::check_strip_nops): New function.
6916 (selftest::test_location_wrappers): New function.
6917 (selftest::tree_c_tests): Call it.
6918 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6919 (maybe_wrap_with_location): New decl.
6920 (EXPR_LOCATION_WRAPPER_P): New macro.
6921 (location_wrapper_p): New inline function.
6922 (tree_strip_any_location_wrapper): New inline function.
6924 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6927 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6928 stack_realign_offset for the largest alignment of stack slot
6930 (ix86_find_max_used_stack_alignment): New function.
6931 (ix86_finalize_stack_frame_flags): Use it. Set
6932 max_used_stack_alignment if we don't realign stack.
6933 * config/i386/i386.h (machine_function): Add
6934 max_used_stack_alignment.
6936 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6938 * config/arm/arm.opt (-mbranch-cost): New option.
6939 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6942 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6945 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6946 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6948 2018-01-10 Richard Biener <rguenther@suse.de>
6951 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6952 early out so it also covers the case where we have a non-NULL
6955 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6957 PR tree-optimization/83753
6958 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6959 for non-strided grouped accesses if the number of elements is 1.
6961 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6964 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6965 * i386.h (TARGET_USE_GATHER): Define.
6966 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6968 2018-01-10 Martin Liska <mliska@suse.cz>
6971 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6972 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6974 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6975 CLEANUP_NO_PARTITIONING is not set.
6977 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6979 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6980 for vectors, as a partial revert of r254296.
6981 * rtl.h (const_vec_p): Delete.
6982 (const_vec_duplicate_p): Don't test for vector CONSTs.
6983 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6984 * expmed.c (make_tree): Likewise.
6987 * common.md (E, F): Use CONSTANT_P instead of checking for
6989 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6990 checking for CONST_VECTOR.
6992 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6995 * predict.c (force_edge_cold): Handle in more sane way edges
6998 2018-01-09 Carl Love <cel@us.ibm.com>
7000 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7002 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7003 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7004 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7005 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7006 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7007 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7008 * config/rs6000/rs6000-protos.h: Add extern defition for
7009 rs6000_generate_float2_double_code.
7010 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7012 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7013 (float2_v2df): Add define_expand.
7015 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7018 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7019 op_mode in the force_to_mode call.
7021 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7023 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7024 instead of checking each element individually.
7025 (aarch64_evpc_uzp): Likewise.
7026 (aarch64_evpc_zip): Likewise.
7027 (aarch64_evpc_ext): Likewise.
7028 (aarch64_evpc_rev): Likewise.
7029 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7030 instead of checking each element individually. Return true without
7032 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7033 whether all selected elements come from the same input, instead of
7034 checking each element individually. Remove calls to gen_rtx_REG,
7035 start_sequence and end_sequence and instead assert that no rtl is
7038 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7040 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7041 order of HIGH and CONST checks.
7043 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7045 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7046 if the destination isn't an SSA_NAME.
7048 2018-01-09 Richard Biener <rguenther@suse.de>
7050 PR tree-optimization/83668
7051 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7053 (canonicalize_loop_form): ... here, renamed from ...
7054 (canonicalize_loop_closed_ssa_form): ... this and amended to
7055 swap successor edges for loop exit blocks to make us use
7056 the RPO order we need for initial schedule generation.
7058 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7060 PR tree-optimization/64811
7061 * match.pd: When optimizing comparisons with Inf, avoid
7062 introducing or losing exceptions from comparisons with NaN.
7064 2018-01-09 Martin Liska <mliska@suse.cz>
7067 * asan.c (shadow_mem_size): Add gcc_assert.
7069 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7071 Don't save registers in main().
7074 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7075 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7076 * config/avr/avr.c (avr_set_current_function): Don't error if
7077 naked, OS_task or OS_main are specified at the same time.
7078 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7080 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7082 * common/config/avr/avr-common.c (avr_option_optimization_table):
7083 Switch on -mmain-is-OS_task for optimizing compilations.
7085 2018-01-09 Richard Biener <rguenther@suse.de>
7087 PR tree-optimization/83572
7088 * graphite.c: Include cfganal.h.
7089 (graphite_transform_loops): Connect infinite loops to exit
7090 and remove fake edges at the end.
7092 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7094 * ipa-inline.c (edge_badness): Revert accidental checkin.
7096 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7099 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7100 symbols; not inline clones.
7102 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7105 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7106 hard registers. Formatting fixes.
7108 PR preprocessor/83722
7109 * gcc.c (try_generate_repro): Pass
7110 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7111 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7114 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7115 Kito Cheng <kito.cheng@gmail.com>
7117 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7118 (riscv_leaf_function_p): Delete.
7119 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7121 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7123 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7125 (do_ifelse): New function.
7126 (do_isel): New function.
7127 (do_sub3): New function.
7128 (do_add3): New function.
7129 (do_load_mask_compare): New function.
7130 (do_overlap_load_compare): New function.
7131 (expand_compare_loop): New function.
7132 (expand_block_compare): Call expand_compare_loop() when appropriate.
7133 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7135 (-mblock-compare-inline-loop-limit): New option.
7137 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7140 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7141 Reverse order of second and third operands in first alternative.
7142 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7143 of first and second elements in UNSPEC_VPERMR vector.
7144 (altivec_expand_vec_perm_le): Likewise.
7146 2018-01-08 Jeff Law <law@redhat.com>
7148 PR rtl-optimizatin/81308
7149 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7150 (process_switch): If group_case_labels makes a change, then set
7152 (pass_convert_switch::execute): If a switch is converted, then
7153 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7155 PR rtl-optimization/81308
7156 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7159 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7161 PR target/83663 - Revert r255946
7162 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7163 generation for cases where splatting a value is not useful.
7164 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7165 across a vec_duplicate and a paradoxical subreg forming a vector
7166 mode to a vec_concat.
7168 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7170 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7171 -march=armv8.3-a variants.
7172 * config/arm/t-multilib: Likewise.
7173 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7175 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7177 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7179 (cceq_ior_compare_complement): Give it a name so I can use it, and
7180 change boolean_or_operator predicate to boolean_operator so it can
7181 be used to generate a crand.
7182 (eqne): New code iterator.
7183 (bd/bd_neg): New code_attrs.
7184 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7185 a single define_insn.
7186 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7187 decrement (bdnzt/bdnzf/bdzt/bdzf).
7188 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7189 with the new names of the branch decrement patterns, and added the
7190 names of the branch decrement conditional patterns.
7192 2018-01-08 Richard Biener <rguenther@suse.de>
7194 PR tree-optimization/83563
7195 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7198 2018-01-08 Richard Biener <rguenther@suse.de>
7201 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7203 2018-01-08 Richard Biener <rguenther@suse.de>
7205 PR tree-optimization/83685
7206 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7207 references to abnormals.
7209 2018-01-08 Richard Biener <rguenther@suse.de>
7212 * dwarf2out.c (output_indirect_strings): Handle empty
7213 skeleton_debug_str_hash.
7214 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7216 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7218 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7219 (emit_store_direct): Likewise.
7220 (arc_trampoline_adjust_address): Likewise.
7221 (arc_asm_trampoline_template): New function.
7222 (arc_initialize_trampoline): Use asm_trampoline_template.
7223 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7224 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7225 * config/arc/arc.md (flush_icache): Delete pattern.
7227 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7229 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7230 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7233 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7236 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7237 by not USED_FOR_TARGET.
7238 (make_pass_resolve_sw_modes): Likewise.
7240 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7242 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7245 2018-01-08 Richard Biener <rguenther@suse.de>
7248 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7250 2018-01-08 Richard Biener <rguenther@suse.de>
7253 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7255 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7258 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7259 basic blocks with a small number of successors.
7260 (convert_control_dep_chain_into_preds): Improve handling of
7262 (dump_predicates): Split apart into...
7263 (dump_pred_chain): ...here...
7264 (dump_pred_info): ...and here.
