1 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6 2018-03-18 Martin Liska <mliska@suse.cz>
8 PR rtl-optimization/84635
9 * regrename.c (build_def_use): Use matches_mode only when
12 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14 PR tree-optimization/84913
15 * tree-vect-loop.c (vectorizable_reduction): Don't try to
16 vectorize chains of COND_EXPRs.
18 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
20 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
22 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
24 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
26 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
28 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
30 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
31 Kito Cheng <kito.cheng@gmail.com>
33 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
34 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
35 (nds32_adjust_reg_alloc_order): New function.
36 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
38 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
40 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
41 nds32_print_operand, nds32_print_operand_address): Use
42 HOST_WIDE_INT_PRINT_DEC instead.
44 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
46 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
48 2018-03-17 Jakub Jelinek <jakub@redhat.com>
51 * config/i386/i386.c (initial_ix86_tune_features,
52 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
54 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
55 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
56 rather than 1u << ix86_tune. Formatting fix.
57 (ix86_option_override_internal): Change ix86_arch_mask from
58 unsigned int to unsigned HOST_WIDE_INT, initialize to
59 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
60 (ix86_function_specific_restore): Likewise.
62 2018-03-16 Jakub Jelinek <jakub@redhat.com>
65 * postreload.c (reload_combine_recognize_pattern): Perform
66 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
67 truncate_int_for_mode the result for the destination's mode.
70 * hsa-gen.c (mem_type_for_type): Fix comment typo.
71 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
73 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
76 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
79 * lra-assigns.c (lra_split_hard_reg_for): Don't use
80 regno_allocno_class_array and sorted_pseudos.
81 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
82 insns where regno is used.
84 2018-03-16 Martin Liska <mliska@suse.cz>
87 * multiple_target.c (create_dispatcher_calls): Redirect
88 reference in the symbol table.
90 2018-03-16 Martin Liska <mliska@suse.cz>
93 * multiple_target.c (create_dispatcher_calls): Redirect also
96 2018-03-16 Jakub Jelinek <jakub@redhat.com>
100 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
101 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
102 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
104 2018-03-16 Julia Koval <julia.koval@intel.com>
106 * doc/invoke.texi (Skylake Server): Add CLWB.
107 Cannonlake): Remove CLWB.
109 2018-03-16 Jakub Jelinek <jakub@redhat.com>
111 PR tree-optimization/84841
112 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
114 (FLOAT_ONE_CONST_TYPE): Define.
115 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
116 (sort_by_operand_rank): Put entries with higher constant_type last
117 rather than first to match comments.
119 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
121 * config/nios2/nios2.md (movsi_internal): Fix thinko in
124 2018-03-15 Jakub Jelinek <jakub@redhat.com>
127 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
128 check and use address of target always.
130 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
133 * config/i386/i386.c (indirect_thunk_needed): Update comments.
134 (indirect_thunk_bnd_needed): Likewise.
135 (indirect_thunks_used): Likewise.
136 (indirect_thunks_bnd_used): Likewise.
137 (indirect_return_needed): New.
138 (indirect_return_bnd_needed): Likewise.
139 (output_indirect_thunk_function): Add a bool argument for
141 (output_indirect_thunk_function): Don't generate alias for
142 function return thunk.
143 (ix86_code_end): Call output_indirect_thunk_function to generate
144 function return thunks.
145 (ix86_output_function_return): Set indirect_return_bnd_needed
146 and indirect_return_needed instead of indirect_thunk_bnd_needed
147 and indirect_thunk_needed.
149 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
151 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
152 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
153 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
155 2018-03-15 David Malcolm <dmalcolm@redhat.com>
156 Paul Hua <paul.hua.gm@gmail.com>
159 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
161 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
163 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
164 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
167 2018-03-15 Tamar Christina <tamar.christina@arm.com>
170 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
171 instead of GET_MODE_SIZE when comparing Units.
173 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
176 * varasm.c (hash_section): Return an unchangeble hash value
177 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
178 Return !aarch64_can_use_per_function_literal_pools_p ().
180 2018-03-15 Jakub Jelinek <jakub@redhat.com>
183 * optabs.c (emit_conditional_move): Pass address of cmode's copy
184 rather than address of cmode as last argument to prepare_cmp_insn.
186 2018-03-15 Julia Koval <julia.koval@intel.com>
188 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
189 F_AVX512VNNI, F_AVX512BITALG): New.
191 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
194 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
195 insn for floating-point loads and stores.
197 2018-03-14 Carl Love <cel@us.ibm.com>
199 * config/rs6000/rs6000-c.c: Add macro definitions for
200 ALTIVEC_BUILTIN_VEC_PERMXOR.
201 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
202 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
203 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
204 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
206 * config/doc/extend.texi: Add prototypes for vec_permxor.
208 2018-03-14 David Malcolm <dmalcolm@redhat.com>
211 * diagnostic-show-locus.c (class layout_point): Convert m_line
212 from int to linenum_type.
213 (line_span::comparator): Use linenum "compare" function when
214 comparing line numbers.
215 (test_line_span): New function.
216 (layout_range::contains_point): Convert param "row" from int to
218 (layout_range::intersects_line_p): Likewise.
219 (layout::will_show_line_p): Likewise.
220 (layout::print_source_line): Likewise.
221 (layout::should_print_annotation_line_p): Likewise.
222 (layout::print_annotation_line): Likewise.
223 (layout::print_leading_fixits): Likewise.
224 (layout::annotation_line_showed_range_p): Likewise.
225 (struct line_corrections): Likewise for field m_row.
226 (line_corrections::line_corrections): Likewise for param "row".
227 (layout::print_trailing_fixits): Likewise.
228 (layout::get_state_at_point): Likewise.
229 (layout::get_x_bound_for_row): Likewise.
230 (layout::print_line): Likewise.
231 (diagnostic_show_locus): Likewise for locals "last_line" and
233 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
234 * input.c (selftest::test_linenum_comparisons): New function.
235 (selftest::input_c_tests): Call it.
236 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
237 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
238 * selftest.h (ASSERT_GT): New macro.
239 (ASSERT_GT_AT): New macro.
240 (ASSERT_LT): New macro.
241 (ASSERT_LT_AT): New macro.
243 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
245 PR rtl-optimization/84780
246 * combine.c (distribute_links): Don't make a link based on pc_rtx.
248 2018-03-14 Martin Liska <mliska@suse.cz>
250 * tree.c (record_node_allocation_statistics): Use
252 (get_stats_node_kind): New function extracted from
253 record_node_allocation_statistics.
254 (free_node): Use get_stats_node_kind.
256 2018-03-14 Richard Biener <rguenther@suse.de>
258 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
259 that the value-set of ANTIC_IN doesn't grow.
262 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
264 (BB_VISITED_WITH_VISITED_SUCCS): New define.
265 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
267 2018-03-14 Julia Koval <julia.koval@intel.com>
269 * config.gcc (icelake-client, icelake-server): New.
271 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
272 (initial_ix86_arch_features): Ditto.
273 (PTA_SKYLAKE): Add SGX.
274 (PTA_ICELAKE): Remove.
275 (PTA_ICELAKE_CLIENT): New.
276 (PTA_ICELAKE_SERVER): New.
277 (ix86_option_override_internal): Split up icelake on icelake client and
279 (get_builtin_code_for_version): Ditto.
280 (fold_builtin_cpu): Ditto.
281 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
282 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
283 * config/i386/i386.h (processor_type): Ditto.
284 * doc/invoke.texi: Ditto.
286 2018-03-14 Jakub Jelinek <jakub@redhat.com>
289 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
290 INTEGER_CST offset, add it together with bitpos / 8 and
291 sign extend based on POINTER_SIZE.
295 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
298 * config/i386/constraints.md (Yc): New register constraint.
299 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
300 Use Yc constraint for alternative 2 of operand 0. Remove
301 preferred_for_speed attribute.
303 2018-03-14 Richard Biener <rguenther@suse.de>
305 PR tree-optimization/84830
306 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
307 with the old one to avoid oscillations.
309 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
312 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
314 (assign_by_spills): Return a flag of reload assignment failure.
315 Do not process the reload assignment failures. Do not spill other
316 reload pseudos if they has the same reg class. Update n if
318 (lra_assign): Add a return arg. Set up from the result of
319 assign_by_spills call.
320 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
321 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
322 usage_insns if it is not NULL.
323 (spill_hard_reg_in_range): New function.
324 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
325 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
327 (lra_assign): Change prototype.
328 * lra.c (lra): Add code to deal with fails by splitting hard reg
331 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
333 * config/riscv/riscv.opt (mrelax): New option.
334 * config/riscv/riscv.c (riscv_file_start): Emit ".option
335 "norelax" when riscv_mrelax is disabled.
336 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
338 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
341 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
342 reassociation for int modes.
344 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
346 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
347 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
349 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
350 * config/aarch64/aarch64-sve.md
351 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
352 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
353 (*extend<mode><Vwide>2): Rename to...
354 (aarch64_sve_extend<mode><Vwide>2): ...this.
355 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
356 renaming the old pattern to...
357 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
359 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
360 define_expand, renaming the old pattern to...
361 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
362 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
363 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
364 account when deciding which SVE instruction the optab should use.
365 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
367 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
369 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
370 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
371 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
372 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
373 (tlsdesc_small_<mode>): Turn a define_expand and use
374 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
375 (tlsdesc_small_advsimd_<mode>): ...this.
376 (tlsdesc_small_sve_<mode>): New pattern.
378 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
380 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
381 (UNSPEC_UMUL_HIGHPART): New constants.
382 (MUL_HIGHPART): New int iteraor.
383 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
384 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
386 (*<su>mul<mode>3_highpart): New define_insn.
388 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
391 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
394 2018-03-13 Martin Liska <mliska@suse.cz>
397 * (sem_item_optimizer::sem_item_optimizer): Initialize new
399 (sem_item_optimizer::~sem_item_optimizer): Release it.
400 (sem_item_optimizer::merge_classes): Register variable aliases.
401 (sem_item_optimizer::fixup_pt_set): New function.
402 (sem_item_optimizer::fixup_points_to_sets): Likewise.
403 * ipa-icf.h: Declare new variables and functions.
405 2018-03-13 Jakub Jelinek <jakub@redhat.com>
408 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
409 integer_pow2p@2 and test integer_pow2p in condition.
410 (A < 0 ? C : 0): Similarly for @1.
413 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
414 characters starting at p contain '\0' character, don't look beyond
418 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
419 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
422 * reg-stack.c (change_stack): Change update_end var from int to
423 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
424 also call set_block_for_insn on the newly added insns and rescan.
427 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
431 * tree.c (stabilize_reference_1): Return save_expr (e) for
432 STATEMENT_LIST even if it doesn't have side-effects.
434 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
436 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
438 2018-03-12 Renlin Li <renlin.li@arm.com>
440 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
441 aarch64_output_scalar_simd_mov_immediate.
443 2018-03-12 Martin Sebor <msebor@redhat.com>
445 PR tree-optimization/83456
446 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
447 for perfectly overlapping calls to memcpy.
448 (gimple_fold_builtin_memory_chk): Same.
449 (gimple_fold_builtin_strcpy): Handle no-warning.
450 (gimple_fold_builtin_stxcpy_chk): Same.
451 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
453 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
455 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
456 parameter. Use it for SFmode.
457 (rs6000_function_arg_advance_1): Adjust.
458 (rs6000_function_arg): Adjust.
459 (rs6000_gimplify_va_arg): Pass false for that new parameter.
461 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
463 PR rtl-optimization/84169
464 PR rtl-optimization/84780
465 * combine.c (can_combine_p): Check for a 2-insn combination whether
466 the destination register is used between the two insns, too.
468 2018-03-12 Richard Biener <rguenther@suse.de>
470 PR tree-optimization/84803
471 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
472 for refs DR analysis didn't process.
474 2018-03-12 Richard Biener <rguenther@suse.de>
476 PR tree-optimization/84777
477 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
478 force-vectorize loops ignore whether we are optimizing for size.
480 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
482 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
483 (TARGET_MD_ASM_ADJUST): Define.
