1 2018-03-19 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/84859
4 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5 (cond_if_else_store_replacement): Perform sinking operation on
6 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7 Generalize what a BB with a single eligible store is.
9 2018-03-19 Richard Biener <rguenther@suse.de>
11 PR tree-optimization/84929
12 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
13 chrec_is_positive against non-chrec arg.
15 2018-03-19 Tamar Christina <tamar.christina@arm.com>
18 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
20 2018-03-18 Martin Liska <mliska@suse.cz>
22 PR rtl-optimization/84635
23 * regrename.c (build_def_use): Use matches_mode only when
26 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
28 PR tree-optimization/84913
29 * tree-vect-loop.c (vectorizable_reduction): Don't try to
30 vectorize chains of COND_EXPRs.
32 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
34 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
36 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
38 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
40 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
42 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
44 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
45 Kito Cheng <kito.cheng@gmail.com>
47 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
48 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
49 (nds32_adjust_reg_alloc_order): New function.
50 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
52 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
54 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
55 nds32_print_operand, nds32_print_operand_address): Use
56 HOST_WIDE_INT_PRINT_DEC instead.
58 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
60 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
62 2018-03-17 Jakub Jelinek <jakub@redhat.com>
65 * config/i386/i386.c (initial_ix86_tune_features,
66 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
68 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
69 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
70 rather than 1u << ix86_tune. Formatting fix.
71 (ix86_option_override_internal): Change ix86_arch_mask from
72 unsigned int to unsigned HOST_WIDE_INT, initialize to
73 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
74 (ix86_function_specific_restore): Likewise.
76 2018-03-16 Jakub Jelinek <jakub@redhat.com>
79 * postreload.c (reload_combine_recognize_pattern): Perform
80 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
81 truncate_int_for_mode the result for the destination's mode.
84 * hsa-gen.c (mem_type_for_type): Fix comment typo.
85 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
87 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
90 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
93 * lra-assigns.c (lra_split_hard_reg_for): Don't use
94 regno_allocno_class_array and sorted_pseudos.
95 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
96 insns where regno is used.
98 2018-03-16 Martin Liska <mliska@suse.cz>
101 * multiple_target.c (create_dispatcher_calls): Redirect
102 reference in the symbol table.
104 2018-03-16 Martin Liska <mliska@suse.cz>
107 * multiple_target.c (create_dispatcher_calls): Redirect also
110 2018-03-16 Jakub Jelinek <jakub@redhat.com>
114 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
115 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
116 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
118 2018-03-16 Julia Koval <julia.koval@intel.com>
120 * doc/invoke.texi (Skylake Server): Add CLWB.
121 Cannonlake): Remove CLWB.
123 2018-03-16 Jakub Jelinek <jakub@redhat.com>
125 PR tree-optimization/84841
126 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
128 (FLOAT_ONE_CONST_TYPE): Define.
129 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
130 (sort_by_operand_rank): Put entries with higher constant_type last
131 rather than first to match comments.
133 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
135 * config/nios2/nios2.md (movsi_internal): Fix thinko in
138 2018-03-15 Jakub Jelinek <jakub@redhat.com>
141 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
142 check and use address of target always.
144 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
147 * config/i386/i386.c (indirect_thunk_needed): Update comments.
148 (indirect_thunk_bnd_needed): Likewise.
149 (indirect_thunks_used): Likewise.
150 (indirect_thunks_bnd_used): Likewise.
151 (indirect_return_needed): New.
152 (indirect_return_bnd_needed): Likewise.
153 (output_indirect_thunk_function): Add a bool argument for
155 (output_indirect_thunk_function): Don't generate alias for
156 function return thunk.
157 (ix86_code_end): Call output_indirect_thunk_function to generate
158 function return thunks.
159 (ix86_output_function_return): Set indirect_return_bnd_needed
160 and indirect_return_needed instead of indirect_thunk_bnd_needed
161 and indirect_thunk_needed.
163 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
165 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
166 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
167 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
169 2018-03-15 David Malcolm <dmalcolm@redhat.com>
170 Paul Hua <paul.hua.gm@gmail.com>
173 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
175 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
177 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
178 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
181 2018-03-15 Tamar Christina <tamar.christina@arm.com>
184 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
185 instead of GET_MODE_SIZE when comparing Units.
187 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
190 * varasm.c (hash_section): Return an unchangeble hash value
191 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
192 Return !aarch64_can_use_per_function_literal_pools_p ().
194 2018-03-15 Jakub Jelinek <jakub@redhat.com>
197 * optabs.c (emit_conditional_move): Pass address of cmode's copy
198 rather than address of cmode as last argument to prepare_cmp_insn.
200 2018-03-15 Julia Koval <julia.koval@intel.com>
202 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
203 F_AVX512VNNI, F_AVX512BITALG): New.
205 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
208 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
209 insn for floating-point loads and stores.
211 2018-03-14 Carl Love <cel@us.ibm.com>
213 * config/rs6000/rs6000-c.c: Add macro definitions for
214 ALTIVEC_BUILTIN_VEC_PERMXOR.
215 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
216 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
217 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
218 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
220 * config/doc/extend.texi: Add prototypes for vec_permxor.
222 2018-03-14 David Malcolm <dmalcolm@redhat.com>
225 * diagnostic-show-locus.c (class layout_point): Convert m_line
226 from int to linenum_type.
227 (line_span::comparator): Use linenum "compare" function when
228 comparing line numbers.
229 (test_line_span): New function.
230 (layout_range::contains_point): Convert param "row" from int to
232 (layout_range::intersects_line_p): Likewise.
233 (layout::will_show_line_p): Likewise.
234 (layout::print_source_line): Likewise.
235 (layout::should_print_annotation_line_p): Likewise.
236 (layout::print_annotation_line): Likewise.
237 (layout::print_leading_fixits): Likewise.
238 (layout::annotation_line_showed_range_p): Likewise.
239 (struct line_corrections): Likewise for field m_row.
240 (line_corrections::line_corrections): Likewise for param "row".
241 (layout::print_trailing_fixits): Likewise.
242 (layout::get_state_at_point): Likewise.
243 (layout::get_x_bound_for_row): Likewise.
244 (layout::print_line): Likewise.
245 (diagnostic_show_locus): Likewise for locals "last_line" and
247 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
248 * input.c (selftest::test_linenum_comparisons): New function.
249 (selftest::input_c_tests): Call it.
250 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
251 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
252 * selftest.h (ASSERT_GT): New macro.
253 (ASSERT_GT_AT): New macro.
254 (ASSERT_LT): New macro.
255 (ASSERT_LT_AT): New macro.
257 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
259 PR rtl-optimization/84780
260 * combine.c (distribute_links): Don't make a link based on pc_rtx.
262 2018-03-14 Martin Liska <mliska@suse.cz>
264 * tree.c (record_node_allocation_statistics): Use
266 (get_stats_node_kind): New function extracted from
267 record_node_allocation_statistics.
268 (free_node): Use get_stats_node_kind.
270 2018-03-14 Richard Biener <rguenther@suse.de>
272 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
273 that the value-set of ANTIC_IN doesn't grow.
276 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
278 (BB_VISITED_WITH_VISITED_SUCCS): New define.
279 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
281 2018-03-14 Julia Koval <julia.koval@intel.com>
283 * config.gcc (icelake-client, icelake-server): New.
285 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
286 (initial_ix86_arch_features): Ditto.
287 (PTA_SKYLAKE): Add SGX.
288 (PTA_ICELAKE): Remove.
289 (PTA_ICELAKE_CLIENT): New.
290 (PTA_ICELAKE_SERVER): New.
291 (ix86_option_override_internal): Split up icelake on icelake client and
293 (get_builtin_code_for_version): Ditto.
294 (fold_builtin_cpu): Ditto.
295 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
296 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
297 * config/i386/i386.h (processor_type): Ditto.
298 * doc/invoke.texi: Ditto.
300 2018-03-14 Jakub Jelinek <jakub@redhat.com>
303 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
304 INTEGER_CST offset, add it together with bitpos / 8 and
305 sign extend based on POINTER_SIZE.
309 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
312 * config/i386/constraints.md (Yc): New register constraint.
313 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
314 Use Yc constraint for alternative 2 of operand 0. Remove
315 preferred_for_speed attribute.
317 2018-03-14 Richard Biener <rguenther@suse.de>
319 PR tree-optimization/84830
320 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
321 with the old one to avoid oscillations.
323 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
326 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
328 (assign_by_spills): Return a flag of reload assignment failure.
329 Do not process the reload assignment failures. Do not spill other
330 reload pseudos if they has the same reg class. Update n if
332 (lra_assign): Add a return arg. Set up from the result of
333 assign_by_spills call.
334 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
335 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
336 usage_insns if it is not NULL.
337 (spill_hard_reg_in_range): New function.
338 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
339 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
341 (lra_assign): Change prototype.
342 * lra.c (lra): Add code to deal with fails by splitting hard reg
345 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
347 * config/riscv/riscv.opt (mrelax): New option.
348 * config/riscv/riscv.c (riscv_file_start): Emit ".option
349 "norelax" when riscv_mrelax is disabled.
350 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
352 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
355 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
356 reassociation for int modes.
358 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
360 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
361 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
363 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
364 * config/aarch64/aarch64-sve.md
365 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
366 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
367 (*extend<mode><Vwide>2): Rename to...
368 (aarch64_sve_extend<mode><Vwide>2): ...this.
369 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
370 renaming the old pattern to...
371 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
373 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
374 define_expand, renaming the old pattern to...
375 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
376 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
377 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
378 account when deciding which SVE instruction the optab should use.
379 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
381 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
383 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
384 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
385 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
386 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
387 (tlsdesc_small_<mode>): Turn a define_expand and use
388 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
389 (tlsdesc_small_advsimd_<mode>): ...this.
390 (tlsdesc_small_sve_<mode>): New pattern.
392 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
394 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
395 (UNSPEC_UMUL_HIGHPART): New constants.
396 (MUL_HIGHPART): New int iteraor.
397 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
398 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
400 (*<su>mul<mode>3_highpart): New define_insn.
402 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
405 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
408 2018-03-13 Martin Liska <mliska@suse.cz>
411 * (sem_item_optimizer::sem_item_optimizer): Initialize new
413 (sem_item_optimizer::~sem_item_optimizer): Release it.
414 (sem_item_optimizer::merge_classes): Register variable aliases.
415 (sem_item_optimizer::fixup_pt_set): New function.
416 (sem_item_optimizer::fixup_points_to_sets): Likewise.
417 * ipa-icf.h: Declare new variables and functions.
419 2018-03-13 Jakub Jelinek <jakub@redhat.com>
422 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
423 integer_pow2p@2 and test integer_pow2p in condition.
424 (A < 0 ? C : 0): Similarly for @1.
427 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
428 characters starting at p contain '\0' character, don't look beyond
432 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
433 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
436 * reg-stack.c (change_stack): Change update_end var from int to
437 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
438 also call set_block_for_insn on the newly added insns and rescan.
441 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
445 * tree.c (stabilize_reference_1): Return save_expr (e) for
446 STATEMENT_LIST even if it doesn't have side-effects.
448 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
450 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
452 2018-03-12 Renlin Li <renlin.li@arm.com>
454 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
455 aarch64_output_scalar_simd_mov_immediate.
457 2018-03-12 Martin Sebor <msebor@redhat.com>
459 PR tree-optimization/83456
460 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
461 for perfectly overlapping calls to memcpy.
462 (gimple_fold_builtin_memory_chk): Same.
463 (gimple_fold_builtin_strcpy): Handle no-warning.
464 (gimple_fold_builtin_stxcpy_chk): Same.
465 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
467 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
469 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
470 parameter. Use it for SFmode.
471 (rs6000_function_arg_advance_1): Adjust.
472 (rs6000_function_arg): Adjust.
473 (rs6000_gimplify_va_arg): Pass false for that new parameter.
475 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
477 PR rtl-optimization/84169
478 PR rtl-optimization/84780
479 * combine.c (can_combine_p): Check for a 2-insn combination whether
480 the destination register is used between the two insns, too.
482 2018-03-12 Richard Biener <rguenther@suse.de>
484 PR tree-optimization/84803
485 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
486 for refs DR analysis didn't process.
488 2018-03-12 Richard Biener <rguenther@suse.de>
490 PR tree-optimization/84777
491 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
492 force-vectorize loops ignore whether we are optimizing for size.
494 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
496 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
497 (TARGET_MD_ASM_ADJUST): Define.
499 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
500 Kito Cheng <kito.cheng@gmail.com>
501 Chung-Ju Wu <jasonwucj@gmail.com>
503 * config/nds32/nds32.c (nds32_compute_stack_frame,
504 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
505 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
506 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
507 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
508 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
509 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
510 * config/nds32/nds32.md (prologue, epilogue): Use macro
511 NDS32_V3PUSH_AVAILABLE_P to do checking.
513 2018-03-11 Jakub Jelinek <jakub@redhat.com>
516 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
517 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
518 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
519 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
520 addition of most attributes on !orig_type_die or the attribute not
521 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
523 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
524 Chung-Ju Wu <jasonwucj@gmail.com>
526 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
528 * config/nds32/nds32.opt (mvh): New option.
530 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
531 Chung-Ju Wu <jasonwucj@gmail.com>
533 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
535 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
536 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
539 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
540 Chung-Ju Wu <jasonwucj@gmail.com>
542 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
544 * config/nds32/nds32-multiple.md (strlensi): New pattern.
545 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
547 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
548 Kito Cheng <kito.cheng@gmail.com>
549 Chung-Ju Wu <jasonwucj@gmail.com>
551 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
552 UNSPEC_FFMISM and UNSPEC_FLMISM.
553 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
554 for ffb, ffmism and flmism.
555 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
556 (unspec_ffmism): Ditto.
557 (unspec_flmism): Ditto.
558 (nds32_expand_builtin_impl): Check if string extension is available.
559 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
560 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
562 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
565 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
568 * lra-assigns.c (assign_by_spills): Return a flag of reload
569 assignment failure. Do not process the reload assignment
570 failures. Do not spill other reload pseudos if they has the same
572 (lra_assign): Add a return arg. Set up from the result of
573 assign_by_spills call.
574 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
575 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
576 usage_insns if it is not NULL.
577 (spill_hard_reg_in_range): New function.
578 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
579 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
581 (lra_assign): Change prototype.
582 * lra.c (lra): Add code to deal with fails by splitting hard reg
585 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
588 * config/i386/i386.opt: Replace Enforcment with Enforcement.
590 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
593 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
594 (dw_val_node): Add val_symbolic_view.
595 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
596 (symview_upper_bound): New.
597 (new_line_info_table): Initialize symviews_since_reset.
598 (dwarf2out_source_line): Count symviews_since_reset and set
600 (dw_val_equal_p): Handle symview.
601 (add_AT_symview): New.
602 (print_dw_val): Handle symview.
603 (attr_checksum, attr_checksum_ordered): Likewise.
604 (same_dw_val_p, size_of_die): Likewise.
605 (value_format, output_die): Likewise.
606 (add_high_low_attributes): Use add_AT_symview for entry_view.
607 (dwarf2out_finish): Reset symview_upper_bound, clear
610 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
613 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
614 Add strict argument and use it.
615 (rs6000_split_multireg_move): Update for new strict argument.
616 (mem_operand_gpr): Disallow all non-offsettable addresses.
617 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
619 2018-03-09 Jakub Jelinek <jakub@redhat.com>
622 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
623 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
624 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
627 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
628 decl, use remap_type if we want to use the type.
630 2018-03-09 Martin Sebor <msebor@redhat.com>
632 PR tree-optimization/84526
633 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
635 (builtin_access::generic_overlap): Be prepared to handle non-array
638 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
640 PR rtl-optimization/84682
641 * lra-constraints.c (process_address_1): Check is_address flag
642 for address constraints.
643 (process_alt_operands): Likewise.
644 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
645 preprocess_constraints.
646 * recog.h (preprocess_constraints): Add oploc parameter.
648 * recog.c (preprocess_constraints): Test address_operand for
649 CT_ADDRESS constraints.
651 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
654 * lra-assigns.c (assign_by_spills): Return a flag of reload
655 assignment failure. Do not process the reload assignment
656 failures. Do not spill other reload pseudos if they has the same
658 (lra_assign): Add a return arg. Set up from the result of
659 assign_by_spills call.
660 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
661 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
662 usage_insns if it is not NULL.
663 (spill_hard_reg_in_range): New function.
664 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
665 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
667 (lra_assign): Change prototype.
668 * lra.c (lra): Add code to deal with fails by splitting hard reg
671 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
674 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
675 Accept complain bool parameter. Only emit errors if it is true.
676 (arm_parse_cpu_option_name): Likewise.
677 (arm_target_thumb_only): Adjust callers of the above.
678 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
679 prototype to take a default true bool parameter.
680 (arm_parse_arch_option_name): Likewise.
682 2018-03-09 David Malcolm <dmalcolm@redhat.com>
683 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
687 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
688 * configure: Regenerate.
689 * configure.ac ("linker --version-script option"): New.
690 ("linker soname option"): New.
692 2018-03-09 Richard Biener <rguenther@suse.de>
694 PR tree-optimization/84775
695 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
696 immediate uses of predicate stmts and mark them modified.
699 PR tree-optimization/84178
700 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
702 (version_loop_for_if_conversion): Delay update_ssa call.
703 (tree_if_conversion): Delay update_ssa until after predicate
706 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
709 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
710 when the function accesses prior frames.
712 2018-03-08 Jakub Jelinek <jakub@redhat.com>
715 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
716 gen_llsym, otherwise call maybe_gen_llsym.
719 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
720 has ',' character inside of it.
722 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
725 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
726 as clobbering CC_REGNUM.
728 2018-03-08 Richard Biener <rguenther@suse.de>
731 * tree-scalar-evolution.c: Include tree-into-ssa.h.
732 (follow_copies_to_constant): Do not follow SSA names registered
735 2018-03-08 Richard Biener <rguenther@suse.de>
737 PR tree-optimization/84178
738 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
740 (version_loop_for_if_conversion): Delay update_ssa call.
741 (tree_if_conversion): Delay update_ssa until after predicate
744 2018-03-08 David Malcolm <dmalcolm@redhat.com>
746 PR tree-optimization/84178
747 * tree-if-conv.c (release_bb_predicate): Remove the
748 the assertion that the stmts have NULL use_ops.
749 Discard the statements, asserting that they haven't
750 yet been added to a BB.
752 2018-03-08 Richard Biener <rguenther@suse.de>
754 PR tree-optimization/84746
755 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
756 (phi_translate): Pass in destination ANTIC_OUT set.
757 (phi_translate_1): Likewise. For a simplified result lookup
758 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
759 (phi_translate_set): Adjust.
760 (do_pre_regular_insertion): Likewise.
761 (do_pre_partial_partial_insertion): Likewise.
763 2018-03-08 Martin Liska <mliska@suse.cz>
765 PR gcov-profile/84735
766 * doc/gcov.texi: Document usage of profile files.
767 * gcov-io.h: Document changes in the format.
769 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
773 * dwarf2out.c (struct dw_line_info_table): Update comments for
775 (FORCE_RESET_NEXT_VIEW): New.
776 (FORCE_RESETTING_VIEW_P): New.
777 (RESETTING_VIEW_P): Check for -1 too.
778 (ZERO_VIEW_P): Likewise.
779 (new_line_info_table): Force-reset next view.
780 (dwarf2out_begin_function): Likewise.
781 (dwarf2out_source_line): Simplify zero_view_p initialization.
782 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
783 view directly. Omit view when omitting .loc at line 0.
785 2018-03-08 Jakub Jelinek <jakub@redhat.com>
787 PR tree-optimization/84740
788 * tree-switch-conversion.c (process_switch): Call build_constructors
789 only if info.phi_count is non-zero.
791 PR tree-optimization/84739
792 * tree-tailcall.c (find_tail_calls): Check call arguments against
793 DECL_ARGUMENTS (current_function_decl) rather than
794 DECL_ARGUMENTS (func) when checking for tail recursion.
796 2018-03-07 Jakub Jelinek <jakub@redhat.com>
798 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
799 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
800 Volker Reichelt's entry and add entries for people that perform
801 GCC fuzzy testing and report numerous bugs.
803 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
806 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
807 readonly data in sdata, if that is disabled.
808 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
809 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
810 -mreadonly-in-sdata option.
812 2018-03-07 Martin Sebor <msebor@redhat.com>
814 PR tree-optimization/84468
815 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
816 basic block when looking for nul assignment.
818 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
821 * except.h (output_function_exception_table): Adjust prototype.
822 * except.c (output_function_exception_table): Remove FNNAME parameter
823 and add SECTION parameter. Ouput one part of the table at a time.
824 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
825 the first part of the exception table and emit unwind directives.
826 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
827 (i386_pe_seh_cold_init): Likewise.
828 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
829 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
830 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
831 (ix86_output_call_insn): Emit a nop in one more case for SEH.
832 * config/i386/winnt.c: Include except.h.
833 (struct seh_frame_state): Add reg_offset, after_prologue and
834 in_cold_section fields.
835 (i386_pe_seh_end_prologue): Set seh->after_prologue.
836 (i386_pe_seh_cold_init): New function.
837 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
838 to seh->in_cold_section.
839 (seh_emit_push): Record the offset of the push.
840 (seh_emit_save): Record the offet of the save.
841 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
842 Test seh->after_prologue to disregard the epilogue.
843 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
844 (i386_pe_end_cold_function): New function.
846 2018-03-07 Jakub Jelinek <jakub@redhat.com>
849 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
850 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
853 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
855 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
856 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
859 * multiple_target.c: Include tree-inline.h and intl.h.
860 (expand_target_clones): Diagnose and fail if node->definition and
861 !tree_versionable_function_p (node->decl).
863 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
865 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
867 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
868 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
869 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
871 2018-03-06 Jakub Jelinek <jakub@redhat.com>
874 * combine.c (try_combine): Use reg_or_subregno instead of handling
875 just paradoxical SUBREGs and REGs.
877 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
879 * config/arc/arc.c (arc_finalize_pic): Remove function.
880 (arc_must_save_register): We use single base PIC register, remove
881 checks to save/restore the PIC register.
882 (arc_expand_prologue): Likewise.
883 * config/arc/arc-protos.h (arc_set_default_type_attributes):
885 (arc_verify_short): Likewise.
886 (arc_attr_type): Likewise.
887 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
888 (walk_stores): Likewise.
889 (arc_address_cost): Make it static.
890 (arc_verify_short): Likewise.
891 (branch_dest): Likewise.
892 (arc_attr_type): Likewise.
893 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
894 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
895 (arc_final_prescan_insn): Remove inserting the nops due to
896 hardware hazards. It is done in reorg step.
897 (insn_length_variant_t): Remove.
898 (insn_length_parameters_t): Likewise.
899 (arc_insn_length_parameters): Likewise.
900 (arc_get_insn_variants): Likewise.
901 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
903 2018-03-06 Jakub Jelinek <jakub@redhat.com>
906 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
909 PR tree-optimization/84687
910 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
912 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
914 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
916 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
917 Rename to ppc_speculation_barrier.
918 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
919 __builtin_ppc_speculation_barrier.
921 2018-03-05 Jakub Jelinek <jakub@redhat.com>
924 * combine.c (combine_simplify_rtx): Don't try to simplify if
925 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
928 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
930 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
931 to 32 bytes when compiling for POWER9.
933 2018-03-05 Jakub Jelinek <jakub@redhat.com>
936 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
937 regparm >= 3 with no arg reg available also for calls with
938 flag_force_indirect_call. Pass decl to ix86_function_regparm.
941 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
943 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
945 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
948 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
950 2018-03-05 Richard Biener <rguenther@suse.de>
952 PR tree-optimization/84486
953 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
954 When inserting a __builtin_assume_aligned call set the LHS
955 SSA name alignment info accordingly.
957 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
959 PR tree-optimization/84114
960 * config/aarch64/aarch64.c (aarch64_reassociation_width)
961 Avoid reassociation of FLOAT_MODE addition.
963 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
965 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
966 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
967 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
968 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
969 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
970 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
971 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
973 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
974 __builtin_ia32_wbinvd): New builtins.
975 (SPECIAL_ARGS2): New.
976 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
977 (SPECIAL_ARGS2): New.
978 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
979 (ix86_valid_target_attribute_inner_p): Ditto.
980 (ix86_init_mmx_sse_builtins): Add special_args2.
981 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
982 TARGET_WBNOINVD_P): New.
983 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
984 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
985 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
986 * config/i386/immintrin.h (_wbinvd): New intrinsic.
987 * config/i386/pconfigintrin.h: New file.
988 * config/i386/wbnoinvdintrin.h: Ditto.
989 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
990 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
992 2018-03-05 Richard Biener <rguenther@suse.de>
994 PR tree-optimization/84670
995 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
997 (BB_VISITED_WITH_VISITED_SUCCS): New define.
998 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
999 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1000 doesn't grow if all successors (recursively) were visited at least
1003 2018-03-05 Richard Biener <rguenther@suse.de>
1005 PR tree-optimization/84650
1006 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1007 if executed in the loop pipeline.
1009 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1011 * doc/configfiles.texi (Configuration Files): Move info about
1012 conditionalizing $target-protos.h to...
1013 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1014 differs from $target-protos.h.
1016 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1017 Chung-Ju Wu <jasonwucj@gmail.com>
1019 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1020 * config/nds32/nds32-multiple.md (setmemsi): Define.
1021 * config/nds32/nds32-memory-manipulation.c
1022 (nds32_gen_dup_4_byte_to_word_value): New.
1023 (emit_setmem_word_loop): New.
1024 (emit_setmem_byte_loop): New.
1025 (nds32_expand_setmem_loop): New.
1026 (nds32_expand_setmem_loop_v3m): New.
1027 (nds32_expand_setmem_unroll): New.
1028 (nds32_expand_setmem): New.
1030 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1031 Chung-Ju Wu <jasonwucj@gmail.com>
1033 * config/nds32/nds32-memory-manipulation.c
1034 (nds32_emit_load_store): New.
1035 (nds32_emit_post_inc_load_store): New.
1036 (nds32_emit_mem_move): New.
1037 (nds32_emit_mem_move_block): New.
1038 (nds32_expand_movmemsi_loop_unknown_size): New.
1039 (nds32_expand_movmemsi_loop_known_size): New.
1040 (nds32_expand_movmemsi_loop): New.
1041 (nds32_expand_movmemsi_unroll): New.
1042 (nds32_expand_movmemqi): Rename ...
1043 (nds32_expand_movmemsi): ... to this.
1044 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1045 (movmemsi): ... to this.
1046 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1047 (nds32_expand_movmemsi): ... to this.
1049 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1050 Monk Chiang <sh.chiang04@gmail.com>
1051 Chung-Ju Wu <jasonwucj@gmail.com>
1053 * config/nds32/nds32-protos.h
1054 (nds32_expand_load_multiple): New arguments.
1055 (nds32_expand_store_multiple): Ditto.
1056 (nds32_valid_multiple_load_store): Rename ...
1057 (nds32_valid_multiple_load_store_p): ... to this.
1058 * config/nds32/nds32-memory-manipulation.c
1059 (nds32_expand_load_multiple): Refine implementation.
1060 (nds32_expand_store_multiple): Ditto.
1061 * config/nds32/nds32-multiple.md
1062 (load_multiple): Update nds32_expand_load_multiple interface.
1063 (store_multiple): Update nds32_expand_store_multiple interface.
1064 * config/nds32/nds32-predicates.c
1065 (nds32_valid_multiple_load_store): Rename ...
1066 (nds32_valid_multiple_load_store_p): ... to this and refine
1068 * config/nds32/predicates.md
1069 (nds32_load_multiple_and_update_address_operation): New predicate.
1070 (nds32_store_multiple_and_update_address_operation): New predicate.
1072 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1073 Chung-Ju Wu <jasonwucj@gmail.com>
1075 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1076 (combo): New attribute.
1077 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1079 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1081 * config/nds32/nds32.opt: Change -mcmodel= default value.
1083 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1084 Monk Chiang <sh.chiang04@gmail.com>
1085 Chung-Ju Wu <jasonwucj@gmail.com>
1087 * config/nds32/constants.md (unspec_element): New enum.
1088 * config/nds32/constraints.md (Umw): New constraint.
1089 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1090 * config/nds32/nds32-intrinsic.md: Likewise.
1091 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1092 (nds32_valid_smw_lwm_base_p): New.
1093 (nds32_output_smw_single_word): New.
1094 (nds32_output_lmw_single_word): New.
1095 (nds32_expand_unaligned_load): New.
1096 (nds32_expand_unaligned_store): New.
1097 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1098 (nds32_output_smw_single_word): Declare.
1099 (nds32_output_lmw_single_word): Declare.
1100 (nds32_expand_unaligned_load): Declare.
1101 (nds32_expand_unaligned_store): Declare.
1102 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1103 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1104 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1105 NDS32_BUILTIN_UASTORE_DW.
1106 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1109 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1110 Kito Cheng <kito.cheng@gmail.com>
1111 Chung-Ju Wu <jasonwucj@gmail.com>
1113 * config/nds32/nds32-intrinsic.c
1114 (nds32_expand_builtin_null_ftype_reg): Delete.
1115 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1116 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1117 (nds32_read_argument): New.
1118 (nds32_legitimize_target): Ditto.
1119 (nds32_legitimize_argument): Ditto.
1120 (nds32_check_constant_argument): Ditto.
1121 (nds32_expand_unop_builtin): Ditto.
1122 (nds32_expand_unopimm_builtin): Ditto.
1123 (nds32_expand_binop_builtin): Ditto.
1124 (nds32_builtin_decl_impl): Ditto.
1125 (builtin_description): Ditto.
1126 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1127 (nds32_init_builtins_impl): Ditto.
1128 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1129 (nds32_builtin_decl): New.
1130 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1131 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1133 2018-03-02 Jeff Law <law@redhat.com>
1135 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1136 (redundant_insn, fill_simple_delay_slots): Likewise.
1137 (fill_slots_from_thread): Likewise.
1138 * resource.c (mark_referenced_resources): Likewise.
1139 (mark_set_resources, find_dead_or_set_registers): Likewise.
1141 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1143 * substring-locations.h (format_warning_va): Formatting fix for
1145 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1147 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1148 * substring-locations.c: Include intl.h.
1149 (format_warning_va): Turned into small wrapper around
1150 format_warning_n_va, renamed to ...
1151 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1152 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1154 (format_warning_at_substring_n): New function.
1155 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1156 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1157 format_warning_at_substring with just a shorter name instead of
1158 const function pointer.
1159 (fmtwarn_n): New function.
1160 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1161 appropriate, get rid of all the fmtstr temporaries, move conditionals
1162 with G_() wrapped string literals directly into fmtwarn arguments,
1163 cast dir.len to (int), formatting fixes.
1165 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1167 * doc/invoke.texi: Remove "Cilk Plus" references.
1169 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1170 Richard Biener <rguenther@suse.de>
1173 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1174 for error or warning attributes if CALL_FROM_THUNK_P is set.
1177 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1180 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1181 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1184 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1185 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1187 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1188 instead of -1U in last predictors element's probability member.
1190 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1193 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1194 arguments if they are comparable.
1196 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1198 PR tree-optimization/84634
1199 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1200 masks and masked_loop_p with a single loop_masks, making sure it's
1201 null for bb vectorization.