7265 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7266 (can_chain_union_be_invalidated_p): Improve check for invalidation
7268 (uninit_uses_cannot_happen): Avoid unnecessary if
7269 convert_control_dep_chain_into_preds yielded nothing.
7271 2018-01-06 Martin Sebor <msebor@redhat.com>
7273 PR tree-optimization/83640
7274 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7275 subtracting negative offset from size.
7276 (builtin_access::overlap): Adjust offset bounds of the access to fall
7277 within the size of the object if possible.
7279 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7281 PR rtl-optimization/83699
7282 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7283 extract_bit_field_as_subreg to cases in which the extracted
7284 value is also a vector.
7286 * lra-constraints.c (process_alt_operands): Test for the equivalence
7287 substitutions when detecting a possible reload cycle.
7289 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7292 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7293 by default if flag_selective_schedling{,2}. Formatting fixes.
7295 PR rtl-optimization/83682
7296 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7297 if it has non-VECTOR_MODE element mode.
7298 (vec_duplicate_p): Likewise.
7301 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7302 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7304 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7307 * config/i386/i386-builtin.def
7308 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7309 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7310 Require also OPTION_MASK_ISA_AVX512F in addition to
7311 OPTION_MASK_ISA_GFNI.
7312 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7313 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7314 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7315 to OPTION_MASK_ISA_GFNI.
7316 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7317 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7318 OPTION_MASK_ISA_AVX512BW.
7319 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7320 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7321 addition to OPTION_MASK_ISA_GFNI.
7322 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7323 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7324 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7325 to OPTION_MASK_ISA_GFNI.
7326 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7327 a requirement for all ISAs rather than any of them with a few
7329 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7331 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7332 bitmasks to be enabled with 3 exceptions, instead of requiring any
7333 enabled ISA with lots of exceptions.
7334 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7335 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7336 Change avx512bw in isa attribute to avx512f.
7337 * config/i386/sgxintrin.h: Add license boilerplate.
7338 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7339 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7340 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7341 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7343 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7344 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7345 temporarily sse2 rather than sse if not enabled already.
7348 * config/i386/sse.md (VI248_VLBW): Rename to ...
7349 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7350 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7351 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7352 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7353 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7354 mode iterator instead of VI248_VLBW.
7356 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7358 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7359 (record_modified): Skip clobbers; add debug output.
7360 (param_change_prob): Use sreal frequencies.
7362 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7364 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7365 punt for user-aligned variables.
7367 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7369 * tree-chrec.c (chrec_contains_symbols): Return true for
7372 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7375 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7376 of (x|y) == x for BICS pattern.
7378 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7380 PR tree-optimization/83605
7381 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7382 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7385 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7387 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7388 * config/epiphany/rtems.h: New file.
7390 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7391 Uros Bizjak <ubizjak@gmail.com>
7394 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7395 QIreg_operand instead of register_operand predicate.
7396 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7397 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7398 comments instead of -fmitigate[-_]rop.
7400 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7403 * cgraphunit.c (symbol_table::compile): Switch to text_section
7404 before calling assembly_start debug hook.
7405 * run-rtl-passes.c (run_rtl_passes): Likewise.
7408 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7410 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7411 range_int_cst_p rather than !symbolic_range_p before calling
7412 extract_range_from_multiplicative_op_1.
7414 2018-01-04 Jeff Law <law@redhat.com>
7416 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7417 redundant test in assertion.
7419 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7421 * doc/rtl.texi: Document machine_mode wrapper classes.
7423 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7425 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7428 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7430 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7431 the VEC_PERM_EXPR fold to fail.
7433 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7436 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7437 to switched_sections.
7439 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7442 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7445 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7448 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7449 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7451 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7454 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7455 is BLKmode and bitpos not zero or mode change is needed.
7457 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7460 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7463 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7466 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7467 instead of MULT rtx. Update all corresponding splitters.
7469 (*ssub<modesuffix>): Ditto.
7471 (*cmp_sadd_di): Update split patterns.
7472 (*cmp_sadd_si): Ditto.
7473 (*cmp_sadd_sidi): Ditto.
7474 (*cmp_ssub_di): Ditto.
7475 (*cmp_ssub_si): Ditto.
7476 (*cmp_ssub_sidi): Ditto.
7477 * config/alpha/predicates.md (const23_operand): New predicate.
7478 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7479 Look for ASHIFT, not MULT inner operand.
7480 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7482 2018-01-04 Martin Liska <mliska@suse.cz>
7484 PR gcov-profile/83669
7485 * gcov.c (output_intermediate_file): Add version to intermediate
7487 * doc/gcov.texi: Document new field 'version' in intermediate
7488 file format. Fix location of '-k' option of gcov command.
7490 2018-01-04 Martin Liska <mliska@suse.cz>
7493 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7495 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7497 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7499 2018-01-03 Martin Sebor <msebor@redhat.com>
7501 PR tree-optimization/83655
7502 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7503 checking calls with invalid arguments.
7505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7507 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7508 (vectorizable_mask_load_store): Delete.
7509 (vectorizable_call): Return false for masked loads and stores.
7510 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7511 instead of gimple_assign_rhs1.
7512 (vectorizable_load): Handle IFN_MASK_LOAD.
7513 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7517 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7519 (vectorizable_mask_load_store): ...here.
7520 (vectorizable_load): ...and here.
7522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7524 * tree-vect-stmts.c (vect_build_all_ones_mask)
7525 (vect_build_zero_merge_argument): New functions, split out from...
7526 (vectorizable_load): ...here.
7528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7530 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7532 (vectorizable_mask_load_store): ...here.
7533 (vectorizable_store): ...and here.
7535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7537 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7539 (vectorizable_mask_load_store): ...here.
7541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7543 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7544 (vect_model_store_cost): Take a vec_load_store_type instead of a
7546 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7547 (vect_model_store_cost): Take a vec_load_store_type instead of a
7549 (vectorizable_mask_load_store): Update accordingly.
7550 (vectorizable_store): Likewise.
7551 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7555 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7556 IFN_MASK_LOAD calls here rather than...
7557 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7560 Alan Hayward <alan.hayward@arm.com>
7561 David Sherwood <david.sherwood@arm.com>
7563 * expmed.c (extract_bit_field_1): For vector extracts,
7564 fall back to extract_bit_field_as_subreg if vec_extract
7567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7572 they are variable or constant sized.
7573 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7574 slots for constant-sized data.
7576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7577 Alan Hayward <alan.hayward@arm.com>
7578 David Sherwood <david.sherwood@arm.com>
7580 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7581 handling COND_EXPRs with boolean comparisons, try to find a better
7582 basis for the mask type than the boolean itself.
7584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7586 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7587 is calculated and how it can be overridden.
7588 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7589 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7591 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7595 Alan Hayward <alan.hayward@arm.com>
7596 David Sherwood <david.sherwood@arm.com>
7598 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7599 Remove the mode argument.
7600 (aarch64_simd_valid_immediate): Remove the mode and inverse
7602 * config/aarch64/iterators.md (bitsize): New iterator.
7603 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7604 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7605 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7606 aarch64_simd_valid_immediate.
7607 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7608 (aarch64_reg_or_bic_imm): Likewise.
7609 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7610 with an insn_type enum and msl with a modifier_type enum.
7611 Replace element_width with a scalar_mode. Change the shift
7612 to unsigned int. Add constructors for scalar_float_mode and
7613 scalar_int_mode elements.
7614 (aarch64_vect_float_const_representable_p): Delete.
7615 (aarch64_can_const_movi_rtx_p)
7616 (aarch64_simd_scalar_immediate_valid_for_move)
7617 (aarch64_simd_make_constant): Update call to
7618 aarch64_simd_valid_immediate.
7619 (aarch64_advsimd_valid_immediate_hs): New function.
7620 (aarch64_advsimd_valid_immediate): Likewise.
7621 (aarch64_simd_valid_immediate): Remove mode and inverse
7622 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7623 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7624 and aarch64_float_const_representable_p on the result.