485 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
486 Kito Cheng <kito.cheng@gmail.com>
487 Chung-Ju Wu <jasonwucj@gmail.com>
489 * config/nds32/nds32.c (nds32_compute_stack_frame,
490 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
491 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
492 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
493 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
494 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
495 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
496 * config/nds32/nds32.md (prologue, epilogue): Use macro
497 NDS32_V3PUSH_AVAILABLE_P to do checking.
499 2018-03-11 Jakub Jelinek <jakub@redhat.com>
502 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
503 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
504 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
505 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
506 addition of most attributes on !orig_type_die or the attribute not
507 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
509 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
510 Chung-Ju Wu <jasonwucj@gmail.com>
512 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
514 * config/nds32/nds32.opt (mvh): New option.
516 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
517 Chung-Ju Wu <jasonwucj@gmail.com>
519 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
521 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
522 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
525 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
526 Chung-Ju Wu <jasonwucj@gmail.com>
528 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
530 * config/nds32/nds32-multiple.md (strlensi): New pattern.
531 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
533 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
534 Kito Cheng <kito.cheng@gmail.com>
535 Chung-Ju Wu <jasonwucj@gmail.com>
537 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
538 UNSPEC_FFMISM and UNSPEC_FLMISM.
539 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
540 for ffb, ffmism and flmism.
541 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
542 (unspec_ffmism): Ditto.
543 (unspec_flmism): Ditto.
544 (nds32_expand_builtin_impl): Check if string extension is available.
545 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
546 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
548 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
551 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
554 * lra-assigns.c (assign_by_spills): Return a flag of reload
555 assignment failure. Do not process the reload assignment
556 failures. Do not spill other reload pseudos if they has the same
558 (lra_assign): Add a return arg. Set up from the result of
559 assign_by_spills call.
560 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
561 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
562 usage_insns if it is not NULL.
563 (spill_hard_reg_in_range): New function.
564 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
565 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
567 (lra_assign): Change prototype.
568 * lra.c (lra): Add code to deal with fails by splitting hard reg
571 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
574 * config/i386/i386.opt: Replace Enforcment with Enforcement.
576 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
579 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
580 (dw_val_node): Add val_symbolic_view.
581 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
582 (symview_upper_bound): New.
583 (new_line_info_table): Initialize symviews_since_reset.
584 (dwarf2out_source_line): Count symviews_since_reset and set
586 (dw_val_equal_p): Handle symview.
587 (add_AT_symview): New.
588 (print_dw_val): Handle symview.
589 (attr_checksum, attr_checksum_ordered): Likewise.
590 (same_dw_val_p, size_of_die): Likewise.
591 (value_format, output_die): Likewise.
592 (add_high_low_attributes): Use add_AT_symview for entry_view.
593 (dwarf2out_finish): Reset symview_upper_bound, clear
596 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
599 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
600 Add strict argument and use it.
601 (rs6000_split_multireg_move): Update for new strict argument.
602 (mem_operand_gpr): Disallow all non-offsettable addresses.
603 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
605 2018-03-09 Jakub Jelinek <jakub@redhat.com>
608 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
609 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
610 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
613 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
614 decl, use remap_type if we want to use the type.
616 2018-03-09 Martin Sebor <msebor@redhat.com>
618 PR tree-optimization/84526
619 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
621 (builtin_access::generic_overlap): Be prepared to handle non-array
624 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
626 PR rtl-optimization/84682
627 * lra-constraints.c (process_address_1): Check is_address flag
628 for address constraints.
629 (process_alt_operands): Likewise.
630 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
631 preprocess_constraints.
632 * recog.h (preprocess_constraints): Add oploc parameter.
634 * recog.c (preprocess_constraints): Test address_operand for
635 CT_ADDRESS constraints.
637 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
640 * lra-assigns.c (assign_by_spills): Return a flag of reload
641 assignment failure. Do not process the reload assignment
642 failures. Do not spill other reload pseudos if they has the same
644 (lra_assign): Add a return arg. Set up from the result of
645 assign_by_spills call.
646 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
647 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
648 usage_insns if it is not NULL.
649 (spill_hard_reg_in_range): New function.
650 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
651 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
653 (lra_assign): Change prototype.
654 * lra.c (lra): Add code to deal with fails by splitting hard reg
657 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
660 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
661 Accept complain bool parameter. Only emit errors if it is true.
662 (arm_parse_cpu_option_name): Likewise.
663 (arm_target_thumb_only): Adjust callers of the above.
664 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
665 prototype to take a default true bool parameter.
666 (arm_parse_arch_option_name): Likewise.
668 2018-03-09 David Malcolm <dmalcolm@redhat.com>
669 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
673 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
674 * configure: Regenerate.
675 * configure.ac ("linker --version-script option"): New.
676 ("linker soname option"): New.
678 2018-03-09 Richard Biener <rguenther@suse.de>
680 PR tree-optimization/84775
681 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
682 immediate uses of predicate stmts and mark them modified.
685 PR tree-optimization/84178
686 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
688 (version_loop_for_if_conversion): Delay update_ssa call.
689 (tree_if_conversion): Delay update_ssa until after predicate
692 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
695 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
696 when the function accesses prior frames.
698 2018-03-08 Jakub Jelinek <jakub@redhat.com>
701 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
702 gen_llsym, otherwise call maybe_gen_llsym.
705 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
706 has ',' character inside of it.
708 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
711 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
712 as clobbering CC_REGNUM.
714 2018-03-08 Richard Biener <rguenther@suse.de>
717 * tree-scalar-evolution.c: Include tree-into-ssa.h.
718 (follow_copies_to_constant): Do not follow SSA names registered
721 2018-03-08 Richard Biener <rguenther@suse.de>
723 PR tree-optimization/84178
724 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
726 (version_loop_for_if_conversion): Delay update_ssa call.
727 (tree_if_conversion): Delay update_ssa until after predicate
730 2018-03-08 David Malcolm <dmalcolm@redhat.com>
732 PR tree-optimization/84178
733 * tree-if-conv.c (release_bb_predicate): Remove the
734 the assertion that the stmts have NULL use_ops.
735 Discard the statements, asserting that they haven't
736 yet been added to a BB.
738 2018-03-08 Richard Biener <rguenther@suse.de>
740 PR tree-optimization/84746
741 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
742 (phi_translate): Pass in destination ANTIC_OUT set.
743 (phi_translate_1): Likewise. For a simplified result lookup
744 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
745 (phi_translate_set): Adjust.
746 (do_pre_regular_insertion): Likewise.
747 (do_pre_partial_partial_insertion): Likewise.
749 2018-03-08 Martin Liska <mliska@suse.cz>
751 PR gcov-profile/84735
752 * doc/gcov.texi: Document usage of profile files.
753 * gcov-io.h: Document changes in the format.
755 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
759 * dwarf2out.c (struct dw_line_info_table): Update comments for
761 (FORCE_RESET_NEXT_VIEW): New.
762 (FORCE_RESETTING_VIEW_P): New.
763 (RESETTING_VIEW_P): Check for -1 too.
764 (ZERO_VIEW_P): Likewise.
765 (new_line_info_table): Force-reset next view.
766 (dwarf2out_begin_function): Likewise.
767 (dwarf2out_source_line): Simplify zero_view_p initialization.
768 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
769 view directly. Omit view when omitting .loc at line 0.
771 2018-03-08 Jakub Jelinek <jakub@redhat.com>
773 PR tree-optimization/84740
774 * tree-switch-conversion.c (process_switch): Call build_constructors
775 only if info.phi_count is non-zero.
777 PR tree-optimization/84739
778 * tree-tailcall.c (find_tail_calls): Check call arguments against
779 DECL_ARGUMENTS (current_function_decl) rather than
780 DECL_ARGUMENTS (func) when checking for tail recursion.
782 2018-03-07 Jakub Jelinek <jakub@redhat.com>
784 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
785 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
786 Volker Reichelt's entry and add entries for people that perform
787 GCC fuzzy testing and report numerous bugs.
789 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
792 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
793 readonly data in sdata, if that is disabled.
794 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
795 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
796 -mreadonly-in-sdata option.
798 2018-03-07 Martin Sebor <msebor@redhat.com>
800 PR tree-optimization/84468
801 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
802 basic block when looking for nul assignment.
804 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
807 * except.h (output_function_exception_table): Adjust prototype.
808 * except.c (output_function_exception_table): Remove FNNAME parameter
809 and add SECTION parameter. Ouput one part of the table at a time.
810 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
811 the first part of the exception table and emit unwind directives.
812 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
813 (i386_pe_seh_cold_init): Likewise.
814 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
815 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
816 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
817 (ix86_output_call_insn): Emit a nop in one more case for SEH.
818 * config/i386/winnt.c: Include except.h.
819 (struct seh_frame_state): Add reg_offset, after_prologue and
820 in_cold_section fields.
821 (i386_pe_seh_end_prologue): Set seh->after_prologue.
822 (i386_pe_seh_cold_init): New function.
823 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
824 to seh->in_cold_section.
825 (seh_emit_push): Record the offset of the push.
826 (seh_emit_save): Record the offet of the save.
827 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
828 Test seh->after_prologue to disregard the epilogue.
829 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
830 (i386_pe_end_cold_function): New function.
832 2018-03-07 Jakub Jelinek <jakub@redhat.com>
835 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
836 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
839 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
841 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
842 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
845 * multiple_target.c: Include tree-inline.h and intl.h.
846 (expand_target_clones): Diagnose and fail if node->definition and
847 !tree_versionable_function_p (node->decl).
849 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
851 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
853 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
854 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
855 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
857 2018-03-06 Jakub Jelinek <jakub@redhat.com>
860 * combine.c (try_combine): Use reg_or_subregno instead of handling
861 just paradoxical SUBREGs and REGs.
863 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
865 * config/arc/arc.c (arc_finalize_pic): Remove function.
866 (arc_must_save_register): We use single base PIC register, remove
867 checks to save/restore the PIC register.
868 (arc_expand_prologue): Likewise.
869 * config/arc/arc-protos.h (arc_set_default_type_attributes):
871 (arc_verify_short): Likewise.
872 (arc_attr_type): Likewise.
873 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
874 (walk_stores): Likewise.
875 (arc_address_cost): Make it static.
876 (arc_verify_short): Likewise.
877 (branch_dest): Likewise.
878 (arc_attr_type): Likewise.
879 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
880 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
881 (arc_final_prescan_insn): Remove inserting the nops due to
882 hardware hazards. It is done in reorg step.
883 (insn_length_variant_t): Remove.
884 (insn_length_parameters_t): Likewise.
885 (arc_insn_length_parameters): Likewise.
886 (arc_get_insn_variants): Likewise.
887 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
889 2018-03-06 Jakub Jelinek <jakub@redhat.com>
892 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
895 PR tree-optimization/84687
896 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
898 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
900 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
902 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
903 Rename to ppc_speculation_barrier.
904 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
905 __builtin_ppc_speculation_barrier.
907 2018-03-05 Jakub Jelinek <jakub@redhat.com>
910 * combine.c (combine_simplify_rtx): Don't try to simplify if
911 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
914 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
916 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
917 to 32 bytes when compiling for POWER9.
919 2018-03-05 Jakub Jelinek <jakub@redhat.com>
922 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
923 regparm >= 3 with no arg reg available also for calls with
924 flag_force_indirect_call. Pass decl to ix86_function_regparm.
927 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
929 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
931 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
934 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
936 2018-03-05 Richard Biener <rguenther@suse.de>
938 PR tree-optimization/84486
939 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
940 When inserting a __builtin_assume_aligned call set the LHS
941 SSA name alignment info accordingly.
943 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
945 PR tree-optimization/84114
946 * config/aarch64/aarch64.c (aarch64_reassociation_width)
947 Avoid reassociation of FLOAT_MODE addition.
949 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
951 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
952 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
953 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
954 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
955 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
956 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
957 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
959 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
960 __builtin_ia32_wbinvd): New builtins.
961 (SPECIAL_ARGS2): New.
962 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
963 (SPECIAL_ARGS2): New.
964 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
965 (ix86_valid_target_attribute_inner_p): Ditto.
966 (ix86_init_mmx_sse_builtins): Add special_args2.
967 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
968 TARGET_WBNOINVD_P): New.
969 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
970 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
971 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
972 * config/i386/immintrin.h (_wbinvd): New intrinsic.
973 * config/i386/pconfigintrin.h: New file.
974 * config/i386/wbnoinvdintrin.h: Ditto.
975 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
976 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
978 2018-03-05 Richard Biener <rguenther@suse.de>
980 PR tree-optimization/84670
981 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
983 (BB_VISITED_WITH_VISITED_SUCCS): New define.