1203 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1205 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1206 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1207 loop->force_vectorize to check whether there is no alias.
1209 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1212 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1214 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1216 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1217 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1218 instead of a loop around prev_real_insn.
1219 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1223 * config/i386/i386.c (ix86_print_operand): Use conditional
1224 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1227 2018-03-02 Richard Biener <rguenther@suse.de>
1229 PR tree-optimization/84427
1230 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1231 (bitmap_set_subtract_values): Rewrite to handle multiple
1234 (prune_clobbered_mems): Likewise.
1235 (phi_translate): Take edge instead of pred/phiblock.
1236 (phi_translate_1): Likewise.
1237 (phi_translate_set): Likewise. Insert all translated
1238 exprs for a value into the set, keeping possibly multiple
1239 expressions per value.
1240 (compute_antic_aux): Adjust for phi_translate changes.
1241 When intersecting union the expressions and prune those
1242 not in the final value set, keeping possibly multiple
1243 expressions per value. Do not use value-insertion
1244 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1245 all expressions. Add verification that the value-sets
1246 only shrink during iteration.
1247 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1248 (do_pre_regular_insertion): Likewise.
1249 (do_pre_partial_partial_insertion): Likewise.
1251 2018-03-02 Richard Biener <rguenther@suse.de>
1254 * config/darwin.c (saved_debug_info_level): New static global.
1255 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1256 (darwin_asm_lto_end): Restore debug info generation settings.
1258 2018-03-01 Martin Liska <mliska@suse.cz>
1261 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1264 2018-03-01 Richard Biener <rguenther@suse.de>
1267 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1268 type annotation with LTO when debug was disabled at compile-time.
1270 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1272 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1274 (mips_final_postscan_insn): Likewise.
1276 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1278 PR rtl-optimization/84528
1279 * alias.c (init_alias_target): Add commentary.
1280 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1281 a unique base value if the frame pointer is not eliminated
1282 to the stack pointer.
1284 2018-03-01 Tom de Vries <tom@codesourcery.com>
1286 PR rtl-optimization/83327
1287 * lra-int.h (hard_regs_spilled_into): Declare.
1288 * lra.c (hard_regs_spilled_into): Define.
1289 (init_reg_info): Init hard_regs_spilled_into.
1290 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1291 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1292 (process_bb_lives): Handle hard_regs_spilled_into.
1293 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1294 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1296 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1298 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1299 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1300 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1301 * config/rs6000/aix72.h: New file.
1303 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1305 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1306 instead of warning_at with conditional singular and plural messages
1310 * stor-layout.c (update_alignment_for_field): For
1311 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1312 && !DECL_PACKED (field), do the alignment update, just use
1313 only desired_align instead of MAX (type_align, desired_align)
1315 (place_field): Don't do known_align < desired_align handling
1316 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1317 is non-NULL, instead do it after rli->prev_field handling and
1318 only if not within a bitfield word. For DECL_PACKED (field)
1319 use type_align of BITS_PER_UNIT.
1321 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1323 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1324 superfluous parentheses and trailing spaces.
1326 2018-02-28 Richard Biener <rguenther@suse.de>
1328 PR tree-optimization/84584
1329 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1330 SCoPs with fake exit edge.
1332 2018-02-28 Martin Liska <mliska@suse.cz>
1335 * timevar.c (timer::print): Fix format to properly print 100%
1338 2018-02-28 Richard Biener <rguenther@suse.de>
1341 * genmatch.c (capture_info::walk_match): Do not mark
1342 captured expressions without operands as expr_p given
1343 they act more like predicates and should be subject to
1344 "lost tail" side-effect preserving.
1346 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1348 PR rtl-optimization/81611
1349 * auto-inc-dec.c (attempt_change): Move dead note from
1350 mem_insn if it's the next use of regno
1351 (find_address): Take address use of reg holding
1352 non-incremented value. Add parm to limit search to the named
1354 (merge_in_block): Attempt to use a mem insn that is the next
1355 use of the original regno.
1357 2018-02-27 Martin Sebor <msebor@redhat.com>
1360 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1361 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1363 2018-02-27 Martin Sebor <msebor@redhat.com>
1365 PR translation/84207
1366 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1367 n argument to unsigned HOST_WIDE_INT.
1368 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1369 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1370 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1371 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1373 2018-02-27 Richard Biener <rguenther@suse.de>
1375 PR tree-optimization/84512
1376 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1377 Do not use the estimate returned from record_stmt_cost for
1378 the scalar iteration cost but sum properly using add_stmt_cost.
1380 2018-02-27 Richard Biener <rguenther@suse.de>
1382 PR tree-optimization/84466
1383 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1384 Adjust last change to less strictly validate use operands.
1386 2018-02-27 Martin Liska <mliska@suse.cz>
1388 PR gcov-profile/84548
1389 * gcov.c (process_file): Allow partial overlap and consider it
1390 also as group functions.
1391 (output_lines): Properly calculate range of lines for a group.
1393 2018-02-27 Martin Liska <mliska@suse.cz>
1395 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1396 'ggc' suffixes. Change first column width.
1397 (timer::print): Fix formatting of the column.
1399 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1401 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1402 preserve inline entry blocks for the sake of debug inline
1403 entry point markers alone.
1404 (remove_unused_locals): Suggest in comments a better place to
1405 force the preservation of inline entry blocks that are
1406 otherwise unused, but do not preserve them.
1408 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1410 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1412 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1415 * config/i386/constraints.md (Bs): Replace
1416 ix86_indirect_branch_register with
1417 TARGET_INDIRECT_BRANCH_REGISTER.
1419 * config/i386/i386.md (indirect_jump): Likewise.
1420 (tablejump): Likewise.
1421 (*sibcall_memory): Likewise.
1422 (*sibcall_value_memory): Likewise.
1423 Peepholes of indirect call and jump via memory: Likewise.
1424 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1425 (*sibcall_value_GOT_32): Likewise.
1426 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1427 (GOT_memory_operand): Likewise.
1428 (call_insn_operand): Likewise.
1429 (sibcall_insn_operand): Likewise.
1430 (GOT32_symbol_operand): Likewise.
1431 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1433 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1435 PR rtl-optimization/83496
1436 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1437 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1438 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1439 redundant insn, if any.
1440 (relax_delay_slots): Likewise.
1441 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1443 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1445 PR tree-optimization/83965
1446 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1447 that grouped statements are part of a reduction chain. Return
1448 true if the statement is not marked as a reduction itself but
1450 (vect_recog_dot_prod_pattern): Don't check whether the statement
1451 is part of a group here.
1452 (vect_recog_sad_pattern): Likewise.
1453 (vect_recog_widen_sum_pattern): Likewise.
1455 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1458 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1460 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1463 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1465 (ix86_output_indirect_function_return): New prototype.
1466 (ix86_split_simple_return_pop_internal): Likewise.
1467 * config/i386/i386.c (indirect_return_via_cx): New.
1468 (indirect_return_via_cx_bnd): Likewise.
1469 (indirect_thunk_name): Handle return va CX_REG.
1470 (output_indirect_thunk_function): Create alias for
1471 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1472 (ix86_output_indirect_jmp): Remove the bool argument.
1473 (ix86_output_indirect_function_return): New function.
1474 (ix86_split_simple_return_pop_internal): Likewise.
1475 * config/i386/i386.md (*indirect_jump): Don't pass false
1476 to ix86_output_indirect_jmp.
1477 (*tablejump_1): Likewise.
1478 (simple_return_pop_internal): Change it to define_insn_and_split.
1479 Call ix86_split_simple_return_pop_internal to split it for
1481 (simple_return_indirect_internal): Call
1482 ix86_output_indirect_function_return instead of
1483 ix86_output_indirect_jmp.
1485 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1488 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1489 memset and value initialization afterwards.
1491 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1493 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1495 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1498 * common/config/aarch64/aarch64-common.c
1499 (aarch_option_optimization_table[]): Switch
1500 off fomit-frame-pointer
1502 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1503 Chung-Ju Wu <jasonwucj@gmail.com>
1505 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1507 (store_multiple): Ditto.
1509 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1511 * config.gcc: Add --with-cpu support for nds32 target.
1512 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1513 * config/nds32/nds32.opt: Add -mcpu= option.
1515 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1517 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1518 isel=yes): Warn for these deprecated options.
1520 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1522 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1523 ISA_2_5_MASKS_EMBEDDED.
1525 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1527 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1528 p->max as pointers rather than using iterative_hash_expr.
1530 2018-02-23 Carl Love <cel@us.ibm.com>
1532 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1533 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1535 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1536 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1537 P8V_BUILTIN_VEC_VUNSIGNED2.
1539 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1542 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1543 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1545 * lra-constraints.c (curr_insn_transform): Set up
1546 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1548 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1549 * lra-eliminations.c (spill_pseudos): Ditto.
1550 (process_insn_for_elimination): Ditto.
1551 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1552 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1553 LRA_NON_CLOBBERED_ALT.
1555 2018-02-22 Martin Sebor <msebor@redhat.com>
1557 PR tree-optimization/84480
1558 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1559 to maybe_diag_stxncpy_trunc. Call it.
1560 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1561 from gimple_fold_builtin_strcpy. Print inlining stack.
1562 (handle_builtin_stxncpy): Print inlining stack.
1563 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1565 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1568 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1569 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1570 and -fcheck-pointer-bounds are used together.
1571 (indirect_thunk_prefix): New enum.
1572 (indirect_thunk_need_prefix): New function.
1573 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1574 "_nt" instead of "_bnd" for NOTRACK prefix.
1575 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1576 (output_indirect_thunk_function): Likewise.
1578 (ix86_code_end): Update output_indirect_thunk_function calls.
1579 (ix86_output_indirect_branch_via_reg): Replace
1580 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1581 (ix86_output_indirect_branch_via_push): Likewise.
1582 (ix86_output_function_return): Likewise.
1583 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1584 incompatible with -fcf-protection=branch and
1585 -fcheck-pointer-bounds.
1587 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1590 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1591 Change gcc_assert call to output_operand_lossage.
1593 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1595 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1597 2018-02-22 DJ Delorie <dj@redhat.com>
1598 Sebastian Perta <sebastian.perta@renesas.com>
1599 Oleg Endo <olegendo@gcc.gnu.org>
1601 * config/rx/rx.c (rx_rtx_costs): New function.
1602 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1604 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1606 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1608 2018-02-22 Martin Liska <mliska@suse.cz>
1611 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1612 Add "native" as a possible value.
1614 2018-02-22 Martin Liska <mliska@suse.cz>
1617 * config/i386/i386.c (ix86_option_override_internal):
1618 Add "native" as a possible value for -march and -mtune.
1620 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1623 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1624 to all type variants.
1626 PR tree-optimization/84503
1627 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1628 width as info->bitpos + info->bitsize - start.
1629 (merged_store_group::merge_overlapping): Simplify width computation.
1630 (check_no_overlap): New function.
1631 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1632 start + width and last_order of the group, fail if check_no_overlap
1634 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1635 to group if check_no_overlap fails.
1637 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1639 * config/rs6000/altivec.md: Delete contraint arguments to
1640 define_expand, define_split, and define_peephole2, and in
1641 define_insn_and_split if always unused.
1642 * config/rs6000/darwin.md: Ditto.
1643 * config/rs6000/dfp.md: Ditto.
1644 * config/rs6000/rs6000.md: Ditto.
1645 * config/rs6000/sync.md: Ditto.
1646 * config/rs6000/vector.md: Ditto.
1647 * config/rs6000/vsx.md: Ditto.
1649 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1651 * config/rs6000/altivec.md: Write output control strings as braced
1652 blocks instead of double-quoted strings.
1653 * config/rs6000/darwin.md: Ditto.
1654 * config/rs6000/rs6000.md: Ditto.
1655 * config/rs6000/vector.md: Ditto.
1656 * config/rs6000/vsx.md: Ditto.
1658 2018-02-21 Jason Merrill <jason@redhat.com>
1660 PR c++/84314 - ICE with templates and fastcall attribute.
1661 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1663 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1665 * ipa-cp.c (determine_versionability): Fix comment typos.
1667 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1670 * ipa-cp.c (determine_versionability): Do not version functions caling
1673 2018-02-21 Martin Liska <mliska@suse.cz>
1676 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1677 Add "native" as a possible value.
1678 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1679 the macro when native cpu detection is available.
1681 2018-02-21 Martin Liska <mliska@suse.cz>
1684 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1685 Add "native" as a possible value.
1686 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1687 when native cpu detection is available.
1689 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1690 Martin Sebor <msebor@redhat.com>
1692 PR tree-optimization/84478
1693 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1695 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1696 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1697 support which is conservatively correct, for 2 only stay conservative
1698 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1699 argument to the 2 argument get_range_strlen, adjust 6 arg
1700 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1702 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1703 (gimple_fold_builtin_strlen): Pass true as last argument to
1706 2018-02-20 Martin Sebor <msebor@redhat.com>
1709 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1710 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1711 (builtin_memref::builtin_memref): Factor out parts into
1712 set_base_and_offset and call it.
1714 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1717 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1718 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1719 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1720 search at the associated MODE_INT.
1722 2018-02-20 Jeff Law <law@redhat.com>
1725 PR tree-optimization/81592
1727 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1728 for range data rather than using global data.
1729 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1730 range data rather than using global data.
1731 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1732 pass it to children as needed.
1733 (struct directive::fmtresult): Similarly.
1734 (struct directive::set_width): Similarly.
1735 (struct directive::set_precision): Similarly.
1736 (format_integer, format_directive, parse_directive): Similarly.
1737 (format_none): Accept unnamed vr_values parameter.
1738 (format_percent, format_floating, format_character): Similarly.
1739 (format_string, format_plain): Similarly.
1740 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1741 the EVRP range analyzer for range data rather than using global data.
1742 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1743 gimple-ssa-evrp-analyze.h
1744 (class sprintf_dom_walker): Add after_dom_children member function.
1745 Add evrp_range_analyzer member.
1746 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1747 range analyzer as needed.
1748 (sprintf_dom_walker::after_dom_children): New member function.
1749 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1751 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1752 (evrp_range_analyzer::pop_to_marker): Likewise.
1754 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1756 PR tree-optimization/84419
1757 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1758 with the required type if its current type is compatible but
1761 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1764 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1765 after vectorization.
1767 2018-02-20 Martin Liska <mliska@suse.cz>
1770 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1771 possible values if we don't have a hint.
1773 2018-02-20 Martin Liska <mliska@suse.cz>
1777 * final.c (shorten_branches): Build align_tab array with one
1779 * opts.c (finish_options): Add alignment option limit check.
1780 (MAX_CODE_ALIGN): Likewise.
1781 (MAX_CODE_ALIGN_VALUE): Likewise.
1782 * doc/invoke.texi: Document maximum allowed option value for
1783 all -falign-* options.
1785 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1788 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1789 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1790 * var-tracking.c (emit_note_insn_var_location): Remove all references
1791 to NOTE_INSN_CALL_ARG_LOCATION.
1792 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1793 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1794 Use copy_rtx_if_shared.
1795 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1796 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1797 (dwarf2out_var_location): Remove handling of
1798 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1800 * final.c (final_scan_insn): Remove all references to
1801 NOTE_INSN_CALL_ARG_LOCATION.
1802 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1803 before dumping final insns.
1804 * except.c (emit_note_eh_region_end): Remove all references to
1805 NOTE_INSN_CALL_ARG_LOCATION.
1806 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1807 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1808 * config/arc/arc.c (hwloop_optimize): Likewise.
1809 * config/arm/arm.c (create_fix_barrier): Likewise.
1810 * config/s390/s390.c (s390_chunkify_start): Likewise.
1811 * config/sh/sh.c (find_barrier): Likewise.
1812 * config/i386/i386.c (rest_of_insert_endbranch,
1813 ix86_seh_fixup_eh_fallthru): Likewise.
1814 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1815 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1816 * config/frv/frv.c (frv_function_prologue): Likewise.
1817 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1819 (note_outside_basic_block_p): Remove all references to
1820 NOTE_INSN_CALL_ARG_LOCATION.
1821 * gengtype.c (adjust_field_rtx_def): Likewise.
1822 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1824 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1825 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1828 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1831 PR tree-optimization/84452
1832 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1833 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1836 2018-02-19 Martin Liska <mliska@suse.cz>
1839 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1841 2018-02-19 Martin Liska <mliska@suse.cz>
1842 Richard Sandiford <richard.sandiford@linaro.org>
1844 PR tree-optimization/82491
1845 * gimple-fold.c (get_base_constructor): Make earlier bail out
1848 2018-02-19 Carl Love <cel@us.ibm.com>
1850 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1851 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1853 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1854 P8V_BUILTIN_VEC_NEG.
1856 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1858 * config/rl78/rl78.md (movdf): New define expand.
1860 2018-02-19 Martin Liska <mliska@suse.cz>
1863 * doc/invoke.texi: Fix typo.
1864 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1866 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1868 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1869 handle rs6000_single_float and rs6000_double_float specially for
1872 2018-02-16 Jeff Law <law@redhat.com>
1874 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1877 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1880 * ipa-inline.c (inline_small_functions): Fix a typo.
1882 2018-02-16 Nathan Sidwell <nathan@acm.org>
1884 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1886 2018-02-16 Carl Love <cel@us.ibm.com>
1888 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1889 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1890 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1891 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1892 expansion to P8V_BUILTIN_VEC_FLOAT2.
1894 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1896 PR rtl-optimization/70023
1897 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1898 src_regno into account.
1900 2018-02-16 Carl Love <cel@us.ibm.com>
1902 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1903 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1904 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1905 * config/rs6000/rs6000.c: Remove case statements for
1906 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1907 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1908 and P9V_BUILTIN_VEC_VINSERT4B.
1909 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1910 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1911 * config/rs6000/vsx.md:
1912 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1915 2018-02-16 Carl Love <cel@us.ibm.com>
1917 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1919 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1921 * config/rs6000/rs6000-c.c: Add the definitions for
1922 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1923 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1924 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1925 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1926 definition for insert4b and define insn *insert3b_internal.
1927 * doc/extend.texi: Add documentation for vec_extract4b.
1929 2018-02-16 Nathan Sidwell <nathan@acm.org>
1931 * doc/extend.texi (Backwards Compatibility): Mention friend
1932 injection. Note for-scope is deprecated.
1933 * doc/invoke.texi (-ffriend-injection): Deprecate.
1935 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1937 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1938 that moved to I2, also allow destinations that are a paradoxical
1939 subreg (instead of a normal reg).
1941 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1944 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1947 2018-02-16 Richard Biener <rguenther@suse.de>
1949 PR tree-optimization/84037
1950 PR tree-optimization/84016
1952 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1953 Adjust vec_construct for the fact we need additional higher latency
1954 128bit inserts for AVX256 and AVX512 vector builds.
1955 (ix86_add_stmt_cost): Scale vector construction cost for
1958 2018-02-16 Richard Biener <rguenther@suse.de>
1960 PR tree-optimization/84417
1961 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1962 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1963 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1965 2018-02-16 Martin Liska <mliska@suse.cz>
1968 * internal-fn.def (ASAN_CHECK): Set proper flags.
1969 (ASAN_MARK): Likewise.
1971 2018-02-16 Julia Koval <julia.koval@intel.com>
1973 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1974 from PTA_CANNONLAKE.
1976 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1979 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1980 Use ++iter rather than iter++ for std::list iterators.
1981 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1982 defer deleting them until all nodes in the forest are processed. Do
1983 free even leaf nodes. Change to_process into auto_vec.
1986 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1987 * vec.h (vec_default_construct): Use memset instead of placement new
1988 if BROKEN_VALUE_INITIALIZATION is defined.
1989 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1990 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1993 PR rtl-optimization/83723
1994 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1995 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1996 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1998 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2000 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2002 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2004 PR rtl-optimization/81443
2005 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2006 from inner REGs to paradoxical SUBREGs.
2008 2018-02-16 Richard Biener <rguenther@suse.de>
2010 PR tree-optimization/84399
2011 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2012 For operands we can analyze at their definition make sure we can
2013 analyze them at each use as well.
2015 2018-02-16 Richard Biener <rguenther@suse.de>
2017 PR tree-optimization/84190
2018 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2019 volatile accesses if the decl isn't volatile.
2021 2018-02-15 Jason Merrill <jason@redhat.com>
2023 PR c++/84314 - ICE with templates and fastcall attribute.
2024 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2025 TYPE_CANONICAL on an existing type.
2027 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2029 PR tree-optimization/84383
2030 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2031 dstoff nor call operand_equal_p if dstbase is NULL.
2033 PR tree-optimization/84334
2034 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2035 also a CONSTANT_CLASS_P, punt.
2037 2018-02-14 Jim Wilson <jimw@sifive.com>
2039 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2040 first SMALL_OPERAND check. New local min_second_step. Move assert
2041 to where locals are set. Add TARGET_RVC support.
2042 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2044 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2046 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2048 2018-02-14 Martin Sebor <msebor@redhat.com>
2050 PR tree-optimization/83698
2051 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2052 arrays constrain the offset range to their bounds.
2053 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2054 (builtin_access::overlap): Avoid setting the size of overlap if it's
2056 (maybe_diag_overlap): Also consider arrays when deciding what values
2057 of offsets to include in diagnostics.
2059 2018-02-14 Martin Sebor <msebor@redhat.com>
2062 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2063 that correspond to the kind of a declaration.
2065 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2068 * config/pa/pa.md: Load address of PIC label using the linkage table
2069 if the label is nonlocal.
2071 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2073 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2074 warning message if user requests -maltivec=be.
2075 * doc/invoke.texi: Document deprecation of -maltivec=be.
2077 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2080 * config/rs6000/rs6000-c.c: Update definitions for
2081 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2082 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2084 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2087 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2088 add _get_ssp intrinsics. Remove argument from
2089 __builtin_ia32_rdssp[d|q].
2090 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2091 * config/i386/i386-builtin.def: Remove argument from
2092 __builtin_ia32_rdssp[d|q].
2093 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2094 ix86_expand_special_args_builtin for _rdssp[d|q].
2095 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2096 Clear register before usage.
2097 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2098 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2100 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2102 PR tree-optimization/84357
2103 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2104 operand 1 of an ARRAY_REF too.
2106 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2109 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2110 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2112 (set_of_reg): New struct.
2113 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2114 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2115 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2117 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2118 Split into bitclr, bitset, bitinvert patterns if appropriate.
2119 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2120 use rx_fuse_in_memory_bitop.
2121 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2122 to named insn, correct maximum insn length.
2124 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2127 * machmode.def: Define a complex mode for PARTIAL_INT.
2128 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2130 * doc/rtl.texi: Document CSPImode.
2131 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2133 (msp430_hard_regno_nregs_with_padding): Likewise.
2135 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2138 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2140 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2142 PR rtl-optimization/84169
2143 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2144 we generated a parallel as new i3 and we split that to new i2 and i3
2145 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2146 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2147 those to i2, not i1. Partially rewrite this scan code.
2149 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2152 * stor-layout.c (place_field): For variable length fields, adjust
2153 offset_align afterwards not just based on the field's alignment,
2154 but also on the size.
2157 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2158 of exps and logs in the use_exp2 case.
2160 2018-02-13 Jeff Law <law@redhat.com>
2162 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2165 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2168 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2172 * common.opt (gas-loc-support, gas-locview-support): New.
2173 (ginline-points, ginternal-reset-location-views): New.
2174 * doc/invoke.texi: Document them. Use @itemx where intended.
2175 (gvariable-location-views): Adjust.
2176 * target.def (reset_location_view): New.
2177 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2178 (TARGET_RESET_LOCATION_VIEW): New.
2179 * doc/tm.texi: Rebuilt.
2180 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2181 (dwarf2out_default_as_locview_support): New.
2182 (output_asm_line_debug_info): Use option variables.
2183 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2184 (output_loc_list): Likewise.
2185 (add_high_low_attributes): Check option variables.
2186 Don't output entry view attribute in strict mode.
2187 (gen_inlined_subroutine_die): Check option variables.
2188 (dwarf2out_inline_entry): Likewise.
2189 (init_sections_and_labels): Likewise.
2190 (dwarf2out_early_finish): Likewise.
2191 (maybe_reset_location_view): New, from...
2192 (dwarf2out_var_location): ... here. Call it.
2193 * debug.h (dwarf2out_default_as_loc_support): Declare.
2194 (dwarf2out_default_as_locview_support): Declare.
2195 * hooks.c (hook_int_rtx_insn_0): New.
2196 * hooks.h (hook_int_rtx_insn_0): Declare.
2197 * toplev.c (process_options): Take -gas-loc-support and
2198 -gas-locview-support from dwarf2out. Enable
2199 -gvariable-location-views by default only with locview
2200 assembler support. Enable -ginternal-reset-location-views by
2201 default only if the target defines the corresponding hook.
2202 Enable -ginline-points by default if location views are
2203 enabled; force it disabled if statement frontiers are
2205 * tree-inline.c (expand_call_inline): Check option variables.
2206 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2208 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2210 PR tree-optimization/84321
2211 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2212 handling. Also check whether the anti-range contains any values
2213 that satisfy the mask; switch to a VR_RANGE if not.
2215 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2218 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2220 2018-02-13 Martin Jambor <mjambor@suse.cz>
2223 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2224 of call statements, also set location of a load to a temporary.
2226 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2228 * config/rl78/rl78.c (add_vector_labels): New function.
2229 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2230 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2231 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2232 which checks that no arguments are passed.
2233 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2234 * doc/extend.texi: Documentation for the new attribute.
2236 2018-02-13 Andreas Schwab <schwab@suse.de>
2238 * config/riscv/linux.h (CPP_SPEC): Define.
2240 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2243 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2244 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2245 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2246 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2247 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2248 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2249 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2250 temporarily for AES and PCLMUL builtins.
2252 PR tree-optimization/84339
2253 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2254 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2258 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2259 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2260 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2262 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2264 * omp-simd-clone.h: New file.
2265 * omp-simd-clone.c: Include omp-simd-clone.h.
2266 (expand_simd_clones): No longer static.
2267 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2268 cgraph.h and omp-simd-clone.h.
2269 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2270 (vect_recog_widen_shift_pattern): Formatting fix.
2271 (vect_pattern_recog_1): Don't check optab for calls.
2274 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2275 operands[2] into a REG before using gen_lowpart on it.
2277 2018-02-12 Jeff Law <law@redhat.com>
2280 * config/sh/sh.c (find_barrier): Consider a sibling call
2283 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2284 successfully back substituting a reg.
2286 2018-02-12 Richard Biener <rguenther@suse.de>
2288 PR tree-optimization/84037
2289 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2290 parameter, move visited init to caller.
2291 (vect_slp_analyze_operations): Separate cost from validity
2292 check, initialize visited once for all instances.
2293 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2295 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2297 (vect_model_promotion_demotion_cost): Likewise.
2298 (vectorizable_bswap): Guard cost modeling with !slp_node
2299 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2301 (vectorizable_call): Likewise.
2302 (vectorizable_conversion): Likewise.
2303 (vectorizable_assignment): Likewise.
2304 (vectorizable_shift): Likewise.
2305 (vectorizable_operation): Likewise.
2306 (vectorizable_store): Likewise.
2307 (vectorizable_load): Likewise.
2308 (vectorizable_condition): Likewise.
2309 (vectorizable_comparison): Likewise.
2311 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2314 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2315 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2316 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2318 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2321 * params.def (inline-min-speedup): Increase from 8 to 15.
2322 (max-inline-insns-auto): Decrease from 40 to 30.
2323 * ipa-split.c (consider_split): Add some buffer for function to
2324 be considered inlining candidate.
2325 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2328 2018-02-12 Richard Biener <rguenther@suse.de>
2330 PR tree-optimization/84037
2331 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2332 matched stmts if we cannot swap the non-matched ones.
2334 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2336 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2337 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2338 _mm_maskz_scalef_round_ss): New intrinsics.
2339 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2340 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2341 __builtin_ia32_scalefss_round): Remove.
2342 (__builtin_ia32_scalefsd_mask_round,
2343 __builtin_ia32_scalefss_mask_round): New intrinsics.
2344 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2345 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2346 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2347 "<round_constraint>")): Changed to ...
2348 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2349 "<round_scalar_constraint>")): ... this.
2350 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2351 %0, %1, %2<round_op3>}"): Changed to ...
2352 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2353 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2354 %2<round_scalar_mask_op3>}"): ... this.
2355 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2357 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2359 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2360 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2361 (_mm_maskz_sqrt_round_ss): New intrinsics.
2362 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2363 (__builtin_ia32_sqrtsd_mask_round)
2364 (__builtin_ia32_sqrtss_mask_round): New builtins.
2365 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2366 (__builtin_ia32_sqrtss_round): Remove.
2367 (__builtin_ia32_sqrtsd_mask_round)
2368 (__builtin_ia32_sqrtss_mask_round): New builtins.
2369 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2370 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2371 ((match_operand:VF_128 1 "vector_operand"
2372 "xBm,<round_constraint>")): Changed to ...
2373 ((match_operand:VF_128 1 "vector_operand"
2374 "xBm,<round_scalar_constraint>")): ... this.
2375 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2376 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2377 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2378 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2379 %<iptr>1<round_scalar_mask_op3>}): ... this.
2380 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2381 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2383 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2386 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2387 Cast vec_cmpeq result to correct type.
2388 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2389 Cast vec_cmpgt result to correct type.
2391 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2393 * final.c (final_scan_insn_1): Renamed from...
2394 (final_scan_insn): ... this. New wrapper, to recover
2395 seen from the outermost call in recursive ones.
2396 * config/sparc/sparc.c (output_return): Drop seen from call.
2397 (output_sibcall): Likewise.
2398 * config/visium/visium.c (output_branch): Likewise.
2400 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2402 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2405 2018-02-10 Alan Modra <amodra@gmail.com>
2408 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2409 Specify LR as an input.
2411 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2414 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2415 remove_member_access_dummy_vars): New functions.
2416 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2417 lower_omp_1, execute_lower_omp): Use them.
2419 PR rtl-optimization/84308
2420 * shrink-wrap.c (spread_components): Release todo vector.
2422 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2424 PR rtl-optimization/57193
2425 * ira-color.c (struct allocno_color_data): Add member
2426 conflict_allocno_hard_prefs.
2427 (update_conflict_allocno_hard_prefs): New.
2428 (bucket_allocno_compare_func): Add a preference based on
2429 conflict_allocno_hard_prefs.
2430 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2431 (color_allocnos): Remove a dead code. Initiate
2432 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2434 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2437 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2438 constraint from =wa to wa. Avoid a subreg on the output operand,
2439 instead use a pseudo and subreg it in a move.
2440 (p9_xxbrd_<mode>): Changed to ...
2441 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2442 (p9_xxbrd_v2df): New expander.
2443 (p9_xxbrw_<mode>): Changed to ...
2444 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2445 (p9_xxbrw_v4sf): New expander.
2447 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2449 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2451 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2454 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2455 multiply in 32-bit mode.
2456 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2457 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2460 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2462 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2463 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2464 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2465 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2467 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2470 * dwarf2out.c (is_trivial_indirect_ref): New function.
2471 (dwarf2out_late_global_decl): Do not generate a location
2472 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2473 and that are not defined in the current unit.