7625 (aarch64_output_simd_mov_immediate): Remove mode argument.
7626 Update call to aarch64_simd_valid_immediate and use of
7627 simd_immediate_info.
7628 (aarch64_output_scalar_simd_mov_immediate): Update call
7631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7632 Alan Hayward <alan.hayward@arm.com>
7633 David Sherwood <david.sherwood@arm.com>
7635 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7636 (mode_nunits): Likewise CONST_MODE_NUNITS.
7637 * machmode.def (ADJUST_NUNITS): Document.
7638 * genmodes.c (mode_data::need_nunits_adj): New field.
7639 (blank_mode): Update accordingly.
7640 (adj_nunits): New variable.
7641 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7643 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7644 listed in adj_nunits.
7645 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7646 listed in adj_nunits. Don't emit case statements for such modes.
7647 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7648 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7649 nothing if adj_nunits is nonnull.
7650 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7651 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7652 (emit_mode_fbit): Update use of print_maybe_const_decl.
7653 (emit_move_size): Likewise. Treat the array as non-const
7655 (emit_mode_adjustments): Handle adj_nunits.
7657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7659 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7660 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7661 (VECTOR_MODES): Use it.
7662 (make_vector_modes): Take the prefix as an argument.
7664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7665 Alan Hayward <alan.hayward@arm.com>
7666 David Sherwood <david.sherwood@arm.com>
7668 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7669 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7670 for MODE_VECTOR_BOOL.
7671 * machmode.def (VECTOR_BOOL_MODE): Document.
7672 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7673 (make_vector_bool_mode): New function.
7674 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7676 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7677 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7679 * stor-layout.c (int_mode_for_mode): Likewise.
7680 * tree.c (build_vector_type_for_mode): Likewise.
7681 * varasm.c (output_constant_pool_2): Likewise.
7682 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7683 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7684 for MODE_VECTOR_BOOL.
7685 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7686 of mode class checks.
7687 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7688 instead of a list of mode class checks.
7689 (expand_vector_scalar_condition): Likewise.
7690 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7693 Alan Hayward <alan.hayward@arm.com>
7694 David Sherwood <david.sherwood@arm.com>
7696 * machmode.h (mode_size): Change from unsigned short to
7698 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7699 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7700 or if measurement_type is not polynomial.
7701 (fixed_size_mode::includes_p): Check for constant-sized modes.
7702 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7703 return a poly_uint16 rather than an unsigned short.
7704 (emit_mode_size): Change the type of mode_size from unsigned short
7705 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7706 (emit_mode_adjustments): Cope with polynomial vector sizes.
7707 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7709 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7711 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7712 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7713 * caller-save.c (setup_save_areas): Likewise.
7714 (replace_reg_with_saved_mem): Likewise.
7715 * calls.c (emit_library_call_value_1): Likewise.
7716 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7717 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7718 (gen_lowpart_for_combine): Likewise.
7719 * convert.c (convert_to_integer_1): Likewise.
7720 * cse.c (equiv_constant, cse_insn): Likewise.
7721 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7722 (cselib_subst_to_values): Likewise.
7723 * dce.c (word_dce_process_block): Likewise.
7724 * df-problems.c (df_word_lr_mark_ref): Likewise.
7725 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7726 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7727 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7728 (rtl_for_decl_location): Likewise.
7729 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7730 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7731 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7732 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7733 (expand_expr_real_1): Likewise.
7734 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7735 (pad_below): Likewise.
7736 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7737 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7738 * ira.c (get_subreg_tracking_sizes): Likewise.
7739 * ira-build.c (ira_create_allocno_objects): Likewise.
7740 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7741 (ira_sort_regnos_for_alter_reg): Likewise.
7742 * ira-costs.c (record_operand_costs): Likewise.
7743 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7744 (resolve_simple_move): Likewise.
7745 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7746 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7747 (lra_constraints): Likewise.
7748 (CONST_POOL_OK_P): Reject variable-sized modes.
7749 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7750 (add_pseudo_to_slot, lra_spill): Likewise.
7751 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7752 * optabs-query.c (get_best_extraction_insn): Likewise.
7753 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7754 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7755 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7756 * recog.c (offsettable_address_addr_space_p): Likewise.
7757 * regcprop.c (maybe_mode_change): Likewise.
7758 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7759 * regrename.c (build_def_use): Likewise.
7760 * regstat.c (dump_reg_info): Likewise.
7761 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7762 (find_reloads, find_reloads_subreg_address): Likewise.
7763 * reload1.c (eliminate_regs_1): Likewise.
7764 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7765 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7766 (simplify_binary_operation_1, simplify_subreg): Likewise.
7767 * targhooks.c (default_function_arg_padding): Likewise.
7768 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7769 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7770 (verify_gimple_assign_ternary): Likewise.
7771 * tree-inline.c (estimate_move_cost): Likewise.
7772 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7773 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7774 (get_address_cost_ainc): Likewise.
7775 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7776 (vect_supportable_dr_alignment): Likewise.
7777 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7778 (vectorizable_reduction): Likewise.
7779 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7780 (vectorizable_operation, vectorizable_load): Likewise.
7781 * tree.c (build_same_sized_truth_vector_type): Likewise.
7782 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7783 * var-tracking.c (emit_note_insn_var_location): Likewise.
7784 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7785 (ADDR_VEC_ALIGN): Likewise.
7787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7788 Alan Hayward <alan.hayward@arm.com>
7789 David Sherwood <david.sherwood@arm.com>
7791 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7793 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7794 or if measurement_type is polynomial.
7795 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7796 * combine.c (make_extraction): Likewise.
7797 * dse.c (find_shift_sequence): Likewise.
7798 * dwarf2out.c (mem_loc_descriptor): Likewise.
7799 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7800 (extract_bit_field, extract_low_bits): Likewise.
7801 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7802 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7803 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7804 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7805 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7806 * reload.c (find_reloads): Likewise.
7807 * reload1.c (alter_reg): Likewise.
7808 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7809 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7810 * tree-if-conv.c (predicate_mem_writes): Likewise.
7811 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7812 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7813 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7814 * valtrack.c (dead_debug_insert_temp): Likewise.
7815 * varasm.c (mergeable_constant_section): Likewise.
7816 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7819 Alan Hayward <alan.hayward@arm.com>
7820 David Sherwood <david.sherwood@arm.com>
7822 * expr.c (expand_assignment): Cope with polynomial mode sizes
7823 when assigning to a CONCAT.
7825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7826 Alan Hayward <alan.hayward@arm.com>
7827 David Sherwood <david.sherwood@arm.com>
7829 * machmode.h (mode_precision): Change from unsigned short to
7831 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7833 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7834 or if measurement_type is not polynomial.
7835 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7836 in which the mode is already known to be a scalar_int_mode.
7837 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7838 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7840 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7841 for GET_MODE_PRECISION.
7842 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7843 for GET_MODE_PRECISION.
7844 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7846 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7847 (expand_field_assignment, make_extraction): Likewise.
7848 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7849 (get_last_value): Likewise.
7850 * convert.c (convert_to_integer_1): Likewise.
7851 * cse.c (cse_insn): Likewise.
7852 * expr.c (expand_expr_real_1): Likewise.
7853 * lra-constraints.c (simplify_operand_subreg): Likewise.
7854 * optabs-query.c (can_atomic_load_p): Likewise.
7855 * optabs.c (expand_atomic_load): Likewise.
7856 (expand_atomic_store): Likewise.
7857 * ree.c (combine_reaching_defs): Likewise.
7858 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7859 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7860 * tree.h (type_has_mode_precision_p): Likewise.
7861 * ubsan.c (instrument_si_overflow): Likewise.
7863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7864 Alan Hayward <alan.hayward@arm.com>
7865 David Sherwood <david.sherwood@arm.com>
7867 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7868 polynomial numbers of units.