984 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
985 (compute_antic_aux): Only assert the number of values in ANTIC_IN
986 doesn't grow if all successors (recursively) were visited at least
989 2018-03-05 Richard Biener <rguenther@suse.de>
991 PR tree-optimization/84650
992 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
993 if executed in the loop pipeline.
995 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
997 * doc/configfiles.texi (Configuration Files): Move info about
998 conditionalizing $target-protos.h to...
999 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1000 differs from $target-protos.h.
1002 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1003 Chung-Ju Wu <jasonwucj@gmail.com>
1005 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1006 * config/nds32/nds32-multiple.md (setmemsi): Define.
1007 * config/nds32/nds32-memory-manipulation.c
1008 (nds32_gen_dup_4_byte_to_word_value): New.
1009 (emit_setmem_word_loop): New.
1010 (emit_setmem_byte_loop): New.
1011 (nds32_expand_setmem_loop): New.
1012 (nds32_expand_setmem_loop_v3m): New.
1013 (nds32_expand_setmem_unroll): New.
1014 (nds32_expand_setmem): New.
1016 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1017 Chung-Ju Wu <jasonwucj@gmail.com>
1019 * config/nds32/nds32-memory-manipulation.c
1020 (nds32_emit_load_store): New.
1021 (nds32_emit_post_inc_load_store): New.
1022 (nds32_emit_mem_move): New.
1023 (nds32_emit_mem_move_block): New.
1024 (nds32_expand_movmemsi_loop_unknown_size): New.
1025 (nds32_expand_movmemsi_loop_known_size): New.
1026 (nds32_expand_movmemsi_loop): New.
1027 (nds32_expand_movmemsi_unroll): New.
1028 (nds32_expand_movmemqi): Rename ...
1029 (nds32_expand_movmemsi): ... to this.
1030 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1031 (movmemsi): ... to this.
1032 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1033 (nds32_expand_movmemsi): ... to this.
1035 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1036 Monk Chiang <sh.chiang04@gmail.com>
1037 Chung-Ju Wu <jasonwucj@gmail.com>
1039 * config/nds32/nds32-protos.h
1040 (nds32_expand_load_multiple): New arguments.
1041 (nds32_expand_store_multiple): Ditto.
1042 (nds32_valid_multiple_load_store): Rename ...
1043 (nds32_valid_multiple_load_store_p): ... to this.
1044 * config/nds32/nds32-memory-manipulation.c
1045 (nds32_expand_load_multiple): Refine implementation.
1046 (nds32_expand_store_multiple): Ditto.
1047 * config/nds32/nds32-multiple.md
1048 (load_multiple): Update nds32_expand_load_multiple interface.
1049 (store_multiple): Update nds32_expand_store_multiple interface.
1050 * config/nds32/nds32-predicates.c
1051 (nds32_valid_multiple_load_store): Rename ...
1052 (nds32_valid_multiple_load_store_p): ... to this and refine
1054 * config/nds32/predicates.md
1055 (nds32_load_multiple_and_update_address_operation): New predicate.
1056 (nds32_store_multiple_and_update_address_operation): New predicate.
1058 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1059 Chung-Ju Wu <jasonwucj@gmail.com>
1061 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1062 (combo): New attribute.
1063 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1065 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1067 * config/nds32/nds32.opt: Change -mcmodel= default value.
1069 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1070 Monk Chiang <sh.chiang04@gmail.com>
1071 Chung-Ju Wu <jasonwucj@gmail.com>
1073 * config/nds32/constants.md (unspec_element): New enum.
1074 * config/nds32/constraints.md (Umw): New constraint.
1075 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1076 * config/nds32/nds32-intrinsic.md: Likewise.
1077 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1078 (nds32_valid_smw_lwm_base_p): New.
1079 (nds32_output_smw_single_word): New.
1080 (nds32_output_lmw_single_word): New.
1081 (nds32_expand_unaligned_load): New.
1082 (nds32_expand_unaligned_store): New.
1083 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1084 (nds32_output_smw_single_word): Declare.
1085 (nds32_output_lmw_single_word): Declare.
1086 (nds32_expand_unaligned_load): Declare.
1087 (nds32_expand_unaligned_store): Declare.
1088 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1089 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1090 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1091 NDS32_BUILTIN_UASTORE_DW.
1092 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1095 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1096 Kito Cheng <kito.cheng@gmail.com>
1097 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config/nds32/nds32-intrinsic.c
1100 (nds32_expand_builtin_null_ftype_reg): Delete.
1101 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1102 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1103 (nds32_read_argument): New.
1104 (nds32_legitimize_target): Ditto.
1105 (nds32_legitimize_argument): Ditto.
1106 (nds32_check_constant_argument): Ditto.
1107 (nds32_expand_unop_builtin): Ditto.
1108 (nds32_expand_unopimm_builtin): Ditto.
1109 (nds32_expand_binop_builtin): Ditto.
1110 (nds32_builtin_decl_impl): Ditto.
1111 (builtin_description): Ditto.
1112 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1113 (nds32_init_builtins_impl): Ditto.
1114 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1115 (nds32_builtin_decl): New.
1116 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1117 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1119 2018-03-02 Jeff Law <law@redhat.com>
1121 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1122 (redundant_insn, fill_simple_delay_slots): Likewise.
1123 (fill_slots_from_thread): Likewise.
1124 * resource.c (mark_referenced_resources): Likewise.
1125 (mark_set_resources, find_dead_or_set_registers): Likewise.
1127 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1129 * substring-locations.h (format_warning_va): Formatting fix for
1131 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1133 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1134 * substring-locations.c: Include intl.h.
1135 (format_warning_va): Turned into small wrapper around
1136 format_warning_n_va, renamed to ...
1137 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1138 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1140 (format_warning_at_substring_n): New function.
1141 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1142 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1143 format_warning_at_substring with just a shorter name instead of
1144 const function pointer.
1145 (fmtwarn_n): New function.
1146 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1147 appropriate, get rid of all the fmtstr temporaries, move conditionals
1148 with G_() wrapped string literals directly into fmtwarn arguments,
1149 cast dir.len to (int), formatting fixes.
1151 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1153 * doc/invoke.texi: Remove "Cilk Plus" references.
1155 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1156 Richard Biener <rguenther@suse.de>
1159 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1160 for error or warning attributes if CALL_FROM_THUNK_P is set.
1163 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1166 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1167 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1170 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1171 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1173 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1174 instead of -1U in last predictors element's probability member.
1176 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1179 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1180 arguments if they are comparable.
1182 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1184 PR tree-optimization/84634
1185 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1186 masks and masked_loop_p with a single loop_masks, making sure it's
1187 null for bb vectorization.
1189 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1191 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1192 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1193 loop->force_vectorize to check whether there is no alias.
1195 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1198 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1200 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1202 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1203 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1204 instead of a loop around prev_real_insn.
1205 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1209 * config/i386/i386.c (ix86_print_operand): Use conditional
1210 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1213 2018-03-02 Richard Biener <rguenther@suse.de>
1215 PR tree-optimization/84427
1216 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1217 (bitmap_set_subtract_values): Rewrite to handle multiple
1220 (prune_clobbered_mems): Likewise.
1221 (phi_translate): Take edge instead of pred/phiblock.
1222 (phi_translate_1): Likewise.
1223 (phi_translate_set): Likewise. Insert all translated
1224 exprs for a value into the set, keeping possibly multiple
1225 expressions per value.
1226 (compute_antic_aux): Adjust for phi_translate changes.
1227 When intersecting union the expressions and prune those
1228 not in the final value set, keeping possibly multiple
1229 expressions per value. Do not use value-insertion
1230 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1231 all expressions. Add verification that the value-sets
1232 only shrink during iteration.
1233 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1234 (do_pre_regular_insertion): Likewise.
1235 (do_pre_partial_partial_insertion): Likewise.
1237 2018-03-02 Richard Biener <rguenther@suse.de>
1240 * config/darwin.c (saved_debug_info_level): New static global.
1241 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1242 (darwin_asm_lto_end): Restore debug info generation settings.
1244 2018-03-01 Martin Liska <mliska@suse.cz>
1247 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1250 2018-03-01 Richard Biener <rguenther@suse.de>
1253 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1254 type annotation with LTO when debug was disabled at compile-time.
1256 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1258 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1260 (mips_final_postscan_insn): Likewise.
1262 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1264 PR rtl-optimization/84528
1265 * alias.c (init_alias_target): Add commentary.
1266 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1267 a unique base value if the frame pointer is not eliminated
1268 to the stack pointer.
1270 2018-03-01 Tom de Vries <tom@codesourcery.com>
1272 PR rtl-optimization/83327
1273 * lra-int.h (hard_regs_spilled_into): Declare.
1274 * lra.c (hard_regs_spilled_into): Define.
1275 (init_reg_info): Init hard_regs_spilled_into.
1276 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1277 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1278 (process_bb_lives): Handle hard_regs_spilled_into.
1279 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1280 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1282 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1284 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1285 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1286 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1287 * config/rs6000/aix72.h: New file.
1289 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1291 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1292 instead of warning_at with conditional singular and plural messages
1296 * stor-layout.c (update_alignment_for_field): For
1297 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1298 && !DECL_PACKED (field), do the alignment update, just use
1299 only desired_align instead of MAX (type_align, desired_align)
1301 (place_field): Don't do known_align < desired_align handling
1302 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1303 is non-NULL, instead do it after rli->prev_field handling and
1304 only if not within a bitfield word. For DECL_PACKED (field)
1305 use type_align of BITS_PER_UNIT.
1307 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1309 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1310 superfluous parentheses and trailing spaces.
1312 2018-02-28 Richard Biener <rguenther@suse.de>
1314 PR tree-optimization/84584
1315 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1316 SCoPs with fake exit edge.
1318 2018-02-28 Martin Liska <mliska@suse.cz>
1321 * timevar.c (timer::print): Fix format to properly print 100%
1324 2018-02-28 Richard Biener <rguenther@suse.de>
1327 * genmatch.c (capture_info::walk_match): Do not mark
1328 captured expressions without operands as expr_p given
1329 they act more like predicates and should be subject to
1330 "lost tail" side-effect preserving.
1332 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1334 PR rtl-optimization/81611
1335 * auto-inc-dec.c (attempt_change): Move dead note from
1336 mem_insn if it's the next use of regno
1337 (find_address): Take address use of reg holding
1338 non-incremented value. Add parm to limit search to the named
1340 (merge_in_block): Attempt to use a mem insn that is the next
1341 use of the original regno.
1343 2018-02-27 Martin Sebor <msebor@redhat.com>
1346 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1347 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1349 2018-02-27 Martin Sebor <msebor@redhat.com>
1351 PR translation/84207
1352 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1353 n argument to unsigned HOST_WIDE_INT.
1354 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1355 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1356 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1357 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1359 2018-02-27 Richard Biener <rguenther@suse.de>
1361 PR tree-optimization/84512
1362 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1363 Do not use the estimate returned from record_stmt_cost for
1364 the scalar iteration cost but sum properly using add_stmt_cost.
1366 2018-02-27 Richard Biener <rguenther@suse.de>
1368 PR tree-optimization/84466
1369 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1370 Adjust last change to less strictly validate use operands.
1372 2018-02-27 Martin Liska <mliska@suse.cz>
1374 PR gcov-profile/84548
1375 * gcov.c (process_file): Allow partial overlap and consider it
1376 also as group functions.
1377 (output_lines): Properly calculate range of lines for a group.
1379 2018-02-27 Martin Liska <mliska@suse.cz>
1381 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1382 'ggc' suffixes. Change first column width.
1383 (timer::print): Fix formatting of the column.
1385 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1387 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1388 preserve inline entry blocks for the sake of debug inline
1389 entry point markers alone.
1390 (remove_unused_locals): Suggest in comments a better place to
1391 force the preservation of inline entry blocks that are
1392 otherwise unused, but do not preserve them.
1394 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1396 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1398 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1401 * config/i386/constraints.md (Bs): Replace
1402 ix86_indirect_branch_register with
1403 TARGET_INDIRECT_BRANCH_REGISTER.
1405 * config/i386/i386.md (indirect_jump): Likewise.
1406 (tablejump): Likewise.
1407 (*sibcall_memory): Likewise.
1408 (*sibcall_value_memory): Likewise.
1409 Peepholes of indirect call and jump via memory: Likewise.
1410 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1411 (*sibcall_value_GOT_32): Likewise.