2475 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2477 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2478 instead of a libcall for UNORDERED.
2480 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2483 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2484 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2486 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2489 * config/s390/s390.c (s390_set_current_function): Invoke
2490 s390_indirect_branch_settings also if fndecl didn't change.
2492 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2494 * config/rs6000/rs6000.md (blockage): Set length to zero.
2496 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2498 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2500 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2503 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2504 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2508 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2509 PARALLEL incoming that failed vt_get_decl_and_offset check.
2512 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2513 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2515 (get_variable_section): For decls in named .bss* sections pass true as
2516 second argument to bss_initializer_p.
2518 2018-02-09 Marek Polacek <polacek@redhat.com>
2519 Jakub Jelinek <jakub@redhat.com>
2522 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2523 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2524 Sync some changes from cxx_fold_indirect_ref.
2526 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2528 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2530 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2531 (BLOCK_INLINE_ENTRY_LABEL): New.
2532 (dwarf2out_var_location): Disregard inline entry markers.
2533 (inline_entry_data): New struct.
2534 (inline_entry_data_hasher): New hashtable type.
2535 (inline_entry_data_hasher::hash): New.
2536 (inline_entry_data_hasher::equal): New.
2537 (inline_entry_data_table): New variable.
2538 (add_high_low_attributes): Add DW_AT_entry_pc and
2539 DW_AT_GNU_entry_view attributes if a pending entry is found
2540 in inline_entry_data_table. Add old entry_pc attribute only
2541 if debug nonbinding markers are disabled.
2542 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2543 markers are enabled.
2544 (block_within_block_p, dwarf2out_inline_entry): New.
2545 (dwarf2out_finish): Check that no entries remained in
2546 inline_entry_data_table.
2547 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2548 (final_scan_insn, notice_source_line): Likewise.
2549 (rest_of_clean_state): Skip inline entry markers.
2550 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2552 * gimple.c (gimple_build_debug_inline_entry): New.
2553 * gimple.h (enum gimple_debug_subcode): Add
2554 GIMPLE_DEBUG_INLINE_ENTRY.
2555 (gimple_build_debug_inline_entry): Declare.
2556 (gimple_debug_inline_entry_p): New.
2557 (gimple_debug_nonbind_marker_p): Adjust.
2558 * insn-notes.def (INLINE_ENTRY): New.
2559 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2560 inline entry marker notes.
2561 (print_insn): Likewise.
2562 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2563 (INSN_DEBUG_MARKER_KIND): Likewise.
2564 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2565 * tree-inline.c (expand_call_inline): Build and insert
2566 debug_inline_entry stmt.
2567 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2568 inline entry blocks early, if nonbind markers are enabled.
2569 (dump_scope_block): Dump fragment info.
2570 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2571 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2572 (gimple_build_debug_inline_entry): New.
2573 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2574 Enable/disable inline entry points too.
2575 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2576 (DEBUG_INSN): Describe inline entry markers.
2578 * common.opt (gvariable-location-views): New.
2579 (gvariable-location-views=incompat5): New.
2580 * config.in: Rebuilt.
2581 * configure: Rebuilt.
2582 * configure.ac: Test assembler for view support.
2583 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2584 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2585 * dwarf2out.c (var_loc_view): New typedef.
2586 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2587 (dwarf2out_locviews_in_attribute): New.
2588 (dwarf2out_locviews_in_loclist): New.
2589 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2590 (enum dw_line_info_opcode): Add LI_adv_address.
2591 (struct dw_line_info_table): Add view.
2592 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2593 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2594 (zero_view_p): New variable.
2595 (ZERO_VIEW_P): New macro.
2596 (output_asm_line_debug_info): New.
2597 (struct var_loc_node): Add view.
2598 (add_AT_view_list, AT_loc_list): New.
2599 (add_var_loc_to_decl): Add view param. Test it against last.
2600 (new_loc_list): Add view params. Record them.
2601 (AT_loc_list_ptr): Handle loc and view lists.
2602 (view_list_to_loc_list_val_node): New.
2603 (print_dw_val): Handle dw_val_class_view_list.
2604 (size_of_die): Likewise.
2605 (value_format): Likewise.
2606 (loc_list_has_views): New.
2607 (gen_llsym): Set vl_symbol too.
2608 (maybe_gen_llsym, skip_loc_list_entry): New.
2609 (dwarf2out_maybe_output_loclist_view_pair): New.
2610 (output_loc_list): Output view list or entries too.
2611 (output_view_list_offset): New.
2612 (output_die): Handle dw_val_class_view_list.
2613 (output_dwarf_version): New.
2614 (output_compilation_unit_header): Use it.
2615 (output_skeleton_debug_sections): Likewise.
2616 (output_rnglists, output_line_info): Likewise.
2617 (output_pubnames, output_aranges): Update version comments.
2618 (output_one_line_info_table): Output view numbers in asm comments.
2619 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2620 Call maybe_gen_llsym.
2621 (loc_list_from_tree_1): Adjust.
2622 (add_AT_location_description): Create view list attribute if
2623 needed, check it's absent otherwise.
2624 (convert_cfa_to_fb_loc_list): Adjust.
2625 (maybe_emit_file): Call output_asm_line_debug_info for test.
2626 (dwarf2out_var_location): Reset views as needed. Precompute
2627 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2628 attribute. Set view.
2629 (new_line_info_table): Reset next view.
2630 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2631 (dwarf2out_source_line): Likewise. Output view resets and labels to
2632 the assembler, or select appropriate line info opcodes.
2633 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2634 (optimize_string_length): Catch it. Adjust.
2635 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2636 dw_val_class_view_list, and remove it if no longer needed.
2637 (hash_loc_list): Hash view numbers.
2638 (loc_list_hasher::equal): Compare them.
2639 (optimize_location_lists): Check whether a view list symbol is
2640 needed, and whether the locview attribute is present, and
2641 whether they match. Remove the locview attribute if no longer
2643 (index_location_lists): Call skip_loc_list_entry for test.
2644 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2645 Use output_dwarf_version.
2646 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2647 (struct dw_val_node): Add val_view_list.
2648 * final.c (SEEN_NEXT_VIEW): New.
2649 (set_next_view_needed): New.
2650 (clear_next_view_needed): New.
2651 (maybe_output_next_view): New.
2652 (final_start_function): Rename to...
2653 (final_start_function_1): ... this. Take pointer to FIRST,
2654 add SEEN parameter. Emit param bindings in the initial view.
2655 (final_start_function): Reintroduce SEEN-less interface.
2656 (final): Rename to...
2657 (final_1): ... this. Take SEEN parameter. Output final pending
2658 next view at the end.
2659 (final): Reintroduce seen-less interface.
2660 (final_scan_insn): Output pending next view before switching
2661 sections or ending a block. Mark the next view as needed when
2662 outputting variable locations. Notify debug backend of section
2663 changes, and of location view changes.
2664 (rest_of_handle_final): Adjust.
2665 * toplev.c (process_options): Autodetect value for debug variable
2666 location views option. Warn on incompat5 without -gdwarf-5.
2667 * doc/invoke.texi (gvariable-location-views): New.
2668 (gvariable-location-views=incompat5): New.
2669 (gno-variable-location-views): New.
2671 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2673 PR tree-optimization/84136
2674 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2675 that the result of find_edge is non-NULL.
2677 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2680 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2681 storing integer register in SImode. Fix cost of 256 and 512
2682 byte aligned SSE register store.
2684 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2686 * config/i386/i386.c (ix86_multiplication_cost): Fix
2687 multiplication cost for TARGET_AVX512DQ.
2689 2018-02-08 Marek Polacek <polacek@redhat.com>
2691 PR tree-optimization/84238
2692 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2695 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2697 PR tree-optimization/84265
2698 * tree-vect-stmts.c (vectorizable_store): Don't treat
2699 VMAT_CONTIGUOUS accesses as grouped.
2700 (vectorizable_load): Likewise.
2702 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2704 PR tree-optimization/81635
2705 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2706 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2707 (test_round_for_mask): New functions.
2708 (wide_int_cc_tests): Call test_round_for_mask.
2709 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2710 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2711 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2712 range returned by get_range_info.
2714 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2717 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2718 * symtab.c: Include builtins.h
2719 (symtab_node::output_to_lto_symbol_table_p): Move here
2720 from lto-streamer-out.c:output_symbol_p.
2721 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2722 (output_symbol_p): Move all logic to symtab.c
2723 (produce_symtab): Update.
2725 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2727 * config/s390/s390-opts.h (enum indirect_branch): Define.
2728 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2729 (s390_indirect_branch_via_thunk)
2730 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2731 (enum s390_indirect_branch_type): Define.
2732 * config/s390/s390.c (struct s390_frame_layout, struct
2733 machine_function): Remove.
2734 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2735 (indirect_branch_table_label_no, indirect_branch_table_name):
2737 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2738 (enum s390_indirect_branch_option): Define.
2739 (s390_return_addr_from_memory): New function.
2740 (s390_handle_string_attribute): New function.
2741 (s390_attribute_table): Add new attribute handler.
2742 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2743 (s390_indirect_branch_via_thunk): New function.
2744 (s390_indirect_branch_via_inline_thunk): New function.
2745 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2746 sibling call optimization for non z10 compiles.
2747 (s390_emit_call): Force indirect branch target to be a single
2748 register. Add r1 clobber for non-z10 compiles.
2749 (s390_emit_epilogue): Emit return jump via return_use expander.
2750 (s390_reorg): Handle JUMP_INSNs as execute targets.
2751 (s390_option_override_internal): Perform validity checks for the
2752 new command line options.
2753 (s390_indirect_branch_attrvalue): New function.
2754 (s390_indirect_branch_settings): New function.
2755 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2756 (s390_output_indirect_thunk_function): New function.
2757 (s390_code_end): Implement target hook.
2758 (s390_case_values_threshold): Implement target hook.
2759 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2761 * config/s390/s390.h (struct s390_frame_layout)
2762 (struct machine_function): Move here from s390.c.
2763 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2764 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2765 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2766 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2767 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2768 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2769 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2770 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2771 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2772 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2773 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2774 (mnemonic attribute): Add values which aren't recognized
2776 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2777 pattern for branch conversion. Fix mnemonic attribute.
2778 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2779 indirect branch via thunk if requested.
2780 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2781 ("*indirect_jump"): Disable for branch conversion using out of
2783 ("indirect_jump_via_thunk<mode>_z10")
2784 ("indirect_jump_via_thunk<mode>")
2785 ("indirect_jump_via_inlinethunk<mode>_z10")
2786 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2787 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2788 ("casesi_jump_via_inlinethunk<mode>_z10")
2789 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2790 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2791 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2792 ("*indirect2_jump"): Disable for branch conversion.
2793 ("casesi_jump"): Turn into expander and expand patterns for branch
2795 ("return_use"): New expander.
2796 ("*return"): Emit return via thunk and rename it to ...
2797 ("*return<mode>"): ... this one.
2798 * config/s390/s390.opt: Add new options and and enum for the
2801 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2803 * lra-constraints.c (match_reload): Unconditionally use
2804 gen_lowpart_SUBREG, rather than selecting between that
2805 and equivalent gen_rtx_SUBREG code.
2807 2018-02-08 Richard Biener <rguenther@suse.de>
2809 PR tree-optimization/84233
2810 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2811 changed flag instead of boguously re-using phi_inserted.
2813 2018-02-08 Martin Jambor <mjambor@suse.cz>
2815 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2816 static local variables.
2818 2018-02-08 Richard Biener <rguenther@suse.de>
2820 PR tree-optimization/84278
2821 * tree-vect-stmts.c (vectorizable_store): When looking for
2822 smaller vector types to perform grouped strided loads/stores
2823 make sure the mode is supported by the target.
2824 (vectorizable_load): Likewise.
2826 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2828 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2829 Increase LDP/STP opportunities by adding adjacent callee-saves.
2831 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2833 PR rtl-optimization/84068
2834 PR rtl-optimization/83459
2835 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2837 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2839 PR tree-optimization/84224
2840 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2841 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2844 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2847 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2848 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2849 count, remove one USE.
2851 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2853 * doc/install.texi (Configuration): Document the
2854 --with-long-double-format={ibm,ieee} PowerPC configuration
2858 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2859 Convert from define_expand to be define_insn_and_split. Rework
2860 float/double/_Float128 conversions to QI/HI/SImode to work with
2861 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2862 conversions to QI/HImode types did a store and then a load to
2863 truncate the value. For conversions to VSX registers, don't split
2864 the insn, instead emit the code directly. Use the code iterator
2865 any_fix to combine signed and unsigned conversions.
2866 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2867 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2868 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2869 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2870 (fix_<mode>di2_hw): Likewise.
2871 (fixuns_<mode>di2_hw): Likewise.
2872 (fix_<mode>si2_hw): Likewise.
2873 (fixuns_<mode>si2_hw): Likewise.
2874 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2875 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2876 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2877 fix<uns>_trunc<SFDF:mode>si2_p8.
2878 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2880 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2881 (fix<uns>_<mode>_mem): Likewise.
2882 (fctiw<u>z_<mode>_mem): Likewise.
2883 (fix<uns>_<mode>_mem): Likewise.
2884 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2885 the register allocator from doing a direct move to the GPRs to do
2886 a store, and instead use the ISA 3.0 store byte/half-word from
2887 vector register instruction. For IEEE 128-bit floating point,
2888 also optimize stores of 32-bit ints.
2889 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2891 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2893 * genextract.c (push_pathstr_operand): New function to support
2895 (walk_rtx): Call push_pathstr_operand.
2896 (print_path): Support [a-zA-Z].
2898 2018-02-07 Richard Biener <rguenther@suse.de>
2900 PR tree-optimization/84037
2901 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2902 (cse_and_gimplify_to_preheader): Declare.
2903 (vect_get_place_in_interleaving_chain): Likewise.
2904 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2906 (_loop_vec_info::~_loop_vec_info): Delete it.
2907 (cse_and_gimplify_to_preheader): New function.
2908 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2909 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2910 (vectorizable_load): Likewise. For grouped stores always base
2911 the IV on the first element.
2912 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2913 condition before gimplifying.
2915 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2917 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2918 *DIV_EXPR and *MOD_EXPR.
2920 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2923 * config/i386/i386.c (ix86_option_override_internal): Mask out
2924 the CF_SET bit when checking -fcf-protection.
2926 2018-02-07 Tom de Vries <tom@codesourcery.com>
2929 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2932 2018-02-07 Richard Biener <rguenther@suse.de>
2934 PR tree-optimization/84204
2935 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2938 PR tree-optimization/84205
2939 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2940 special-case isl_ast_op_zdiv_r.
2942 PR tree-optimization/84223
2943 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2944 Only add conditions from within the region.
2945 (gather_bbs::after_dom_children): Adjust.
2947 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2950 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2951 * config/avr/avr.md: Only post-reload split REG-REG moves if
2952 either register is GENERAL_REG_P.
2954 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2956 PR tree-optimization/84235
2957 * tree-ssa-scopedtables.c
2958 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2959 if the subtraction is performed in floating point type where NaNs are
2960 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2961 build 1. Formatting fix.
2963 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2966 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2967 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2968 and skip it regardless of bb boundaries. Use CALL_P macro,
2969 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2970 unnecessarily, formatting fix.
2972 2018-02-06 Michael Collison <michael.collison@arm.com>
2974 * config/arm/thumb2.md:
2975 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2976 (*thumb_mov_notscc): Ditto.
2978 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2981 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2982 unsigned_fix, not "s".
2984 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2986 * configure.ac (gcc_fn_eh_frame_ro): New function.
2987 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2988 correct .eh_frame permissions.
2989 * configure: Regenerate.
2991 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2993 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2996 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2998 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2999 Display warning message for -mno-speculate-indirect-jumps.
3001 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3003 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3005 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3007 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3009 PR tree-optimization/84225
3010 * tree-eh.c (find_trapping_overflow): Only call
3011 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3013 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3016 * config/i386/i386.c: Reimplement the check of possible options
3017 -mibt/-mshstk conbination. Change error messages.
3018 * doc/invoke.texi: Fix a typo: remove extra '='.
3020 2018-02-06 Marek Polacek <polacek@redhat.com>
3022 PR tree-optimization/84228
3023 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3025 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3028 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3029 emitted arch directives.
3030 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3031 __ARM_FEATURE_COPROC before changing architectures.
3033 2018-02-06 Richard Biener <rguenther@suse.de>
3035 * config/i386/i386.c (print_reg): Fix typo.
3036 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3038 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3040 * configure: Regenerate.
3042 2018-02-05 Martin Sebor <msebor@redhat.com>
3044 PR tree-optimization/83369
3045 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3048 2018-02-05 Martin Liska <mliska@suse.cz>
3050 * doc/invoke.texi: Cherry-pick upstream r323995.
3052 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3054 * ira.c (ira_init_register_move_cost): Adjust comment.
3056 2018-02-05 Martin Liska <mliska@suse.cz>
3058 PR gcov-profile/84137
3059 * doc/gcov.texi: Fix typo in documentation.
3061 2018-02-05 Martin Liska <mliska@suse.cz>
3063 PR gcov-profile/83879
3064 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3065 using dlopen functionality.
3067 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3069 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3070 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3071 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3072 _mm_maskz_range_round_ss): New intrinsics.
3073 (__builtin_ia32_rangesd128_round)
3074 (__builtin_ia32_rangess128_round): Remove.
3075 (__builtin_ia32_rangesd128_mask_round,
3076 __builtin_ia32_rangess128_mask_round): New builtins.
3077 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3078 __builtin_ia32_rangess128_round): Remove.
3079 (__builtin_ia32_rangesd128_mask_round,
3080 __builtin_ia32_rangess128_mask_round): New builtins.
3081 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3082 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3083 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3084 "<round_saeonly_constraint>")): Changed to ...
3085 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3086 "<round_saeonly_scalar_constraint>")): ... this.
3087 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3088 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3089 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3090 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3091 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3093 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3095 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3097 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3098 Remove all values except native, 8540 and 8548.
3100 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3102 * config/i386/i386.c (ix86_output_function_return): Pass
3103 INVALID_REGNUM, instead of -1, as invalid register number to
3104 indirect_thunk_name and output_indirect_thunk.
3106 2018-02-02 Julia Koval <julia.koval@intel.com>
3108 * config.gcc: Add -march=icelake.
3109 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3110 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3111 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3112 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3113 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3114 (processor_target_table): Add icelake.
3115 (ix86_option_override_internal): Handle new PTAs.
3116 (get_builtin_code_for_version): Handle icelake.
3117 (M_INTEL_COREI7_ICELAKE): New.
3118 (fold_builtin_cpu): Handle icelake.
3119 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3120 * doc/invoke.texi: Add -march=icelake.
3122 2018-02-02 Julia Koval <julia.koval@intel.com>
3124 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3125 to wide_int_bitmask.
3126 * wide-int-bitmask.h: New.
3128 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3131 * config/i386/i386.md: Replace Pmode with word_mode in
3132 builtin_setjmp_setup and builtin_longjmp to support x32.
3134 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3138 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3140 (rs6000_supported_cpu_names): New static variable.
3141 (linux_cpu_translation_table): Likewise.
3142 (elf_platform) <cpu>: Define new static variable and use it.
3143 Translate kernel AT_PLATFORM name to canonical name if needed.
3144 Error if platform name is unknown.
3146 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3149 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3151 2018-02-01 Jeff Law <law@redhat.com>
3154 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3155 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3156 the scratch if RELEASE_VIA_POP is false.
3157 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3158 If we have to save a temporary register, decrement SIZE appropriately.
3159 Pass new arguments to release_scratch_register_on_entry.
3160 (ix86_adjust_stack_and_probe): Likewise.
3161 (ix86_emit_probe_stack_range): Pass new arguments to
3162 release_scratch_register_on_entry.
3164 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3166 PR rtl-optimization/84157
3167 * combine.c (change_zero_ext): Use REG_P predicate in
3168 front of HARD_REGISTER_P predicate.
3170 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3172 * config/avr/avr.c (avr_option_override): Move disabling of
3173 -fdelete-null-pointer-checks to...
3174 * common/config/avr/avr-common.c (avr_option_optimization_table):
3177 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3179 PR tree-optimization/81635
3180 * tree-data-ref.c (split_constant_offset_1): For types that
3181 wrap on overflow, try to use range info to prove that wrapping
3184 2018-02-01 Renlin Li <renlin.li@arm.com>
3187 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3189 (aarch64_register_move_cost): Likewise.
3190 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3192 (REG_CLASS_NAMES): Likewise.
3193 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3194 TAILCALL_ADDR_REGS. Remove IP registers.
3195 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3197 2018-02-01 Richard Biener <rguenther@suse.de>
3199 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3200 for specifying RPO order and allow NULL for that.
3201 * domwalk.c (dom_walker::dom_walker): Likewise.
3202 (dom_walker::walk): Handle NULL RPO order.
3203 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3205 (rewrite_update_dom_walker): Likewise.
3206 (mark_def_dom_walker): Likewise.
3208 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3210 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3211 (aarch64_maybe_expand_sve_subreg_move): Declare.
3212 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3213 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3215 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3216 that are semantically a reverse operation.
3217 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3218 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3219 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3221 (aarch64_can_change_mode_class): For big-endian, forbid changes
3222 between two SVE modes if they have different element sizes.
3224 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3226 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3227 the TImode handling for big-endian targets.
3229 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3231 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3232 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3234 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3235 Remove BSWAP handing for big-endian targets and use the form of
3236 LD1RQ appropariate for the mode.
3238 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3240 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3241 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3244 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3247 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3248 check for operands that need to go through aarch64_sve_reload_be.
3250 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3252 PR tree-optimization/81661
3253 PR tree-optimization/84117
3254 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3255 * tree-eh.c: Include gimplify.h.
3256 (find_trapping_overflow, replace_trapping_overflow,
3257 rewrite_to_non_trapping_overflow): New functions.
3258 * tree-vect-loop.c: Include tree-eh.h.
3259 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3260 * tree-data-ref.c: Include tree-eh.h.
3261 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3263 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3265 PR rtl-optimization/84123
3266 * combine.c (change_zero_ext): Check if hard register satisfies
3267 can_change_dest_mode before calling gen_lowpart_SUBREG.
3269 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3272 * ira.c (ira_init_register_move_cost): Remove assert.
3274 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3276 PR rtl-optimization/84071
3277 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3278 * doc/tm.texi: Regenerate.
3280 2018-01-31 Richard Biener <rguenther@suse.de>
3282 PR tree-optimization/84132
3283 * tree-data-ref.c (analyze_miv_subscript): Properly
3284 check whether evolution_function_is_affine_multivariate_p
3285 before calling gcd_of_steps_may_divide_p.
3287 2018-01-31 Julia Koval <julia.koval@intel.com>
3290 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3291 * config/i386/i386.md (rdpid_rex64) New.
3292 (rdpid): Make 32bit only.
3294 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3297 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3298 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3300 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3303 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3305 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3307 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3309 PR rtl-optimization/84071
3310 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3311 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3313 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3315 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3316 (arc_attribute_table): Add 'aux' attribute.
3317 (arc_in_small_data_p): Consider aux like variables.
3318 (arc_is_aux_reg_p): New function.
3319 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3320 (arc_get_aux_arg): New function.
3321 (prepare_move_operands): Handle aux-register access.
3322 (arc_handle_aux_attribute): New function.
3323 * doc/extend.texi (ARC Variable attributes): Add subsection.
3325 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3327 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3328 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3329 (arc_attribute_table): Add 'uncached' attribute.
3330 (arc_print_operand): Print '.di' flag for uncached memory
3332 (arc_in_small_data_p): Do not consider for small data the uncached
3334 (arc_is_uncached_mem_p): New function.
3335 * config/arc/predicates.md (compact_store_memory_operand): Check
3336 for uncached memory accesses.
3337 (nonvol_nonimm_operand): Likewise.
3338 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3340 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3343 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3344 falign-loops=): Add Optimization flag.
3346 2018-01-30 Jeff Law <law@redhat.com>
3349 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3350 INT_REGISTERS_SAVED. Check it prior to calling
3351 get_scratch_register_on_entry.
3352 (ix86_adjust_stack_and_probe): Similarly.
3353 (ix86_emit_probe_stack_range): Similarly.
3354 (ix86_expand_prologue): Corresponding changes.
3356 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3359 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3360 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3362 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3365 * lra-constraints.c (curr_insn_transform): Process AND in the
3368 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3370 PR rtl-optimization/83986
3371 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3372 dependence against last_pending_memory_flush in addition to
3375 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3377 PR tree-optimization/81611
3378 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3381 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3384 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3387 2018-01-30 Richard Biener <rguenther@suse.de>
3388 Jakub Jelinek <jakub@redhat.com>
3390 PR tree-optimization/84111
3391 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3392 inner loops added during recursion, as they don't have up-to-date
3395 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3398 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3399 (can_inline_edge_by_limits_p): ... here.
3400 (can_early_inline_edge_p, check_callers,
3401 update_caller_keys, update_callee_keys, recursive_inlining,
3402 add_new_edges_to_heap, speculation_useful_p,
3403 inline_small_functions,
3404 inline_small_functions, flatten_function,
3405 inline_to_all_callers_1): Update.
3407 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3409 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3412 2018-01-30 Richard Biener <rguenther@suse.de>
3414 PR tree-optimization/83008
3415 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3416 invariant and constant vector uses in stmts when they need
3419 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3422 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3423 * configure: Regenerate.
3425 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3427 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3429 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3430 Use gen_rtx_REG rather than gen_lowpart.
3432 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3434 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3435 rather than 0 when creating partial subregs.
3437 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3439 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3442 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3445 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3446 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3447 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3448 flags. This restores the settings used before the 2017-07-24.
3449 Turning off pre increment/decrement/modify allows IVOPTS to
3450 optimize DF/SF loops where the index is an int.
3452 2018-01-29 Richard Biener <rguenther@suse.de>
3453 Kelvin Nilsen <kelvin@gcc.gnu.org>
3456 * tree-vect-stmts.c (vectorizable_call): Don't call
3457 targetm.vectorize_builtin_md_vectorized_function if callee is
3460 2018-01-22 Carl Love <cel@us.ibm.com>
3462 * doc/extend.tex: Fix typo in second arg in
3463 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3465 2018-01-29 Richard Biener <rguenther@suse.de>
3467 PR tree-optimization/84086
3468 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3469 (flush_ssaname_freelist): When SSA names were released reset
3470 the SCEV hash table.
3472 2018-01-29 Richard Biener <rguenther@suse.de>
3474 PR tree-optimization/84057
3475 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3476 removed paths when removing edges.
3478 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3480 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3481 -mfunction-return=@var{choice}.
3483 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3486 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3488 (layout::print_source_line): Likewise.
3489 (test_get_line_width_without_trailing_whitespace): Add test cases.
3491 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3494 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3497 2018-01-26 Jim Wilson <jimw@sifive.com>
3499 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3501 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3504 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3506 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3507 and CMP + SUB-immediate -> SUBS.
3509 2018-01-26 Martin Sebor <msebor@redhat.com>
3511 PR tree-optimization/83896
3512 * tree-ssa-strlen.c (get_string_len): Rename...
3513 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3514 Avoid assuming length is constant.
3515 (handle_char_store): Use HOST_WIDE_INT for string length.
3517 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3520 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3521 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3523 2018-01-26 Richard Biener <rguenther@suse.de>
3525 PR rtl-optimization/84003
3526 * dse.c (record_store): Only record redundant stores when
3527 the earlier store aliases at least all accesses the later one does.
3529 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3531 PR rtl-optimization/83985
3532 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3533 REG_CFA_RESTORE insns.
3534 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3535 REG_CFA_RESTORE insns here.
3538 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3539 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3541 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3543 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3544 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3545 (arc_init): Likewise.
3546 (arc_override_options): Likewise.
3547 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3549 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3551 * config/arc/arc.h (TARGET_DBNZ): Define.
3552 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3553 properly set the tune attribute.
3554 (dbnz): Use TARGET_DBNZ guard.
3555 * config/arc/arc.opt (mtune): Add core3 option.
3557 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3559 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3560 recognize new pic like addresses.
3561 (arc_delegitimize_address): Clean up.
3563 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3565 * config/arc/arc-arches.def: Option mrf16 valid for all
3567 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3568 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3569 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3570 * config/arc/arc-tables.opt: Regenerate.
3571 * config/arc/arc.c (arc_conditional_register_usage): Handle
3572 reduced register file case.
3573 (arc_file_start): Set must have build attributes.
3574 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3576 * config/arc/arc.opt (mrf16): Add new option.
3577 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3578 * config/arc/genmultilib.awk: Handle new mrf16 option.
3579 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3580 * config/arc/t-multilib: Regenerate.
3581 * doc/invoke.texi (ARC Options): Document mrf16 option.
3583 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3585 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3586 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3587 (arc_attribute_table): Add 'secure_call' attribute.
3588 (arc_print_operand): Print secure call operand.
3589 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3591 (arc_is_secure_call_p): New function. * config/arc/arc.md
3592 (call_i): Add support for sjli instruction.
3593 (call_value_i): Likewise.
3594 * config/arc/constraints.md (Csc): New constraint.
3596 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3597 John Eric Martin <John.Martin@emmicro-us.com>
3599 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3600 * config/arc/arc.c (_arc_jli_section): New struct.
3601 (arc_jli_section): New type.
3602 (rc_jli_sections): New static variable.
3603 (arc_handle_jli_attribute): New function.
3604 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3605 (arc_file_end): New function.
3606 (TARGET_ASM_FILE_END): Define.
3607 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3608 (arc_add_jli_section): New function.
3609 (jli_call_scan): Likewise.
3610 (arc_reorg): Call jli_call_scan.
3611 (arc_output_addsi): Remove 'S' from printing asm operand.
3612 (arc_is_jli_call_p): New function.
3613 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3615 (movhi_insn): Likewise.
3616 (movsi_insn): Likewise.
3617 (movsi_set_cc_insn): Likewise.
3618 (loadqi_update): Likewise.
3619 (load_zeroextendqisi_update): Likewise.
3620 (load_signextendqisi_update): Likewise.
3621 (loadhi_update): Likewise.
3622 (load_zeroextendhisi_update): Likewise.
3623 (load_signextendhisi_update): Likewise.
3624 (loadsi_update): Likewise.
3625 (loadsf_update): Likewise.
3626 (movsicc_insn): Likewise.
3627 (bset_insn): Likewise.
3628 (bxor_insn): Likewise.
3629 (bclr_insn): Likewise.
3630 (bmsk_insn): Likewise.
3631 (bicsi3_insn): Likewise.
3632 (cmpsi_cc_c_insn): Likewise.
3633 (movsi_ne): Likewise.
3634 (movsi_cond_exec): Likewise.
3635 (clrsbsi2): Likewise.
3646 (call_i): Remove 'S' asm letter, add jli instruction.
3647 (call_value_i): Likewise.
3648 * config/arc/arc.op (mjli-always): New option.
3649 * config/arc/constraints.md (Cji): New constraint.
3650 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3652 (subsf3_fpx): Likewise.
3653 (mulsf3_fpx): Likewise.
3654 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3656 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3658 * doc/invoke.texi (ARC): Document mjli-always option.
3660 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3662 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3663 avoid addition with 0 and use incw and decw where possible.