7869 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7870 (valid_vector_subparts_p): New function.
7871 (build_vector_type): Remove temporary shim and take the number
7872 of units as a poly_uint64 rather than an int.
7873 (build_opaque_vector_type): Take the number of units as a
7874 poly_uint64 rather than an int.
7875 * tree.c (build_vector_from_ctor): Handle polynomial
7876 TYPE_VECTOR_SUBPARTS.
7877 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7878 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7879 (build_vector_from_val): If the number of units is variable,
7880 use build_vec_duplicate_cst for constant operands and
7881 VEC_DUPLICATE_EXPR otherwise.
7882 (make_vector_type): Remove temporary is_constant ().
7883 (build_vector_type, build_opaque_vector_type): Take the number of
7884 units as a poly_uint64 rather than an int.
7885 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7887 * cfgexpand.c (expand_debug_expr): Likewise.
7888 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7889 (store_constructor, expand_expr_real_1): Likewise.
7890 (const_scalar_mask_from_tree): Likewise.
7891 * fold-const-call.c (fold_const_reduction): Likewise.
7892 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7893 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7894 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7895 (fold_relational_const): Likewise.
7896 (native_interpret_vector): Likewise. Change the size from an
7897 int to an unsigned int.
7898 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7899 TYPE_VECTOR_SUBPARTS.
7900 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7901 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7902 duplicating a non-constant operand into a variable-length vector.
7903 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7904 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7905 * ipa-icf.c (sem_variable::equals): Likewise.
7906 * match.pd: Likewise.
7907 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7908 * print-tree.c (print_node): Likewise.
7909 * stor-layout.c (layout_type): Likewise.
7910 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7911 * tree-cfg.c (verify_gimple_comparison): Likewise.
7912 (verify_gimple_assign_binary): Likewise.
7913 (verify_gimple_assign_ternary): Likewise.
7914 (verify_gimple_assign_single): Likewise.
7915 * tree-pretty-print.c (dump_generic_node): Likewise.
7916 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7917 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7918 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7919 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7920 (vect_shift_permute_load_chain): Likewise.
7921 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7922 (expand_vector_condition, optimize_vector_constructor): Likewise.
7923 (lower_vec_perm, get_compute_type): Likewise.
7924 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7925 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7926 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7927 (vect_recog_mask_conversion_pattern): Likewise.
7928 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7929 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7930 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7931 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7932 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7933 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7934 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7935 (supportable_widening_operation): Likewise.
7936 (supportable_narrowing_operation): Likewise.
7937 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7939 * varasm.c (output_constant): Likewise.
7941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7942 Alan Hayward <alan.hayward@arm.com>
7943 David Sherwood <david.sherwood@arm.com>
7945 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7946 so that both the length == 3 and length != 3 cases set up their
7947 own permute vectors. Add comments explaining why we know the
7948 number of elements is constant.
7949 (vect_permute_load_chain): Likewise.
7951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7952 Alan Hayward <alan.hayward@arm.com>
7953 David Sherwood <david.sherwood@arm.com>
7955 * machmode.h (mode_nunits): Change from unsigned char to
7957 (ONLY_FIXED_SIZE_MODES): New macro.
7958 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7959 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7960 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7962 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7963 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7964 or if measurement_type is not polynomial.
7965 * genmodes.c (ZERO_COEFFS): New macro.
7966 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7968 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7969 Use ZERO_COEFFS when emitting initializers.
7970 * data-streamer.h (bp_pack_poly_value): New function.
7971 (bp_unpack_poly_value): Likewise.
7972 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7973 for GET_MODE_NUNITS.
7974 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7975 for GET_MODE_NUNITS.
7976 * tree.c (make_vector_type): Remove temporary shim and make
7977 the real function take the number of units as a poly_uint64
7979 (build_vector_type_for_mode): Handle polynomial nunits.
7980 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7981 * emit-rtl.c (const_vec_series_p_1): Likewise.
7982 (gen_rtx_CONST_VECTOR): Likewise.
7983 * fold-const.c (test_vec_duplicate_folding): Likewise.
7984 * genrecog.c (validate_pattern): Likewise.
7985 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7986 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7987 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7988 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7989 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7990 * rtlanal.c (subreg_get_info): Likewise.
7991 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7992 (vect_grouped_load_supported): Likewise.
7993 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7994 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7995 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7996 (simplify_const_unary_operation, simplify_binary_operation_1)
7997 (simplify_const_binary_operation, simplify_ternary_operation)
7998 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7999 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8000 instead of CONST_VECTOR_NUNITS.
8001 * varasm.c (output_constant_pool_2): Likewise.
8002 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8003 explicit-encoded elements in the XVEC for variable-length vectors.
8005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8007 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8010 Alan Hayward <alan.hayward@arm.com>
8011 David Sherwood <david.sherwood@arm.com>
8013 * coretypes.h (fixed_size_mode): Declare.
8014 (fixed_size_mode_pod): New typedef.
8015 * builtins.h (target_builtins::x_apply_args_mode)
8016 (target_builtins::x_apply_result_mode): Change type to
8017 fixed_size_mode_pod.
8018 * builtins.c (apply_args_size, apply_result_size, result_vector)
8019 (expand_builtin_apply_args_1, expand_builtin_apply)
8020 (expand_builtin_return): Update accordingly.
8022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8024 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8025 * cselib.c (cselib_hash_rtx): Likewise.
8026 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8027 CONST_VECTOR encoding.
8029 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8030 Jeff Law <law@redhat.com>
8033 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8034 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8035 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8036 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8039 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8040 explicitly probe *sp in a noreturn function if there were any callee
8041 register saves or frame pointer is needed.
8043 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8046 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8047 BLKmode for ternary, binary or unary expressions.
8050 * var-tracking.c (delete_vta_debug_insn): New inline function.
8051 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8052 insns from get_insns () to NULL instead of each bb separately.
8053 Use delete_vta_debug_insn. No longer static.
8054 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8055 delete_vta_debug_insns callers.
8056 * rtl.h (delete_vta_debug_insns): Declare.
8057 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8058 instead of variable_tracking_main.
8060 2018-01-03 Martin Sebor <msebor@redhat.com>
8062 PR tree-optimization/83603
8063 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8064 arguments past the endof the argument list in functions declared
8065 without a prototype.
8066 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8067 Avoid checking when arguments are null.
8069 2018-01-03 Martin Sebor <msebor@redhat.com>
8072 * doc/extend.texi (attribute const): Fix a typo.
8073 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8074 issuing -Wsuggest-attribute for void functions.
8076 2018-01-03 Martin Sebor <msebor@redhat.com>
8078 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8079 offset_int::from instead of wide_int::to_shwi.
8080 (maybe_diag_overlap): Remove assertion.
8081 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8082 * gimple-ssa-sprintf.c (format_directive): Same.
8083 (parse_directive): Same.
8084 (sprintf_dom_walker::compute_format_length): Same.
8085 (try_substitute_return_value): Same.
8087 2018-01-03 Jeff Law <law@redhat.com>
8090 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8091 non-constant residual for zero at runtime and avoid probing in
8092 that case. Reorganize code for trailing problem to mirror handling
8095 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8097 PR tree-optimization/83501
8098 * tree-ssa-strlen.c (get_string_cst): New.
8099 (handle_char_store): Call get_string_cst.
8101 2018-01-03 Martin Liska <mliska@suse.cz>
8103 PR tree-optimization/83593
8104 * tree-ssa-strlen.c: Include tree-cfg.h.
8105 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8106 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8107 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8109 (strlen_dom_walker::before_dom_children): Call
8110 gimple_purge_dead_eh_edges. Dump tranformation with details
8112 (strlen_dom_walker::before_dom_children): Update call by adding
8113 new argument cleanup_eh.
8114 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8116 2018-01-03 Martin Liska <mliska@suse.cz>
8119 * cif-code.def (VARIADIC_THUNK): New enum value.