1412 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1413 (GOT_memory_operand): Likewise.
1414 (call_insn_operand): Likewise.
1415 (sibcall_insn_operand): Likewise.
1416 (GOT32_symbol_operand): Likewise.
1417 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1419 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1421 PR rtl-optimization/83496
1422 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1423 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1424 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1425 redundant insn, if any.
1426 (relax_delay_slots): Likewise.
1427 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1429 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1431 PR tree-optimization/83965
1432 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1433 that grouped statements are part of a reduction chain. Return
1434 true if the statement is not marked as a reduction itself but
1436 (vect_recog_dot_prod_pattern): Don't check whether the statement
1437 is part of a group here.
1438 (vect_recog_sad_pattern): Likewise.
1439 (vect_recog_widen_sum_pattern): Likewise.
1441 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1444 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1446 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1449 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1451 (ix86_output_indirect_function_return): New prototype.
1452 (ix86_split_simple_return_pop_internal): Likewise.
1453 * config/i386/i386.c (indirect_return_via_cx): New.
1454 (indirect_return_via_cx_bnd): Likewise.
1455 (indirect_thunk_name): Handle return va CX_REG.
1456 (output_indirect_thunk_function): Create alias for
1457 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1458 (ix86_output_indirect_jmp): Remove the bool argument.
1459 (ix86_output_indirect_function_return): New function.
1460 (ix86_split_simple_return_pop_internal): Likewise.
1461 * config/i386/i386.md (*indirect_jump): Don't pass false
1462 to ix86_output_indirect_jmp.
1463 (*tablejump_1): Likewise.
1464 (simple_return_pop_internal): Change it to define_insn_and_split.
1465 Call ix86_split_simple_return_pop_internal to split it for
1467 (simple_return_indirect_internal): Call
1468 ix86_output_indirect_function_return instead of
1469 ix86_output_indirect_jmp.
1471 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1474 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1475 memset and value initialization afterwards.
1477 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1479 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1481 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1484 * common/config/aarch64/aarch64-common.c
1485 (aarch_option_optimization_table[]): Switch
1486 off fomit-frame-pointer
1488 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1489 Chung-Ju Wu <jasonwucj@gmail.com>
1491 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1493 (store_multiple): Ditto.
1495 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1497 * config.gcc: Add --with-cpu support for nds32 target.
1498 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1499 * config/nds32/nds32.opt: Add -mcpu= option.
1501 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1503 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1504 isel=yes): Warn for these deprecated options.
1506 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1508 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1509 ISA_2_5_MASKS_EMBEDDED.
1511 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1513 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1514 p->max as pointers rather than using iterative_hash_expr.
1516 2018-02-23 Carl Love <cel@us.ibm.com>
1518 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1519 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1521 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1522 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1523 P8V_BUILTIN_VEC_VUNSIGNED2.
1525 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1528 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1529 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1531 * lra-constraints.c (curr_insn_transform): Set up
1532 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1534 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1535 * lra-eliminations.c (spill_pseudos): Ditto.
1536 (process_insn_for_elimination): Ditto.
1537 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1538 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1539 LRA_NON_CLOBBERED_ALT.
1541 2018-02-22 Martin Sebor <msebor@redhat.com>
1543 PR tree-optimization/84480
1544 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1545 to maybe_diag_stxncpy_trunc. Call it.
1546 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1547 from gimple_fold_builtin_strcpy. Print inlining stack.
1548 (handle_builtin_stxncpy): Print inlining stack.
1549 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1551 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1554 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1555 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1556 and -fcheck-pointer-bounds are used together.
1557 (indirect_thunk_prefix): New enum.
1558 (indirect_thunk_need_prefix): New function.
1559 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1560 "_nt" instead of "_bnd" for NOTRACK prefix.
1561 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1562 (output_indirect_thunk_function): Likewise.
1564 (ix86_code_end): Update output_indirect_thunk_function calls.
1565 (ix86_output_indirect_branch_via_reg): Replace
1566 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1567 (ix86_output_indirect_branch_via_push): Likewise.
1568 (ix86_output_function_return): Likewise.
1569 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1570 incompatible with -fcf-protection=branch and
1571 -fcheck-pointer-bounds.
1573 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1576 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1577 Change gcc_assert call to output_operand_lossage.
1579 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1581 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1583 2018-02-22 DJ Delorie <dj@redhat.com>
1584 Sebastian Perta <sebastian.perta@renesas.com>
1585 Oleg Endo <olegendo@gcc.gnu.org>
1587 * config/rx/rx.c (rx_rtx_costs): New function.
1588 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1590 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1592 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1594 2018-02-22 Martin Liska <mliska@suse.cz>
1597 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1598 Add "native" as a possible value.
1600 2018-02-22 Martin Liska <mliska@suse.cz>
1603 * config/i386/i386.c (ix86_option_override_internal):
1604 Add "native" as a possible value for -march and -mtune.
1606 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1609 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1610 to all type variants.
1612 PR tree-optimization/84503
1613 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1614 width as info->bitpos + info->bitsize - start.
1615 (merged_store_group::merge_overlapping): Simplify width computation.
1616 (check_no_overlap): New function.
1617 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1618 start + width and last_order of the group, fail if check_no_overlap
1620 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1621 to group if check_no_overlap fails.
1623 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1625 * config/rs6000/altivec.md: Delete contraint arguments to
1626 define_expand, define_split, and define_peephole2, and in
1627 define_insn_and_split if always unused.
1628 * config/rs6000/darwin.md: Ditto.
1629 * config/rs6000/dfp.md: Ditto.
1630 * config/rs6000/rs6000.md: Ditto.
1631 * config/rs6000/sync.md: Ditto.
1632 * config/rs6000/vector.md: Ditto.
1633 * config/rs6000/vsx.md: Ditto.
1635 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1637 * config/rs6000/altivec.md: Write output control strings as braced
1638 blocks instead of double-quoted strings.
1639 * config/rs6000/darwin.md: Ditto.
1640 * config/rs6000/rs6000.md: Ditto.
1641 * config/rs6000/vector.md: Ditto.
1642 * config/rs6000/vsx.md: Ditto.
1644 2018-02-21 Jason Merrill <jason@redhat.com>
1646 PR c++/84314 - ICE with templates and fastcall attribute.
1647 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1649 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1651 * ipa-cp.c (determine_versionability): Fix comment typos.
1653 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1656 * ipa-cp.c (determine_versionability): Do not version functions caling
1659 2018-02-21 Martin Liska <mliska@suse.cz>
1662 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1663 Add "native" as a possible value.
1664 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1665 the macro when native cpu detection is available.
1667 2018-02-21 Martin Liska <mliska@suse.cz>
1670 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1671 Add "native" as a possible value.
1672 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1673 when native cpu detection is available.
1675 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1676 Martin Sebor <msebor@redhat.com>
1678 PR tree-optimization/84478
1679 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1681 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1682 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1683 support which is conservatively correct, for 2 only stay conservative
1684 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1685 argument to the 2 argument get_range_strlen, adjust 6 arg
1686 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1688 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1689 (gimple_fold_builtin_strlen): Pass true as last argument to
1692 2018-02-20 Martin Sebor <msebor@redhat.com>
1695 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1696 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1697 (builtin_memref::builtin_memref): Factor out parts into
1698 set_base_and_offset and call it.
1700 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1703 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1704 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1705 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1706 search at the associated MODE_INT.
1708 2018-02-20 Jeff Law <law@redhat.com>
1711 PR tree-optimization/81592
1713 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1714 for range data rather than using global data.
1715 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1716 range data rather than using global data.
1717 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1718 pass it to children as needed.
1719 (struct directive::fmtresult): Similarly.
1720 (struct directive::set_width): Similarly.
1721 (struct directive::set_precision): Similarly.
1722 (format_integer, format_directive, parse_directive): Similarly.
1723 (format_none): Accept unnamed vr_values parameter.
1724 (format_percent, format_floating, format_character): Similarly.
1725 (format_string, format_plain): Similarly.
1726 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1727 the EVRP range analyzer for range data rather than using global data.
1728 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1729 gimple-ssa-evrp-analyze.h
1730 (class sprintf_dom_walker): Add after_dom_children member function.
1731 Add evrp_range_analyzer member.
1732 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1733 range analyzer as needed.
1734 (sprintf_dom_walker::after_dom_children): New member function.
1735 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1737 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1738 (evrp_range_analyzer::pop_to_marker): Likewise.
1740 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1742 PR tree-optimization/84419
1743 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1744 with the required type if its current type is compatible but
1747 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1750 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1751 after vectorization.
1753 2018-02-20 Martin Liska <mliska@suse.cz>
1756 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1757 possible values if we don't have a hint.
1759 2018-02-20 Martin Liska <mliska@suse.cz>
1763 * final.c (shorten_branches): Build align_tab array with one
1765 * opts.c (finish_options): Add alignment option limit check.
1766 (MAX_CODE_ALIGN): Likewise.
1767 (MAX_CODE_ALIGN_VALUE): Likewise.
1768 * doc/invoke.texi: Document maximum allowed option value for
1769 all -falign-* options.
1771 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1774 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1775 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1776 * var-tracking.c (emit_note_insn_var_location): Remove all references
1777 to NOTE_INSN_CALL_ARG_LOCATION.
1778 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1779 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1780 Use copy_rtx_if_shared.
1781 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1782 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1783 (dwarf2out_var_location): Remove handling of
1784 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1786 * final.c (final_scan_insn): Remove all references to
1787 NOTE_INSN_CALL_ARG_LOCATION.
1788 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1789 before dumping final insns.
1790 * except.c (emit_note_eh_region_end): Remove all references to
1791 NOTE_INSN_CALL_ARG_LOCATION.
1792 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1793 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1794 * config/arc/arc.c (hwloop_optimize): Likewise.
1795 * config/arm/arm.c (create_fix_barrier): Likewise.
1796 * config/s390/s390.c (s390_chunkify_start): Likewise.
1797 * config/sh/sh.c (find_barrier): Likewise.
1798 * config/i386/i386.c (rest_of_insert_endbranch,
1799 ix86_seh_fixup_eh_fallthru): Likewise.
1800 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1801 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1802 * config/frv/frv.c (frv_function_prologue): Likewise.
1803 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1805 (note_outside_basic_block_p): Remove all references to
1806 NOTE_INSN_CALL_ARG_LOCATION.
1807 * gengtype.c (adjust_field_rtx_def): Likewise.
1808 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1810 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1811 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1814 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1817 PR tree-optimization/84452
1818 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1819 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1822 2018-02-19 Martin Liska <mliska@suse.cz>
1825 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1827 2018-02-19 Martin Liska <mliska@suse.cz>
1828 Richard Sandiford <richard.sandiford@linaro.org>
1830 PR tree-optimization/82491
1831 * gimple-fold.c (get_base_constructor): Make earlier bail out
1834 2018-02-19 Carl Love <cel@us.ibm.com>
1836 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1837 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1839 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1840 P8V_BUILTIN_VEC_NEG.
1842 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1844 * config/rl78/rl78.md (movdf): New define expand.
1846 2018-02-19 Martin Liska <mliska@suse.cz>
1849 * doc/invoke.texi: Fix typo.
1850 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1852 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1854 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1855 handle rs6000_single_float and rs6000_double_float specially for
1858 2018-02-16 Jeff Law <law@redhat.com>
1860 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1863 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1866 * ipa-inline.c (inline_small_functions): Fix a typo.
1868 2018-02-16 Nathan Sidwell <nathan@acm.org>
1870 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1872 2018-02-16 Carl Love <cel@us.ibm.com>
1874 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1875 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1876 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1877 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1878 expansion to P8V_BUILTIN_VEC_FLOAT2.
1880 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1882 PR rtl-optimization/70023
1883 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1884 src_regno into account.
1886 2018-02-16 Carl Love <cel@us.ibm.com>
1888 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1889 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1890 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1891 * config/rs6000/rs6000.c: Remove case statements for
1892 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1893 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1894 and P9V_BUILTIN_VEC_VINSERT4B.
1895 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1896 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1897 * config/rs6000/vsx.md:
1898 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1901 2018-02-16 Carl Love <cel@us.ibm.com>
1903 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1905 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1907 * config/rs6000/rs6000-c.c: Add the definitions for
1908 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1909 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1910 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1911 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1912 definition for insert4b and define insn *insert3b_internal.