3665 2018-01-26 Richard Biener <rguenther@suse.de>
3667 PR tree-optimization/81082
3668 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3669 association if it requires casting to unsigned.
3670 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3671 from fold_plusminus_mult_expr to catch important cases late when
3672 range info is available.
3674 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3676 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3677 * configure.ac (hidden_linkonce): New test.
3678 * configure: Regenerate.
3679 * config.in: Regenerate.
3681 2018-01-26 Julia Koval <julia.koval@intel.com>
3683 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3684 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3685 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3686 _mm_mask_bitshuffle_epi64_mask): Fix type.
3687 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3688 USI_FTYPE_V4DI_V4DI_USI): Remove.
3689 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3690 __builtin_ia32_vpshufbitqmb256_mask,
3691 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3692 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3693 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3695 2018-01-26 Alan Modra <amodra@gmail.com>
3698 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3699 UNSPEC_VBPERMQ. Sort other unspecs.
3701 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3703 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3705 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3708 * predict.c (drop_profile): Do not push/pop cfun; update also
3710 (handle_missing_profiles): Fix logic looking for zero profiles.
3712 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3715 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3716 on functions with #pragma omp declare simd or functions with simd
3718 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3719 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3720 Remove trailing \n from warning_at calls.
3722 2018-01-25 Tom de Vries <tom@codesourcery.com>
3725 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3726 for neutered workers.
3728 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3731 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3732 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3734 2018-01-24 Jeff Law <law@redhat.com>
3737 * i386.c (get_probe_interval): Move to earlier point.
3738 (ix86_compute_frame_layout): If -fstack-clash-protection and
3739 the frame is larger than the probe interval, then use pushes
3740 to save registers rather than reg->mem moves.
3741 (ix86_expand_prologue): Remove conditional for int_registers_saved
3744 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3747 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3748 min/max for never referenced object.
3750 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3753 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3755 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3756 attributes from DECL_ATTRIBUTES (decl) without affecting
3757 DECL_ATTRIBUTES (current_function_decl).
3758 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3759 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3761 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3763 PR tree-optimization/83979
3764 * fold-const.c (fold_comparison): Use constant_boolean_node
3765 instead of boolean_{true,false}_node.
3767 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3769 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3772 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3774 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3775 Simplify the clause that sets the length attribute.
3776 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3777 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3778 clause that sets the length attribute.
3779 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3781 2018-01-24 Tom de Vries <tom@codesourcery.com>
3784 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3785 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3786 Add strict parameter.
3787 (prevent_branch_around_nothing): Insert dummy insn between branch to
3788 label and label with no ptx insn inbetween.
3789 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3791 2018-01-24 Tom de Vries <tom@codesourcery.com>
3794 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3795 for neutered threads in warp.
3796 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3798 2018-01-24 Richard Biener <rguenther@suse.de>
3800 PR tree-optimization/83176
3801 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3804 2018-01-24 Richard Biener <rguenther@suse.de>
3806 PR tree-optimization/82819
3807 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3808 code generating pluses that are no-ops in the target precision.
3810 2018-01-24 Richard Biener <rguenther@suse.de>
3813 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3815 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3817 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3818 to merge probabilities.
3819 * predict.c (probably_never_executed): Also mark as cold functions
3820 with global 0 profile and guessed local profile.
3821 * profile-count.c (profile_probability::combine_with_count): New
3823 * profile-count.h (profile_probability::operator*,
3824 profile_probability::operator*=, profile_probability::operator/,
3825 profile_probability::operator/=): Reduce precision to adjusted
3826 and set value to guessed on contradictory divisions.
3827 (profile_probability::combine_with_freq): Remove.
3828 (profile_probability::combine_wiht_count): Declare.
3829 (profile_count::force_nonzero):: Set to adjusted.
3830 (profile_count::probability_in):: Set quality to adjusted.
3831 * tree-ssa-tail-merge.c (replace_block_by): Use
3834 2018-01-23 Andrew Waterman <andrew@sifive.com>
3835 Jim Wilson <jimw@sifive.com>
3837 * config/riscv/riscv.c (riscv_stack_boundary): New.
3838 (riscv_option_override): Set riscv_stack_boundary. Handle
3839 riscv_preferred_stack_boundary_arg.
3840 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3841 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3842 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3843 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3844 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3845 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3847 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3850 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3851 of struct ix86_frame.
3852 (ix86_expand_epilogue): Likewise. Add a local variable for
3853 the reg_save_offset field in struct ix86_frame.
3855 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3857 PR tree-optimization/82604
3858 * tree-loop-distribution.c (enum partition_kind): New enum item
3859 PKIND_PARTIAL_MEMSET.
3860 (partition_builtin_p): Support above new enum item.
3861 (generate_code_for_partition): Ditto.
3862 (compute_access_range): Differentiate cases that equality can be
3863 proven at all loops, the innermost loops or no loops.
3864 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3865 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3866 (finalize_partitions, distribute_loop): Don't fuse partition of
3867 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3868 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3871 2018-01-23 Martin Liska <mliska@suse.cz>
3873 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3874 order to ignore the predictor.
3875 (PRED_POLYMORPHIC_CALL): Likewise.
3876 (PRED_RECURSIVE_CALL): Likewise.
3878 2018-01-23 Martin Liska <mliska@suse.cz>
3880 * tree-profile.c (tree_profiling): Print function header to
3881 aware reader which function we are working on.
3882 * value-prof.c (gimple_find_values_to_profile): Do not print
3883 not interesting value histograms.
3885 2018-01-23 Martin Liska <mliska@suse.cz>
3887 * profile-count.h (enum profile_quality): Add
3888 profile_uninitialized as the first value. Do not number values
3889 as they are zero based.
3890 (profile_count::verify): Update sanity check.
3891 (profile_probability::verify): Likewise.
3893 2018-01-23 Nathan Sidwell <nathan@acm.org>
3895 * doc/invoke.texi (ffor-scope): Deprecate.
3897 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3899 PR tree-optimization/83510
3900 * domwalk.c (set_all_edges_as_executable): New function.
3901 (dom_walker::dom_walker): Convert bool param
3902 "skip_unreachable_blocks" to enum reachability. Move setup of
3903 edge flags to set_all_edges_as_executable and only do it when
3904 reachability is REACHABLE_BLOCKS.
3905 * domwalk.h (enum dom_walker::reachability): New enum.
3906 (dom_walker::dom_walker): Convert bool param
3907 "skip_unreachable_blocks" to enum reachability.
3908 (set_all_edges_as_executable): New decl.
3909 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3910 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3912 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3913 but converting true to REACHABLE_BLOCKS.
3914 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3916 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3917 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3918 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3920 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3921 if check_all_array_refs will be called.
3923 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3925 * tree.c (selftest::test_location_wrappers): Add more test
3928 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3930 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3931 (selftest::test_bit_in_range): Likewise.
3933 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3936 * doc/sourcebuild.texi (vect_float): Say that the selector
3937 only describes the situation when -funsafe-math-optimizations is on.
3938 (vect_float_strict): Document.
3940 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3942 PR tree-optimization/83965
3943 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3944 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3945 instead of checking only for a reduction.
3946 (vect_recog_widen_sum_pattern): Likewise.
3948 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3950 * predict.c (probably_never_executed): Only use precise profile info.
3951 (compute_function_frequency): Skip after inlining hack since we now
3952 have quality checking.
3954 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3956 * profile-count.h (profile_probability::very_unlikely,
3957 profile_probability::unlikely, profile_probability::even): Set
3958 precision to guessed.
3960 2018-01-23 Richard Biener <rguenther@suse.de>
3962 PR tree-optimization/83963
3963 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3964 Properly terminate dominator walk when crossing the exit edge not
3965 when visiting its source block.
3967 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3970 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3971 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3973 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3975 PR tree-optimization/83957
3976 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3977 semicolon after for body surrounded by braces.
3979 PR tree-optimization/83081
3980 * profile-count.h (profile_probability::split): New method.
3981 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3982 Use profile_probability::split.
3983 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3984 when splitting a single conditional jump into 2.
3986 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3988 PR tree-optimization/69452
3989 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3994 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3995 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3996 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3998 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4000 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4001 * config/rl78/rl78.md (movdi): New define_expand.
4002 * config/rl78/rl78.c (rl78_split_movdi): New function.
4004 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4007 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4009 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4010 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4011 128-bit to produce an UNSPEC move to get the double word with the
4012 signbit and then a shift directly to do signbit.
4013 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4014 implementation with a new version that just does either a direct
4015 move or a regular move. Move memory interface to separate insns.
4016 Move insns so they are next to the expander.
4017 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4018 with signbit move. Split big and little endian case.
4019 (signbit<mode>2_dm_mem_le): Likewise.
4020 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4021 (signbit<mode>2_dm2): Likewise.
4023 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4025 * config/rl78/rl78.md (anddi3): New define_expand.
4027 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4029 * config/rl78/rl78.md (umindi3): New define_expand.
4031 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4033 * config/rl78/rl78.md (smindi3): New define_expand.
4035 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4037 * config/rl78/rl78.md (smaxdi3): New define_expand.
4039 2018-01-22 Carl Love <cel@us.ibm.com>
4041 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4042 LVX_V1TI): Add macro expansion.
4043 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4044 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4045 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4046 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4047 Change check to determine if the instruction is a byte reversing
4048 entry. Fix typo in comment.
4049 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4050 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4051 Add def_builtin calls for new builtins.
4052 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4053 Add define_insn expansion.
4055 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4057 * config/rl78/rl78.md (umaxdi3): New define_expand.
4059 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4061 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4062 for non-QImode registers.
4064 2018-01-22 Richard Biener <rguenther@suse.de>
4066 PR tree-optimization/83963
4067 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4068 including the loop exit block.
4069 (scop_detection::merge_sese): Likewise.
4070 (scop_detection::add_scop): Do it here instead.
4072 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4074 * doc/sourcebuild.texi (arm_softfloat): Document.
4076 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4079 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4080 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4081 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4083 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4084 David Edelsohn <dje.gcc@gmail.com>
4087 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4088 Change "crset eq" to "crset 2".
4089 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4090 (*call_indirect_aix<mode>_nospec): Likewise.
4091 (*call_value_indirect_aix<mode>_nospec): Likewise.
4092 (*call_indirect_elfv2<mode>_nospec): Likewise.
4093 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4094 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4095 change assembly output from . to $.
4096 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4097 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4098 (*tablejump<mode>_internal1_nospec): Likewise.
4100 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4103 * config/sh/sh_optimize_sett_clrt.cc:
4104 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4106 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4108 PR tree-optimization/83940
4109 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4110 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4111 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4112 use it to pass back the definition type.
4113 (vect_check_store_rhs): Likewise rhs_dt_out.
4114 (vect_build_gather_load_calls): Add a mask_dt argument and use
4115 it instead of a call to vect_is_simple_use.
4116 (vectorizable_store): Update calls to vect_check_load_store_mask
4117 and vect_check_store_rhs. Use the dt returned by the latter instead
4118 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4119 instead of calls to vect_is_simple_use. Pass the scalar rather
4120 than the vector operand to vect_is_simple_use when handling
4121 second and subsequent copies of an rhs value.
4122 (vectorizable_load): Update calls to vect_check_load_store_mask
4123 and vect_build_gather_load_calls. Use the cached mask_dt and
4124 gs_info.offset_dt instead of calls to vect_is_simple_use.
4126 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4129 * tree-emutls.c: Include gimplify.h.
4130 (lower_emutls_2): New function.
4131 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4132 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4133 it before further processing.
4136 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4137 UINTVAL (trueop1) instead of INTVAL (op1).
4139 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4143 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4144 INCOMING_FRAME_SP_OFFSET if not defined.
4145 (scan_trace): Add ENTRY argument. If true and
4146 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4147 emit a note to adjust the CFA offset.
4148 (create_cfi_notes): Adjust scan_trace callers.
4149 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4150 INCOMING_FRAME_SP_OFFSET in the CIE.
4151 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4152 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4154 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4155 * doc/tm.texi: Regenerated.
4157 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4159 PR rtl-optimization/83147
4160 * lra-constraints.c (remove_inheritance_pseudos): Use
4161 lra_substitute_pseudo_within_insn.
4163 2018-01-19 Tom de Vries <tom@codesourcery.com>
4164 Cesar Philippidis <cesar@codesourcery.com>
4167 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4169 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4172 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4173 spaces for function labels.
4175 2018-01-19 Martin Liska <mliska@suse.cz>
4177 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4178 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4179 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4180 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4181 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4182 (PRED_CONST_RETURN): Change from 69 to 65.
4183 (PRED_NULL_RETURN): Change from 91 to 71.
4184 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4185 (PRED_LOOP_GUARD): Change from 66 to 73.
4187 2018-01-19 Martin Liska <mliska@suse.cz>
4189 * predict.c (predict_insn_def): Add new assert.
4190 (struct branch_predictor): Change type to signed integer.
4191 (test_prediction_value_range): Amend test to cover
4193 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4194 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4195 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4196 (PRED_LOOP_IV_COMPARE): Likewise.
4197 * predict.h (PROB_UNINITIALIZED): Define new constant.
4199 2018-01-19 Martin Liska <mliska@suse.cz>
4201 * predict.c (dump_prediction): Add new format for
4202 analyze_brprob.py script which is enabled with -details
4204 * profile-count.h (precise_p): New function.
4206 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4208 PR tree-optimization/83922
4209 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4210 there are no statements that need masking.
4211 (vect_active_double_reduction_p): New function.
4212 (vect_analyze_loop_operations): Use it when handling phis that
4213 are not in the loop header.
4215 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4217 PR tree-optimization/83914
4218 * tree-vect-loop.c (vectorizable_induction): Don't convert
4219 init_expr or apply the peeling adjustment for inductions
4220 that are nested within the vectorized loop.
4222 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4224 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4227 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4231 * function.h (gimplify_parameters): Add gimple_seq * argument.
4232 * function.c: Include gimple.h and options.h.
4233 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4234 for the added local temporaries if needed.
4235 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4236 if there are any parameter cleanups, wrap whole body into a
4237 try/finally with the cleanups.
4239 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4242 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4243 Use GET_MODE_CLASS for scalar floating point.
4245 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4249 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4250 Fix call of call_cgraph_insertion_hooks.
4252 2018-01-18 Martin Sebor <msebor@redhat.com>
4254 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4256 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4259 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4262 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4265 * common.opt: (-ffile-prefix-map): New option.
4266 * opts.c (common_handle_option): Defer it.
4267 * opts-global.c (handle_common_deferred_options): Handle it.
4268 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4269 * file-prefix-map.h: New file.
4270 (remap_debug_filename, add_debug_prefix_map): ...here.
4271 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4272 * final.c (debug_prefix_map, add_debug_prefix_map
4273 remap_debug_filename): Move to...
4274 * file-prefix-map.c: New file.
4275 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4276 generalize, get rid of alloca(), use strrchr() instead of strchr().
4277 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4278 Implement in terms of add_prefix_map().
4279 (remap_macro_filename, remap_debug_filename): Implement in term of
4281 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4282 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4283 * dbxout.c: Include file-prefix-map.h.
4284 * varasm.c: Likewise.
4285 * vmsdbgout.c: Likewise.
4286 * xcoffout.c: Likewise.
4287 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4288 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4289 * doc/invoke.texi (-ffile-prefix-map): Document.
4290 (-fdebug-prefix-map): Update description.
4292 2018-01-18 Martin Liska <mliska@suse.cz>
4294 * config/i386/i386.c (indirect_thunk_name): Document that also
4296 (output_indirect_thunk): Document why both instructions
4297 (pause and lfence) are generated.
4299 2018-01-18 Richard Biener <rguenther@suse.de>
4301 PR tree-optimization/83887
4302 * graphite-scop-detection.c
4303 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4304 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4305 (scop_detection::merge_sese): Re-implement with a flood-fill
4306 algorithm that properly finds a SESE region if it exists.
4308 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4311 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4312 pointer_diff optimizations use view_convert instead of convert.
4314 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4316 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4317 Generate different code for -mno-speculate-indirect-jumps.
4318 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4319 (*call_indirect_aix<mode>): Disable for
4320 -mno-speculate-indirect-jumps.
4321 (*call_indirect_aix<mode>_nospec): New define_insn.
4322 (*call_value_indirect_aix<mode>): Disable for
4323 -mno-speculate-indirect-jumps.
4324 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4325 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4326 -mno-speculate-indirect-jumps.
4327 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4329 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4331 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4332 long double type, set the flags for noting the default long double
4333 type, even if we don't pass or return a long double type.
4335 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4338 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4341 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4343 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4344 support for merge[hl].
4345 (fold_mergehl_helper): New helper function.
4346 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4347 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4348 (altivec_vmrglw_direct): Add xxmrglw insn.
4350 2018-01-17 Andrew Waterman <andrew@sifive.com>
4352 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4353 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4355 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4358 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4359 call the lto_location_cache before reading the
4360 DECL_SOURCE_LOCATION of the types.
4362 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4363 Richard Sandiford <richard.sandiford@linaro.org>
4365 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4366 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4367 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4368 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4369 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4371 * config/aarch64/constraints.md (aarch64_movti_operand):
4373 * config/aarch64/predicates.md (Uti): Add new constraint.
4375 2018-01-17 Carl Love <cel@us.ibm.com>
4377 * config/rs6000/vsx.md (define_expand xl_len_r,
4378 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4379 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4381 (define_expand, define_insn): Move the shift left from the
4382 define_insn to the define_expand for lxvl and stxvl instructions.
4383 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4384 and XL_LEN_R definitions to PURE.
4386 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4388 * config/i386/i386.c (indirect_thunk_name): Declare regno
4389 as unsigned int. Compare regno with INVALID_REGNUM.
4390 (output_indirect_thunk): Ditto.
4391 (output_indirect_thunk_function): Ditto.
4392 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4393 in the call to output_indirect_thunk_function.
4395 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4398 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4399 rather than the size of inner_type to determine the stack slot size
4400 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4402 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4405 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4408 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4410 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4411 endian Linux systems to optionally enable multilibs for selecting
4412 the long double type if the user configured an explicit type.
4413 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4414 have no long double multilibs if not defined.
4415 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4416 warn if the user used -mabi={ieee,ibm}longdouble and we built
4417 multilibs for long double.
4418 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4419 appropriate multilib option.
4420 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4422 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4423 for building long double multilibs.
4424 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4426 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4428 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4431 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4433 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4436 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4439 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4442 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4444 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4445 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4447 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4449 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4450 different rtl trees depending on TARGET_64BIT.
4451 (rs6000_gen_lvx): Likewise.
4453 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4455 * config/visium/visium.md (nop): Tweak comment.
4456 (hazard_nop): Likewise.
4458 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4460 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4461 -mspeculate-indirect-jumps.
4462 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4463 for -mno-speculate-indirect-jumps.
4464 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4465 (*call_value_indirect_elfv2<mode>): Disable for
4466 -mno-speculate-indirect-jumps.
4467 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4468 (indirect_jump): Emit different RTL for
4469 -mno-speculate-indirect-jumps.
4470 (*indirect_jump<mode>): Disable for
4471 -mno-speculate-indirect-jumps.
4472 (*indirect_jump<mode>_nospec): New define_insn.
4473 (tablejump): Emit different RTL for
4474 -mno-speculate-indirect-jumps.
4475 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4476 (tablejumpsi_nospec): New define_expand.
4477 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4478 (tablejumpdi_nospec): New define_expand.
4479 (*tablejump<mode>_internal1): Disable for
4480 -mno-speculate-indirect-jumps.
4481 (*tablejump<mode>_internal1_nospec): New define_insn.
4482 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4485 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4487 * caller-save.c (insert_save): Drop unnecessary parameter. All
4490 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4491 Richard Biener <rguenth@suse.de>
4494 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4495 return early, inline manually is_gimple_sizepos. Make sure if we
4496 call gimplify_expr we don't end up with a gimple constant.
4497 * tree.c (variably_modified_type_p): Don't return true for
4498 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4499 * gimplify.h (is_gimple_sizepos): Remove.
4501 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4503 PR tree-optimization/83857
4504 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4505 vectorizable_live_operation for pure SLP statements.
4506 (vectorizable_live_operation): Handle PHIs.
4508 2018-01-16 Richard Biener <rguenther@suse.de>
4510 PR tree-optimization/83867
4511 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4512 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4514 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4517 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4518 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4519 If off is not INTEGER_CST, issue a may not be aligned warning
4520 rather than isn't aligned. Use isn%'t rather than isn't.
4521 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4523 <case MULT_EXPR>: Improve the case when bottom and one of the
4524 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4525 operand, in that case check if the other operand is multiple of
4526 bottom divided by the INTEGER_CST operand.
4528 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4531 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4532 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4533 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4534 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4535 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4536 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4537 (pa_function_arg_size): New function.
4539 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4541 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4542 in a separate statement.
4544 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4546 PR tree-optimization/83847
4547 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4548 group gathers and scatters.
4550 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4552 PR rtl-optimization/86620
4553 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4555 PR rtl-optimization/83213
4556 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4557 to last if both are JUMP_INSNs.
4559 PR tree-optimization/83843
4560 * gimple-ssa-store-merging.c
4561 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4562 store_immediate_info for bswap/nop orig_stores.
4564 2018-01-15 Andrew Waterman <andrew@sifive.com>
4566 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4568 <UDIV>: Increase cost if !TARGET_DIV.
4570 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4572 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4573 (define_attr "cr_logical_3op"): New.
4574 (cceq_ior_compare): Adjust.
4575 (cceq_ior_compare_complement): Adjust.
4576 (*cceq_rev_compare): Adjust.
4577 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4578 (is_cracked_insn): Adjust.
4579 (insn_must_be_first_in_group): Adjust.
4580 * config/rs6000/40x.md: Adjust.
4581 * config/rs6000/440.md: Adjust.
4582 * config/rs6000/476.md: Adjust.
4583 * config/rs6000/601.md: Adjust.
4584 * config/rs6000/603.md: Adjust.
4585 * config/rs6000/6xx.md: Adjust.
4586 * config/rs6000/7450.md: Adjust.
4587 * config/rs6000/7xx.md: Adjust.
4588 * config/rs6000/8540.md: Adjust.
4589 * config/rs6000/cell.md: Adjust.
4590 * config/rs6000/e300c2c3.md: Adjust.
4591 * config/rs6000/e500mc.md: Adjust.
4592 * config/rs6000/e500mc64.md: Adjust.
4593 * config/rs6000/e5500.md: Adjust.
4594 * config/rs6000/e6500.md: Adjust.
4595 * config/rs6000/mpc.md: Adjust.
4596 * config/rs6000/power4.md: Adjust.
4597 * config/rs6000/power5.md: Adjust.
4598 * config/rs6000/power6.md: Adjust.
4599 * config/rs6000/power7.md: Adjust.
4600 * config/rs6000/power8.md: Adjust.
4601 * config/rs6000/power9.md: Adjust.
4602 * config/rs6000/rs64.md: Adjust.
4603 * config/rs6000/titan.md: Adjust.
4605 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4607 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4608 ix86_indirect_branch_register logic.
4610 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4612 * config/i386/constraints.md (Bs): Update
4613 ix86_indirect_branch_register check. Don't check
4614 ix86_indirect_branch_register with GOT_memory_operand.
4616 * config/i386/predicates.md (GOT_memory_operand): Don't check
4617 ix86_indirect_branch_register here.
4618 (GOT32_symbol_operand): Likewise.
4620 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4622 * config/i386/predicates.md (constant_call_address_operand):
4623 Rewrite ix86_indirect_branch_register logic.
4624 (sibcall_insn_operand): Likewise.
4626 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4628 * config/i386/constraints.md (Bs): Replace
4629 ix86_indirect_branch_thunk_register with
4630 ix86_indirect_branch_register.
4632 * config/i386/i386.md (indirect_jump): Likewise.
4633 (tablejump): Likewise.
4634 (*sibcall_memory): Likewise.
4635 (*sibcall_value_memory): Likewise.
4636 Peepholes of indirect call and jump via memory: Likewise.
4637 * config/i386/i386.opt: Likewise.
4638 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4639 (GOT_memory_operand): Likewise.
4640 (call_insn_operand): Likewise.
4641 (sibcall_insn_operand): Likewise.
4642 (GOT32_symbol_operand): Likewise.
4644 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4647 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4648 type rather than type addr's type points to.
4649 (expand_omp_atomic_mutex): Likewise.
4650 (expand_omp_atomic): Likewise.
4652 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4655 * config/i386/i386.c (output_indirect_thunk_function): Use
4656 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4657 for __x86_return_thunk.
4659 2018-01-15 Richard Biener <rguenther@suse.de>
4662 * expmed.c (extract_bit_field_1): Fix typo.
4664 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4667 * config/arm/iterators.md (VF): New mode iterator.
4668 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4669 Remove integer-related logic from pattern.
4670 (neon_vabd<mode>_3): Likewise.
4672 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4675 * common.opt (fstrict-overflow): No longer an alias.
4676 (fwrapv-pointer): New option.
4677 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4678 also for pointer types based on flag_wrapv_pointer.
4679 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4680 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4681 opts->x_flag_wrapv got set.
4682 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4683 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4684 POINTER_TYPE_OVERFLOW_UNDEFINED.
4685 * match.pd: Likewise in address comparison pattern.
4686 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4688 2018-01-15 Richard Biener <rguenther@suse.de>
4691 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4692 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4693 Reset type names to their identifier if their TYPE_DECL doesn't
4694 have linkage (and thus is used for ODR and devirt).
4695 (save_debug_info_for_decl): Remove.
4696 (save_debug_info_for_type): Likewise.
4697 (add_tree_to_fld_list): Adjust.
4698 * tree-pretty-print.c (dump_generic_node): Make dumping of
4699 type names more robust.
4701 2018-01-15 Richard Biener <rguenther@suse.de>
4703 * BASE-VER: Bump to 8.0.1.
4705 2018-01-14 Martin Sebor <msebor@redhat.com>
4708 * builtins.c (check_access): Avoid warning when the no-warning bit
4711 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4713 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4714 * ira-color (allocno_hard_regs_compare): Likewise.
4716 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4719 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4720 Use .pushsection/.popsection.
4722 2018-01-14 Martin Sebor <msebor@redhat.com>
4725 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4727 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4729 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4730 entry from extra_headers.
4731 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4732 extra_headers, make the list bitwise identical to the i?86-*-* one.
4734 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4736 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4737 -mcmodel=large with -mindirect-branch=thunk,
4738 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4739 -mfunction-return=thunk-extern.
4740 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4741 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4742 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4744 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4746 * config/i386/i386.c (print_reg): Print the name of the full
4747 integer register without '%'.
4748 (ix86_print_operand): Handle 'V'.
4749 * doc/extend.texi: Document 'V' modifier.
4751 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4753 * config/i386/constraints.md (Bs): Disallow memory operand for
4754 -mindirect-branch-register.
4756 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4757 (GOT_memory_operand): Likewise.
4758 (call_insn_operand): Likewise.
4759 (sibcall_insn_operand): Likewise.
4760 (GOT32_symbol_operand): Likewise.
4761 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4762 for -mindirect-branch-register.
4763 (tablejump): Likewise.
4764 (*sibcall_memory): Likewise.
4765 (*sibcall_value_memory): Likewise.
4766 Disallow peepholes of indirect call and jump via memory for
4767 -mindirect-branch-register.
4768 (*call_pop): Replace m with Bw.
4769 (*call_value_pop): Likewise.
4770 (*sibcall_pop_memory): Replace m with Bs.
4771 * config/i386/i386.opt (mindirect-branch-register): New option.
4772 * doc/invoke.texi: Document -mindirect-branch-register option.
4774 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4776 * config/i386/i386-protos.h (ix86_output_function_return): New.
4777 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4778 set function_return_type.
4779 (indirect_thunk_name): Add ret_p to indicate thunk for function
4781 (output_indirect_thunk_function): Pass false to
4782 indirect_thunk_name.
4783 (ix86_output_indirect_branch_via_reg): Likewise.
4784 (ix86_output_indirect_branch_via_push): Likewise.
4785 (output_indirect_thunk_function): Create alias for function
4786 return thunk if regno < 0.
4787 (ix86_output_function_return): New function.
4788 (ix86_handle_fndecl_attribute): Handle function_return.
4789 (ix86_attribute_table): Add function_return.
4790 * config/i386/i386.h (machine_function): Add
4791 function_return_type.
4792 * config/i386/i386.md (simple_return_internal): Use
4793 ix86_output_function_return.
4794 (simple_return_internal_long): Likewise.
4795 * config/i386/i386.opt (mfunction-return=): New option.
4796 (indirect_branch): Mention -mfunction-return=.
4797 * doc/extend.texi: Document function_return function attribute.
4798 * doc/invoke.texi: Document -mfunction-return= option.
4800 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4802 * config/i386/i386-opts.h (indirect_branch): New.
4803 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4804 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4805 with local indirect jump when converting indirect call and jump.
4806 (ix86_set_indirect_branch_type): New.
4807 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4808 (indirectlabelno): New.
4809 (indirect_thunk_needed): Likewise.
4810 (indirect_thunk_bnd_needed): Likewise.
4811 (indirect_thunks_used): Likewise.
4812 (indirect_thunks_bnd_used): Likewise.
4813 (INDIRECT_LABEL): Likewise.
4814 (indirect_thunk_name): Likewise.
4815 (output_indirect_thunk): Likewise.
4816 (output_indirect_thunk_function): Likewise.
4817 (ix86_output_indirect_branch_via_reg): Likewise.
4818 (ix86_output_indirect_branch_via_push): Likewise.
4819 (ix86_output_indirect_branch): Likewise.
4820 (ix86_output_indirect_jmp): Likewise.
4821 (ix86_code_end): Call output_indirect_thunk_function if needed.
4822 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4824 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4825 (ix86_attribute_table): Add indirect_branch.
4826 * config/i386/i386.h (machine_function): Add indirect_branch_type
4827 and has_local_indirect_jump.
4828 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4830 (tablejump): Likewise.
4831 (*indirect_jump): Use ix86_output_indirect_jmp.
4832 (*tablejump_1): Likewise.
4833 (simple_return_indirect_internal): Likewise.
4834 * config/i386/i386.opt (mindirect-branch=): New option.
4835 (indirect_branch): New.
4838 (thunk-inline): Likewise.
4839 (thunk-extern): Likewise.
4840 * doc/extend.texi: Document indirect_branch function attribute.
4841 * doc/invoke.texi: Document -mindirect-branch= option.
4843 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4846 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4848 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4850 * ipa-inline.c (want_inline_small_function_p): Return false if
4851 inlining has already failed with CIF_FINAL_ERROR.
4852 (update_caller_keys): Call want_inline_small_function_p before
4854 (update_callee_keys): Likewise.
4856 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4858 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4860 (rs6000_quadword_masked_address_p): Likewise.
4861 (quad_aligned_load_p): Likewise.
4862 (quad_aligned_store_p): Likewise.
4863 (const_load_sequence_p): Add comment to describe the outer-most loop.
4864 (mimic_memory_attributes_and_flags): New function.
4865 (rs6000_gen_stvx): Likewise.
4866 (replace_swapped_aligned_store): Likewise.
4867 (rs6000_gen_lvx): Likewise.
4868 (replace_swapped_aligned_load): Likewise.