8120 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8123 2018-01-03 Jan Beulich <jbeulich@suse.com>
8125 * sse.md (mov<mode>_internal): Tighten condition for when to use
8126 vmovdqu<ssescalarsize> for TI and OI modes.
8128 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8130 Update copyright years.
8132 2018-01-03 Martin Liska <mliska@suse.cz>
8135 * ipa-visibility.c (function_and_variable_visibility): Skip
8136 functions with noipa attribure.
8138 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8140 * gcc.c (process_command): Update copyright notice dates.
8141 * gcov-dump.c (print_version): Ditto.
8142 * gcov.c (print_version): Ditto.
8143 * gcov-tool.c (print_version): Ditto.
8144 * gengtype.c (create_file): Ditto.
8145 * doc/cpp.texi: Bump @copying's copyright year.
8146 * doc/cppinternals.texi: Ditto.
8147 * doc/gcc.texi: Ditto.
8148 * doc/gccint.texi: Ditto.
8149 * doc/gcov.texi: Ditto.
8150 * doc/install.texi: Ditto.
8151 * doc/invoke.texi: Ditto.
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8155 * vector-builder.h (vector_builder::m_full_nelts): Change from
8156 unsigned int to poly_uint64.
8157 (vector_builder::full_nelts): Update prototype accordingly.
8158 (vector_builder::new_vector): Likewise.
8159 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8160 (vector_builder::operator ==): Likewise.
8161 (vector_builder::finalize): Likewise.
8162 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8163 Take the number of elements as a poly_uint64 rather than an
8165 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8166 from unsigned int to poly_uint64.
8167 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8168 (vec_perm_indices::new_vector): Likewise.
8169 (vec_perm_indices::length): Likewise.
8170 (vec_perm_indices::nelts_per_input): Likewise.
8171 (vec_perm_indices::input_nelts): Likewise.
8172 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8173 number of elements per input as a poly_uint64 rather than an
8174 unsigned int. Use the original encoding for variable-length
8175 vectors, rather than clamping each individual element.
8176 For the second and subsequent elements in each pattern,
8177 clamp the step and base before clamping their sum.
8178 (vec_perm_indices::series_p): Handle polynomial element counts.
8179 (vec_perm_indices::all_in_range_p): Likewise.
8180 (vec_perm_indices_to_tree): Likewise.
8181 (vec_perm_indices_to_rtx): Likewise.
8182 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8183 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8184 (tree_vector_builder::new_binary_operation): Handle polynomial
8185 element counts. Return false if we need to know the number
8186 of elements at compile time.
8187 * fold-const.c (fold_vec_perm): Punt if the number of elements
8188 isn't known at compile time.
8190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8192 * vec-perm-indices.h (vec_perm_builder): Change element type
8193 from HOST_WIDE_INT to poly_int64.
8194 (vec_perm_indices::element_type): Update accordingly.
8195 (vec_perm_indices::clamp): Handle polynomial element_types.
8196 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8197 (vec_perm_indices::all_in_range_p): Likewise.
8198 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8200 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8201 polynomial vec_perm_indices element types.
8202 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8203 * fold-const.c (fold_vec_perm): Likewise.
8204 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8205 * tree-vect-generic.c (lower_vec_perm): Likewise.
8206 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8207 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8208 element type to HOST_WIDE_INT.
8210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8211 Alan Hayward <alan.hayward@arm.com>
8212 David Sherwood <david.sherwood@arm.com>
8214 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8215 rather than an int. Use plus_constant.
8216 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8217 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8220 Alan Hayward <alan.hayward@arm.com>
8221 David Sherwood <david.sherwood@arm.com>
8223 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8224 a HOST_WIDE_INT to a poly_int64.
8226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8227 Alan Hayward <alan.hayward@arm.com>
8228 David Sherwood <david.sherwood@arm.com>
8230 * calls.c (load_register_parameters): Cope with polynomial
8231 mode sizes. Require a constant size for BLKmode parameters
8232 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8233 forces a parameter to be padded at the lsb end in order to
8234 fill a complete number of words, require the parameter size
8235 to be ordered wrt UNITS_PER_WORD.
8237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8238 Alan Hayward <alan.hayward@arm.com>
8239 David Sherwood <david.sherwood@arm.com>
8241 * reload1.c (spill_stack_slot_width): Change element type
8242 from unsigned int to poly_uint64_pod.
8243 (alter_reg): Treat mode sizes as polynomial.
8245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8246 Alan Hayward <alan.hayward@arm.com>
8247 David Sherwood <david.sherwood@arm.com>
8249 * reload.c (complex_word_subreg_p): New function.
8250 (reload_inner_reg_of_subreg, push_reload): Use it.
8252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8253 Alan Hayward <alan.hayward@arm.com>
8254 David Sherwood <david.sherwood@arm.com>
8256 * lra-constraints.c (process_alt_operands): Reject matched
8257 operands whose sizes aren't ordered.
8258 (match_reload): Refer to this check here.
8260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8261 Alan Hayward <alan.hayward@arm.com>
8262 David Sherwood <david.sherwood@arm.com>
8264 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8265 that the mode size is in the set {1, 2, 4, 8, 16}.
8267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8268 Alan Hayward <alan.hayward@arm.com>
8269 David Sherwood <david.sherwood@arm.com>
8271 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8272 Use plus_constant instead of gen_rtx_PLUS.
8274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8275 Alan Hayward <alan.hayward@arm.com>
8276 David Sherwood <david.sherwood@arm.com>
8278 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8279 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8280 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8281 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8282 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8283 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8284 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8285 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8286 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8287 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8289 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8290 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8291 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8292 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8293 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8294 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8295 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8296 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8297 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8298 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8300 * expr.c (emit_move_resolve_push): Treat the input and result
8301 of PUSH_ROUNDING as a poly_int64.
8302 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8303 (emit_push_insn): Likewise.
8304 * lra-eliminations.c (mark_not_eliminable): Likewise.
8305 * recog.c (push_operand): Likewise.
8306 * reload1.c (elimination_effects): Likewise.
8307 * rtlanal.c (nonzero_bits1): Likewise.
8308 * calls.c (store_one_arg): Likewise. Require the padding to be
8309 known at compile time.
8311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8312 Alan Hayward <alan.hayward@arm.com>
8313 David Sherwood <david.sherwood@arm.com>
8315 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8316 Use plus_constant instead of gen_rtx_PLUS.
8318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8319 Alan Hayward <alan.hayward@arm.com>
8320 David Sherwood <david.sherwood@arm.com>
8322 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8326 Alan Hayward <alan.hayward@arm.com>
8327 David Sherwood <david.sherwood@arm.com>
8329 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8330 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8331 via stack temporaries. Treat the mode size as polynomial too.
8333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8334 Alan Hayward <alan.hayward@arm.com>
8335 David Sherwood <david.sherwood@arm.com>
8337 * expr.c (expand_expr_real_2): When handling conversions involving
8338 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8339 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8340 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8349 Alan Hayward <alan.hayward@arm.com>
8350 David Sherwood <david.sherwood@arm.com>
8352 * combine.c (can_change_dest_mode): Handle polynomial
8353 REGMODE_NATURAL_SIZE.
8354 * expmed.c (store_bit_field_1): Likewise.
8355 * expr.c (store_constructor): Likewise.
8356 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8357 and polynomial REGMODE_NATURAL_SIZE.
8358 (gen_lowpart_common): Likewise.
8359 * reginfo.c (record_subregs_of_mode): Likewise.
8360 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8367 numbers of elements.
8369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8370 Alan Hayward <alan.hayward@arm.com>
8371 David Sherwood <david.sherwood@arm.com>
8373 * match.pd: Cope with polynomial numbers of vector elements.
8375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8376 Alan Hayward <alan.hayward@arm.com>
8377 David Sherwood <david.sherwood@arm.com>
8379 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8380 in a POINTER_PLUS_EXPR.