1913 * doc/extend.texi: Add documentation for vec_extract4b.
1915 2018-02-16 Nathan Sidwell <nathan@acm.org>
1917 * doc/extend.texi (Backwards Compatibility): Mention friend
1918 injection. Note for-scope is deprecated.
1919 * doc/invoke.texi (-ffriend-injection): Deprecate.
1921 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1923 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1924 that moved to I2, also allow destinations that are a paradoxical
1925 subreg (instead of a normal reg).
1927 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1930 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1933 2018-02-16 Richard Biener <rguenther@suse.de>
1935 PR tree-optimization/84037
1936 PR tree-optimization/84016
1938 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1939 Adjust vec_construct for the fact we need additional higher latency
1940 128bit inserts for AVX256 and AVX512 vector builds.
1941 (ix86_add_stmt_cost): Scale vector construction cost for
1944 2018-02-16 Richard Biener <rguenther@suse.de>
1946 PR tree-optimization/84417
1947 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1948 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1949 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1951 2018-02-16 Martin Liska <mliska@suse.cz>
1954 * internal-fn.def (ASAN_CHECK): Set proper flags.
1955 (ASAN_MARK): Likewise.
1957 2018-02-16 Julia Koval <julia.koval@intel.com>
1959 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1960 from PTA_CANNONLAKE.
1962 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1965 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1966 Use ++iter rather than iter++ for std::list iterators.
1967 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1968 defer deleting them until all nodes in the forest are processed. Do
1969 free even leaf nodes. Change to_process into auto_vec.
1972 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1973 * vec.h (vec_default_construct): Use memset instead of placement new
1974 if BROKEN_VALUE_INITIALIZATION is defined.
1975 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1976 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1979 PR rtl-optimization/83723
1980 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1981 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1982 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1984 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1986 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1988 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1990 PR rtl-optimization/81443
1991 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1992 from inner REGs to paradoxical SUBREGs.
1994 2018-02-16 Richard Biener <rguenther@suse.de>
1996 PR tree-optimization/84399
1997 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1998 For operands we can analyze at their definition make sure we can
1999 analyze them at each use as well.
2001 2018-02-16 Richard Biener <rguenther@suse.de>
2003 PR tree-optimization/84190
2004 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2005 volatile accesses if the decl isn't volatile.
2007 2018-02-15 Jason Merrill <jason@redhat.com>
2009 PR c++/84314 - ICE with templates and fastcall attribute.
2010 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2011 TYPE_CANONICAL on an existing type.
2013 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2015 PR tree-optimization/84383
2016 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2017 dstoff nor call operand_equal_p if dstbase is NULL.
2019 PR tree-optimization/84334
2020 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2021 also a CONSTANT_CLASS_P, punt.
2023 2018-02-14 Jim Wilson <jimw@sifive.com>
2025 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2026 first SMALL_OPERAND check. New local min_second_step. Move assert
2027 to where locals are set. Add TARGET_RVC support.
2028 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2030 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2032 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2034 2018-02-14 Martin Sebor <msebor@redhat.com>
2036 PR tree-optimization/83698
2037 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2038 arrays constrain the offset range to their bounds.
2039 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2040 (builtin_access::overlap): Avoid setting the size of overlap if it's
2042 (maybe_diag_overlap): Also consider arrays when deciding what values
2043 of offsets to include in diagnostics.
2045 2018-02-14 Martin Sebor <msebor@redhat.com>
2048 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2049 that correspond to the kind of a declaration.
2051 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2054 * config/pa/pa.md: Load address of PIC label using the linkage table
2055 if the label is nonlocal.
2057 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2059 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2060 warning message if user requests -maltivec=be.
2061 * doc/invoke.texi: Document deprecation of -maltivec=be.
2063 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2066 * config/rs6000/rs6000-c.c: Update definitions for
2067 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2068 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2070 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2073 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2074 add _get_ssp intrinsics. Remove argument from
2075 __builtin_ia32_rdssp[d|q].
2076 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2077 * config/i386/i386-builtin.def: Remove argument from
2078 __builtin_ia32_rdssp[d|q].
2079 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2080 ix86_expand_special_args_builtin for _rdssp[d|q].
2081 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2082 Clear register before usage.
2083 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2084 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2086 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2088 PR tree-optimization/84357
2089 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2090 operand 1 of an ARRAY_REF too.
2092 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2095 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2096 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2098 (set_of_reg): New struct.
2099 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2100 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2101 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2103 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2104 Split into bitclr, bitset, bitinvert patterns if appropriate.
2105 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2106 use rx_fuse_in_memory_bitop.
2107 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2108 to named insn, correct maximum insn length.
2110 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2113 * machmode.def: Define a complex mode for PARTIAL_INT.
2114 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2116 * doc/rtl.texi: Document CSPImode.
2117 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2119 (msp430_hard_regno_nregs_with_padding): Likewise.
2121 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2124 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2126 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2128 PR rtl-optimization/84169
2129 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2130 we generated a parallel as new i3 and we split that to new i2 and i3
2131 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2132 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2133 those to i2, not i1. Partially rewrite this scan code.
2135 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2138 * stor-layout.c (place_field): For variable length fields, adjust
2139 offset_align afterwards not just based on the field's alignment,
2140 but also on the size.
2143 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2144 of exps and logs in the use_exp2 case.
2146 2018-02-13 Jeff Law <law@redhat.com>
2148 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2151 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2154 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2158 * common.opt (gas-loc-support, gas-locview-support): New.
2159 (ginline-points, ginternal-reset-location-views): New.
2160 * doc/invoke.texi: Document them. Use @itemx where intended.
2161 (gvariable-location-views): Adjust.
2162 * target.def (reset_location_view): New.
2163 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2164 (TARGET_RESET_LOCATION_VIEW): New.
2165 * doc/tm.texi: Rebuilt.
2166 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2167 (dwarf2out_default_as_locview_support): New.
2168 (output_asm_line_debug_info): Use option variables.
2169 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2170 (output_loc_list): Likewise.
2171 (add_high_low_attributes): Check option variables.
2172 Don't output entry view attribute in strict mode.
2173 (gen_inlined_subroutine_die): Check option variables.
2174 (dwarf2out_inline_entry): Likewise.
2175 (init_sections_and_labels): Likewise.
2176 (dwarf2out_early_finish): Likewise.
2177 (maybe_reset_location_view): New, from...
2178 (dwarf2out_var_location): ... here. Call it.
2179 * debug.h (dwarf2out_default_as_loc_support): Declare.
2180 (dwarf2out_default_as_locview_support): Declare.
2181 * hooks.c (hook_int_rtx_insn_0): New.
2182 * hooks.h (hook_int_rtx_insn_0): Declare.
2183 * toplev.c (process_options): Take -gas-loc-support and
2184 -gas-locview-support from dwarf2out. Enable
2185 -gvariable-location-views by default only with locview
2186 assembler support. Enable -ginternal-reset-location-views by
2187 default only if the target defines the corresponding hook.
2188 Enable -ginline-points by default if location views are
2189 enabled; force it disabled if statement frontiers are
2191 * tree-inline.c (expand_call_inline): Check option variables.
2192 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2194 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2196 PR tree-optimization/84321
2197 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2198 handling. Also check whether the anti-range contains any values
2199 that satisfy the mask; switch to a VR_RANGE if not.
2201 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2204 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2206 2018-02-13 Martin Jambor <mjambor@suse.cz>
2209 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2210 of call statements, also set location of a load to a temporary.
2212 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2214 * config/rl78/rl78.c (add_vector_labels): New function.
2215 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2216 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2217 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2218 which checks that no arguments are passed.
2219 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2220 * doc/extend.texi: Documentation for the new attribute.
2222 2018-02-13 Andreas Schwab <schwab@suse.de>
2224 * config/riscv/linux.h (CPP_SPEC): Define.
2226 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2229 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2230 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2231 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2232 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2233 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2234 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2235 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2236 temporarily for AES and PCLMUL builtins.
2238 PR tree-optimization/84339
2239 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2240 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2244 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2245 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2246 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2248 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2250 * omp-simd-clone.h: New file.
2251 * omp-simd-clone.c: Include omp-simd-clone.h.
2252 (expand_simd_clones): No longer static.
2253 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2254 cgraph.h and omp-simd-clone.h.
2255 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2256 (vect_recog_widen_shift_pattern): Formatting fix.
2257 (vect_pattern_recog_1): Don't check optab for calls.
2260 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2261 operands[2] into a REG before using gen_lowpart on it.
2263 2018-02-12 Jeff Law <law@redhat.com>
2266 * config/sh/sh.c (find_barrier): Consider a sibling call
2269 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2270 successfully back substituting a reg.
2272 2018-02-12 Richard Biener <rguenther@suse.de>
2274 PR tree-optimization/84037
2275 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2276 parameter, move visited init to caller.
2277 (vect_slp_analyze_operations): Separate cost from validity
2278 check, initialize visited once for all instances.
2279 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2281 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2283 (vect_model_promotion_demotion_cost): Likewise.
2284 (vectorizable_bswap): Guard cost modeling with !slp_node
2285 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2287 (vectorizable_call): Likewise.
2288 (vectorizable_conversion): Likewise.
2289 (vectorizable_assignment): Likewise.
2290 (vectorizable_shift): Likewise.
2291 (vectorizable_operation): Likewise.
2292 (vectorizable_store): Likewise.
2293 (vectorizable_load): Likewise.
2294 (vectorizable_condition): Likewise.
2295 (vectorizable_comparison): Likewise.
2297 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2300 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2301 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2302 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2304 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2307 * params.def (inline-min-speedup): Increase from 8 to 15.
2308 (max-inline-insns-auto): Decrease from 40 to 30.
2309 * ipa-split.c (consider_split): Add some buffer for function to
2310 be considered inlining candidate.
2311 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2314 2018-02-12 Richard Biener <rguenther@suse.de>
2316 PR tree-optimization/84037
2317 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2318 matched stmts if we cannot swap the non-matched ones.
2320 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2322 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2323 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2324 _mm_maskz_scalef_round_ss): New intrinsics.
2325 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2326 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2327 __builtin_ia32_scalefss_round): Remove.
2328 (__builtin_ia32_scalefsd_mask_round,
2329 __builtin_ia32_scalefss_mask_round): New intrinsics.
2330 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2331 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2332 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2333 "<round_constraint>")): Changed to ...
2334 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2335 "<round_scalar_constraint>")): ... this.
2336 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2337 %0, %1, %2<round_op3>}"): Changed to ...
2338 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2339 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2340 %2<round_scalar_mask_op3>}"): ... this.
2341 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2343 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2345 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2346 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2347 (_mm_maskz_sqrt_round_ss): New intrinsics.
2348 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2349 (__builtin_ia32_sqrtsd_mask_round)
2350 (__builtin_ia32_sqrtss_mask_round): New builtins.
2351 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2352 (__builtin_ia32_sqrtss_round): Remove.
2353 (__builtin_ia32_sqrtsd_mask_round)
2354 (__builtin_ia32_sqrtss_mask_round): New builtins.
2355 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2356 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2357 ((match_operand:VF_128 1 "vector_operand"
2358 "xBm,<round_constraint>")): Changed to ...
2359 ((match_operand:VF_128 1 "vector_operand"
2360 "xBm,<round_scalar_constraint>")): ... this.
2361 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2362 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2363 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2364 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2365 %<iptr>1<round_scalar_mask_op3>}): ... this.
2366 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2367 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2369 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2372 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2373 Cast vec_cmpeq result to correct type.
2374 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2375 Cast vec_cmpgt result to correct type.
2377 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2379 * final.c (final_scan_insn_1): Renamed from...
2380 (final_scan_insn): ... this. New wrapper, to recover
2381 seen from the outermost call in recursive ones.
2382 * config/sparc/sparc.c (output_return): Drop seen from call.
2383 (output_sibcall): Likewise.
2384 * config/visium/visium.c (output_branch): Likewise.
2386 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2388 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2391 2018-02-10 Alan Modra <amodra@gmail.com>
2394 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2395 Specify LR as an input.
2397 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2400 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2401 remove_member_access_dummy_vars): New functions.
2402 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2403 lower_omp_1, execute_lower_omp): Use them.
2405 PR rtl-optimization/84308
2406 * shrink-wrap.c (spread_components): Release todo vector.