4869 (replace_swapped_load_constant): Capitalize argument name in
4870 comment describing this function.
4871 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4872 and stores that access quad-word aligned addresses and replace
4873 with stvx or lvx instructions when appropriate.
4874 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4875 New function prototype.
4876 (rs6000_quadword_masked_address_p): Likewise.
4877 (rs6000_gen_lvx): Likewise.
4878 (rs6000_gen_stvx): Likewise.
4879 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4880 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4881 when memory address is aligned.
4882 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4883 this split to select lvx instruction when memory address is aligned.
4884 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4885 instruction when memory address is aligned.
4886 (*vsx_le_perm_load_v16qi): Likewise.
4887 (four unnamed splitters): Modify to select the stvx instruction
4888 when memory is aligned.
4890 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4892 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4893 which appears in the queue multiple times.
4895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4896 Alan Hayward <alan.hayward@arm.com>
4897 David Sherwood <david.sherwood@arm.com>
4899 * tree-vectorizer.h (vec_lower_bound): New structure.
4900 (_loop_vec_info): Add check_nonzero and lower_bounds.
4901 (LOOP_VINFO_CHECK_NONZERO): New macro.
4902 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4903 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4904 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4905 fields. Make seg_len the distance travelled, not including the
4907 (dr_direction_indicator): Declare.
4908 (dr_zero_step_indicator): Likewise.
4909 (dr_known_forward_stride_p): Likewise.
4910 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4912 (runtime_alias_check_p): Allow runtime alias checks with
4914 (operator ==): Compare access_size and align.
4915 (prune_runtime_alias_test_list): Rework for new distinction between
4916 the access_size and seg_len.
4917 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4919 (get_segment_min_max): New function.
4920 (create_intersect_range_checks): Use it.
4921 (dr_step_indicator): New function.
4922 (dr_direction_indicator): Likewise.
4923 (dr_zero_step_indicator): Likewise.
4924 (dr_known_forward_stride_p): Likewise.
4925 * tree-loop-distribution.c (data_ref_segment_size): Return
4926 DR_STEP * (niters - 1).
4927 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4929 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4930 (vect_preserves_scalar_order_p): New function, split out from...
4931 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4932 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4933 (vect_vfa_access_size): New function.
4934 (vect_vfa_align): Likewise.
4935 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4936 (dump_lower_bound): New function.
4937 (vect_check_lower_bound): Likewise.
4938 (vect_small_gap_p): Likewise.
4939 (vectorizable_with_step_bound_p): Likewise.
4940 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4941 depencies if the vectorization factor is 1. Convert the checks
4942 for nonzero steps into checks on the bounds of DR_STEP. Try using
4943 a bunds check for variable steps if the minimum required step is
4944 relatively small. Update calls to the dr_with_seg_len
4945 constructor and to vect_compile_time_alias.
4946 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4948 (vect_loop_versioning): Call it.
4949 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4951 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4954 Alan Hayward <alan.hayward@arm.com>
4955 David Sherwood <david.sherwood@arm.com>
4957 * doc/sourcebuild.texi (vect_scatter_store): Document.
4958 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4960 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4962 * genopinit.c (main): Add supports_vec_scatter_store and
4963 supports_vec_scatter_store_cached to target_optabs.
4964 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4965 IFN_MASK_SCATTER_STORE.
4966 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4968 * internal-fn.h (internal_store_fn_p): Declare.
4969 (internal_fn_stored_value_index): Likewise.
4970 * internal-fn.c (scatter_store_direct): New macro.
4971 (expand_scatter_store_optab_fn): New function.
4972 (direct_scatter_store_optab_supported_p): New macro.
4973 (internal_store_fn_p): New function.
4974 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4975 IFN_MASK_SCATTER_STORE.
4976 (internal_fn_mask_index): Likewise.
4977 (internal_fn_stored_value_index): New function.
4978 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4980 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4981 * optabs-query.c (supports_vec_scatter_store_p): New function.
4982 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4983 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4984 true for scatter stores.
4985 (vect_gather_scatter_fn_p): Handle scatter stores too.
4986 (vect_check_gather_scatter): Consider using scatter stores if
4987 supports_vec_scatter_store_p.
4988 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4990 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4991 internal_fn_stored_value_index.
4992 (check_load_store_masking): Handle scatter stores too.
4993 (vect_get_store_rhs): Make public.
4994 (vectorizable_call): Use internal_store_fn_p.
4995 (vectorizable_store): Handle scatter store internal functions.
4996 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4997 when deciding whether the end of the group has been reached.
4998 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4999 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5000 (mask_scatter_store<mode>): New insns.
5002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5003 Alan Hayward <alan.hayward@arm.com>
5004 David Sherwood <david.sherwood@arm.com>
5006 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5007 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5008 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5010 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5011 Use vect_truncate_gather_scatter_offset if we can't treat the
5012 operation as a normal gather load or scatter store.
5013 (get_group_load_store_type): Take the gather_scatter_info
5014 as argument. Try using a gather load or scatter store for
5015 single-element groups.
5016 (get_load_store_type): Update calls to get_group_load_store_type
5017 and vect_use_strided_gather_scatters_p.
5019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5020 Alan Hayward <alan.hayward@arm.com>
5021 David Sherwood <david.sherwood@arm.com>
5023 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5024 optional tree argument.
5025 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5027 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5028 but continue to use the current value as a fallback.
5029 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5030 to compare the updates.
5031 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5032 (get_load_store_type): Use it when handling a strided access.
5033 (vect_get_strided_load_store_ops): New function.
5034 (vect_get_data_ptr_increment): Likewise.
5035 (vectorizable_load): Handle strided gather loads. Always pass
5036 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5039 Alan Hayward <alan.hayward@arm.com>
5040 David Sherwood <david.sherwood@arm.com>
5042 * doc/md.texi (gather_load@var{m}): Document.
5043 (mask_gather_load@var{m}): Likewise.
5044 * genopinit.c (main): Add supports_vec_gather_load and
5045 supports_vec_gather_load_cached to target_optabs.
5046 * optabs-tree.c (init_tree_optimization_optabs): Use
5047 ggc_cleared_alloc to allocate target_optabs.
5048 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5049 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5051 * internal-fn.h (internal_load_fn_p): Declare.
5052 (internal_gather_scatter_fn_p): Likewise.
5053 (internal_fn_mask_index): Likewise.
5054 (internal_gather_scatter_fn_supported_p): Likewise.
5055 * internal-fn.c (gather_load_direct): New macro.
5056 (expand_gather_load_optab_fn): New function.
5057 (direct_gather_load_optab_supported_p): New macro.
5058 (direct_internal_fn_optab): New function.
5059 (internal_load_fn_p): Likewise.
5060 (internal_gather_scatter_fn_p): Likewise.
5061 (internal_fn_mask_index): Likewise.
5062 (internal_gather_scatter_fn_supported_p): Likewise.
5063 * optabs-query.c (supports_at_least_one_mode_p): New function.
5064 (supports_vec_gather_load_p): Likewise.
5065 * optabs-query.h (supports_vec_gather_load_p): Declare.
5066 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5067 and memory_type field.
5068 (NUM_PATTERNS): Bump to 15.
5069 * tree-vect-data-refs.c: Include internal-fn.h.
5070 (vect_gather_scatter_fn_p): New function.
5071 (vect_describe_gather_scatter_call): Likewise.
5072 (vect_check_gather_scatter): Try using internal functions for
5073 gather loads. Recognize existing calls to a gather load function.
5074 (vect_analyze_data_refs): Consider using gather loads if
5075 supports_vec_gather_load_p.
5076 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5077 (vect_get_gather_scatter_offset_type): Likewise.
5078 (vect_convert_mask_for_vectype): Likewise.
5079 (vect_add_conversion_to_patterm): Likewise.
5080 (vect_try_gather_scatter_pattern): Likewise.
5081 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5082 (vect_vect_recog_func_ptrs): Add it.
5083 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5084 internal_fn_mask_index and internal_gather_scatter_fn_p.
5085 (check_load_store_masking): Take the gather_scatter_info as an
5086 argument and handle gather loads.
5087 (vect_get_gather_scatter_ops): New function.
5088 (vectorizable_call): Check internal_load_fn_p.
5089 (vectorizable_load): Likewise. Handle gather load internal
5091 (vectorizable_store): Update call to check_load_store_masking.
5092 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5093 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5094 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5095 (aarch64_gather_scale_operand_d): New predicates.
5096 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5097 (mask_gather_load<mode>): New insns.
5099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5100 Alan Hayward <alan.hayward@arm.com>
5101 David Sherwood <david.sherwood@arm.com>
5103 * optabs.def (fold_left_plus_optab): New optab.
5104 * doc/md.texi (fold_left_plus_@var{m}): Document.
5105 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5106 * internal-fn.c (fold_left_direct): Define.
5107 (expand_fold_left_optab_fn): Likewise.
5108 (direct_fold_left_optab_supported_p): Likewise.
5109 * fold-const-call.c (fold_const_fold_left): New function.
5110 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5111 * tree-parloops.c (valid_reduction_p): New function.
5112 (gather_scalar_reductions): Use it.
5113 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5114 (vect_finish_replace_stmt): Declare.
5115 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5116 (needs_fold_left_reduction_p): New function, split out from...
5117 (vect_is_simple_reduction): ...here. Accept reductions that
5118 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5119 (vect_force_simple_reduction): Also store the reduction type in
5120 the assignment's STMT_VINFO_REDUC_TYPE.
5121 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5122 (merge_with_identity): New function.
5123 (vect_expand_fold_left): Likewise.
5124 (vectorize_fold_left_reduction): Likewise.
5125 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5126 scalar phi in place for it. Check for target support and reject
5127 cases that would reassociate the operation. Defer the transform
5128 phase to vectorize_fold_left_reduction.
5129 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5130 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5131 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5135 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5136 call to ifc_temp_var.
5138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5139 Alan Hayward <alan.hayward@arm.com>
5140 David Sherwood <david.sherwood@arm.com>
5142 * target.def (legitimize_address_displacement): Take the original
5143 offset as a poly_int.
5144 * targhooks.h (default_legitimize_address_displacement): Update
5146 * targhooks.c (default_legitimize_address_displacement): Likewise.
5147 * doc/tm.texi: Regenerate.
5148 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5149 as an argument, moving assert of ad->disp == ad->disp_term to...
5150 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5151 Try calling targetm.legitimize_address_displacement before expanding
5152 the address rather than afterwards, and adjust for the new interface.
5153 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5154 Match the new hook interface. Handle SVE addresses.
5155 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5160 * Makefile.in (OBJS): Add early-remat.o.
5161 * target.def (select_early_remat_modes): New hook.
5162 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5163 * doc/tm.texi: Regenerate.
5164 * targhooks.h (default_select_early_remat_modes): Declare.
5165 * targhooks.c (default_select_early_remat_modes): New function.
5166 * timevar.def (TV_EARLY_REMAT): New timevar.
5167 * passes.def (pass_early_remat): New pass.
5168 * tree-pass.h (make_pass_early_remat): Declare.
5169 * early-remat.c: New file.
5170 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5172 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5175 Alan Hayward <alan.hayward@arm.com>
5176 David Sherwood <david.sherwood@arm.com>
5178 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5179 vfm1 with a bound_epilog parameter.
5180 (vect_do_peeling): Update calls accordingly, and move the prologue
5181 call earlier in the function. Treat the base bound_epilog as 0 for
5182 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5183 this base when peeling for gaps.
5184 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5185 with fully-masked loops.
5186 (vect_estimate_min_profitable_iters): Handle the single peeled
5187 iteration in that case.
5189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5190 Alan Hayward <alan.hayward@arm.com>
5191 David Sherwood <david.sherwood@arm.com>
5193 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5194 single-element interleaving even if the size is not a power of 2.
5195 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5196 accesses for single-element interleaving if the group size is
5199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5200 Alan Hayward <alan.hayward@arm.com>
5201 David Sherwood <david.sherwood@arm.com>
5203 * doc/md.texi (fold_extract_last_@var{m}): Document.
5204 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5205 * optabs.def (fold_extract_last_optab): New optab.
5206 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5207 * internal-fn.c (fold_extract_direct): New macro.
5208 (expand_fold_extract_optab_fn): Likewise.
5209 (direct_fold_extract_optab_supported_p): Likewise.
5210 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5211 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5212 EXTRACT_LAST_REDUCTION.
5213 (get_initial_def_for_reduction): Do not create an initial vector
5214 for EXTRACT_LAST_REDUCTION reductions.
5215 (vectorizable_reduction): Leave the scalar phi in place for
5216 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5217 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5218 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5219 transform phase to vectorizable_condition.
5220 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5222 (vect_finish_stmt_generation): ...here.
5223 (vect_finish_replace_stmt): New function.
5224 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5225 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5227 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5230 Alan Hayward <alan.hayward@arm.com>
5231 David Sherwood <david.sherwood@arm.com>
5233 * doc/md.texi (extract_last_@var{m}): Document.
5234 * optabs.def (extract_last_optab): New optab.
5235 * internal-fn.def (EXTRACT_LAST): New internal function.
5236 * internal-fn.c (cond_unary_direct): New macro.
5237 (expand_cond_unary_optab_fn): Likewise.
5238 (direct_cond_unary_optab_supported_p): Likewise.
5239 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5240 loops using EXTRACT_LAST.
5241 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5242 (extract_last_<mode>): ...this optab.
5243 (vec_extract<mode><Vel>): Update accordingly.
5245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5246 Alan Hayward <alan.hayward@arm.com>
5247 David Sherwood <david.sherwood@arm.com>
5249 * target.def (empty_mask_is_expensive): New hook.
5250 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5251 * doc/tm.texi: Regenerate.
5252 * targhooks.h (default_empty_mask_is_expensive): Declare.
5253 * targhooks.c (default_empty_mask_is_expensive): New function.
5254 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5255 if the target says that empty masks are expensive.
5256 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5258 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5261 Alan Hayward <alan.hayward@arm.com>
5262 David Sherwood <david.sherwood@arm.com>
5264 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5265 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5266 (vect_use_loop_mask_for_alignment_p): New function.
5267 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5268 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5269 niters_skip argument. Make sure that the first niters_skip elements
5270 of the first iteration are inactive.
5271 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5272 Update call to vect_set_loop_masks_directly.
5273 (get_misalign_in_elems): New function, split out from...
5274 (vect_gen_prolog_loop_niters): ...here.
5275 (vect_update_init_of_dr): Take a code argument that specifies whether
5276 the adjustment should be added or subtracted.
5277 (vect_update_init_of_drs): Likewise.
5278 (vect_prepare_for_masked_peels): New function.
5279 (vect_do_peeling): Skip prologue peeling if we're using a mask
5280 instead. Update call to vect_update_inits_of_drs.
5281 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5283 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5284 alignment. Do not include the number of peeled iterations in
5285 the minimum threshold in that case.
5286 (vectorizable_induction): Adjust the start value down by
5287 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5288 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5289 Take the number of skipped iterations into account when calculating
5291 * tree-vect-stmts.c (vect_gen_while_not): New function.
5293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5294 Alan Hayward <alan.hayward@arm.com>
5295 David Sherwood <david.sherwood@arm.com>
5297 * doc/sourcebuild.texi (vect_fully_masked): Document.
5298 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5300 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5302 (vect_analyze_loop_2): ...here. Don't check the vectorization
5303 factor against the number of loop iterations if the loop is
5306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5307 Alan Hayward <alan.hayward@arm.com>
5308 David Sherwood <david.sherwood@arm.com>
5310 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5311 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5312 (dump_groups): Update accordingly.
5313 (iv_use::mem_type): New member variable.
5314 (address_p): New function.
5315 (record_use): Add a mem_type argument and initialize the new
5317 (record_group_use): Add a mem_type argument. Use address_p.
5318 Remove obsolete null checks of base_object. Update call to record_use.
5319 (find_interesting_uses_op): Update call to record_group_use.
5320 (find_interesting_uses_cond): Likewise.
5321 (find_interesting_uses_address): Likewise.
5322 (get_mem_type_for_internal_fn): New function.
5323 (find_address_like_use): Likewise.
5324 (find_interesting_uses_stmt): Try find_address_like_use before
5325 calling find_interesting_uses_op.
5326 (addr_offset_valid_p): Use the iv mem_type field as the type
5327 of the addressed memory.
5328 (add_autoinc_candidates): Likewise.
5329 (get_address_cost): Likewise.
5330 (split_small_address_groups_p): Use address_p.
5331 (split_address_groups): Likewise.
5332 (add_iv_candidate_for_use): Likewise.
5333 (autoinc_possible_for_pair): Likewise.
5334 (rewrite_groups): Likewise.
5335 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5336 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5337 (get_alias_ptr_type_for_ptr_address): New function.
5338 (rewrite_use_address): Rewrite address uses in calls that were
5339 identified by find_address_like_use.
5341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5342 Alan Hayward <alan.hayward@arm.com>
5343 David Sherwood <david.sherwood@arm.com>
5345 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5347 * gimple-expr.h (is_gimple_addressable: Likewise.
5348 * gimple-expr.c (is_gimple_address): Likewise.
5349 * internal-fn.c (expand_call_mem_ref): New function.
5350 (expand_mask_load_optab_fn): Use it.
5351 (expand_mask_store_optab_fn): Likewise.
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 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5358 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5359 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5360 (cond_umax@var{mode}): Document.
5361 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5362 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5363 (cond_umin_optab, cond_umax_optab): New optabs.
5364 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5365 (COND_IOR, COND_XOR): New internal functions.
5366 * internal-fn.h (get_conditional_internal_fn): Declare.
5367 * internal-fn.c (cond_binary_direct): New macro.
5368 (expand_cond_binary_optab_fn): Likewise.
5369 (direct_cond_binary_optab_supported_p): Likewise.
5370 (get_conditional_internal_fn): New function.
5371 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5372 Cope with reduction statements that are vectorized as calls rather
5374 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5375 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5376 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5377 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5378 (UNSPEC_COND_EOR): New unspecs.
5379 (optab): Add mappings for them.
5380 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5381 (sve_int_op, sve_fp_op): New int attributes.
5383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5384 Alan Hayward <alan.hayward@arm.com>
5385 David Sherwood <david.sherwood@arm.com>
5387 * optabs.def (while_ult_optab): New optab.
5388 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5389 * internal-fn.def (WHILE_ULT): New internal function.
5390 * internal-fn.h (direct_internal_fn_supported_p): New override
5391 that takes two types as argument.
5392 * internal-fn.c (while_direct): New macro.
5393 (expand_while_optab_fn): New function.
5394 (convert_optab_supported_p): Likewise.
5395 (direct_while_optab_supported_p): New macro.
5396 * wide-int.h (wi::udiv_ceil): New function.
5397 * tree-vectorizer.h (rgroup_masks): New structure.
5398 (vec_loop_masks): New typedef.
5399 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5401 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5402 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5403 (vect_max_vf): New function.
5404 (slpeel_make_loop_iterate_ntimes): Delete.
5405 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5406 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5407 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5408 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5409 internal-fn.h, stor-layout.h and optabs-query.h.
5410 (vect_set_loop_mask): New function.
5411 (add_preheader_seq): Likewise.
5412 (add_header_seq): Likewise.
5413 (interleave_supported_p): Likewise.
5414 (vect_maybe_permute_loop_masks): Likewise.
5415 (vect_set_loop_masks_directly): Likewise.
5416 (vect_set_loop_condition_masked): Likewise.
5417 (vect_set_loop_condition_unmasked): New function, split out from
5418 slpeel_make_loop_iterate_ntimes.
5419 (slpeel_make_loop_iterate_ntimes): Rename to..
5420 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5421 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5422 (vect_do_peeling): Update call accordingly.
5423 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5425 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5426 mask_compare_type, can_fully_mask_p and fully_masked_p.
5427 (release_vec_loop_masks): New function.
5428 (_loop_vec_info): Use it to free the loop masks.
5429 (can_produce_all_loop_masks_p): New function.
5430 (vect_get_max_nscalars_per_iter): Likewise.
5431 (vect_verify_full_masking): Likewise.
5432 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5433 retries, and free the mask rgroups before retrying. Check loop-wide
5434 reasons for disallowing fully-masked loops. Make the final decision
5435 about whether use a fully-masked loop or not.
5436 (vect_estimate_min_profitable_iters): Do not assume that peeling
5437 for the number of iterations will be needed for fully-masked loops.
5438 (vectorizable_reduction): Disable fully-masked loops.
5439 (vectorizable_live_operation): Likewise.
5440 (vect_halve_mask_nunits): New function.
5441 (vect_double_mask_nunits): Likewise.
5442 (vect_record_loop_mask): Likewise.
5443 (vect_get_loop_mask): Likewise.
5444 (vect_transform_loop): Handle the case in which the final loop
5445 iteration might handle a partial vector. Call vect_set_loop_condition
5446 instead of slpeel_make_loop_iterate_ntimes.
5447 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5448 (check_load_store_masking): New function.
5449 (prepare_load_store_mask): Likewise.
5450 (vectorizable_store): Handle fully-masked loops.
5451 (vectorizable_load): Likewise.
5452 (supportable_widening_operation): Use vect_halve_mask_nunits for
5454 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5455 (vect_gen_while): New function.
5456 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5457 (aarch64_uqdec<mode>): New insn.
5459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5460 Alan Hayward <alan.hayward@arm.com>
5461 David Sherwood <david.sherwood@arm.com>
5463 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5464 (reduc_xor_scal_optab): New optabs.
5465 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5466 (reduc_xor_scal_@var{m}): Document.
5467 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5468 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5470 * fold-const-call.c (fold_const_call): Handle them.
5471 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5472 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5473 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5474 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5475 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5476 (UNSPEC_XORV): New unspecs.
5477 (optab): Add entries for them.
5478 (BITWISEV): New int iterator.
5479 (bit_reduc_op): New int attributes.
5481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5482 Alan Hayward <alan.hayward@arm.com>
5483 David Sherwood <david.sherwood@arm.com>
5485 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5486 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5487 * optabs.def (vec_shl_insert_optab): New optab.
5488 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5489 (duplicate_and_interleave): Likewise.
5490 * tree-vect-loop.c: Include internal-fn.h.
5491 (neutral_op_for_slp_reduction): New function, split out from
5492 get_initial_defs_for_reduction.
5493 (get_initial_def_for_reduction): Handle option 2 for variable-length
5494 vectors by loading the neutral value into a vector and then shifting
5495 the initial value into element 0.
5496 (get_initial_defs_for_reduction): Replace the code argument with
5497 the neutral value calculated by neutral_op_for_slp_reduction.
5498 Use gimple_build_vector for constant-length vectors.
5499 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5500 but the first group_size elements have a neutral value.
5501 Use duplicate_and_interleave otherwise.
5502 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5503 Update call to get_initial_defs_for_reduction. Handle SLP
5504 reductions for variable-length vectors by creating one vector
5505 result for each scalar result, with the elements associated
5506 with other scalar results stubbed out with the neutral value.
5507 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5508 Require IFN_VEC_SHL_INSERT for double reductions on
5509 variable-length vectors, or SLP reductions that have
5510 a neutral value. Require can_duplicate_and_interleave_p
5511 support for variable-length unchained SLP reductions if there
5512 is no neutral value, such as for MIN/MAX reductions. Also require
5513 the number of vector elements to be a multiple of the number of
5514 SLP statements when doing variable-length unchained SLP reductions.
5515 Update call to vect_create_epilog_for_reduction.
5516 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5517 and remove initial values.
5518 (duplicate_and_interleave): Make public.
5519 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5520 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5523 Alan Hayward <alan.hayward@arm.com>
5524 David Sherwood <david.sherwood@arm.com>
5526 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5527 (can_duplicate_and_interleave_p): New function.
5528 (vect_get_and_check_slp_defs): Take the vector of statements
5529 rather than just the current one. Remove excess parentheses.
5530 Restriction rejectinon of vect_constant_def and vect_external_def
5531 for variable-length vectors to boolean types, or types for which
5532 can_duplicate_and_interleave_p is false.
5533 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5534 (duplicate_and_interleave): New function.
5535 (vect_get_constant_vectors): Use gimple_build_vector for
5536 constant-length vectors and suitable variable-length constant
5537 vectors. Use duplicate_and_interleave for other variable-length
5538 vectors. Don't defer the update when inserting new statements.
5540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5541 Alan Hayward <alan.hayward@arm.com>
5542 David Sherwood <david.sherwood@arm.com>
5544 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5545 min_profitable_iters doesn't go negative.
5547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5548 Alan Hayward <alan.hayward@arm.com>
5549 David Sherwood <david.sherwood@arm.com>
5551 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5552 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5553 * optabs.def (vec_mask_load_lanes_optab): New optab.
5554 (vec_mask_store_lanes_optab): Likewise.
5555 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5556 (MASK_STORE_LANES): Likewise.
5557 * internal-fn.c (mask_load_lanes_direct): New macro.
5558 (mask_store_lanes_direct): Likewise.
5559 (expand_mask_load_optab_fn): Handle masked operations.
5560 (expand_mask_load_lanes_optab_fn): New macro.
5561 (expand_mask_store_optab_fn): Handle masked operations.
5562 (expand_mask_store_lanes_optab_fn): New macro.
5563 (direct_mask_load_lanes_optab_supported_p): Likewise.
5564 (direct_mask_store_lanes_optab_supported_p): Likewise.
5565 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5567 (vect_load_lanes_supported): Likewise.
5568 * tree-vect-data-refs.c (strip_conversion): New function.
5569 (can_group_stmts_p): Likewise.
5570 (vect_analyze_data_ref_accesses): Use it instead of checking
5571 for a pair of assignments.
5572 (vect_store_lanes_supported): Take a masked_p parameter.
5573 (vect_load_lanes_supported): Likewise.
5574 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5575 vect_store_lanes_supported and vect_load_lanes_supported.
5576 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5577 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5578 parameter. Don't allow gaps for masked accesses.
5579 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5580 and vect_load_lanes_supported.
5581 (get_load_store_type): Take a masked_p parameter and update
5582 call to get_group_load_store_type.
5583 (vectorizable_store): Update call to get_load_store_type.
5584 Handle IFN_MASK_STORE_LANES.
5585 (vectorizable_load): Update call to get_load_store_type.
5586 Handle IFN_MASK_LOAD_LANES.
5588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5589 Alan Hayward <alan.hayward@arm.com>
5590 David Sherwood <david.sherwood@arm.com>
5592 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5594 * config/aarch64/aarch64-protos.h
5595 (aarch64_sve_struct_memory_operand_p): Declare.
5596 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5597 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5598 (VPRED, vpred): Handle SVE structure modes.
5599 * config/aarch64/constraints.md (Utx): New constraint.
5600 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5601 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5602 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5603 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5604 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5605 structure modes. Split into pieces after RA.
5606 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5607 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5609 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5610 SVE structure modes.
5611 (aarch64_classify_address): Likewise.
5612 (sizetochar): Move earlier in file.
5613 (aarch64_print_operand): Handle SVE register lists.
5614 (aarch64_array_mode): New function.
5615 (aarch64_sve_struct_memory_operand_p): Likewise.
5616 (TARGET_ARRAY_MODE): Redefine.
5618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5619 Alan Hayward <alan.hayward@arm.com>
5620 David Sherwood <david.sherwood@arm.com>
5622 * target.def (array_mode): New target hook.
5623 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5624 * doc/tm.texi: Regenerate.
5625 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5626 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5627 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5629 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5633 Alan Hayward <alan.hayward@arm.com>
5634 David Sherwood <david.sherwood@arm.com>
5636 * fold-const.c (fold_binary_loc): Check the argument types
5637 rather than the result type when testing for a vector operation.
5639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5641 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5642 * doc/tm.texi: Regenerate.
5644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5645 Alan Hayward <alan.hayward@arm.com>
5646 David Sherwood <david.sherwood@arm.com>
5648 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5649 (sve): Document new AArch64 extension.
5650 * doc/md.texi (w): Extend the description of the AArch64
5651 constraint to include SVE vectors.
5652 (Upl, Upa): Document new AArch64 predicate constraints.
5653 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5655 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5656 (msve-vector-bits=): New option.
5657 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5658 SVE when these are disabled.
5659 (sve): New extension.
5660 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5661 modes. Adjust their number of units based on aarch64_sve_vg.
5662 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5663 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5664 aarch64_addr_query_type.
5665 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5666 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5667 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5668 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5669 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5670 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5671 (aarch64_simd_imm_zero_p): Delete.
5672 (aarch64_check_zero_based_sve_index_immediate): Declare.
5673 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5674 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5675 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5676 (aarch64_sve_float_mul_immediate_p): Likewise.
5677 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5679 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5680 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5681 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5682 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5683 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5684 (aarch64_regmode_natural_size): Likewise.
5685 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5686 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5688 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5689 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5690 for VG and the SVE predicate registers.
5691 (V_ALIASES): Add a "z"-prefixed alias.
5692 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5693 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5694 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5695 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5696 (REG_CLASS_NAMES): Add entries for them.
5697 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5698 and the predicate registers.
5699 (aarch64_sve_vg): Declare.
5700 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5701 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5702 (REGMODE_NATURAL_SIZE): Define.
5703 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5705 * config/aarch64/aarch64.c: Include cfgrtl.h.
5706 (simd_immediate_info): Add a constructor for series vectors,
5707 and an associated step field.
5708 (aarch64_sve_vg): New variable.
5709 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5710 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5711 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5712 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5713 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5714 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5715 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5716 (aarch64_get_mask_mode): New functions.
5717 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5718 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5719 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5720 predicate modes and predicate registers. Explicitly restrict
5721 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5722 to store a vector mode if it is recognized by
5723 aarch64_classify_vector_mode.
5724 (aarch64_regmode_natural_size): New function.
5725 (aarch64_hard_regno_caller_save_mode): Return the original mode
5727 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5728 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5729 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5730 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5732 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5733 does not overlap dest if the function is frame-related. Handle
5735 (aarch64_split_add_offset): New function.
5736 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5737 them aarch64_add_offset.
5738 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5739 and update call to aarch64_sub_sp.
5740 (aarch64_add_cfa_expression): New function.
5741 (aarch64_expand_prologue): Pass extra temporary registers to the
5742 functions above. Handle the case in which we need to emit new
5743 DW_CFA_expressions for registers that were originally saved
5744 relative to the stack pointer, but now have to be expressed
5745 relative to the frame pointer.
5746 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5748 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5749 IP0 and IP1 values for SVE frames.
5750 (aarch64_expand_vec_series): New function.
5751 (aarch64_expand_sve_widened_duplicate): Likewise.
5752 (aarch64_expand_sve_const_vector): Likewise.
5753 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5754 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5755 into the register, rather than emitting a SET directly.
5756 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5757 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5758 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5759 (offset_9bit_signed_scaled_p): New functions.
5760 (aarch64_replicate_bitmask_imm): New function.
5761 (aarch64_bitmask_imm): Use it.
5762 (aarch64_cannot_force_const_mem): Reject expressions involving
5763 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5764 (aarch64_classify_index): Handle SVE indices, by requiring
5765 a plain register index with a scale that matches the element size.
5766 (aarch64_classify_address): Handle SVE addresses. Assert that
5767 the mode of the address is VOIDmode or an integer mode.
5768 Update call to aarch64_classify_symbol.
5769 (aarch64_classify_symbolic_expression): Update call to
5770 aarch64_classify_symbol.