8382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8383 Alan Hayward <alan.hayward@arm.com>
8384 David Sherwood <david.sherwood@arm.com>
8386 * omp-simd-clone.c (simd_clone_subparts): New function.
8387 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8388 (ipa_simd_modify_function_body): Likewise.
8390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8391 Alan Hayward <alan.hayward@arm.com>
8392 David Sherwood <david.sherwood@arm.com>
8394 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8395 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8396 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8397 (expand_vector_condition, vector_element): Likewise.
8398 (subparts_gt): New function.
8399 (get_compute_type): Use subparts_gt.
8400 (count_type_subparts): Delete.
8401 (expand_vector_operations_1): Use subparts_gt instead of
8402 count_type_subparts.
8404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8405 Alan Hayward <alan.hayward@arm.com>
8406 David Sherwood <david.sherwood@arm.com>
8408 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8409 (vect_compile_time_alias): ...this new function. Do the calculation
8410 on poly_ints rather than trees.
8411 (vect_prune_runtime_alias_test_list): Update call accordingly.
8413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8414 Alan Hayward <alan.hayward@arm.com>
8415 David Sherwood <david.sherwood@arm.com>
8417 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8419 (vect_schedule_slp_instance): Likewise.
8421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8422 Alan Hayward <alan.hayward@arm.com>
8423 David Sherwood <david.sherwood@arm.com>
8425 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8426 constant and extern definitions for variable-length vectors.
8427 (vect_get_constant_vectors): Note that the number of units
8428 is known to be constant.
8430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8431 Alan Hayward <alan.hayward@arm.com>
8432 David Sherwood <david.sherwood@arm.com>
8434 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8435 of units as polynomial. Choose between WIDE and NARROW based
8438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8439 Alan Hayward <alan.hayward@arm.com>
8440 David Sherwood <david.sherwood@arm.com>
8442 * tree-vect-stmts.c (simd_clone_subparts): New function.
8443 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-stmts.c (vectorizable_call): Treat the number of
8450 vectors as polynomial. Use build_index_vector for
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-stmts.c (get_load_store_type): Treat the number of
8458 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8459 for variable-length vectors.
8460 (vectorizable_mask_load_store): Treat the number of units as
8461 polynomial, asserting that it is constant if the condition has
8462 already been enforced.
8463 (vectorizable_store, vectorizable_load): Likewise.
8465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8466 Alan Hayward <alan.hayward@arm.com>
8467 David Sherwood <david.sherwood@arm.com>
8469 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8470 of units as polynomial. Punt if we can't tell at compile time
8471 which vector contains the final result.
8473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8474 Alan Hayward <alan.hayward@arm.com>
8475 David Sherwood <david.sherwood@arm.com>
8477 * tree-vect-loop.c (vectorizable_induction): Treat the number
8478 of units as polynomial. Punt on SLP inductions. Use an integer
8479 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8480 cast of such a series for variable-length floating-point
8483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8484 Alan Hayward <alan.hayward@arm.com>
8485 David Sherwood <david.sherwood@arm.com>
8487 * tree.h (build_index_vector): Declare.
8488 * tree.c (build_index_vector): New function.
8489 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8490 of units as polynomial, forcibly converting it to a constant if
8491 vectorizable_reduction has already enforced the condition.
8492 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8493 to create a {1,2,3,...} vector.
8494 (vectorizable_reduction): Treat the number of units as polynomial.
8495 Choose vectype_in based on the largest scalar element size rather
8496 than the smallest number of units. Enforce the restrictions
8499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8500 Alan Hayward <alan.hayward@arm.com>
8501 David Sherwood <david.sherwood@arm.com>
8503 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8504 number of units as polynomial.
8506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8507 Alan Hayward <alan.hayward@arm.com>
8508 David Sherwood <david.sherwood@arm.com>
8510 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8511 * target.def (autovectorize_vector_sizes): Return the vector sizes
8512 by pointer, using vector_sizes rather than a bitmask.
8513 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8514 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8515 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8517 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8518 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8519 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8520 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8521 * omp-general.c (omp_max_vf): Likewise.
8522 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8523 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8524 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8525 * tree-vect-slp.c (vect_slp_bb): Likewise.
8526 * doc/tm.texi: Regenerate.
8527 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8529 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8530 the vector size as a poly_uint64 rather than an unsigned int.
8531 (current_vector_size): Change from an unsigned int to a poly_uint64.
8532 (get_vectype_for_scalar_type): Update accordingly.
8533 * tree.h (build_truth_vector_type): Take the size and number of
8534 units as a poly_uint64 rather than an unsigned int.
8535 (build_vector_type): Add a temporary overload that takes
8536 the number of units as a poly_uint64 rather than an unsigned int.
8537 * tree.c (make_vector_type): Likewise.
8538 (build_truth_vector_type): Take the number of units as a poly_uint64
8539 rather than an unsigned int.
8541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8542 Alan Hayward <alan.hayward@arm.com>
8543 David Sherwood <david.sherwood@arm.com>
8545 * target.def (get_mask_mode): Take the number of units and length
8546 as poly_uint64s rather than unsigned ints.
8547 * targhooks.h (default_get_mask_mode): Update accordingly.
8548 * targhooks.c (default_get_mask_mode): Likewise.
8549 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8550 * doc/tm.texi: Regenerate.
8552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8553 Alan Hayward <alan.hayward@arm.com>
8554 David Sherwood <david.sherwood@arm.com>
8556 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8557 * omp-general.c (omp_max_vf): Likewise.
8558 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8559 (expand_omp_simd): Handle polynomial safelen.
8560 * omp-low.c (omplow_simd_context): Add a default constructor.
8561 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8562 (lower_rec_simd_input_clauses): Update accordingly.
8563 (lower_rec_input_clauses): Likewise.
8565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8566 Alan Hayward <alan.hayward@arm.com>
8567 David Sherwood <david.sherwood@arm.com>
8569 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8570 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8571 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8572 (vect_analyze_slp_cost): Likewise.
8573 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8574 (vect_model_load_cost): Likewise.
8576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8577 Alan Hayward <alan.hayward@arm.com>
8578 David Sherwood <david.sherwood@arm.com>
8580 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8581 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8582 from an unsigned int * to a poly_uint64_pod *.
8583 (calculate_unrolling_factor): New function.
8584 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8587 Alan Hayward <alan.hayward@arm.com>
8588 David Sherwood <david.sherwood@arm.com>
8590 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8591 from an unsigned int to a poly_uint64.
8592 (_loop_vec_info::slp_unrolling_factor): Likewise.
8593 (_loop_vec_info::vectorization_factor): Change from an int
8595 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8596 (vect_get_num_vectors): New function.
8597 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8598 (vect_get_num_copies): Use vect_get_num_vectors.
8599 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8600 to an unsigned int *.
8601 (vect_analyze_data_refs): Change min_vf from an int * to a
8603 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8604 than an unsigned HOST_WIDE_INT.
8605 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8606 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8607 to an unsigned int *.
8608 (vect_analyze_data_ref_dependences): Likewise.
8609 (vect_compute_data_ref_alignment): Handle polynomial vf.
8610 (vect_enhance_data_refs_alignment): Likewise.
8611 (vect_prune_runtime_alias_test_list): Likewise.
8612 (vect_shift_permute_load_chain): Likewise.
8613 (vect_supportable_dr_alignment): Likewise.
8614 (dependence_distance_ge_vf): Take the vectorization factor as a
8615 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8616 (vect_analyze_data_refs): Change min_vf from an int * to a
8618 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8619 vfm1 as a poly_uint64 rather than an int. Make the same change
8620 for the returned bound_scalar.
8621 (vect_gen_vector_loop_niters): Handle polynomial vf.
8622 (vect_do_peeling): Likewise. Update call to
8623 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8624 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8626 * tree-vect-loop.c (vect_determine_vectorization_factor)
8627 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8628 (vect_get_known_peeling_cost): Likewise.