2408 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2410 PR rtl-optimization/57193
2411 * ira-color.c (struct allocno_color_data): Add member
2412 conflict_allocno_hard_prefs.
2413 (update_conflict_allocno_hard_prefs): New.
2414 (bucket_allocno_compare_func): Add a preference based on
2415 conflict_allocno_hard_prefs.
2416 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2417 (color_allocnos): Remove a dead code. Initiate
2418 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2420 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2423 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2424 constraint from =wa to wa. Avoid a subreg on the output operand,
2425 instead use a pseudo and subreg it in a move.
2426 (p9_xxbrd_<mode>): Changed to ...
2427 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2428 (p9_xxbrd_v2df): New expander.
2429 (p9_xxbrw_<mode>): Changed to ...
2430 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2431 (p9_xxbrw_v4sf): New expander.
2433 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2435 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2437 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2440 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2441 multiply in 32-bit mode.
2442 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2443 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2446 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2448 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2449 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2450 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2451 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2453 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2456 * dwarf2out.c (is_trivial_indirect_ref): New function.
2457 (dwarf2out_late_global_decl): Do not generate a location
2458 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2459 and that are not defined in the current unit.
2461 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2463 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2464 instead of a libcall for UNORDERED.
2466 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2469 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2470 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2472 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2475 * config/s390/s390.c (s390_set_current_function): Invoke
2476 s390_indirect_branch_settings also if fndecl didn't change.
2478 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2480 * config/rs6000/rs6000.md (blockage): Set length to zero.
2482 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2484 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2486 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2489 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2490 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2494 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2495 PARALLEL incoming that failed vt_get_decl_and_offset check.
2498 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2499 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2501 (get_variable_section): For decls in named .bss* sections pass true as
2502 second argument to bss_initializer_p.
2504 2018-02-09 Marek Polacek <polacek@redhat.com>
2505 Jakub Jelinek <jakub@redhat.com>
2508 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2509 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2510 Sync some changes from cxx_fold_indirect_ref.
2512 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2514 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2516 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2517 (BLOCK_INLINE_ENTRY_LABEL): New.
2518 (dwarf2out_var_location): Disregard inline entry markers.
2519 (inline_entry_data): New struct.
2520 (inline_entry_data_hasher): New hashtable type.
2521 (inline_entry_data_hasher::hash): New.
2522 (inline_entry_data_hasher::equal): New.
2523 (inline_entry_data_table): New variable.
2524 (add_high_low_attributes): Add DW_AT_entry_pc and
2525 DW_AT_GNU_entry_view attributes if a pending entry is found
2526 in inline_entry_data_table. Add old entry_pc attribute only
2527 if debug nonbinding markers are disabled.
2528 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2529 markers are enabled.
2530 (block_within_block_p, dwarf2out_inline_entry): New.
2531 (dwarf2out_finish): Check that no entries remained in
2532 inline_entry_data_table.
2533 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2534 (final_scan_insn, notice_source_line): Likewise.
2535 (rest_of_clean_state): Skip inline entry markers.
2536 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2538 * gimple.c (gimple_build_debug_inline_entry): New.
2539 * gimple.h (enum gimple_debug_subcode): Add
2540 GIMPLE_DEBUG_INLINE_ENTRY.
2541 (gimple_build_debug_inline_entry): Declare.
2542 (gimple_debug_inline_entry_p): New.
2543 (gimple_debug_nonbind_marker_p): Adjust.
2544 * insn-notes.def (INLINE_ENTRY): New.
2545 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2546 inline entry marker notes.
2547 (print_insn): Likewise.
2548 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2549 (INSN_DEBUG_MARKER_KIND): Likewise.
2550 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2551 * tree-inline.c (expand_call_inline): Build and insert
2552 debug_inline_entry stmt.
2553 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2554 inline entry blocks early, if nonbind markers are enabled.
2555 (dump_scope_block): Dump fragment info.
2556 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2557 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2558 (gimple_build_debug_inline_entry): New.
2559 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2560 Enable/disable inline entry points too.
2561 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2562 (DEBUG_INSN): Describe inline entry markers.
2564 * common.opt (gvariable-location-views): New.
2565 (gvariable-location-views=incompat5): New.
2566 * config.in: Rebuilt.
2567 * configure: Rebuilt.
2568 * configure.ac: Test assembler for view support.
2569 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2570 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2571 * dwarf2out.c (var_loc_view): New typedef.
2572 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2573 (dwarf2out_locviews_in_attribute): New.
2574 (dwarf2out_locviews_in_loclist): New.
2575 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2576 (enum dw_line_info_opcode): Add LI_adv_address.
2577 (struct dw_line_info_table): Add view.
2578 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2579 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2580 (zero_view_p): New variable.
2581 (ZERO_VIEW_P): New macro.
2582 (output_asm_line_debug_info): New.
2583 (struct var_loc_node): Add view.
2584 (add_AT_view_list, AT_loc_list): New.
2585 (add_var_loc_to_decl): Add view param. Test it against last.
2586 (new_loc_list): Add view params. Record them.
2587 (AT_loc_list_ptr): Handle loc and view lists.
2588 (view_list_to_loc_list_val_node): New.
2589 (print_dw_val): Handle dw_val_class_view_list.
2590 (size_of_die): Likewise.
2591 (value_format): Likewise.
2592 (loc_list_has_views): New.
2593 (gen_llsym): Set vl_symbol too.
2594 (maybe_gen_llsym, skip_loc_list_entry): New.
2595 (dwarf2out_maybe_output_loclist_view_pair): New.
2596 (output_loc_list): Output view list or entries too.
2597 (output_view_list_offset): New.
2598 (output_die): Handle dw_val_class_view_list.
2599 (output_dwarf_version): New.
2600 (output_compilation_unit_header): Use it.
2601 (output_skeleton_debug_sections): Likewise.
2602 (output_rnglists, output_line_info): Likewise.
2603 (output_pubnames, output_aranges): Update version comments.
2604 (output_one_line_info_table): Output view numbers in asm comments.
2605 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2606 Call maybe_gen_llsym.
2607 (loc_list_from_tree_1): Adjust.
2608 (add_AT_location_description): Create view list attribute if
2609 needed, check it's absent otherwise.
2610 (convert_cfa_to_fb_loc_list): Adjust.
2611 (maybe_emit_file): Call output_asm_line_debug_info for test.
2612 (dwarf2out_var_location): Reset views as needed. Precompute
2613 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2614 attribute. Set view.
2615 (new_line_info_table): Reset next view.
2616 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2617 (dwarf2out_source_line): Likewise. Output view resets and labels to
2618 the assembler, or select appropriate line info opcodes.
2619 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2620 (optimize_string_length): Catch it. Adjust.
2621 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2622 dw_val_class_view_list, and remove it if no longer needed.
2623 (hash_loc_list): Hash view numbers.
2624 (loc_list_hasher::equal): Compare them.
2625 (optimize_location_lists): Check whether a view list symbol is
2626 needed, and whether the locview attribute is present, and
2627 whether they match. Remove the locview attribute if no longer
2629 (index_location_lists): Call skip_loc_list_entry for test.
2630 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2631 Use output_dwarf_version.
2632 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2633 (struct dw_val_node): Add val_view_list.
2634 * final.c (SEEN_NEXT_VIEW): New.
2635 (set_next_view_needed): New.
2636 (clear_next_view_needed): New.
2637 (maybe_output_next_view): New.
2638 (final_start_function): Rename to...
2639 (final_start_function_1): ... this. Take pointer to FIRST,
2640 add SEEN parameter. Emit param bindings in the initial view.
2641 (final_start_function): Reintroduce SEEN-less interface.
2642 (final): Rename to...
2643 (final_1): ... this. Take SEEN parameter. Output final pending
2644 next view at the end.
2645 (final): Reintroduce seen-less interface.
2646 (final_scan_insn): Output pending next view before switching
2647 sections or ending a block. Mark the next view as needed when
2648 outputting variable locations. Notify debug backend of section
2649 changes, and of location view changes.
2650 (rest_of_handle_final): Adjust.
2651 * toplev.c (process_options): Autodetect value for debug variable
2652 location views option. Warn on incompat5 without -gdwarf-5.
2653 * doc/invoke.texi (gvariable-location-views): New.
2654 (gvariable-location-views=incompat5): New.
2655 (gno-variable-location-views): New.
2657 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2659 PR tree-optimization/84136
2660 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2661 that the result of find_edge is non-NULL.
2663 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2666 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2667 storing integer register in SImode. Fix cost of 256 and 512
2668 byte aligned SSE register store.
2670 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2672 * config/i386/i386.c (ix86_multiplication_cost): Fix
2673 multiplication cost for TARGET_AVX512DQ.
2675 2018-02-08 Marek Polacek <polacek@redhat.com>
2677 PR tree-optimization/84238
2678 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2681 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2683 PR tree-optimization/84265
2684 * tree-vect-stmts.c (vectorizable_store): Don't treat
2685 VMAT_CONTIGUOUS accesses as grouped.
2686 (vectorizable_load): Likewise.
2688 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2690 PR tree-optimization/81635
2691 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2692 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2693 (test_round_for_mask): New functions.
2694 (wide_int_cc_tests): Call test_round_for_mask.
2695 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2696 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2697 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2698 range returned by get_range_info.
2700 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2703 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2704 * symtab.c: Include builtins.h
2705 (symtab_node::output_to_lto_symbol_table_p): Move here
2706 from lto-streamer-out.c:output_symbol_p.
2707 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2708 (output_symbol_p): Move all logic to symtab.c
2709 (produce_symtab): Update.
2711 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2713 * config/s390/s390-opts.h (enum indirect_branch): Define.
2714 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2715 (s390_indirect_branch_via_thunk)
2716 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2717 (enum s390_indirect_branch_type): Define.
2718 * config/s390/s390.c (struct s390_frame_layout, struct
2719 machine_function): Remove.
2720 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2721 (indirect_branch_table_label_no, indirect_branch_table_name):
2723 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2724 (enum s390_indirect_branch_option): Define.
2725 (s390_return_addr_from_memory): New function.
2726 (s390_handle_string_attribute): New function.
2727 (s390_attribute_table): Add new attribute handler.
2728 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2729 (s390_indirect_branch_via_thunk): New function.
2730 (s390_indirect_branch_via_inline_thunk): New function.
2731 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2732 sibling call optimization for non z10 compiles.
2733 (s390_emit_call): Force indirect branch target to be a single
2734 register. Add r1 clobber for non-z10 compiles.
2735 (s390_emit_epilogue): Emit return jump via return_use expander.
2736 (s390_reorg): Handle JUMP_INSNs as execute targets.
2737 (s390_option_override_internal): Perform validity checks for the
2738 new command line options.
2739 (s390_indirect_branch_attrvalue): New function.
2740 (s390_indirect_branch_settings): New function.
2741 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2742 (s390_output_indirect_thunk_function): New function.
2743 (s390_code_end): Implement target hook.
2744 (s390_case_values_threshold): Implement target hook.
2745 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2747 * config/s390/s390.h (struct s390_frame_layout)
2748 (struct machine_function): Move here from s390.c.
2749 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2750 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2751 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2752 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2753 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2754 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2755 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2756 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2757 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2758 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2759 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2760 (mnemonic attribute): Add values which aren't recognized
2762 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2763 pattern for branch conversion. Fix mnemonic attribute.
2764 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2765 indirect branch via thunk if requested.
2766 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2767 ("*indirect_jump"): Disable for branch conversion using out of
2769 ("indirect_jump_via_thunk<mode>_z10")
2770 ("indirect_jump_via_thunk<mode>")
2771 ("indirect_jump_via_inlinethunk<mode>_z10")
2772 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2773 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2774 ("casesi_jump_via_inlinethunk<mode>_z10")
2775 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2776 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2777 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2778 ("*indirect2_jump"): Disable for branch conversion.
2779 ("casesi_jump"): Turn into expander and expand patterns for branch
2781 ("return_use"): New expander.
2782 ("*return"): Emit return via thunk and rename it to ...
2783 ("*return<mode>"): ... this one.
2784 * config/s390/s390.opt: Add new options and and enum for the
2787 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2789 * lra-constraints.c (match_reload): Unconditionally use
2790 gen_lowpart_SUBREG, rather than selecting between that
2791 and equivalent gen_rtx_SUBREG code.
2793 2018-02-08 Richard Biener <rguenther@suse.de>
2795 PR tree-optimization/84233
2796 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2797 changed flag instead of boguously re-using phi_inserted.