5771 (aarch64_const_vec_all_in_range_p): New function.
5772 (aarch64_print_vector_float_operand): Likewise.
5773 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5774 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5775 and the FP immediates 1.0 and 0.5.
5776 (aarch64_print_address_internal): Handle SVE addresses.
5777 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5778 (aarch64_regno_regclass): Handle predicate registers.
5779 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5781 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5782 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5783 (aarch64_convert_sve_vector_bits): New function.
5784 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5785 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5787 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5788 Handle SVE vector and predicate modes. Accept VL-based constants
5789 that need only one temporary register, and VL offsets that require
5790 no temporary registers.
5791 (aarch64_conditional_register_usage): Mark the predicate registers
5792 as fixed if SVE isn't available.
5793 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5794 Return true for SVE vector and predicate modes.
5795 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5796 rather than an unsigned int. Handle SVE modes.
5797 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5799 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5801 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5802 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5803 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5804 (aarch64_sve_float_mul_immediate_p): New functions.
5805 (aarch64_sve_valid_immediate): New function.
5806 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5807 Explicitly reject structure modes. Check for INDEX constants.
5808 Handle PTRUE and PFALSE constants.
5809 (aarch64_check_zero_based_sve_index_immediate): New function.
5810 (aarch64_simd_imm_zero_p): Delete.
5811 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5812 vector modes. Accept constants in the range of CNT[BHWD].
5813 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5814 ask for an Advanced SIMD mode.
5815 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5816 (aarch64_simd_vector_alignment): Handle SVE predicates.
5817 (aarch64_vectorize_preferred_vector_alignment): New function.
5818 (aarch64_simd_vector_alignment_reachable): Use it instead of
5820 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5821 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5823 (MAX_VECT_LEN): Delete.
5824 (expand_vec_perm_d): Add a vec_flags field.
5825 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5826 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5827 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5829 (aarch64_evpc_rev): Rename to...
5830 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5831 (aarch64_evpc_rev_global): New function.
5832 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5833 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5835 (aarch64_evpc_sve_tbl): New function.
5836 (aarch64_expand_vec_perm_const_1): Update after rename of
5837 aarch64_evpc_rev. Handle SVE permutes too, trying
5838 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5839 than aarch64_evpc_tbl.
5840 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5841 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5842 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5843 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5844 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5845 (aarch64_expand_sve_vcond): New functions.
5846 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5847 of aarch64_vector_mode_p.
5848 (aarch64_dwarf_poly_indeterminate_value): New function.
5849 (aarch64_compute_pressure_classes): Likewise.
5850 (aarch64_can_change_mode_class): Likewise.
5851 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5852 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5853 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5854 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5855 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5856 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5857 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5858 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5860 (Dn, Dl, Dr): Accept const as well as const_vector.
5861 (Dz): Likewise. Compare against CONST0_RTX.
5862 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5863 of "vector" where appropriate.
5864 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5865 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5866 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5867 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5868 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5869 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5870 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5871 (v_int_equiv): Extend to SVE modes.
5872 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5874 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5875 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5876 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5877 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5878 (SVE_COND_FP_CMP): New int iterators.
5879 (perm_hilo): Handle the new unpack unspecs.
5880 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5882 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5883 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5884 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5885 (aarch64_equality_operator, aarch64_constant_vector_operand)
5886 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5887 (aarch64_sve_nonimmediate_operand): Likewise.
5888 (aarch64_sve_general_operand): Likewise.
5889 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5890 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5891 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5892 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5893 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5894 (aarch64_sve_float_arith_immediate): Likewise.
5895 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5896 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5897 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5898 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5899 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5900 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5901 (aarch64_sve_float_arith_operand): Likewise.
5902 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5903 (aarch64_sve_float_mul_operand): Likewise.
5904 (aarch64_sve_vec_perm_operand): Likewise.
5905 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5906 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5907 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5908 as well as const_vector.
5909 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5910 in file. Use CONST0_RTX and CONSTM1_RTX.
5911 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5912 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5913 Use aarch64_simd_imm_zero.
5914 * config/aarch64/aarch64-sve.md: New file.
5915 * config/aarch64/aarch64.md: Include it.
5916 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5917 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5918 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5919 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5920 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5921 (sve): New attribute.
5922 (enabled): Disable instructions with the sve attribute unless
5924 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5925 aarch64_expand_mov_immediate.
5926 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5927 CNT[BHSD] immediates.
5928 (movti): Split CONST_POLY_INT moves into two halves.
5929 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5930 Split additions that need a temporary here if the destination
5931 is the stack pointer.
5932 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5933 (*add<mode>3_poly_1): New instruction.
5934 (set_clobber_cc): New expander.
5936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5938 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5939 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5940 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5941 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5942 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5943 Change innermode from fixed_mode_size to machine_mode.
5944 (simplify_subreg): Update call accordingly. Handle a constant-sized
5945 subreg of a variable-length CONST_VECTOR.
5947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5948 Alan Hayward <alan.hayward@arm.com>
5949 David Sherwood <david.sherwood@arm.com>
5951 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5952 (add_offset_to_base): New function, split out from...
5953 (create_mem_ref): ...here. When handling a scale other than 1,
5954 check first whether the address is valid without the offset.
5955 Add it into the base if so, leaving the index and scale as-is.
5957 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5960 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5961 fold_for_warn before checking if arg2 is INTEGER_CST.
5963 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5965 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5966 (store_multiple_operation): Delete.
5967 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5968 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5969 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5970 guarded by TARGET_STRING.
5971 (rs6000_output_load_multiple): Delete.
5972 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5973 OPTION_MASK_STRING / TARGET_STRING handling.
5974 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5975 (const rs6000_opt_masks) <"string">: Change mask to 0.
5976 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5977 (MASK_STRING): Delete.
5978 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5980 (load_multiple): Delete.
5987 (store_multiple): Delete.
5994 (movmemsi_8reg): Delete.
5995 (corresponding unnamed define_insn): Delete.
5996 (movmemsi_6reg): Delete.
5997 (corresponding unnamed define_insn): Delete.
5998 (movmemsi_4reg): Delete.
5999 (corresponding unnamed define_insn): Delete.
6000 (movmemsi_2reg): Delete.
6001 (corresponding unnamed define_insn): Delete.
6002 (movmemsi_1reg): Delete.
6003 (corresponding unnamed define_insn): Delete.
6004 * config/rs6000/rs6000.opt (mno-string): New.
6005 (mstring): Replace by deprecation warning stub.
6006 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6008 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6010 * regrename.c (regrename_do_replace): If replacing the same
6011 reg multiple times, try to reuse last created gen_raw_REG.
6014 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6015 main to workaround a bug in GDB.
6017 2018-01-12 Tom de Vries <tom@codesourcery.com>
6020 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6022 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6024 PR rtl-optimization/80481
6025 * ira-color.c (get_cap_member): New function.
6026 (allocnos_conflict_by_live_ranges_p): Use it.
6027 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6028 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6030 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6033 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6034 (*saddl_se_1): Ditto.
6036 (*ssubl_se_1): Ditto.
6038 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6040 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6041 rather than wi::to_widest for DR_INITs.
6042 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6043 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6044 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6046 (vect_analyze_group_access_1): Note that here.
6048 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6050 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6051 polynomial type sizes.
6053 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6055 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6056 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6057 (gimple_add_tmp_var): Likewise.
6059 2018-01-12 Martin Liska <mliska@suse.cz>
6061 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6062 (gimple_alloc_sizes): Likewise.
6063 (dump_gimple_statistics): Use PRIu64 in printf format.
6064 * gimple.h: Change uint64_t to int.
6066 2018-01-12 Martin Liska <mliska@suse.cz>
6068 * tree-core.h: Use uint64_t instead of int.
6069 * tree.c (tree_node_counts): Likewise.
6070 (tree_node_sizes): Likewise.
6071 (dump_tree_statistics): Use PRIu64 in printf format.
6073 2018-01-12 Martin Liska <mliska@suse.cz>
6075 * Makefile.in: As qsort_chk is implemented in vec.c, add
6076 vec.o to linkage of gencfn-macros.
6077 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6078 passing the info to record_node_allocation_statistics.
6079 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6081 * ggc-common.c (struct ggc_usage): Add operator== and use
6082 it in operator< and compare function.
6083 * mem-stats.h (struct mem_usage): Likewise.
6084 * vec.c (struct vec_usage): Remove operator< and compare
6085 function. Can be simply inherited.
6087 2018-01-12 Martin Jambor <mjambor@suse.cz>
6090 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6091 * tree-ssa-math-opts.c: Include domwalk.h.
6092 (convert_mult_to_fma_1): New function.
6093 (fma_transformation_info): New type.
6094 (fma_deferring_state): Likewise.
6095 (cancel_fma_deferring): New function.
6096 (result_of_phi): Likewise.
6097 (last_fma_candidate_feeds_initial_phi): Likewise.
6098 (convert_mult_to_fma): Added deferring logic, split actual
6099 transformation to convert_mult_to_fma_1.
6100 (math_opts_dom_walker): New type.
6101 (math_opts_dom_walker::after_dom_children): New method, body moved
6102 here from pass_optimize_widening_mul::execute, added deferring logic
6104 (pass_optimize_widening_mul::execute): Moved most of code to
6105 math_opts_dom_walker::after_dom_children.
6106 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6107 * config/i386/i386.c (ix86_option_override_internal): Added
6108 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6110 2018-01-12 Richard Biener <rguenther@suse.de>
6113 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6114 inline instance vars.
6116 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6119 * config/rx/rx.c (rx_is_restricted_memory_address):
6122 2018-01-12 Richard Biener <rguenther@suse.de>
6124 PR tree-optimization/80846
6125 * target.def (split_reduction): New target hook.
6126 * targhooks.c (default_split_reduction): New function.
6127 * targhooks.h (default_split_reduction): Declare.
6128 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6129 target requests first reduce vectors by combining low and high
6131 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6132 (get_vectype_for_scalar_type_and_size): Export.
6133 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6134 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6135 * doc/tm.texi: Regenerate.
6136 * config/i386/i386.c (ix86_split_reduction): Implement
6137 TARGET_VECTORIZE_SPLIT_REDUCTION.
6139 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6142 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6143 in PIC mode except for TARGET_VXWORKS_RTP.
6144 * config/sparc/sparc.c: Include cfgrtl.h.
6145 (TARGET_INIT_PIC_REG): Define.
6146 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6147 (sparc_pic_register_p): New predicate.
6148 (sparc_legitimate_address_p): Use it.
6149 (sparc_legitimize_pic_address): Likewise.
6150 (sparc_delegitimize_address): Likewise.
6151 (sparc_mode_dependent_address_p): Likewise.
6152 (gen_load_pcrel_sym): Remove 4th parameter.
6153 (load_got_register): Adjust call to above. Remove obsolete stuff.
6154 (sparc_expand_prologue): Do not call load_got_register here.
6155 (sparc_flat_expand_prologue): Likewise.
6156 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6157 (sparc_use_pseudo_pic_reg): New function.
6158 (sparc_init_pic_reg): Likewise.
6159 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6160 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6162 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6164 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6165 Add item for branch_cost.
6167 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6169 PR rtl-optimization/83565
6170 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6171 not extend the result to a larger mode for rotate operations.
6172 (num_sign_bit_copies1): Likewise.
6174 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6177 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6179 Use values-Xc.o for -pedantic.
6180 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6182 2018-01-12 Martin Liska <mliska@suse.cz>
6185 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6187 (possible_polymorphic_call_targets): Use it.
6188 (ipa_devirt): Likewise.
6190 2018-01-12 Martin Liska <mliska@suse.cz>
6192 * profile-count.h (enum profile_quality): Use 0 as invalid
6193 enum value of profile_quality.
6195 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6197 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6198 -mext-string options.
6200 2018-01-12 Richard Biener <rguenther@suse.de>
6202 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6203 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6204 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6206 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6208 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6210 * configure.ac (--with-long-double-format): Add support for the
6211 configuration option to change the default long double format on
6213 * config.gcc (powerpc*-linux*-*): Likewise.
6214 * configure: Regenerate.
6215 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6216 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6217 used without modification.
6219 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6221 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6222 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6223 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6224 MISC_BUILTIN_SPEC_BARRIER.
6225 (rs6000_init_builtins): Likewise.
6226 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6228 (speculation_barrier): New define_insn.
6229 * doc/extend.texi: Document __builtin_speculation_barrier.
6231 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6234 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6235 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6236 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6238 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6239 integral modes instead of "ss" and "sd".
6240 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6241 vectors with 32-bit and 64-bit elements.
6242 (vecdupssescalarmodesuffix): New mode attribute.
6243 (vec_dup<mode>): Use it.
6245 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6248 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6249 frame if argument is passed on stack.
6251 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6254 * ree.c (combine_reaching_defs): Optimize also
6255 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6256 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6258 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6261 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6263 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6266 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6267 after they are computed.
6269 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6271 PR tree-optimization/83695
6272 * gimple-loop-linterchange.cc
6273 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6274 reset cached scev information after interchange.
6275 (pass_linterchange::execute): Remove call to scev_reset_htab.
6277 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6279 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6280 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6281 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6282 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6283 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6284 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6285 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6286 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6287 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6288 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6289 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6290 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6291 (V_lane_reg): Likewise.
6292 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6294 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6295 (vfmal_lane_low<mode>_intrinsic,
6296 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6297 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6298 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6299 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6300 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6301 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6303 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6305 * config/arm/arm-cpus.in (fp16fml): New feature.
6306 (ALL_SIMD): Add fp16fml.
6307 (armv8.2-a): Add fp16fml as an option.
6308 (armv8.3-a): Likewise.
6309 (armv8.4-a): Add fp16fml as part of fp16.
6310 * config/arm/arm.h (TARGET_FP16FML): Define.
6311 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6313 * config/arm/arm-modes.def (V2HF): Define.
6314 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6315 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6316 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6317 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6318 vfmsl_low, vfmsl_high): New set of builtins.
6319 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6320 (vfml_op): New code attribute.
6321 (VFMLHALVES): New int iterator.
6322 (VFML, VFMLSEL): New mode attributes.
6323 (V_reg): Define mapping for V2HF.
6324 (V_hi, V_lo): New mode attributes.
6325 (VF_constraint): Likewise.
6326 (vfml_half, vfml_half_selector): New int attributes.
6327 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6329 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6330 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6332 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6333 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6334 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6335 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6337 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6338 Document new effective target and option set.
6340 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6342 * config/arm/arm-cpus.in (armv8_4): New feature.
6343 (ARMv8_4a): New fgroup.
6344 (armv8.4-a): New arch.
6345 * config/arm/arm-tables.opt: Regenerate.
6346 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6347 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6348 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6349 Add matching rules for -march=armv8.4-a and extensions.
6350 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6352 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6355 * config/rx/rx.md (BW): New mode attribute.
6356 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6358 2018-01-11 Richard Biener <rguenther@suse.de>
6360 PR tree-optimization/83435
6361 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6362 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6363 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6365 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6366 Alan Hayward <alan.hayward@arm.com>
6367 David Sherwood <david.sherwood@arm.com>
6369 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6371 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6372 (aarch64_print_address_internal): Use it to check for a zero offset.
6374 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6375 Alan Hayward <alan.hayward@arm.com>
6376 David Sherwood <david.sherwood@arm.com>
6378 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6379 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6380 Return a poly_int64 rather than a HOST_WIDE_INT.
6381 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6382 rather than a HOST_WIDE_INT.
6383 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6384 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6385 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6386 final_offset from HOST_WIDE_INT to poly_int64.
6387 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6388 to_constant when getting the number of units in an Advanced SIMD
6390 (aarch64_builtin_vectorized_function): Check for a constant number
6392 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6394 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6395 attribute instead of GET_MODE_NUNITS.
6396 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6397 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6398 GET_MODE_SIZE for fixed-size registers.
6399 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6400 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6401 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6402 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6403 (aarch64_print_operand, aarch64_print_address_internal)
6404 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6405 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6406 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6407 Handle polynomial GET_MODE_SIZE.
6408 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6409 wider than SImode without modification.
6410 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6411 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6412 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6413 passing and returning SVE modes.
6414 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6415 rather than GEN_INT.
6416 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6417 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6418 (aarch64_allocate_and_probe_stack_space): Likewise.
6419 (aarch64_layout_frame): Cope with polynomial offsets.
6420 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6421 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6423 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6424 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6425 poly_int64 rather than a HOST_WIDE_INT.
6426 (aarch64_get_separate_components, aarch64_process_components)
6427 (aarch64_expand_prologue, aarch64_expand_epilogue)
6428 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6429 (aarch64_anchor_offset): New function, split out from...
6430 (aarch64_legitimize_address): ...here.
6431 (aarch64_builtin_vectorization_cost): Handle polynomial
6432 TYPE_VECTOR_SUBPARTS.
6433 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6435 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6436 number of elements from the PARALLEL rather than the mode.
6437 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6438 rather than GET_MODE_BITSIZE.
6439 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6440 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6441 (aarch64_expand_vec_perm_const_1): Handle polynomial
6442 d->perm.length () and d->perm elements.
6443 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6444 Apply to_constant to d->perm elements.
6445 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6446 polynomial CONST_VECTOR_NUNITS.
6447 (aarch64_move_pointer): Take amount as a poly_int64 rather
6449 (aarch64_progress_pointer): Avoid temporary variable.
6450 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6451 the mode attribute instead of GET_MODE.
6453 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6454 Alan Hayward <alan.hayward@arm.com>
6455 David Sherwood <david.sherwood@arm.com>
6457 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6458 x exists before using it.
6459 (aarch64_add_constant_internal): Rename to...
6460 (aarch64_add_offset_1): ...this. Replace regnum with separate
6461 src and dest rtxes. Handle the case in which they're different,
6462 including when the offset is zero. Replace scratchreg with an rtx.
6463 Use 2 additions if there is no spare register into which we can
6464 move a 16-bit constant.
6465 (aarch64_add_constant): Delete.
6466 (aarch64_add_offset): Replace reg with separate src and dest
6467 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6468 Use aarch64_add_offset_1.
6469 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6470 an rtx rather than an int. Take the delta as a poly_int64
6471 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6472 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6473 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6474 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6475 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6477 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6478 aarch64_add_constant.
6480 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6482 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6483 Use scalar_float_mode.
6485 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6487 * config/aarch64/aarch64-simd.md
6488 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6489 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6490 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6491 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6492 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6493 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6494 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6495 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6496 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6497 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6499 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6502 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6503 targ_options->x_arm_arch_string is non NULL.
6505 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6507 * config/aarch64/aarch64.h
6508 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6510 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6513 * expmed.c (emit_store_flag_force): Swap if const op0
6514 and change VOIDmode to mode of op0.
6516 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6518 PR rtl-optimization/83761
6519 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6520 than bytes to mode_for_size.
6522 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6525 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6526 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6529 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6532 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6533 when in layout mode.
6534 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6535 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6538 2018-01-10 Michael Collison <michael.collison@arm.com>
6540 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6541 * config/aarch64/aarch64-option-extension.def: Add
6542 AARCH64_OPT_EXTENSION of 'fp16fml'.
6543 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6544 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6545 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6546 * config/aarch64/constraints.md (Ui7): New constraint.
6547 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6548 (VFMLA_SEL_W): Ditto.
6551 (VFMLA16_LOW): New int iterator.
6552 (VFMLA16_HIGH): Ditto.
6553 (UNSPEC_FMLAL): New unspec.
6554 (UNSPEC_FMLSL): Ditto.
6555 (UNSPEC_FMLAL2): Ditto.
6556 (UNSPEC_FMLSL2): Ditto.
6557 (f16mac): New code attribute.
6558 * config/aarch64/aarch64-simd-builtins.def
6559 (aarch64_fmlal_lowv2sf): Ditto.
6560 (aarch64_fmlsl_lowv2sf): Ditto.
6561 (aarch64_fmlalq_lowv4sf): Ditto.
6562 (aarch64_fmlslq_lowv4sf): Ditto.
6563 (aarch64_fmlal_highv2sf): Ditto.
6564 (aarch64_fmlsl_highv2sf): Ditto.
6565 (aarch64_fmlalq_highv4sf): Ditto.
6566 (aarch64_fmlslq_highv4sf): Ditto.
6567 (aarch64_fmlal_lane_lowv2sf): Ditto.
6568 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6569 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6570 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6571 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6572 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6573 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6574 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6575 (aarch64_fmlal_lane_highv2sf): Ditto.
6576 (aarch64_fmlsl_lane_highv2sf): Ditto.
6577 (aarch64_fmlal_laneq_highv2sf): Ditto.
6578 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6579 (aarch64_fmlalq_lane_highv4sf): Ditto.
6580 (aarch64_fmlsl_lane_highv4sf): Ditto.
6581 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6582 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6583 * config/aarch64/aarch64-simd.md:
6584 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6585 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6586 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6587 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6588 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6589 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6590 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6591 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6592 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6593 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6594 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6595 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6596 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6597 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6598 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6599 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6600 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6601 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6602 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6603 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6604 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6605 (vfmlsl_low_u32): Ditto.
6606 (vfmlalq_low_u32): Ditto.
6607 (vfmlslq_low_u32): Ditto.
6608 (vfmlal_high_u32): Ditto.
6609 (vfmlsl_high_u32): Ditto.
6610 (vfmlalq_high_u32): Ditto.
6611 (vfmlslq_high_u32): Ditto.
6612 (vfmlal_lane_low_u32): Ditto.
6613 (vfmlsl_lane_low_u32): Ditto.
6614 (vfmlal_laneq_low_u32): Ditto.
6615 (vfmlsl_laneq_low_u32): Ditto.
6616 (vfmlalq_lane_low_u32): Ditto.
6617 (vfmlslq_lane_low_u32): Ditto.
6618 (vfmlalq_laneq_low_u32): Ditto.
6619 (vfmlslq_laneq_low_u32): Ditto.
6620 (vfmlal_lane_high_u32): Ditto.
6621 (vfmlsl_lane_high_u32): Ditto.
6622 (vfmlal_laneq_high_u32): Ditto.
6623 (vfmlsl_laneq_high_u32): Ditto.
6624 (vfmlalq_lane_high_u32): Ditto.
6625 (vfmlslq_lane_high_u32): Ditto.
6626 (vfmlalq_laneq_high_u32): Ditto.
6627 (vfmlslq_laneq_high_u32): Ditto.
6628 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6629 (AARCH64_FL_FOR_ARCH8_4): New.
6630 (AARCH64_ISA_F16FML): New ISA flag.
6631 (TARGET_F16FML): New feature flag for fp16fml.
6632 (doc/invoke.texi): Document new fp16fml option.
6634 2018-01-10 Michael Collison <michael.collison@arm.com>
6636 * config/aarch64/aarch64-builtins.c:
6637 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6638 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6639 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6640 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6641 (AARCH64_ISA_SHA3): New ISA flag.
6642 (TARGET_SHA3): New feature flag for sha3.
6643 * config/aarch64/iterators.md (sha512_op): New int attribute.
6644 (CRYPTO_SHA512): New int iterator.
6645 (UNSPEC_SHA512H): New unspec.
6646 (UNSPEC_SHA512H2): Ditto.
6647 (UNSPEC_SHA512SU0): Ditto.
6648 (UNSPEC_SHA512SU1): Ditto.
6649 * config/aarch64/aarch64-simd-builtins.def
6650 (aarch64_crypto_sha512hqv2di): New builtin.
6651 (aarch64_crypto_sha512h2qv2di): Ditto.
6652 (aarch64_crypto_sha512su0qv2di): Ditto.
6653 (aarch64_crypto_sha512su1qv2di): Ditto.
6654 (aarch64_eor3qv8hi): Ditto.
6655 (aarch64_rax1qv2di): Ditto.
6656 (aarch64_xarqv2di): Ditto.
6657 (aarch64_bcaxqv8hi): Ditto.
6658 * config/aarch64/aarch64-simd.md:
6659 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6660 (aarch64_crypto_sha512su0qv2di): Ditto.
6661 (aarch64_crypto_sha512su1qv2di): Ditto.
6662 (aarch64_eor3qv8hi): Ditto.
6663 (aarch64_rax1qv2di): Ditto.
6664 (aarch64_xarqv2di): Ditto.
6665 (aarch64_bcaxqv8hi): Ditto.
6666 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6667 (vsha512h2q_u64): Ditto.
6668 (vsha512su0q_u64): Ditto.
6669 (vsha512su1q_u64): Ditto.
6670 (veor3q_u16): Ditto.
6671 (vrax1q_u64): Ditto.
6673 (vbcaxq_u16): Ditto.
6674 * config/arm/types.md (crypto_sha512): New type attribute.
6675 (crypto_sha3): Ditto.
6676 (doc/invoke.texi): Document new sha3 option.
6678 2018-01-10 Michael Collison <michael.collison@arm.com>
6680 * config/aarch64/aarch64-builtins.c:
6681 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6682 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6683 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6684 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6685 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6686 (AARCH64_ISA_SM4): New ISA flag.
6687 (TARGET_SM4): New feature flag for sm4.
6688 * config/aarch64/aarch64-simd-builtins.def
6689 (aarch64_sm3ss1qv4si): Ditto.
6690 (aarch64_sm3tt1aq4si): Ditto.
6691 (aarch64_sm3tt1bq4si): Ditto.
6692 (aarch64_sm3tt2aq4si): Ditto.
6693 (aarch64_sm3tt2bq4si): Ditto.
6694 (aarch64_sm3partw1qv4si): Ditto.
6695 (aarch64_sm3partw2qv4si): Ditto.
6696 (aarch64_sm4eqv4si): Ditto.
6697 (aarch64_sm4ekeyqv4si): Ditto.
6698 * config/aarch64/aarch64-simd.md:
6699 (aarch64_sm3ss1qv4si): Ditto.
6700 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6701 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6702 (aarch64_sm4eqv4si): Ditto.
6703 (aarch64_sm4ekeyqv4si): Ditto.
6704 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6705 (sm3part_op): Ditto.
6706 (CRYPTO_SM3TT): Ditto.
6707 (CRYPTO_SM3PART): Ditto.
6708 (UNSPEC_SM3SS1): New unspec.
6709 (UNSPEC_SM3TT1A): Ditto.
6710 (UNSPEC_SM3TT1B): Ditto.
6711 (UNSPEC_SM3TT2A): Ditto.
6712 (UNSPEC_SM3TT2B): Ditto.
6713 (UNSPEC_SM3PARTW1): Ditto.
6714 (UNSPEC_SM3PARTW2): Ditto.
6715 (UNSPEC_SM4E): Ditto.
6716 (UNSPEC_SM4EKEY): Ditto.
6717 * config/aarch64/constraints.md (Ui2): New constraint.
6718 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6719 * config/arm/types.md (crypto_sm3): New type attribute.
6720 (crypto_sm4): Ditto.
6721 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6722 (vsm3tt1aq_u32): Ditto.
6723 (vsm3tt1bq_u32): Ditto.
6724 (vsm3tt2aq_u32): Ditto.
6725 (vsm3tt2bq_u32): Ditto.
6726 (vsm3partw1q_u32): Ditto.
6727 (vsm3partw2q_u32): Ditto.
6728 (vsm4eq_u32): Ditto.
6729 (vsm4ekeyq_u32): Ditto.
6730 (doc/invoke.texi): Document new sm4 option.
6732 2018-01-10 Michael Collison <michael.collison@arm.com>
6734 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6735 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6736 (AARCH64_FL_FOR_ARCH8_4): New.
6737 (AARCH64_FL_V8_4): New flag.
6738 (doc/invoke.texi): Document new armv8.4-a option.
6740 2018-01-10 Michael Collison <michael.collison@arm.com>
6742 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6743 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6744 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6745 * config/aarch64/aarch64-option-extension.def: Add
6746 AARCH64_OPT_EXTENSION of 'sha2'.
6747 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6748 (crypto): Disable sha2 and aes if crypto disabled.
6749 (crypto): Enable aes and sha2 if enabled.
6750 (simd): Disable sha2 and aes if simd disabled.
6751 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6753 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6754 (TARGET_SHA2): New feature flag for sha2.
6755 (TARGET_AES): New feature flag for aes.
6756 * config/aarch64/aarch64-simd.md:
6757 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6758 conditional on TARGET_AES.
6759 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6760 (aarch64_crypto_sha1hsi): Make pattern conditional
6762 (aarch64_crypto_sha1hv4si): Ditto.
6763 (aarch64_be_crypto_sha1hv4si): Ditto.
6764 (aarch64_crypto_sha1su1v4si): Ditto.
6765 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6766 (aarch64_crypto_sha1su0v4si): Ditto.
6767 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6768 (aarch64_crypto_sha256su0v4si): Ditto.
6769 (aarch64_crypto_sha256su1v4si): Ditto.
6770 (doc/invoke.texi): Document new aes and sha2 options.
6772 2018-01-10 Martin Sebor <msebor@redhat.com>
6774 PR tree-optimization/83781
6775 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6778 2018-01-11 Martin Sebor <msebor@gmail.com>
6779 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6781 PR tree-optimization/83501
6782 PR tree-optimization/81703
6784 * tree-ssa-strlen.c (get_string_cst): Rename...
6785 (get_string_len): ...to this. Handle global constants.
6786 (handle_char_store): Adjust.
6788 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6789 Jim Wilson <jimw@sifive.com>
6791 * config/riscv/riscv-protos.h (riscv_output_return): New.
6792 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6793 (riscv_attribute_table, riscv_output_return),
6794 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6795 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6796 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6798 (riscv_expand_prologue): Add early return for naked function.
6799 (riscv_expand_epilogue): Likewise.
6800 (riscv_function_ok_for_sibcall): Return false for naked function.
6801 (riscv_set_current_function): New.
6802 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6803 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6804 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6805 * doc/extend.texi (RISC-V Function Attributes): New.
6807 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6809 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6810 check for 128-bit long double before checking TCmode.
6811 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6812 128-bit long doubles before checking TFmode or TCmode.
6813 (FLOAT128_IBM_P): Likewise.
6815 2018-01-10 Martin Sebor <msebor@redhat.com>
6817 PR tree-optimization/83671
6818 * builtins.c (c_strlen): Unconditionally return zero for the empty
6820 Use -Warray-bounds for warnings.
6821 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6822 for non-constant array indices with COMPONENT_REF, arrays of
6823 arrays, and pointers to arrays.
6824 (gimple_fold_builtin_strlen): Determine and set length range for
6825 non-constant character arrays.
6827 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6830 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6833 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6835 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6837 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6840 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6841 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6842 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6843 indexed_or_indirect_operand predicate.
6844 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6845 (*vsx_le_perm_load_v8hi): Likewise.
6846 (*vsx_le_perm_load_v16qi): Likewise.
6847 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6848 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6849 (*vsx_le_perm_store_v8hi): Likewise.
6850 (*vsx_le_perm_store_v16qi): Likewise.
6851 (eight unnamed splitters): Likewise.
6853 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6855 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6856 * config/rs6000/emmintrin.h: Likewise.
6857 * config/rs6000/mmintrin.h: Likewise.
6858 * config/rs6000/xmmintrin.h: Likewise.
6860 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6863 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6865 * tree.c (tree_nop_conversion): Return true for location wrapper
6867 (maybe_wrap_with_location): New function.