8629 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8630 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8631 (vect_transform_loop): Likewise. Use the lowest possible VF when
8632 updating the upper bounds of the loop.
8633 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8635 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8636 polynomial unroll factors.
8637 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8638 (vect_make_slp_decision): Likewise.
8639 (vect_supported_load_permutation_p): Likewise, and polynomial
8641 (vect_analyze_slp_cost): Handle polynomial vf.
8642 (vect_slp_analyze_node_operations): Likewise.
8643 (vect_slp_analyze_bb_1): Likewise.
8644 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8645 than an unsigned HOST_WIDE_INT.
8646 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8647 (vectorizable_load): Handle polynomial vf.
8648 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8650 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8653 Alan Hayward <alan.hayward@arm.com>
8654 David Sherwood <david.sherwood@arm.com>
8656 * match.pd: Handle bit operations involving three constants
8657 and try to fold one pair.
8659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8661 * tree-vect-loop-manip.c: Include gimple-fold.h.
8662 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8663 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8664 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8665 Add a path that uses a step of VF instead of 1, but disable it
8667 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8668 and niters_no_overflow parameters. Update calls to
8669 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8670 Create a new SSA name if the latter choses to use a ste other
8671 than zero, and return it via niters_vector_mult_vf_var.
8672 * tree-vect-loop.c (vect_transform_loop): Update calls to
8673 vect_do_peeling, vect_gen_vector_loop_niters and
8674 slpeel_make_loop_iterate_ntimes.
8675 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8676 (vect_gen_vector_loop_niters): Update declarations after above changes.
8678 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8680 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8681 128-bit round to integer instructions.
8682 (ceil<mode>2): Likewise.
8683 (btrunc<mode>2): Likewise.
8684 (round<mode>2): Likewise.
8686 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8688 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8689 unaligned VSX load/store on P8/P9.
8690 (expand_block_clear): Allow the use of unaligned VSX
8691 load/store on P8/P9.
8693 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8695 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8697 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8698 swap associated with both a load and a store.
8700 2018-01-02 Andrew Waterman <andrew@sifive.com>
8702 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8703 * config/riscv/riscv.md (clear_cache): Use it.
8705 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8707 * web.c: Remove out-of-date comment.
8709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8711 * expr.c (fixup_args_size_notes): Check that any existing
8712 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8713 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8714 (emit_single_push_insn): ...here.
8716 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8718 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8719 (const_vector_encoded_nelts): New function.
8720 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8721 (const_vector_int_elt, const_vector_elt): Declare.
8722 * emit-rtl.c (const_vector_int_elt_1): New function.
8723 (const_vector_elt): Likewise.
8724 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8725 of CONST_VECTOR_ELT.
8727 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8729 * expr.c: Include rtx-vector-builder.h.
8730 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8731 directly on the tree encoding.
8732 (const_vector_from_tree): Likewise.
8733 * optabs.c: Include rtx-vector-builder.h.
8734 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8735 sequence of "u" values.
8736 * vec-perm-indices.c: Include rtx-vector-builder.h.
8737 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8738 directly on the vec_perm_indices encoding.
8740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8742 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8743 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8744 * rtx-vector-builder.h: New file.
8745 * rtx-vector-builder.c: Likewise.
8746 * rtl.h (rtx_def::u2): Add a const_vector field.
8747 (CONST_VECTOR_NPATTERNS): New macro.
8748 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8749 (CONST_VECTOR_DUPLICATE_P): Likewise.
8750 (CONST_VECTOR_STEPPED_P): Likewise.
8751 (CONST_VECTOR_ENCODED_ELT): Likewise.
8752 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8753 (unwrap_const_vec_duplicate): Likewise.
8754 (const_vec_series_p): Check for a non-duplicated vector encoding.
8755 Say that the function only returns true for integer vectors.
8756 * emit-rtl.c: Include rtx-vector-builder.h.
8757 (gen_const_vec_duplicate_1): Delete.
8758 (gen_const_vector): Call gen_const_vec_duplicate instead of
8759 gen_const_vec_duplicate_1.
8760 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8761 (gen_const_vec_duplicate): Use rtx_vector_builder.
8762 (gen_const_vec_series): Likewise.
8763 (gen_rtx_CONST_VECTOR): Likewise.
8764 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8765 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8766 Build a new vector rather than modifying a CONST_VECTOR in-place.
8767 (handle_special_swappables): Update call accordingly.
8768 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8769 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8770 Build a new vector rather than modifying a CONST_VECTOR in-place.
8771 (handle_special_swappables): Update call accordingly.
8773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8775 * simplify-rtx.c (simplify_const_binary_operation): Use
8776 CONST_VECTOR_ELT instead of XVECEXP.
8778 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8780 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8781 the selector elements to be different from the data elements
8782 if the selector is a VECTOR_CST.
8783 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8784 ssizetype for the selector.
8786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8788 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8789 before testing each element individually.
8790 * tree-vect-generic.c (lower_vec_perm): Likewise.
8792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8794 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8795 * selftest-run-tests.c (selftest::run_tests): Call it.
8796 * vector-builder.h (vector_builder::operator ==): New function.
8797 (vector_builder::operator !=): Likewise.
8798 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8799 (vec_perm_indices::all_from_input_p): New function.
8800 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8801 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8802 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8803 instead of reading the VECTOR_CST directly. Detect whether both
8804 vector inputs are the same before constructing the vec_perm_indices,
8805 and update the number of inputs argument accordingly. Use the
8806 utility functions added above. Only construct sel2 if we need to.
8808 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8810 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8811 the broadcast of the low byte.
8812 (expand_mult_highpart): Use an explicit encoding for the permutes.
8813 * optabs-query.c (can_mult_highpart_p): Likewise.
8814 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8815 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8816 (vectorizable_bswap): Likewise.
8817 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8818 explicit encoding for the power-of-2 permutes.
8819 (vect_permute_store_chain): Likewise.
8820 (vect_grouped_load_supported): Likewise.
8821 (vect_permute_load_chain): Likewise.
8823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8825 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8826 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8827 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8828 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8829 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8830 (vect_gen_perm_mask_any): Likewise.
8832 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8834 * int-vector-builder.h: New file.
8835 * vec-perm-indices.h: Include int-vector-builder.h.
8836 (vec_perm_indices): Redefine as an int_vector_builder.
8837 (auto_vec_perm_indices): Delete.
8838 (vec_perm_builder): Redefine as a stand-alone class.
8839 (vec_perm_indices::vec_perm_indices): New function.
8840 (vec_perm_indices::clamp): Likewise.
8841 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8842 (vec_perm_indices::new_vector): New function.
8843 (vec_perm_indices::new_expanded_vector): Update for new
8844 vec_perm_indices class.
8845 (vec_perm_indices::rotate_inputs): New function.
8846 (vec_perm_indices::all_in_range_p): Operate directly on the
8847 encoded form, without computing elided elements.
8848 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8849 encoding. Update for new vec_perm_indices class.
8850 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8851 the given vec_perm_builder.
8852 (expand_vec_perm_var): Update vec_perm_builder constructor.
8853 (expand_mult_highpart): Use vec_perm_builder instead of
8854 auto_vec_perm_indices.
8855 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8856 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8857 or double series encoding as appropriate.
8858 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8859 vec_perm_indices instead of auto_vec_perm_indices.
8860 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8861 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8862 (vect_permute_store_chain): Likewise.
8863 (vect_grouped_load_supported): Likewise.
8864 (vect_permute_load_chain): Likewise.
8865 (vect_shift_permute_load_chain): Likewise.
8866 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8867 (vect_transform_slp_perm_load): Likewise.
8868 (vect_schedule_slp_instance): Likewise.
8869 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8870 (vectorizable_mask_load_store): Likewise.
8871 (vectorizable_bswap): Likewise.
8872 (vectorizable_store): Likewise.
8873 (vectorizable_load): Likewise.
8874 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8875 vec_perm_indices instead of auto_vec_perm_indices. Use
8876 tree_to_vec_perm_builder to read the vector from a tree.