2799 2018-02-08 Martin Jambor <mjambor@suse.cz>
2801 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2802 static local variables.
2804 2018-02-08 Richard Biener <rguenther@suse.de>
2806 PR tree-optimization/84278
2807 * tree-vect-stmts.c (vectorizable_store): When looking for
2808 smaller vector types to perform grouped strided loads/stores
2809 make sure the mode is supported by the target.
2810 (vectorizable_load): Likewise.
2812 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2814 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2815 Increase LDP/STP opportunities by adding adjacent callee-saves.
2817 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2819 PR rtl-optimization/84068
2820 PR rtl-optimization/83459
2821 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2823 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2825 PR tree-optimization/84224
2826 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2827 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2830 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2833 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2834 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2835 count, remove one USE.
2837 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2839 * doc/install.texi (Configuration): Document the
2840 --with-long-double-format={ibm,ieee} PowerPC configuration
2844 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2845 Convert from define_expand to be define_insn_and_split. Rework
2846 float/double/_Float128 conversions to QI/HI/SImode to work with
2847 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2848 conversions to QI/HImode types did a store and then a load to
2849 truncate the value. For conversions to VSX registers, don't split
2850 the insn, instead emit the code directly. Use the code iterator
2851 any_fix to combine signed and unsigned conversions.
2852 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2853 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2854 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2855 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2856 (fix_<mode>di2_hw): Likewise.
2857 (fixuns_<mode>di2_hw): Likewise.
2858 (fix_<mode>si2_hw): Likewise.
2859 (fixuns_<mode>si2_hw): Likewise.
2860 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2861 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2862 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2863 fix<uns>_trunc<SFDF:mode>si2_p8.
2864 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2866 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2867 (fix<uns>_<mode>_mem): Likewise.
2868 (fctiw<u>z_<mode>_mem): Likewise.
2869 (fix<uns>_<mode>_mem): Likewise.
2870 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2871 the register allocator from doing a direct move to the GPRs to do
2872 a store, and instead use the ISA 3.0 store byte/half-word from
2873 vector register instruction. For IEEE 128-bit floating point,
2874 also optimize stores of 32-bit ints.
2875 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2877 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2879 * genextract.c (push_pathstr_operand): New function to support
2881 (walk_rtx): Call push_pathstr_operand.
2882 (print_path): Support [a-zA-Z].
2884 2018-02-07 Richard Biener <rguenther@suse.de>
2886 PR tree-optimization/84037
2887 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2888 (cse_and_gimplify_to_preheader): Declare.
2889 (vect_get_place_in_interleaving_chain): Likewise.
2890 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2892 (_loop_vec_info::~_loop_vec_info): Delete it.
2893 (cse_and_gimplify_to_preheader): New function.
2894 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2895 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2896 (vectorizable_load): Likewise. For grouped stores always base
2897 the IV on the first element.
2898 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2899 condition before gimplifying.
2901 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2903 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2904 *DIV_EXPR and *MOD_EXPR.
2906 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2909 * config/i386/i386.c (ix86_option_override_internal): Mask out
2910 the CF_SET bit when checking -fcf-protection.
2912 2018-02-07 Tom de Vries <tom@codesourcery.com>
2915 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2918 2018-02-07 Richard Biener <rguenther@suse.de>
2920 PR tree-optimization/84204
2921 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2924 PR tree-optimization/84205
2925 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2926 special-case isl_ast_op_zdiv_r.
2928 PR tree-optimization/84223
2929 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2930 Only add conditions from within the region.
2931 (gather_bbs::after_dom_children): Adjust.
2933 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2936 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2937 * config/avr/avr.md: Only post-reload split REG-REG moves if
2938 either register is GENERAL_REG_P.
2940 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2942 PR tree-optimization/84235
2943 * tree-ssa-scopedtables.c
2944 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2945 if the subtraction is performed in floating point type where NaNs are
2946 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2947 build 1. Formatting fix.
2949 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2952 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2953 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2954 and skip it regardless of bb boundaries. Use CALL_P macro,
2955 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2956 unnecessarily, formatting fix.
2958 2018-02-06 Michael Collison <michael.collison@arm.com>
2960 * config/arm/thumb2.md:
2961 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2962 (*thumb_mov_notscc): Ditto.
2964 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2967 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2968 unsigned_fix, not "s".
2970 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2972 * configure.ac (gcc_fn_eh_frame_ro): New function.
2973 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2974 correct .eh_frame permissions.
2975 * configure: Regenerate.
2977 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2979 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2982 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2984 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2985 Display warning message for -mno-speculate-indirect-jumps.
2987 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2989 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2991 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2993 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2995 PR tree-optimization/84225
2996 * tree-eh.c (find_trapping_overflow): Only call
2997 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2999 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3002 * config/i386/i386.c: Reimplement the check of possible options
3003 -mibt/-mshstk conbination. Change error messages.
3004 * doc/invoke.texi: Fix a typo: remove extra '='.
3006 2018-02-06 Marek Polacek <polacek@redhat.com>
3008 PR tree-optimization/84228
3009 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3011 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3014 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3015 emitted arch directives.
3016 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3017 __ARM_FEATURE_COPROC before changing architectures.
3019 2018-02-06 Richard Biener <rguenther@suse.de>
3021 * config/i386/i386.c (print_reg): Fix typo.
3022 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3024 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3026 * configure: Regenerate.
3028 2018-02-05 Martin Sebor <msebor@redhat.com>
3030 PR tree-optimization/83369
3031 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3034 2018-02-05 Martin Liska <mliska@suse.cz>
3036 * doc/invoke.texi: Cherry-pick upstream r323995.
3038 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3040 * ira.c (ira_init_register_move_cost): Adjust comment.
3042 2018-02-05 Martin Liska <mliska@suse.cz>
3044 PR gcov-profile/84137
3045 * doc/gcov.texi: Fix typo in documentation.
3047 2018-02-05 Martin Liska <mliska@suse.cz>
3049 PR gcov-profile/83879
3050 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3051 using dlopen functionality.
3053 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3055 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3056 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3057 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3058 _mm_maskz_range_round_ss): New intrinsics.
3059 (__builtin_ia32_rangesd128_round)
3060 (__builtin_ia32_rangess128_round): Remove.
3061 (__builtin_ia32_rangesd128_mask_round,
3062 __builtin_ia32_rangess128_mask_round): New builtins.
3063 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3064 __builtin_ia32_rangess128_round): Remove.
3065 (__builtin_ia32_rangesd128_mask_round,
3066 __builtin_ia32_rangess128_mask_round): New builtins.
3067 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3068 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3069 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3070 "<round_saeonly_constraint>")): Changed to ...
3071 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3072 "<round_saeonly_scalar_constraint>")): ... this.
3073 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3074 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3075 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3076 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3077 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3079 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3081 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3083 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3084 Remove all values except native, 8540 and 8548.
3086 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3088 * config/i386/i386.c (ix86_output_function_return): Pass
3089 INVALID_REGNUM, instead of -1, as invalid register number to
3090 indirect_thunk_name and output_indirect_thunk.
3092 2018-02-02 Julia Koval <julia.koval@intel.com>
3094 * config.gcc: Add -march=icelake.
3095 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3096 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3097 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3098 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3099 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3100 (processor_target_table): Add icelake.
3101 (ix86_option_override_internal): Handle new PTAs.
3102 (get_builtin_code_for_version): Handle icelake.
3103 (M_INTEL_COREI7_ICELAKE): New.
3104 (fold_builtin_cpu): Handle icelake.
3105 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3106 * doc/invoke.texi: Add -march=icelake.
3108 2018-02-02 Julia Koval <julia.koval@intel.com>
3110 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3111 to wide_int_bitmask.
3112 * wide-int-bitmask.h: New.
3114 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3117 * config/i386/i386.md: Replace Pmode with word_mode in
3118 builtin_setjmp_setup and builtin_longjmp to support x32.
3120 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3124 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3126 (rs6000_supported_cpu_names): New static variable.
3127 (linux_cpu_translation_table): Likewise.
3128 (elf_platform) <cpu>: Define new static variable and use it.
3129 Translate kernel AT_PLATFORM name to canonical name if needed.
3130 Error if platform name is unknown.
3132 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3135 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3137 2018-02-01 Jeff Law <law@redhat.com>
3140 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3141 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3142 the scratch if RELEASE_VIA_POP is false.
3143 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3144 If we have to save a temporary register, decrement SIZE appropriately.
3145 Pass new arguments to release_scratch_register_on_entry.
3146 (ix86_adjust_stack_and_probe): Likewise.
3147 (ix86_emit_probe_stack_range): Pass new arguments to
3148 release_scratch_register_on_entry.
3150 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3152 PR rtl-optimization/84157
3153 * combine.c (change_zero_ext): Use REG_P predicate in
3154 front of HARD_REGISTER_P predicate.
3156 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3158 * config/avr/avr.c (avr_option_override): Move disabling of
3159 -fdelete-null-pointer-checks to...
3160 * common/config/avr/avr-common.c (avr_option_optimization_table):
3163 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3165 PR tree-optimization/81635
3166 * tree-data-ref.c (split_constant_offset_1): For types that
3167 wrap on overflow, try to use range info to prove that wrapping
3170 2018-02-01 Renlin Li <renlin.li@arm.com>
3173 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3175 (aarch64_register_move_cost): Likewise.
3176 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3178 (REG_CLASS_NAMES): Likewise.
3179 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3180 TAILCALL_ADDR_REGS. Remove IP registers.
3181 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3183 2018-02-01 Richard Biener <rguenther@suse.de>
3185 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3186 for specifying RPO order and allow NULL for that.
3187 * domwalk.c (dom_walker::dom_walker): Likewise.
3188 (dom_walker::walk): Handle NULL RPO order.
3189 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3191 (rewrite_update_dom_walker): Likewise.
3192 (mark_def_dom_walker): Likewise.
3194 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3196 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3197 (aarch64_maybe_expand_sve_subreg_move): Declare.
3198 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3199 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3201 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3202 that are semantically a reverse operation.
3203 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3204 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3205 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3207 (aarch64_can_change_mode_class): For big-endian, forbid changes
3208 between two SVE modes if they have different element sizes.
3210 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3212 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3213 the TImode handling for big-endian targets.
3215 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3217 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3218 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3220 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3221 Remove BSWAP handing for big-endian targets and use the form of
3222 LD1RQ appropariate for the mode.
3224 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3226 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3227 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3230 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3233 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3234 check for operands that need to go through aarch64_sve_reload_be.
3236 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3238 PR tree-optimization/81661
3239 PR tree-optimization/84117
3240 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3241 * tree-eh.c: Include gimplify.h.
3242 (find_trapping_overflow, replace_trapping_overflow,
3243 rewrite_to_non_trapping_overflow): New functions.
3244 * tree-vect-loop.c: Include tree-eh.h.
3245 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3246 * tree-data-ref.c: Include tree-eh.h.
3247 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3249 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3251 PR rtl-optimization/84123
3252 * combine.c (change_zero_ext): Check if hard register satisfies
3253 can_change_dest_mode before calling gen_lowpart_SUBREG.
3255 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3258 * ira.c (ira_init_register_move_cost): Remove assert.
3260 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3262 PR rtl-optimization/84071
3263 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3264 * doc/tm.texi: Regenerate.
3266 2018-01-31 Richard Biener <rguenther@suse.de>
3268 PR tree-optimization/84132
3269 * tree-data-ref.c (analyze_miv_subscript): Properly
3270 check whether evolution_function_is_affine_multivariate_p
3271 before calling gcd_of_steps_may_divide_p.
3273 2018-01-31 Julia Koval <julia.koval@intel.com>
3276 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3277 * config/i386/i386.md (rdpid_rex64) New.
3278 (rdpid): Make 32bit only.
3280 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3283 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3284 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3286 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3289 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3291 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3293 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3295 PR rtl-optimization/84071
3296 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3297 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3299 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3301 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3302 (arc_attribute_table): Add 'aux' attribute.
3303 (arc_in_small_data_p): Consider aux like variables.
3304 (arc_is_aux_reg_p): New function.
3305 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3306 (arc_get_aux_arg): New function.
3307 (prepare_move_operands): Handle aux-register access.
3308 (arc_handle_aux_attribute): New function.
3309 * doc/extend.texi (ARC Variable attributes): Add subsection.