6868 (selftest::check_strip_nops): New function.
6869 (selftest::test_location_wrappers): New function.
6870 (selftest::tree_c_tests): Call it.
6871 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6872 (maybe_wrap_with_location): New decl.
6873 (EXPR_LOCATION_WRAPPER_P): New macro.
6874 (location_wrapper_p): New inline function.
6875 (tree_strip_any_location_wrapper): New inline function.
6877 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6880 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6881 stack_realign_offset for the largest alignment of stack slot
6883 (ix86_find_max_used_stack_alignment): New function.
6884 (ix86_finalize_stack_frame_flags): Use it. Set
6885 max_used_stack_alignment if we don't realign stack.
6886 * config/i386/i386.h (machine_function): Add
6887 max_used_stack_alignment.
6889 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6891 * config/arm/arm.opt (-mbranch-cost): New option.
6892 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6895 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6898 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6899 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6901 2018-01-10 Richard Biener <rguenther@suse.de>
6904 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6905 early out so it also covers the case where we have a non-NULL
6908 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6910 PR tree-optimization/83753
6911 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6912 for non-strided grouped accesses if the number of elements is 1.
6914 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6917 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6918 * i386.h (TARGET_USE_GATHER): Define.
6919 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6921 2018-01-10 Martin Liska <mliska@suse.cz>
6924 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6925 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6927 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6928 CLEANUP_NO_PARTITIONING is not set.
6930 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6932 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6933 for vectors, as a partial revert of r254296.
6934 * rtl.h (const_vec_p): Delete.
6935 (const_vec_duplicate_p): Don't test for vector CONSTs.
6936 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6937 * expmed.c (make_tree): Likewise.
6940 * common.md (E, F): Use CONSTANT_P instead of checking for
6942 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6943 checking for CONST_VECTOR.
6945 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6948 * predict.c (force_edge_cold): Handle in more sane way edges
6951 2018-01-09 Carl Love <cel@us.ibm.com>
6953 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6955 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6956 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6957 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6958 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6959 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6960 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6961 * config/rs6000/rs6000-protos.h: Add extern defition for
6962 rs6000_generate_float2_double_code.
6963 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6965 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6966 (float2_v2df): Add define_expand.
6968 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6971 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6972 op_mode in the force_to_mode call.
6974 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6976 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6977 instead of checking each element individually.
6978 (aarch64_evpc_uzp): Likewise.
6979 (aarch64_evpc_zip): Likewise.
6980 (aarch64_evpc_ext): Likewise.
6981 (aarch64_evpc_rev): Likewise.
6982 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6983 instead of checking each element individually. Return true without
6985 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6986 whether all selected elements come from the same input, instead of
6987 checking each element individually. Remove calls to gen_rtx_REG,
6988 start_sequence and end_sequence and instead assert that no rtl is
6991 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6993 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6994 order of HIGH and CONST checks.
6996 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6998 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6999 if the destination isn't an SSA_NAME.
7001 2018-01-09 Richard Biener <rguenther@suse.de>
7003 PR tree-optimization/83668
7004 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7006 (canonicalize_loop_form): ... here, renamed from ...
7007 (canonicalize_loop_closed_ssa_form): ... this and amended to
7008 swap successor edges for loop exit blocks to make us use
7009 the RPO order we need for initial schedule generation.
7011 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7013 PR tree-optimization/64811
7014 * match.pd: When optimizing comparisons with Inf, avoid
7015 introducing or losing exceptions from comparisons with NaN.
7017 2018-01-09 Martin Liska <mliska@suse.cz>
7020 * asan.c (shadow_mem_size): Add gcc_assert.
7022 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7024 Don't save registers in main().
7027 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7028 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7029 * config/avr/avr.c (avr_set_current_function): Don't error if
7030 naked, OS_task or OS_main are specified at the same time.
7031 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7033 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7035 * common/config/avr/avr-common.c (avr_option_optimization_table):
7036 Switch on -mmain-is-OS_task for optimizing compilations.
7038 2018-01-09 Richard Biener <rguenther@suse.de>
7040 PR tree-optimization/83572
7041 * graphite.c: Include cfganal.h.
7042 (graphite_transform_loops): Connect infinite loops to exit
7043 and remove fake edges at the end.
7045 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7047 * ipa-inline.c (edge_badness): Revert accidental checkin.
7049 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7052 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7053 symbols; not inline clones.
7055 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7058 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7059 hard registers. Formatting fixes.
7061 PR preprocessor/83722
7062 * gcc.c (try_generate_repro): Pass
7063 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7064 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7067 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7068 Kito Cheng <kito.cheng@gmail.com>
7070 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7071 (riscv_leaf_function_p): Delete.
7072 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7074 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7076 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7078 (do_ifelse): New function.
7079 (do_isel): New function.
7080 (do_sub3): New function.
7081 (do_add3): New function.
7082 (do_load_mask_compare): New function.
7083 (do_overlap_load_compare): New function.
7084 (expand_compare_loop): New function.
7085 (expand_block_compare): Call expand_compare_loop() when appropriate.
7086 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7088 (-mblock-compare-inline-loop-limit): New option.
7090 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7093 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7094 Reverse order of second and third operands in first alternative.
7095 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7096 of first and second elements in UNSPEC_VPERMR vector.
7097 (altivec_expand_vec_perm_le): Likewise.
7099 2018-01-08 Jeff Law <law@redhat.com>
7101 PR rtl-optimizatin/81308
7102 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7103 (process_switch): If group_case_labels makes a change, then set
7105 (pass_convert_switch::execute): If a switch is converted, then
7106 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7108 PR rtl-optimization/81308
7109 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7112 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7114 PR target/83663 - Revert r255946
7115 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7116 generation for cases where splatting a value is not useful.
7117 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7118 across a vec_duplicate and a paradoxical subreg forming a vector
7119 mode to a vec_concat.
7121 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7123 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7124 -march=armv8.3-a variants.
7125 * config/arm/t-multilib: Likewise.
7126 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7128 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7130 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7132 (cceq_ior_compare_complement): Give it a name so I can use it, and
7133 change boolean_or_operator predicate to boolean_operator so it can
7134 be used to generate a crand.
7135 (eqne): New code iterator.
7136 (bd/bd_neg): New code_attrs.
7137 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7138 a single define_insn.
7139 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7140 decrement (bdnzt/bdnzf/bdzt/bdzf).
7141 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7142 with the new names of the branch decrement patterns, and added the
7143 names of the branch decrement conditional patterns.
7145 2018-01-08 Richard Biener <rguenther@suse.de>
7147 PR tree-optimization/83563
7148 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7151 2018-01-08 Richard Biener <rguenther@suse.de>
7154 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7156 2018-01-08 Richard Biener <rguenther@suse.de>
7158 PR tree-optimization/83685
7159 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7160 references to abnormals.
7162 2018-01-08 Richard Biener <rguenther@suse.de>
7165 * dwarf2out.c (output_indirect_strings): Handle empty
7166 skeleton_debug_str_hash.
7167 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7169 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7171 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7172 (emit_store_direct): Likewise.
7173 (arc_trampoline_adjust_address): Likewise.
7174 (arc_asm_trampoline_template): New function.
7175 (arc_initialize_trampoline): Use asm_trampoline_template.
7176 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7177 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7178 * config/arc/arc.md (flush_icache): Delete pattern.
7180 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7182 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7183 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7186 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7189 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7190 by not USED_FOR_TARGET.
7191 (make_pass_resolve_sw_modes): Likewise.
7193 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7195 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7198 2018-01-08 Richard Biener <rguenther@suse.de>
7201 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7203 2018-01-08 Richard Biener <rguenther@suse.de>
7206 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7208 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7211 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7212 basic blocks with a small number of successors.
7213 (convert_control_dep_chain_into_preds): Improve handling of
7215 (dump_predicates): Split apart into...
7216 (dump_pred_chain): ...here...
7217 (dump_pred_info): ...and here.
7218 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7219 (can_chain_union_be_invalidated_p): Improve check for invalidation
7221 (uninit_uses_cannot_happen): Avoid unnecessary if
7222 convert_control_dep_chain_into_preds yielded nothing.
7224 2018-01-06 Martin Sebor <msebor@redhat.com>
7226 PR tree-optimization/83640
7227 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7228 subtracting negative offset from size.
7229 (builtin_access::overlap): Adjust offset bounds of the access to fall
7230 within the size of the object if possible.
7232 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7234 PR rtl-optimization/83699
7235 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7236 extract_bit_field_as_subreg to cases in which the extracted
7237 value is also a vector.
7239 * lra-constraints.c (process_alt_operands): Test for the equivalence
7240 substitutions when detecting a possible reload cycle.
7242 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7245 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7246 by default if flag_selective_schedling{,2}. Formatting fixes.
7248 PR rtl-optimization/83682
7249 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7250 if it has non-VECTOR_MODE element mode.
7251 (vec_duplicate_p): Likewise.
7254 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7255 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7257 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7260 * config/i386/i386-builtin.def
7261 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7262 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7263 Require also OPTION_MASK_ISA_AVX512F in addition to
7264 OPTION_MASK_ISA_GFNI.
7265 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7266 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7267 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7268 to OPTION_MASK_ISA_GFNI.
7269 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7270 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7271 OPTION_MASK_ISA_AVX512BW.
7272 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7273 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7274 addition to OPTION_MASK_ISA_GFNI.
7275 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7276 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7277 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7278 to OPTION_MASK_ISA_GFNI.
7279 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7280 a requirement for all ISAs rather than any of them with a few
7282 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7284 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7285 bitmasks to be enabled with 3 exceptions, instead of requiring any
7286 enabled ISA with lots of exceptions.
7287 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7288 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7289 Change avx512bw in isa attribute to avx512f.
7290 * config/i386/sgxintrin.h: Add license boilerplate.
7291 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7292 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7293 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7294 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7296 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7297 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7298 temporarily sse2 rather than sse if not enabled already.
7301 * config/i386/sse.md (VI248_VLBW): Rename to ...
7302 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7303 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7304 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7305 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7306 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7307 mode iterator instead of VI248_VLBW.
7309 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7311 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7312 (record_modified): Skip clobbers; add debug output.
7313 (param_change_prob): Use sreal frequencies.
7315 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7317 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7318 punt for user-aligned variables.
7320 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7322 * tree-chrec.c (chrec_contains_symbols): Return true for
7325 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7328 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7329 of (x|y) == x for BICS pattern.
7331 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7333 PR tree-optimization/83605
7334 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7335 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7338 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7340 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7341 * config/epiphany/rtems.h: New file.
7343 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7344 Uros Bizjak <ubizjak@gmail.com>
7347 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7348 QIreg_operand instead of register_operand predicate.
7349 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7350 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7351 comments instead of -fmitigate[-_]rop.
7353 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7356 * cgraphunit.c (symbol_table::compile): Switch to text_section
7357 before calling assembly_start debug hook.
7358 * run-rtl-passes.c (run_rtl_passes): Likewise.
7361 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7363 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7364 range_int_cst_p rather than !symbolic_range_p before calling
7365 extract_range_from_multiplicative_op_1.
7367 2018-01-04 Jeff Law <law@redhat.com>
7369 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7370 redundant test in assertion.
7372 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7374 * doc/rtl.texi: Document machine_mode wrapper classes.
7376 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7378 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7381 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7383 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7384 the VEC_PERM_EXPR fold to fail.
7386 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7389 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7390 to switched_sections.
7392 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7395 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7398 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7401 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7402 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7404 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7407 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7408 is BLKmode and bitpos not zero or mode change is needed.
7410 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7413 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7416 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7419 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7420 instead of MULT rtx. Update all corresponding splitters.
7422 (*ssub<modesuffix>): Ditto.
7424 (*cmp_sadd_di): Update split patterns.
7425 (*cmp_sadd_si): Ditto.
7426 (*cmp_sadd_sidi): Ditto.
7427 (*cmp_ssub_di): Ditto.
7428 (*cmp_ssub_si): Ditto.
7429 (*cmp_ssub_sidi): Ditto.
7430 * config/alpha/predicates.md (const23_operand): New predicate.
7431 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7432 Look for ASHIFT, not MULT inner operand.
7433 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7435 2018-01-04 Martin Liska <mliska@suse.cz>
7437 PR gcov-profile/83669
7438 * gcov.c (output_intermediate_file): Add version to intermediate
7440 * doc/gcov.texi: Document new field 'version' in intermediate
7441 file format. Fix location of '-k' option of gcov command.
7443 2018-01-04 Martin Liska <mliska@suse.cz>
7446 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7448 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7450 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7452 2018-01-03 Martin Sebor <msebor@redhat.com>
7454 PR tree-optimization/83655
7455 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7456 checking calls with invalid arguments.
7458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7460 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7461 (vectorizable_mask_load_store): Delete.
7462 (vectorizable_call): Return false for masked loads and stores.
7463 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7464 instead of gimple_assign_rhs1.
7465 (vectorizable_load): Handle IFN_MASK_LOAD.
7466 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7470 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7472 (vectorizable_mask_load_store): ...here.
7473 (vectorizable_load): ...and here.
7475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7477 * tree-vect-stmts.c (vect_build_all_ones_mask)
7478 (vect_build_zero_merge_argument): New functions, split out from...
7479 (vectorizable_load): ...here.
7481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7483 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7485 (vectorizable_mask_load_store): ...here.
7486 (vectorizable_store): ...and here.
7488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7490 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7492 (vectorizable_mask_load_store): ...here.
7494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7496 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7497 (vect_model_store_cost): Take a vec_load_store_type instead of a
7499 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7500 (vect_model_store_cost): Take a vec_load_store_type instead of a
7502 (vectorizable_mask_load_store): Update accordingly.
7503 (vectorizable_store): Likewise.
7504 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7508 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7509 IFN_MASK_LOAD calls here rather than...
7510 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7513 Alan Hayward <alan.hayward@arm.com>
7514 David Sherwood <david.sherwood@arm.com>
7516 * expmed.c (extract_bit_field_1): For vector extracts,
7517 fall back to extract_bit_field_as_subreg if vec_extract
7520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7521 Alan Hayward <alan.hayward@arm.com>
7522 David Sherwood <david.sherwood@arm.com>
7524 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7525 they are variable or constant sized.
7526 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7527 slots for constant-sized data.
7529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7530 Alan Hayward <alan.hayward@arm.com>
7531 David Sherwood <david.sherwood@arm.com>
7533 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7534 handling COND_EXPRs with boolean comparisons, try to find a better
7535 basis for the mask type than the boolean itself.
7537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7539 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7540 is calculated and how it can be overridden.
7541 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7542 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7544 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7548 Alan Hayward <alan.hayward@arm.com>
7549 David Sherwood <david.sherwood@arm.com>
7551 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7552 Remove the mode argument.
7553 (aarch64_simd_valid_immediate): Remove the mode and inverse
7555 * config/aarch64/iterators.md (bitsize): New iterator.
7556 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7557 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7558 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7559 aarch64_simd_valid_immediate.
7560 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7561 (aarch64_reg_or_bic_imm): Likewise.
7562 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7563 with an insn_type enum and msl with a modifier_type enum.
7564 Replace element_width with a scalar_mode. Change the shift
7565 to unsigned int. Add constructors for scalar_float_mode and
7566 scalar_int_mode elements.
7567 (aarch64_vect_float_const_representable_p): Delete.
7568 (aarch64_can_const_movi_rtx_p)
7569 (aarch64_simd_scalar_immediate_valid_for_move)
7570 (aarch64_simd_make_constant): Update call to
7571 aarch64_simd_valid_immediate.
7572 (aarch64_advsimd_valid_immediate_hs): New function.
7573 (aarch64_advsimd_valid_immediate): Likewise.
7574 (aarch64_simd_valid_immediate): Remove mode and inverse
7575 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7576 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7577 and aarch64_float_const_representable_p on the result.
7578 (aarch64_output_simd_mov_immediate): Remove mode argument.
7579 Update call to aarch64_simd_valid_immediate and use of
7580 simd_immediate_info.
7581 (aarch64_output_scalar_simd_mov_immediate): Update call
7584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7585 Alan Hayward <alan.hayward@arm.com>
7586 David Sherwood <david.sherwood@arm.com>
7588 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7589 (mode_nunits): Likewise CONST_MODE_NUNITS.
7590 * machmode.def (ADJUST_NUNITS): Document.
7591 * genmodes.c (mode_data::need_nunits_adj): New field.
7592 (blank_mode): Update accordingly.
7593 (adj_nunits): New variable.
7594 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7596 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7597 listed in adj_nunits.
7598 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7599 listed in adj_nunits. Don't emit case statements for such modes.
7600 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7601 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7602 nothing if adj_nunits is nonnull.
7603 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7604 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7605 (emit_mode_fbit): Update use of print_maybe_const_decl.
7606 (emit_move_size): Likewise. Treat the array as non-const
7608 (emit_mode_adjustments): Handle adj_nunits.
7610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7612 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7613 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7614 (VECTOR_MODES): Use it.
7615 (make_vector_modes): Take the prefix as an argument.
7617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7618 Alan Hayward <alan.hayward@arm.com>
7619 David Sherwood <david.sherwood@arm.com>
7621 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7622 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7623 for MODE_VECTOR_BOOL.
7624 * machmode.def (VECTOR_BOOL_MODE): Document.
7625 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7626 (make_vector_bool_mode): New function.
7627 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7629 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7630 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7632 * stor-layout.c (int_mode_for_mode): Likewise.
7633 * tree.c (build_vector_type_for_mode): Likewise.
7634 * varasm.c (output_constant_pool_2): Likewise.
7635 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7636 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7637 for MODE_VECTOR_BOOL.
7638 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7639 of mode class checks.
7640 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7641 instead of a list of mode class checks.
7642 (expand_vector_scalar_condition): Likewise.
7643 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7646 Alan Hayward <alan.hayward@arm.com>
7647 David Sherwood <david.sherwood@arm.com>
7649 * machmode.h (mode_size): Change from unsigned short to
7651 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7652 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7653 or if measurement_type is not polynomial.
7654 (fixed_size_mode::includes_p): Check for constant-sized modes.
7655 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7656 return a poly_uint16 rather than an unsigned short.
7657 (emit_mode_size): Change the type of mode_size from unsigned short
7658 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7659 (emit_mode_adjustments): Cope with polynomial vector sizes.
7660 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7662 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7664 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7665 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7666 * caller-save.c (setup_save_areas): Likewise.
7667 (replace_reg_with_saved_mem): Likewise.
7668 * calls.c (emit_library_call_value_1): Likewise.
7669 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7670 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7671 (gen_lowpart_for_combine): Likewise.
7672 * convert.c (convert_to_integer_1): Likewise.
7673 * cse.c (equiv_constant, cse_insn): Likewise.
7674 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7675 (cselib_subst_to_values): Likewise.
7676 * dce.c (word_dce_process_block): Likewise.
7677 * df-problems.c (df_word_lr_mark_ref): Likewise.
7678 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7679 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7680 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7681 (rtl_for_decl_location): Likewise.
7682 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7683 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7684 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7685 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7686 (expand_expr_real_1): Likewise.
7687 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7688 (pad_below): Likewise.
7689 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7690 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7691 * ira.c (get_subreg_tracking_sizes): Likewise.
7692 * ira-build.c (ira_create_allocno_objects): Likewise.
7693 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7694 (ira_sort_regnos_for_alter_reg): Likewise.
7695 * ira-costs.c (record_operand_costs): Likewise.
7696 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7697 (resolve_simple_move): Likewise.
7698 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7699 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7700 (lra_constraints): Likewise.
7701 (CONST_POOL_OK_P): Reject variable-sized modes.
7702 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7703 (add_pseudo_to_slot, lra_spill): Likewise.
7704 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7705 * optabs-query.c (get_best_extraction_insn): Likewise.
7706 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7707 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7708 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7709 * recog.c (offsettable_address_addr_space_p): Likewise.
7710 * regcprop.c (maybe_mode_change): Likewise.
7711 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7712 * regrename.c (build_def_use): Likewise.
7713 * regstat.c (dump_reg_info): Likewise.
7714 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7715 (find_reloads, find_reloads_subreg_address): Likewise.
7716 * reload1.c (eliminate_regs_1): Likewise.
7717 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7718 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7719 (simplify_binary_operation_1, simplify_subreg): Likewise.
7720 * targhooks.c (default_function_arg_padding): Likewise.
7721 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7722 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7723 (verify_gimple_assign_ternary): Likewise.
7724 * tree-inline.c (estimate_move_cost): Likewise.
7725 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7726 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7727 (get_address_cost_ainc): Likewise.
7728 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7729 (vect_supportable_dr_alignment): Likewise.
7730 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7731 (vectorizable_reduction): Likewise.
7732 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7733 (vectorizable_operation, vectorizable_load): Likewise.
7734 * tree.c (build_same_sized_truth_vector_type): Likewise.
7735 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7736 * var-tracking.c (emit_note_insn_var_location): Likewise.
7737 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7738 (ADDR_VEC_ALIGN): Likewise.
7740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7741 Alan Hayward <alan.hayward@arm.com>
7742 David Sherwood <david.sherwood@arm.com>
7744 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7746 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7747 or if measurement_type is polynomial.
7748 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7749 * combine.c (make_extraction): Likewise.
7750 * dse.c (find_shift_sequence): Likewise.
7751 * dwarf2out.c (mem_loc_descriptor): Likewise.
7752 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7753 (extract_bit_field, extract_low_bits): Likewise.
7754 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7755 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7756 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7757 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7758 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7759 * reload.c (find_reloads): Likewise.
7760 * reload1.c (alter_reg): Likewise.
7761 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7762 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7763 * tree-if-conv.c (predicate_mem_writes): Likewise.
7764 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7765 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7766 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7767 * valtrack.c (dead_debug_insert_temp): Likewise.
7768 * varasm.c (mergeable_constant_section): Likewise.
7769 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7772 Alan Hayward <alan.hayward@arm.com>
7773 David Sherwood <david.sherwood@arm.com>
7775 * expr.c (expand_assignment): Cope with polynomial mode sizes
7776 when assigning to a CONCAT.
7778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7779 Alan Hayward <alan.hayward@arm.com>
7780 David Sherwood <david.sherwood@arm.com>
7782 * machmode.h (mode_precision): Change from unsigned short to
7784 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7786 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7787 or if measurement_type is not polynomial.
7788 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7789 in which the mode is already known to be a scalar_int_mode.
7790 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7791 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7793 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7794 for GET_MODE_PRECISION.
7795 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7796 for GET_MODE_PRECISION.
7797 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7799 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7800 (expand_field_assignment, make_extraction): Likewise.
7801 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7802 (get_last_value): Likewise.
7803 * convert.c (convert_to_integer_1): Likewise.
7804 * cse.c (cse_insn): Likewise.
7805 * expr.c (expand_expr_real_1): Likewise.
7806 * lra-constraints.c (simplify_operand_subreg): Likewise.
7807 * optabs-query.c (can_atomic_load_p): Likewise.
7808 * optabs.c (expand_atomic_load): Likewise.
7809 (expand_atomic_store): Likewise.
7810 * ree.c (combine_reaching_defs): Likewise.
7811 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7812 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7813 * tree.h (type_has_mode_precision_p): Likewise.
7814 * ubsan.c (instrument_si_overflow): Likewise.
7816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7817 Alan Hayward <alan.hayward@arm.com>
7818 David Sherwood <david.sherwood@arm.com>
7820 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7821 polynomial numbers of units.
7822 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7823 (valid_vector_subparts_p): New function.
7824 (build_vector_type): Remove temporary shim and take the number
7825 of units as a poly_uint64 rather than an int.
7826 (build_opaque_vector_type): Take the number of units as a
7827 poly_uint64 rather than an int.
7828 * tree.c (build_vector_from_ctor): Handle polynomial
7829 TYPE_VECTOR_SUBPARTS.
7830 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7831 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7832 (build_vector_from_val): If the number of units is variable,
7833 use build_vec_duplicate_cst for constant operands and
7834 VEC_DUPLICATE_EXPR otherwise.
7835 (make_vector_type): Remove temporary is_constant ().
7836 (build_vector_type, build_opaque_vector_type): Take the number of
7837 units as a poly_uint64 rather than an int.
7838 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7840 * cfgexpand.c (expand_debug_expr): Likewise.
7841 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7842 (store_constructor, expand_expr_real_1): Likewise.
7843 (const_scalar_mask_from_tree): Likewise.
7844 * fold-const-call.c (fold_const_reduction): Likewise.
7845 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7846 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7847 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7848 (fold_relational_const): Likewise.
7849 (native_interpret_vector): Likewise. Change the size from an
7850 int to an unsigned int.
7851 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7852 TYPE_VECTOR_SUBPARTS.
7853 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7854 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7855 duplicating a non-constant operand into a variable-length vector.
7856 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7857 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7858 * ipa-icf.c (sem_variable::equals): Likewise.
7859 * match.pd: Likewise.
7860 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7861 * print-tree.c (print_node): Likewise.
7862 * stor-layout.c (layout_type): Likewise.
7863 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7864 * tree-cfg.c (verify_gimple_comparison): Likewise.
7865 (verify_gimple_assign_binary): Likewise.
7866 (verify_gimple_assign_ternary): Likewise.
7867 (verify_gimple_assign_single): Likewise.
7868 * tree-pretty-print.c (dump_generic_node): Likewise.
7869 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7870 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7871 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7872 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7873 (vect_shift_permute_load_chain): Likewise.
7874 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7875 (expand_vector_condition, optimize_vector_constructor): Likewise.
7876 (lower_vec_perm, get_compute_type): Likewise.
7877 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7878 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7879 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7880 (vect_recog_mask_conversion_pattern): Likewise.
7881 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7882 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7883 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7884 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7885 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7886 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7887 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7888 (supportable_widening_operation): Likewise.
7889 (supportable_narrowing_operation): Likewise.
7890 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7892 * varasm.c (output_constant): Likewise.
7894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7895 Alan Hayward <alan.hayward@arm.com>
7896 David Sherwood <david.sherwood@arm.com>
7898 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7899 so that both the length == 3 and length != 3 cases set up their
7900 own permute vectors. Add comments explaining why we know the
7901 number of elements is constant.
7902 (vect_permute_load_chain): Likewise.
7904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7905 Alan Hayward <alan.hayward@arm.com>
7906 David Sherwood <david.sherwood@arm.com>
7908 * machmode.h (mode_nunits): Change from unsigned char to
7910 (ONLY_FIXED_SIZE_MODES): New macro.
7911 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7912 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7913 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7915 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7916 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7917 or if measurement_type is not polynomial.
7918 * genmodes.c (ZERO_COEFFS): New macro.
7919 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7921 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7922 Use ZERO_COEFFS when emitting initializers.
7923 * data-streamer.h (bp_pack_poly_value): New function.
7924 (bp_unpack_poly_value): Likewise.
7925 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7926 for GET_MODE_NUNITS.
7927 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7928 for GET_MODE_NUNITS.
7929 * tree.c (make_vector_type): Remove temporary shim and make
7930 the real function take the number of units as a poly_uint64
7932 (build_vector_type_for_mode): Handle polynomial nunits.
7933 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7934 * emit-rtl.c (const_vec_series_p_1): Likewise.
7935 (gen_rtx_CONST_VECTOR): Likewise.
7936 * fold-const.c (test_vec_duplicate_folding): Likewise.
7937 * genrecog.c (validate_pattern): Likewise.
7938 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7939 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7940 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7941 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7942 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7943 * rtlanal.c (subreg_get_info): Likewise.
7944 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7945 (vect_grouped_load_supported): Likewise.
7946 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7947 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7948 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7949 (simplify_const_unary_operation, simplify_binary_operation_1)
7950 (simplify_const_binary_operation, simplify_ternary_operation)
7951 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7952 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7953 instead of CONST_VECTOR_NUNITS.
7954 * varasm.c (output_constant_pool_2): Likewise.
7955 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7956 explicit-encoded elements in the XVEC for variable-length vectors.
7958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7960 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7963 Alan Hayward <alan.hayward@arm.com>
7964 David Sherwood <david.sherwood@arm.com>
7966 * coretypes.h (fixed_size_mode): Declare.
7967 (fixed_size_mode_pod): New typedef.
7968 * builtins.h (target_builtins::x_apply_args_mode)
7969 (target_builtins::x_apply_result_mode): Change type to
7970 fixed_size_mode_pod.
7971 * builtins.c (apply_args_size, apply_result_size, result_vector)
7972 (expand_builtin_apply_args_1, expand_builtin_apply)
7973 (expand_builtin_return): Update accordingly.
7975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7977 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7978 * cselib.c (cselib_hash_rtx): Likewise.
7979 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7980 CONST_VECTOR encoding.
7982 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7983 Jeff Law <law@redhat.com>
7986 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7987 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7988 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7989 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7992 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7993 explicitly probe *sp in a noreturn function if there were any callee
7994 register saves or frame pointer is needed.
7996 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7999 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8000 BLKmode for ternary, binary or unary expressions.
8003 * var-tracking.c (delete_vta_debug_insn): New inline function.
8004 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8005 insns from get_insns () to NULL instead of each bb separately.
8006 Use delete_vta_debug_insn. No longer static.
8007 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8008 delete_vta_debug_insns callers.
8009 * rtl.h (delete_vta_debug_insns): Declare.
8010 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8011 instead of variable_tracking_main.
8013 2018-01-03 Martin Sebor <msebor@redhat.com>
8015 PR tree-optimization/83603
8016 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8017 arguments past the endof the argument list in functions declared
8018 without a prototype.
8019 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8020 Avoid checking when arguments are null.
8022 2018-01-03 Martin Sebor <msebor@redhat.com>
8025 * doc/extend.texi (attribute const): Fix a typo.
8026 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8027 issuing -Wsuggest-attribute for void functions.
8029 2018-01-03 Martin Sebor <msebor@redhat.com>
8031 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8032 offset_int::from instead of wide_int::to_shwi.
8033 (maybe_diag_overlap): Remove assertion.
8034 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8035 * gimple-ssa-sprintf.c (format_directive): Same.
8036 (parse_directive): Same.
8037 (sprintf_dom_walker::compute_format_length): Same.
8038 (try_substitute_return_value): Same.
8040 2018-01-03 Jeff Law <law@redhat.com>
8043 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8044 non-constant residual for zero at runtime and avoid probing in
8045 that case. Reorganize code for trailing problem to mirror handling
8048 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8050 PR tree-optimization/83501
8051 * tree-ssa-strlen.c (get_string_cst): New.
8052 (handle_char_store): Call get_string_cst.
8054 2018-01-03 Martin Liska <mliska@suse.cz>
8056 PR tree-optimization/83593
8057 * tree-ssa-strlen.c: Include tree-cfg.h.
8058 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8059 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8060 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8062 (strlen_dom_walker::before_dom_children): Call
8063 gimple_purge_dead_eh_edges. Dump tranformation with details
8065 (strlen_dom_walker::before_dom_children): Update call by adding
8066 new argument cleanup_eh.
8067 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8069 2018-01-03 Martin Liska <mliska@suse.cz>
8072 * cif-code.def (VARIADIC_THUNK): New enum value.
8073 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8076 2018-01-03 Jan Beulich <jbeulich@suse.com>
8078 * sse.md (mov<mode>_internal): Tighten condition for when to use
8079 vmovdqu<ssescalarsize> for TI and OI modes.