8877 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8878 vec_perm_builder instead of a vec_perm_indices.
8879 (have_whole_vector_shift): Use vec_perm_builder and
8880 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8881 truncation to calc_vec_perm_mask_for_shift.
8882 (vect_create_epilog_for_reduction): Likewise.
8883 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8884 from auto_vec_perm_indices to vec_perm_indices.
8885 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8886 instead of changing individual elements.
8887 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8888 the vector in d.perm.
8889 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8890 from auto_vec_perm_indices to vec_perm_indices.
8891 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8892 instead of changing individual elements.
8893 (arm_vectorize_vec_perm_const): Use new_vector to install
8894 the vector in d.perm.
8895 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8896 Update vec_perm_builder constructor.
8897 (rs6000_expand_interleave): Likewise.
8898 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8899 (rs6000_expand_interleave): Likewise.
8901 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8903 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8904 to qimode could truncate the indices.
8905 * optabs.c (expand_vec_perm_var): Likewise.
8907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8909 * Makefile.in (OBJS): Add vec-perm-indices.o.
8910 * vec-perm-indices.h: New file.
8911 * vec-perm-indices.c: Likewise.
8912 * target.h (vec_perm_indices): Replace with a forward class
8914 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8915 * optabs.h: Include vec-perm-indices.h.
8916 (expand_vec_perm): Delete.
8917 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8918 (expand_vec_perm_const): Declare.
8919 * target.def (vec_perm_const_ok): Replace with...
8920 (vec_perm_const): ...this new hook.
8921 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8922 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8923 * doc/tm.texi: Regenerate.
8924 * optabs.def (vec_perm_const): Delete.
8925 * doc/md.texi (vec_perm_const): Likewise.
8926 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8927 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8928 expand_vec_perm for constant permutation vectors. Assert that
8929 the mode of variable permutation vectors is the integer equivalent
8930 of the mode that is being permuted.
8931 * optabs-query.h (selector_fits_mode_p): Declare.
8932 * optabs-query.c: Include vec-perm-indices.h.
8933 (selector_fits_mode_p): New function.
8934 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8935 is defined, instead of checking whether the vec_perm_const_optab
8936 exists. Use targetm.vectorize.vec_perm_const instead of
8937 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8938 fit in the vector mode before using a variable permute.
8939 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8940 vec_perm_indices instead of an rtx.
8941 (expand_vec_perm): Replace with...
8942 (expand_vec_perm_const): ...this new function. Take the selector
8943 as a vec_perm_indices rather than an rtx. Also take the mode of
8944 the selector. Update call to shift_amt_for_vec_perm_mask.
8945 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8946 Use vec_perm_indices::new_expanded_vector to expand the original
8947 selector into bytes. Check whether the indices fit in the vector
8948 mode before using a variable permute.
8949 (expand_vec_perm_var): Make global.
8950 (expand_mult_highpart): Use expand_vec_perm_const.
8951 * fold-const.c: Includes vec-perm-indices.h.
8952 * tree-ssa-forwprop.c: Likewise.
8953 * tree-vect-data-refs.c: Likewise.
8954 * tree-vect-generic.c: Likewise.
8955 * tree-vect-loop.c: Likewise.
8956 * tree-vect-slp.c: Likewise.
8957 * tree-vect-stmts.c: Likewise.
8958 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8960 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8961 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8962 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8963 (aarch64_vectorize_vec_perm_const): ...this new function.
8964 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8965 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8966 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8967 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8968 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8969 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8970 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8972 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8973 check for NEON modes.
8974 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8975 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8976 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8977 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8979 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8980 the old VEC_PERM_CONST conditions.
8981 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8982 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8983 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8984 (ia64_vectorize_vec_perm_const_ok): Merge into...
8985 (ia64_vectorize_vec_perm_const): ...this new function.
8986 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8987 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8988 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8989 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8990 * config/mips/mips.c (mips_expand_vec_perm_const)
8991 (mips_vectorize_vec_perm_const_ok): Merge into...
8992 (mips_vectorize_vec_perm_const): ...this new function.
8993 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8994 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8995 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8996 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8997 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8998 (rs6000_expand_vec_perm_const): Delete.
8999 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9001 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9002 (altivec_expand_vec_perm_const_le): Take each operand individually.
9003 Operate on constant selectors rather than rtxes.
9004 (altivec_expand_vec_perm_const): Likewise. Update call to
9005 altivec_expand_vec_perm_const_le.
9006 (rs6000_expand_vec_perm_const): Delete.
9007 (rs6000_vectorize_vec_perm_const_ok): Delete.
9008 (rs6000_vectorize_vec_perm_const): New function.
9009 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9010 an element count and rtx array.
9011 (rs6000_expand_extract_even): Update call accordingly.
9012 (rs6000_expand_interleave): Likewise.
9013 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9014 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9015 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9016 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9017 (rs6000_expand_vec_perm_const): Delete.
9018 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9019 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9020 (altivec_expand_vec_perm_const_le): Take each operand individually.
9021 Operate on constant selectors rather than rtxes.
9022 (altivec_expand_vec_perm_const): Likewise. Update call to
9023 altivec_expand_vec_perm_const_le.
9024 (rs6000_expand_vec_perm_const): Delete.
9025 (rs6000_vectorize_vec_perm_const_ok): Delete.
9026 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9027 reference to the SPE evmerge intructions.
9028 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9029 an element count and rtx array.
9030 (rs6000_expand_extract_even): Update call accordingly.
9031 (rs6000_expand_interleave): Likewise.
9032 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9033 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9035 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9039 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9040 vector mode and that that mode matches the mode of the data
9042 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9043 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9044 directly using expand_vec_perm_1 when forcing selectors into
9046 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9050 * optabs-query.h (can_vec_perm_p): Delete.
9051 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9052 * optabs-query.c (can_vec_perm_p): Split into...
9053 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9054 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9055 particular selector is valid.
9056 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9057 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9058 (vect_grouped_load_supported): Likewise.
9059 (vect_shift_permute_load_chain): Likewise.
9060 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9061 (vect_transform_slp_perm_load): Likewise.
9062 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9063 (vectorizable_bswap): Likewise.
9064 (vect_gen_perm_mask_checked): Likewise.
9065 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9066 implementations of variable permutation vectors into account
9067 when deciding which selector to use.
9068 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9069 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9070 with a false third argument.
9071 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9072 to test whether the constant selector is valid and can_vec_perm_var_p
9073 to test whether a variable selector is valid.
9075 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9077 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9078 * optabs-query.c (can_vec_perm_p): Likewise.
9079 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9080 instead of vec_perm_indices.
9081 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9082 (vect_gen_perm_mask_checked): Likewise,
9083 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9084 (vect_gen_perm_mask_checked): Likewise,
9086 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9088 * optabs-query.h (qimode_for_vec_perm): Declare.
9089 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9090 (qimode_for_vec_perm): ...this new function.
9091 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9093 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9095 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9096 does not have a conditional at the top.
9098 2018-01-02 Richard Biener <rguenther@suse.de>
9100 * ipa-inline.c (big_speedup_p): Fix expression.
9102 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9105 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9108 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9112 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9113 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9114 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9115 cond_taken_branch_cost 3->4.
9117 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9119 PR tree-optimization/83581
9120 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9121 TODO_cleanup_cfg if any changes have been made.
9124 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9125 convert_modes if target mode has the right side, but different mode
9129 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9130 last argument when extracting from CONCAT. If either from_real or
9131 from_imag is NULL, use expansion through memory. If result is not
9132 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9133 the parts directly to inner mode, if even that fails, use expansion
9137 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9138 check for bswap in mode rather than HImode and use that in expand_unop
9141 Copyright (C) 2018 Free Software Foundation, Inc.
9143 Copying and distribution of this file, with or without modification,
9144 are permitted in any medium without royalty provided the copyright
9145 notice and this notice are preserved.