8081 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8083 Update copyright years.
8085 2018-01-03 Martin Liska <mliska@suse.cz>
8088 * ipa-visibility.c (function_and_variable_visibility): Skip
8089 functions with noipa attribure.
8091 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8093 * gcc.c (process_command): Update copyright notice dates.
8094 * gcov-dump.c (print_version): Ditto.
8095 * gcov.c (print_version): Ditto.
8096 * gcov-tool.c (print_version): Ditto.
8097 * gengtype.c (create_file): Ditto.
8098 * doc/cpp.texi: Bump @copying's copyright year.
8099 * doc/cppinternals.texi: Ditto.
8100 * doc/gcc.texi: Ditto.
8101 * doc/gccint.texi: Ditto.
8102 * doc/gcov.texi: Ditto.
8103 * doc/install.texi: Ditto.
8104 * doc/invoke.texi: Ditto.
8106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8108 * vector-builder.h (vector_builder::m_full_nelts): Change from
8109 unsigned int to poly_uint64.
8110 (vector_builder::full_nelts): Update prototype accordingly.
8111 (vector_builder::new_vector): Likewise.
8112 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8113 (vector_builder::operator ==): Likewise.
8114 (vector_builder::finalize): Likewise.
8115 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8116 Take the number of elements as a poly_uint64 rather than an
8118 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8119 from unsigned int to poly_uint64.
8120 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8121 (vec_perm_indices::new_vector): Likewise.
8122 (vec_perm_indices::length): Likewise.
8123 (vec_perm_indices::nelts_per_input): Likewise.
8124 (vec_perm_indices::input_nelts): Likewise.
8125 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8126 number of elements per input as a poly_uint64 rather than an
8127 unsigned int. Use the original encoding for variable-length
8128 vectors, rather than clamping each individual element.
8129 For the second and subsequent elements in each pattern,
8130 clamp the step and base before clamping their sum.
8131 (vec_perm_indices::series_p): Handle polynomial element counts.
8132 (vec_perm_indices::all_in_range_p): Likewise.
8133 (vec_perm_indices_to_tree): Likewise.
8134 (vec_perm_indices_to_rtx): Likewise.
8135 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8136 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8137 (tree_vector_builder::new_binary_operation): Handle polynomial
8138 element counts. Return false if we need to know the number
8139 of elements at compile time.
8140 * fold-const.c (fold_vec_perm): Punt if the number of elements
8141 isn't known at compile time.
8143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8145 * vec-perm-indices.h (vec_perm_builder): Change element type
8146 from HOST_WIDE_INT to poly_int64.
8147 (vec_perm_indices::element_type): Update accordingly.
8148 (vec_perm_indices::clamp): Handle polynomial element_types.
8149 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8150 (vec_perm_indices::all_in_range_p): Likewise.
8151 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8153 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8154 polynomial vec_perm_indices element types.
8155 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8156 * fold-const.c (fold_vec_perm): Likewise.
8157 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8158 * tree-vect-generic.c (lower_vec_perm): Likewise.
8159 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8160 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8161 element type to HOST_WIDE_INT.
8163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8164 Alan Hayward <alan.hayward@arm.com>
8165 David Sherwood <david.sherwood@arm.com>
8167 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8168 rather than an int. Use plus_constant.
8169 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8170 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8173 Alan Hayward <alan.hayward@arm.com>
8174 David Sherwood <david.sherwood@arm.com>
8176 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8177 a HOST_WIDE_INT to a poly_int64.
8179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8180 Alan Hayward <alan.hayward@arm.com>
8181 David Sherwood <david.sherwood@arm.com>
8183 * calls.c (load_register_parameters): Cope with polynomial
8184 mode sizes. Require a constant size for BLKmode parameters
8185 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8186 forces a parameter to be padded at the lsb end in order to
8187 fill a complete number of words, require the parameter size
8188 to be ordered wrt UNITS_PER_WORD.
8190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8191 Alan Hayward <alan.hayward@arm.com>
8192 David Sherwood <david.sherwood@arm.com>
8194 * reload1.c (spill_stack_slot_width): Change element type
8195 from unsigned int to poly_uint64_pod.
8196 (alter_reg): Treat mode sizes as polynomial.
8198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8199 Alan Hayward <alan.hayward@arm.com>
8200 David Sherwood <david.sherwood@arm.com>
8202 * reload.c (complex_word_subreg_p): New function.
8203 (reload_inner_reg_of_subreg, push_reload): Use it.
8205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8206 Alan Hayward <alan.hayward@arm.com>
8207 David Sherwood <david.sherwood@arm.com>
8209 * lra-constraints.c (process_alt_operands): Reject matched
8210 operands whose sizes aren't ordered.
8211 (match_reload): Refer to this check here.
8213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8214 Alan Hayward <alan.hayward@arm.com>
8215 David Sherwood <david.sherwood@arm.com>
8217 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8218 that the mode size is in the set {1, 2, 4, 8, 16}.
8220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8221 Alan Hayward <alan.hayward@arm.com>
8222 David Sherwood <david.sherwood@arm.com>
8224 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8225 Use plus_constant instead of gen_rtx_PLUS.
8227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8228 Alan Hayward <alan.hayward@arm.com>
8229 David Sherwood <david.sherwood@arm.com>
8231 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8232 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8233 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8234 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8235 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8236 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8237 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8238 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8239 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8240 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8242 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8243 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8244 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8245 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8246 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8247 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8248 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8249 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8250 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8251 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8253 * expr.c (emit_move_resolve_push): Treat the input and result
8254 of PUSH_ROUNDING as a poly_int64.
8255 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8256 (emit_push_insn): Likewise.
8257 * lra-eliminations.c (mark_not_eliminable): Likewise.
8258 * recog.c (push_operand): Likewise.
8259 * reload1.c (elimination_effects): Likewise.
8260 * rtlanal.c (nonzero_bits1): Likewise.
8261 * calls.c (store_one_arg): Likewise. Require the padding to be
8262 known at compile time.
8264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8265 Alan Hayward <alan.hayward@arm.com>
8266 David Sherwood <david.sherwood@arm.com>
8268 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8269 Use plus_constant instead of gen_rtx_PLUS.
8271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8272 Alan Hayward <alan.hayward@arm.com>
8273 David Sherwood <david.sherwood@arm.com>
8275 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8279 Alan Hayward <alan.hayward@arm.com>
8280 David Sherwood <david.sherwood@arm.com>
8282 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8283 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8284 via stack temporaries. Treat the mode size as polynomial too.
8286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8287 Alan Hayward <alan.hayward@arm.com>
8288 David Sherwood <david.sherwood@arm.com>
8290 * expr.c (expand_expr_real_2): When handling conversions involving
8291 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8292 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8293 as a poly_uint64 too.
8295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8296 Alan Hayward <alan.hayward@arm.com>
8297 David Sherwood <david.sherwood@arm.com>
8299 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8302 Alan Hayward <alan.hayward@arm.com>
8303 David Sherwood <david.sherwood@arm.com>
8305 * combine.c (can_change_dest_mode): Handle polynomial
8306 REGMODE_NATURAL_SIZE.
8307 * expmed.c (store_bit_field_1): Likewise.
8308 * expr.c (store_constructor): Likewise.
8309 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8310 and polynomial REGMODE_NATURAL_SIZE.
8311 (gen_lowpart_common): Likewise.
8312 * reginfo.c (record_subregs_of_mode): Likewise.
8313 * rtlanal.c (read_modify_subreg_p): Likewise.
8315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8316 Alan Hayward <alan.hayward@arm.com>
8317 David Sherwood <david.sherwood@arm.com>
8319 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8320 numbers of elements.
8322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8323 Alan Hayward <alan.hayward@arm.com>
8324 David Sherwood <david.sherwood@arm.com>
8326 * match.pd: Cope with polynomial numbers of vector elements.
8328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8329 Alan Hayward <alan.hayward@arm.com>
8330 David Sherwood <david.sherwood@arm.com>
8332 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8333 in a POINTER_PLUS_EXPR.
8335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8336 Alan Hayward <alan.hayward@arm.com>
8337 David Sherwood <david.sherwood@arm.com>
8339 * omp-simd-clone.c (simd_clone_subparts): New function.
8340 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8341 (ipa_simd_modify_function_body): Likewise.
8343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8344 Alan Hayward <alan.hayward@arm.com>
8345 David Sherwood <david.sherwood@arm.com>
8347 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8348 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8349 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8350 (expand_vector_condition, vector_element): Likewise.
8351 (subparts_gt): New function.
8352 (get_compute_type): Use subparts_gt.
8353 (count_type_subparts): Delete.
8354 (expand_vector_operations_1): Use subparts_gt instead of
8355 count_type_subparts.
8357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8358 Alan Hayward <alan.hayward@arm.com>
8359 David Sherwood <david.sherwood@arm.com>
8361 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8362 (vect_compile_time_alias): ...this new function. Do the calculation
8363 on poly_ints rather than trees.
8364 (vect_prune_runtime_alias_test_list): Update call accordingly.
8366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8367 Alan Hayward <alan.hayward@arm.com>
8368 David Sherwood <david.sherwood@arm.com>
8370 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8372 (vect_schedule_slp_instance): Likewise.
8374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8375 Alan Hayward <alan.hayward@arm.com>
8376 David Sherwood <david.sherwood@arm.com>
8378 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8379 constant and extern definitions for variable-length vectors.
8380 (vect_get_constant_vectors): Note that the number of units
8381 is known to be constant.
8383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8384 Alan Hayward <alan.hayward@arm.com>
8385 David Sherwood <david.sherwood@arm.com>
8387 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8388 of units as polynomial. Choose between WIDE and NARROW based
8391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8392 Alan Hayward <alan.hayward@arm.com>
8393 David Sherwood <david.sherwood@arm.com>
8395 * tree-vect-stmts.c (simd_clone_subparts): New function.
8396 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8399 Alan Hayward <alan.hayward@arm.com>
8400 David Sherwood <david.sherwood@arm.com>
8402 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8403 vectors as polynomial. Use build_index_vector for
8406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8407 Alan Hayward <alan.hayward@arm.com>
8408 David Sherwood <david.sherwood@arm.com>
8410 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8411 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8412 for variable-length vectors.
8413 (vectorizable_mask_load_store): Treat the number of units as
8414 polynomial, asserting that it is constant if the condition has
8415 already been enforced.
8416 (vectorizable_store, vectorizable_load): Likewise.
8418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8419 Alan Hayward <alan.hayward@arm.com>
8420 David Sherwood <david.sherwood@arm.com>
8422 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8423 of units as polynomial. Punt if we can't tell at compile time
8424 which vector contains the final result.
8426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8427 Alan Hayward <alan.hayward@arm.com>
8428 David Sherwood <david.sherwood@arm.com>
8430 * tree-vect-loop.c (vectorizable_induction): Treat the number
8431 of units as polynomial. Punt on SLP inductions. Use an integer
8432 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8433 cast of such a series for variable-length floating-point
8436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8437 Alan Hayward <alan.hayward@arm.com>
8438 David Sherwood <david.sherwood@arm.com>
8440 * tree.h (build_index_vector): Declare.
8441 * tree.c (build_index_vector): New function.
8442 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8443 of units as polynomial, forcibly converting it to a constant if
8444 vectorizable_reduction has already enforced the condition.
8445 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8446 to create a {1,2,3,...} vector.
8447 (vectorizable_reduction): Treat the number of units as polynomial.
8448 Choose vectype_in based on the largest scalar element size rather
8449 than the smallest number of units. Enforce the restrictions
8452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8453 Alan Hayward <alan.hayward@arm.com>
8454 David Sherwood <david.sherwood@arm.com>
8456 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8457 number of units as polynomial.
8459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8460 Alan Hayward <alan.hayward@arm.com>
8461 David Sherwood <david.sherwood@arm.com>
8463 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8464 * target.def (autovectorize_vector_sizes): Return the vector sizes
8465 by pointer, using vector_sizes rather than a bitmask.
8466 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8467 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8468 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8470 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8471 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8472 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8473 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8474 * omp-general.c (omp_max_vf): Likewise.
8475 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8476 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8477 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8478 * tree-vect-slp.c (vect_slp_bb): Likewise.
8479 * doc/tm.texi: Regenerate.
8480 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8482 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8483 the vector size as a poly_uint64 rather than an unsigned int.
8484 (current_vector_size): Change from an unsigned int to a poly_uint64.
8485 (get_vectype_for_scalar_type): Update accordingly.
8486 * tree.h (build_truth_vector_type): Take the size and number of
8487 units as a poly_uint64 rather than an unsigned int.
8488 (build_vector_type): Add a temporary overload that takes
8489 the number of units as a poly_uint64 rather than an unsigned int.
8490 * tree.c (make_vector_type): Likewise.
8491 (build_truth_vector_type): Take the number of units as a poly_uint64
8492 rather than an unsigned int.
8494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8495 Alan Hayward <alan.hayward@arm.com>
8496 David Sherwood <david.sherwood@arm.com>
8498 * target.def (get_mask_mode): Take the number of units and length
8499 as poly_uint64s rather than unsigned ints.
8500 * targhooks.h (default_get_mask_mode): Update accordingly.
8501 * targhooks.c (default_get_mask_mode): Likewise.
8502 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8503 * doc/tm.texi: Regenerate.
8505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8506 Alan Hayward <alan.hayward@arm.com>
8507 David Sherwood <david.sherwood@arm.com>
8509 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8510 * omp-general.c (omp_max_vf): Likewise.
8511 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8512 (expand_omp_simd): Handle polynomial safelen.
8513 * omp-low.c (omplow_simd_context): Add a default constructor.
8514 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8515 (lower_rec_simd_input_clauses): Update accordingly.
8516 (lower_rec_input_clauses): Likewise.
8518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8519 Alan Hayward <alan.hayward@arm.com>
8520 David Sherwood <david.sherwood@arm.com>
8522 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8523 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8524 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8525 (vect_analyze_slp_cost): Likewise.
8526 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8527 (vect_model_load_cost): Likewise.
8529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8530 Alan Hayward <alan.hayward@arm.com>
8531 David Sherwood <david.sherwood@arm.com>
8533 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8534 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8535 from an unsigned int * to a poly_uint64_pod *.
8536 (calculate_unrolling_factor): New function.
8537 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8540 Alan Hayward <alan.hayward@arm.com>
8541 David Sherwood <david.sherwood@arm.com>
8543 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8544 from an unsigned int to a poly_uint64.
8545 (_loop_vec_info::slp_unrolling_factor): Likewise.
8546 (_loop_vec_info::vectorization_factor): Change from an int
8548 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8549 (vect_get_num_vectors): New function.
8550 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8551 (vect_get_num_copies): Use vect_get_num_vectors.
8552 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8553 to an unsigned int *.
8554 (vect_analyze_data_refs): Change min_vf from an int * to a
8556 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8557 than an unsigned HOST_WIDE_INT.
8558 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8559 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8560 to an unsigned int *.
8561 (vect_analyze_data_ref_dependences): Likewise.
8562 (vect_compute_data_ref_alignment): Handle polynomial vf.
8563 (vect_enhance_data_refs_alignment): Likewise.
8564 (vect_prune_runtime_alias_test_list): Likewise.
8565 (vect_shift_permute_load_chain): Likewise.
8566 (vect_supportable_dr_alignment): Likewise.
8567 (dependence_distance_ge_vf): Take the vectorization factor as a
8568 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8569 (vect_analyze_data_refs): Change min_vf from an int * to a
8571 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8572 vfm1 as a poly_uint64 rather than an int. Make the same change
8573 for the returned bound_scalar.
8574 (vect_gen_vector_loop_niters): Handle polynomial vf.
8575 (vect_do_peeling): Likewise. Update call to
8576 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8577 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8579 * tree-vect-loop.c (vect_determine_vectorization_factor)
8580 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8581 (vect_get_known_peeling_cost): Likewise.
8582 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8583 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8584 (vect_transform_loop): Likewise. Use the lowest possible VF when
8585 updating the upper bounds of the loop.
8586 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8588 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8589 polynomial unroll factors.
8590 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8591 (vect_make_slp_decision): Likewise.
8592 (vect_supported_load_permutation_p): Likewise, and polynomial
8594 (vect_analyze_slp_cost): Handle polynomial vf.
8595 (vect_slp_analyze_node_operations): Likewise.
8596 (vect_slp_analyze_bb_1): Likewise.
8597 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8598 than an unsigned HOST_WIDE_INT.
8599 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8600 (vectorizable_load): Handle polynomial vf.
8601 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8603 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8606 Alan Hayward <alan.hayward@arm.com>
8607 David Sherwood <david.sherwood@arm.com>
8609 * match.pd: Handle bit operations involving three constants
8610 and try to fold one pair.
8612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8614 * tree-vect-loop-manip.c: Include gimple-fold.h.
8615 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8616 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8617 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8618 Add a path that uses a step of VF instead of 1, but disable it
8620 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8621 and niters_no_overflow parameters. Update calls to
8622 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8623 Create a new SSA name if the latter choses to use a ste other
8624 than zero, and return it via niters_vector_mult_vf_var.
8625 * tree-vect-loop.c (vect_transform_loop): Update calls to
8626 vect_do_peeling, vect_gen_vector_loop_niters and
8627 slpeel_make_loop_iterate_ntimes.
8628 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8629 (vect_gen_vector_loop_niters): Update declarations after above changes.
8631 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8633 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8634 128-bit round to integer instructions.
8635 (ceil<mode>2): Likewise.
8636 (btrunc<mode>2): Likewise.
8637 (round<mode>2): Likewise.
8639 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8641 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8642 unaligned VSX load/store on P8/P9.
8643 (expand_block_clear): Allow the use of unaligned VSX
8644 load/store on P8/P9.
8646 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8648 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8650 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8651 swap associated with both a load and a store.
8653 2018-01-02 Andrew Waterman <andrew@sifive.com>
8655 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8656 * config/riscv/riscv.md (clear_cache): Use it.
8658 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8660 * web.c: Remove out-of-date comment.
8662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8664 * expr.c (fixup_args_size_notes): Check that any existing
8665 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8666 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8667 (emit_single_push_insn): ...here.
8669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8671 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8672 (const_vector_encoded_nelts): New function.
8673 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8674 (const_vector_int_elt, const_vector_elt): Declare.
8675 * emit-rtl.c (const_vector_int_elt_1): New function.
8676 (const_vector_elt): Likewise.
8677 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8678 of CONST_VECTOR_ELT.
8680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8682 * expr.c: Include rtx-vector-builder.h.
8683 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8684 directly on the tree encoding.
8685 (const_vector_from_tree): Likewise.
8686 * optabs.c: Include rtx-vector-builder.h.
8687 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8688 sequence of "u" values.
8689 * vec-perm-indices.c: Include rtx-vector-builder.h.
8690 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8691 directly on the vec_perm_indices encoding.
8693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8695 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8696 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8697 * rtx-vector-builder.h: New file.
8698 * rtx-vector-builder.c: Likewise.
8699 * rtl.h (rtx_def::u2): Add a const_vector field.
8700 (CONST_VECTOR_NPATTERNS): New macro.
8701 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8702 (CONST_VECTOR_DUPLICATE_P): Likewise.
8703 (CONST_VECTOR_STEPPED_P): Likewise.
8704 (CONST_VECTOR_ENCODED_ELT): Likewise.
8705 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8706 (unwrap_const_vec_duplicate): Likewise.
8707 (const_vec_series_p): Check for a non-duplicated vector encoding.
8708 Say that the function only returns true for integer vectors.
8709 * emit-rtl.c: Include rtx-vector-builder.h.
8710 (gen_const_vec_duplicate_1): Delete.
8711 (gen_const_vector): Call gen_const_vec_duplicate instead of
8712 gen_const_vec_duplicate_1.
8713 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8714 (gen_const_vec_duplicate): Use rtx_vector_builder.
8715 (gen_const_vec_series): Likewise.
8716 (gen_rtx_CONST_VECTOR): Likewise.
8717 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8718 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8719 Build a new vector rather than modifying a CONST_VECTOR in-place.
8720 (handle_special_swappables): Update call accordingly.
8721 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8722 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8723 Build a new vector rather than modifying a CONST_VECTOR in-place.
8724 (handle_special_swappables): Update call accordingly.
8726 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8728 * simplify-rtx.c (simplify_const_binary_operation): Use
8729 CONST_VECTOR_ELT instead of XVECEXP.
8731 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8733 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8734 the selector elements to be different from the data elements
8735 if the selector is a VECTOR_CST.
8736 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8737 ssizetype for the selector.
8739 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8741 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8742 before testing each element individually.
8743 * tree-vect-generic.c (lower_vec_perm): Likewise.
8745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8747 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8748 * selftest-run-tests.c (selftest::run_tests): Call it.
8749 * vector-builder.h (vector_builder::operator ==): New function.
8750 (vector_builder::operator !=): Likewise.
8751 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8752 (vec_perm_indices::all_from_input_p): New function.
8753 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8754 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8755 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8756 instead of reading the VECTOR_CST directly. Detect whether both
8757 vector inputs are the same before constructing the vec_perm_indices,
8758 and update the number of inputs argument accordingly. Use the
8759 utility functions added above. Only construct sel2 if we need to.
8761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8763 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8764 the broadcast of the low byte.
8765 (expand_mult_highpart): Use an explicit encoding for the permutes.
8766 * optabs-query.c (can_mult_highpart_p): Likewise.
8767 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8768 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8769 (vectorizable_bswap): Likewise.
8770 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8771 explicit encoding for the power-of-2 permutes.
8772 (vect_permute_store_chain): Likewise.
8773 (vect_grouped_load_supported): Likewise.
8774 (vect_permute_load_chain): Likewise.
8776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8778 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8779 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8780 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8781 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8782 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8783 (vect_gen_perm_mask_any): Likewise.
8785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8787 * int-vector-builder.h: New file.
8788 * vec-perm-indices.h: Include int-vector-builder.h.
8789 (vec_perm_indices): Redefine as an int_vector_builder.
8790 (auto_vec_perm_indices): Delete.
8791 (vec_perm_builder): Redefine as a stand-alone class.
8792 (vec_perm_indices::vec_perm_indices): New function.
8793 (vec_perm_indices::clamp): Likewise.
8794 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8795 (vec_perm_indices::new_vector): New function.
8796 (vec_perm_indices::new_expanded_vector): Update for new
8797 vec_perm_indices class.
8798 (vec_perm_indices::rotate_inputs): New function.
8799 (vec_perm_indices::all_in_range_p): Operate directly on the
8800 encoded form, without computing elided elements.
8801 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8802 encoding. Update for new vec_perm_indices class.
8803 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8804 the given vec_perm_builder.
8805 (expand_vec_perm_var): Update vec_perm_builder constructor.
8806 (expand_mult_highpart): Use vec_perm_builder instead of
8807 auto_vec_perm_indices.
8808 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8809 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8810 or double series encoding as appropriate.
8811 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8812 vec_perm_indices instead of auto_vec_perm_indices.
8813 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8814 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8815 (vect_permute_store_chain): Likewise.
8816 (vect_grouped_load_supported): Likewise.
8817 (vect_permute_load_chain): Likewise.
8818 (vect_shift_permute_load_chain): Likewise.
8819 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8820 (vect_transform_slp_perm_load): Likewise.
8821 (vect_schedule_slp_instance): Likewise.
8822 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8823 (vectorizable_mask_load_store): Likewise.
8824 (vectorizable_bswap): Likewise.
8825 (vectorizable_store): Likewise.
8826 (vectorizable_load): Likewise.
8827 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8828 vec_perm_indices instead of auto_vec_perm_indices. Use
8829 tree_to_vec_perm_builder to read the vector from a tree.
8830 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8831 vec_perm_builder instead of a vec_perm_indices.
8832 (have_whole_vector_shift): Use vec_perm_builder and
8833 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8834 truncation to calc_vec_perm_mask_for_shift.
8835 (vect_create_epilog_for_reduction): Likewise.
8836 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8837 from auto_vec_perm_indices to vec_perm_indices.
8838 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8839 instead of changing individual elements.
8840 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8841 the vector in d.perm.
8842 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8843 from auto_vec_perm_indices to vec_perm_indices.
8844 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8845 instead of changing individual elements.
8846 (arm_vectorize_vec_perm_const): Use new_vector to install
8847 the vector in d.perm.
8848 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8849 Update vec_perm_builder constructor.
8850 (rs6000_expand_interleave): Likewise.
8851 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8852 (rs6000_expand_interleave): Likewise.
8854 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8856 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8857 to qimode could truncate the indices.
8858 * optabs.c (expand_vec_perm_var): Likewise.
8860 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8862 * Makefile.in (OBJS): Add vec-perm-indices.o.
8863 * vec-perm-indices.h: New file.
8864 * vec-perm-indices.c: Likewise.
8865 * target.h (vec_perm_indices): Replace with a forward class
8867 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8868 * optabs.h: Include vec-perm-indices.h.
8869 (expand_vec_perm): Delete.
8870 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8871 (expand_vec_perm_const): Declare.
8872 * target.def (vec_perm_const_ok): Replace with...
8873 (vec_perm_const): ...this new hook.
8874 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8875 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8876 * doc/tm.texi: Regenerate.
8877 * optabs.def (vec_perm_const): Delete.
8878 * doc/md.texi (vec_perm_const): Likewise.
8879 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8880 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8881 expand_vec_perm for constant permutation vectors. Assert that
8882 the mode of variable permutation vectors is the integer equivalent
8883 of the mode that is being permuted.
8884 * optabs-query.h (selector_fits_mode_p): Declare.
8885 * optabs-query.c: Include vec-perm-indices.h.
8886 (selector_fits_mode_p): New function.
8887 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8888 is defined, instead of checking whether the vec_perm_const_optab
8889 exists. Use targetm.vectorize.vec_perm_const instead of
8890 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8891 fit in the vector mode before using a variable permute.
8892 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8893 vec_perm_indices instead of an rtx.
8894 (expand_vec_perm): Replace with...
8895 (expand_vec_perm_const): ...this new function. Take the selector
8896 as a vec_perm_indices rather than an rtx. Also take the mode of
8897 the selector. Update call to shift_amt_for_vec_perm_mask.
8898 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8899 Use vec_perm_indices::new_expanded_vector to expand the original
8900 selector into bytes. Check whether the indices fit in the vector
8901 mode before using a variable permute.
8902 (expand_vec_perm_var): Make global.
8903 (expand_mult_highpart): Use expand_vec_perm_const.
8904 * fold-const.c: Includes vec-perm-indices.h.
8905 * tree-ssa-forwprop.c: Likewise.
8906 * tree-vect-data-refs.c: Likewise.
8907 * tree-vect-generic.c: Likewise.
8908 * tree-vect-loop.c: Likewise.
8909 * tree-vect-slp.c: Likewise.
8910 * tree-vect-stmts.c: Likewise.
8911 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8913 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8914 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8915 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8916 (aarch64_vectorize_vec_perm_const): ...this new function.
8917 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8918 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8919 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8920 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8921 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8922 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8923 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8925 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8926 check for NEON modes.
8927 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8928 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8929 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8930 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8932 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8933 the old VEC_PERM_CONST conditions.
8934 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8935 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8936 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8937 (ia64_vectorize_vec_perm_const_ok): Merge into...
8938 (ia64_vectorize_vec_perm_const): ...this new function.
8939 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8940 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8941 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8942 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8943 * config/mips/mips.c (mips_expand_vec_perm_const)
8944 (mips_vectorize_vec_perm_const_ok): Merge into...
8945 (mips_vectorize_vec_perm_const): ...this new function.
8946 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8947 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8948 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8949 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8950 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8951 (rs6000_expand_vec_perm_const): Delete.
8952 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8954 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8955 (altivec_expand_vec_perm_const_le): Take each operand individually.
8956 Operate on constant selectors rather than rtxes.
8957 (altivec_expand_vec_perm_const): Likewise. Update call to
8958 altivec_expand_vec_perm_const_le.
8959 (rs6000_expand_vec_perm_const): Delete.
8960 (rs6000_vectorize_vec_perm_const_ok): Delete.
8961 (rs6000_vectorize_vec_perm_const): New function.
8962 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8963 an element count and rtx array.
8964 (rs6000_expand_extract_even): Update call accordingly.
8965 (rs6000_expand_interleave): Likewise.
8966 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8967 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8968 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8969 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8970 (rs6000_expand_vec_perm_const): Delete.
8971 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8972 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8973 (altivec_expand_vec_perm_const_le): Take each operand individually.
8974 Operate on constant selectors rather than rtxes.
8975 (altivec_expand_vec_perm_const): Likewise. Update call to
8976 altivec_expand_vec_perm_const_le.
8977 (rs6000_expand_vec_perm_const): Delete.
8978 (rs6000_vectorize_vec_perm_const_ok): Delete.
8979 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8980 reference to the SPE evmerge intructions.
8981 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8982 an element count and rtx array.
8983 (rs6000_expand_extract_even): Update call accordingly.
8984 (rs6000_expand_interleave): Likewise.
8985 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8986 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8988 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8992 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8993 vector mode and that that mode matches the mode of the data
8995 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8996 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8997 directly using expand_vec_perm_1 when forcing selectors into
8999 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9001 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9003 * optabs-query.h (can_vec_perm_p): Delete.
9004 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9005 * optabs-query.c (can_vec_perm_p): Split into...
9006 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9007 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9008 particular selector is valid.
9009 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9010 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9011 (vect_grouped_load_supported): Likewise.
9012 (vect_shift_permute_load_chain): Likewise.
9013 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9014 (vect_transform_slp_perm_load): Likewise.
9015 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9016 (vectorizable_bswap): Likewise.
9017 (vect_gen_perm_mask_checked): Likewise.
9018 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9019 implementations of variable permutation vectors into account
9020 when deciding which selector to use.
9021 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9022 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9023 with a false third argument.
9024 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9025 to test whether the constant selector is valid and can_vec_perm_var_p
9026 to test whether a variable selector is valid.
9028 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9030 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9031 * optabs-query.c (can_vec_perm_p): Likewise.
9032 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9033 instead of vec_perm_indices.
9034 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9035 (vect_gen_perm_mask_checked): Likewise,
9036 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9037 (vect_gen_perm_mask_checked): Likewise,
9039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9041 * optabs-query.h (qimode_for_vec_perm): Declare.
9042 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9043 (qimode_for_vec_perm): ...this new function.
9044 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9046 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9048 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9049 does not have a conditional at the top.
9051 2018-01-02 Richard Biener <rguenther@suse.de>
9053 * ipa-inline.c (big_speedup_p): Fix expression.
9055 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9058 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9061 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9065 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9066 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9067 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9068 cond_taken_branch_cost 3->4.
9070 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9072 PR tree-optimization/83581
9073 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9074 TODO_cleanup_cfg if any changes have been made.
9077 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9078 convert_modes if target mode has the right side, but different mode
9082 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9083 last argument when extracting from CONCAT. If either from_real or
9084 from_imag is NULL, use expansion through memory. If result is not
9085 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9086 the parts directly to inner mode, if even that fails, use expansion
9090 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9091 check for bswap in mode rather than HImode and use that in expand_unop
9094 Copyright (C) 2018 Free Software Foundation, Inc.
9096 Copying and distribution of this file, with or without modification,
9097 are permitted in any medium without royalty provided the copyright
9098 notice and this notice are preserved.