3311 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3313 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3314 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3315 (arc_attribute_table): Add 'uncached' attribute.
3316 (arc_print_operand): Print '.di' flag for uncached memory
3318 (arc_in_small_data_p): Do not consider for small data the uncached
3320 (arc_is_uncached_mem_p): New function.
3321 * config/arc/predicates.md (compact_store_memory_operand): Check
3322 for uncached memory accesses.
3323 (nonvol_nonimm_operand): Likewise.
3324 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3326 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3329 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3330 falign-loops=): Add Optimization flag.
3332 2018-01-30 Jeff Law <law@redhat.com>
3335 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3336 INT_REGISTERS_SAVED. Check it prior to calling
3337 get_scratch_register_on_entry.
3338 (ix86_adjust_stack_and_probe): Similarly.
3339 (ix86_emit_probe_stack_range): Similarly.
3340 (ix86_expand_prologue): Corresponding changes.
3342 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3345 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3346 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3348 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3351 * lra-constraints.c (curr_insn_transform): Process AND in the
3354 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3356 PR rtl-optimization/83986
3357 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3358 dependence against last_pending_memory_flush in addition to
3361 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3363 PR tree-optimization/81611
3364 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3367 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3370 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3373 2018-01-30 Richard Biener <rguenther@suse.de>
3374 Jakub Jelinek <jakub@redhat.com>
3376 PR tree-optimization/84111
3377 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3378 inner loops added during recursion, as they don't have up-to-date
3381 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3384 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3385 (can_inline_edge_by_limits_p): ... here.
3386 (can_early_inline_edge_p, check_callers,
3387 update_caller_keys, update_callee_keys, recursive_inlining,
3388 add_new_edges_to_heap, speculation_useful_p,
3389 inline_small_functions,
3390 inline_small_functions, flatten_function,
3391 inline_to_all_callers_1): Update.
3393 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3395 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3398 2018-01-30 Richard Biener <rguenther@suse.de>
3400 PR tree-optimization/83008
3401 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3402 invariant and constant vector uses in stmts when they need
3405 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3408 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3409 * configure: Regenerate.
3411 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3413 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3415 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3416 Use gen_rtx_REG rather than gen_lowpart.
3418 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3420 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3421 rather than 0 when creating partial subregs.
3423 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3425 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3428 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3431 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3432 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3433 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3434 flags. This restores the settings used before the 2017-07-24.
3435 Turning off pre increment/decrement/modify allows IVOPTS to
3436 optimize DF/SF loops where the index is an int.
3438 2018-01-29 Richard Biener <rguenther@suse.de>
3439 Kelvin Nilsen <kelvin@gcc.gnu.org>
3442 * tree-vect-stmts.c (vectorizable_call): Don't call
3443 targetm.vectorize_builtin_md_vectorized_function if callee is
3446 2018-01-22 Carl Love <cel@us.ibm.com>
3448 * doc/extend.tex: Fix typo in second arg in
3449 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3451 2018-01-29 Richard Biener <rguenther@suse.de>
3453 PR tree-optimization/84086
3454 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3455 (flush_ssaname_freelist): When SSA names were released reset
3456 the SCEV hash table.
3458 2018-01-29 Richard Biener <rguenther@suse.de>
3460 PR tree-optimization/84057
3461 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3462 removed paths when removing edges.
3464 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3466 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3467 -mfunction-return=@var{choice}.
3469 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3472 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3474 (layout::print_source_line): Likewise.
3475 (test_get_line_width_without_trailing_whitespace): Add test cases.
3477 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3480 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3483 2018-01-26 Jim Wilson <jimw@sifive.com>
3485 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3487 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3490 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3492 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3493 and CMP + SUB-immediate -> SUBS.
3495 2018-01-26 Martin Sebor <msebor@redhat.com>
3497 PR tree-optimization/83896
3498 * tree-ssa-strlen.c (get_string_len): Rename...
3499 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3500 Avoid assuming length is constant.
3501 (handle_char_store): Use HOST_WIDE_INT for string length.
3503 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3506 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3507 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3509 2018-01-26 Richard Biener <rguenther@suse.de>
3511 PR rtl-optimization/84003
3512 * dse.c (record_store): Only record redundant stores when
3513 the earlier store aliases at least all accesses the later one does.
3515 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3517 PR rtl-optimization/83985
3518 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3519 REG_CFA_RESTORE insns.
3520 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3521 REG_CFA_RESTORE insns here.
3524 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3525 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3527 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3529 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3530 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3531 (arc_init): Likewise.
3532 (arc_override_options): Likewise.
3533 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3535 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3537 * config/arc/arc.h (TARGET_DBNZ): Define.
3538 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3539 properly set the tune attribute.
3540 (dbnz): Use TARGET_DBNZ guard.
3541 * config/arc/arc.opt (mtune): Add core3 option.
3543 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3545 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3546 recognize new pic like addresses.
3547 (arc_delegitimize_address): Clean up.
3549 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3551 * config/arc/arc-arches.def: Option mrf16 valid for all
3553 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3554 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3555 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3556 * config/arc/arc-tables.opt: Regenerate.
3557 * config/arc/arc.c (arc_conditional_register_usage): Handle
3558 reduced register file case.
3559 (arc_file_start): Set must have build attributes.
3560 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3562 * config/arc/arc.opt (mrf16): Add new option.
3563 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3564 * config/arc/genmultilib.awk: Handle new mrf16 option.
3565 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3566 * config/arc/t-multilib: Regenerate.
3567 * doc/invoke.texi (ARC Options): Document mrf16 option.
3569 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3571 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3572 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3573 (arc_attribute_table): Add 'secure_call' attribute.
3574 (arc_print_operand): Print secure call operand.
3575 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3577 (arc_is_secure_call_p): New function. * config/arc/arc.md
3578 (call_i): Add support for sjli instruction.
3579 (call_value_i): Likewise.
3580 * config/arc/constraints.md (Csc): New constraint.
3582 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3583 John Eric Martin <John.Martin@emmicro-us.com>
3585 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3586 * config/arc/arc.c (_arc_jli_section): New struct.
3587 (arc_jli_section): New type.
3588 (rc_jli_sections): New static variable.
3589 (arc_handle_jli_attribute): New function.
3590 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3591 (arc_file_end): New function.
3592 (TARGET_ASM_FILE_END): Define.
3593 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3594 (arc_add_jli_section): New function.
3595 (jli_call_scan): Likewise.
3596 (arc_reorg): Call jli_call_scan.
3597 (arc_output_addsi): Remove 'S' from printing asm operand.
3598 (arc_is_jli_call_p): New function.
3599 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3601 (movhi_insn): Likewise.
3602 (movsi_insn): Likewise.
3603 (movsi_set_cc_insn): Likewise.
3604 (loadqi_update): Likewise.
3605 (load_zeroextendqisi_update): Likewise.
3606 (load_signextendqisi_update): Likewise.
3607 (loadhi_update): Likewise.
3608 (load_zeroextendhisi_update): Likewise.
3609 (load_signextendhisi_update): Likewise.
3610 (loadsi_update): Likewise.
3611 (loadsf_update): Likewise.
3612 (movsicc_insn): Likewise.
3613 (bset_insn): Likewise.
3614 (bxor_insn): Likewise.
3615 (bclr_insn): Likewise.
3616 (bmsk_insn): Likewise.
3617 (bicsi3_insn): Likewise.
3618 (cmpsi_cc_c_insn): Likewise.
3619 (movsi_ne): Likewise.
3620 (movsi_cond_exec): Likewise.
3621 (clrsbsi2): Likewise.
3632 (call_i): Remove 'S' asm letter, add jli instruction.
3633 (call_value_i): Likewise.
3634 * config/arc/arc.op (mjli-always): New option.
3635 * config/arc/constraints.md (Cji): New constraint.
3636 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3638 (subsf3_fpx): Likewise.
3639 (mulsf3_fpx): Likewise.
3640 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3642 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3644 * doc/invoke.texi (ARC): Document mjli-always option.
3646 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3648 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3649 avoid addition with 0 and use incw and decw where possible.
3651 2018-01-26 Richard Biener <rguenther@suse.de>
3653 PR tree-optimization/81082
3654 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3655 association if it requires casting to unsigned.
3656 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3657 from fold_plusminus_mult_expr to catch important cases late when
3658 range info is available.
3660 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3662 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3663 * configure.ac (hidden_linkonce): New test.
3664 * configure: Regenerate.
3665 * config.in: Regenerate.
3667 2018-01-26 Julia Koval <julia.koval@intel.com>
3669 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3670 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3671 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3672 _mm_mask_bitshuffle_epi64_mask): Fix type.
3673 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3674 USI_FTYPE_V4DI_V4DI_USI): Remove.
3675 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3676 __builtin_ia32_vpshufbitqmb256_mask,
3677 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3678 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3679 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3681 2018-01-26 Alan Modra <amodra@gmail.com>
3684 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3685 UNSPEC_VBPERMQ. Sort other unspecs.
3687 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3689 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3691 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3694 * predict.c (drop_profile): Do not push/pop cfun; update also
3696 (handle_missing_profiles): Fix logic looking for zero profiles.
3698 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3701 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3702 on functions with #pragma omp declare simd or functions with simd
3704 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3705 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3706 Remove trailing \n from warning_at calls.
3708 2018-01-25 Tom de Vries <tom@codesourcery.com>
3711 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3712 for neutered workers.
3714 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3717 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3718 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3720 2018-01-24 Jeff Law <law@redhat.com>
3723 * i386.c (get_probe_interval): Move to earlier point.
3724 (ix86_compute_frame_layout): If -fstack-clash-protection and
3725 the frame is larger than the probe interval, then use pushes
3726 to save registers rather than reg->mem moves.
3727 (ix86_expand_prologue): Remove conditional for int_registers_saved
3730 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3733 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3734 min/max for never referenced object.
3736 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3739 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3741 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3742 attributes from DECL_ATTRIBUTES (decl) without affecting
3743 DECL_ATTRIBUTES (current_function_decl).
3744 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3745 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3747 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3749 PR tree-optimization/83979
3750 * fold-const.c (fold_comparison): Use constant_boolean_node
3751 instead of boolean_{true,false}_node.
3753 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3755 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3758 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3760 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3761 Simplify the clause that sets the length attribute.
3762 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3763 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3764 clause that sets the length attribute.
3765 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3767 2018-01-24 Tom de Vries <tom@codesourcery.com>
3770 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3771 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3772 Add strict parameter.
3773 (prevent_branch_around_nothing): Insert dummy insn between branch to
3774 label and label with no ptx insn inbetween.
3775 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3777 2018-01-24 Tom de Vries <tom@codesourcery.com>
3780 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3781 for neutered threads in warp.
3782 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3784 2018-01-24 Richard Biener <rguenther@suse.de>
3786 PR tree-optimization/83176
3787 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3790 2018-01-24 Richard Biener <rguenther@suse.de>
3792 PR tree-optimization/82819
3793 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3794 code generating pluses that are no-ops in the target precision.
3796 2018-01-24 Richard Biener <rguenther@suse.de>
3799 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3801 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3803 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3804 to merge probabilities.
3805 * predict.c (probably_never_executed): Also mark as cold functions
3806 with global 0 profile and guessed local profile.
3807 * profile-count.c (profile_probability::combine_with_count): New
3809 * profile-count.h (profile_probability::operator*,
3810 profile_probability::operator*=, profile_probability::operator/,
3811 profile_probability::operator/=): Reduce precision to adjusted
3812 and set value to guessed on contradictory divisions.
3813 (profile_probability::combine_with_freq): Remove.
3814 (profile_probability::combine_wiht_count): Declare.
3815 (profile_count::force_nonzero):: Set to adjusted.
3816 (profile_count::probability_in):: Set quality to adjusted.
3817 * tree-ssa-tail-merge.c (replace_block_by): Use
3820 2018-01-23 Andrew Waterman <andrew@sifive.com>
3821 Jim Wilson <jimw@sifive.com>
3823 * config/riscv/riscv.c (riscv_stack_boundary): New.
3824 (riscv_option_override): Set riscv_stack_boundary. Handle
3825 riscv_preferred_stack_boundary_arg.
3826 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3827 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3828 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3829 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3830 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3831 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3833 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3836 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3837 of struct ix86_frame.
3838 (ix86_expand_epilogue): Likewise. Add a local variable for
3839 the reg_save_offset field in struct ix86_frame.
3841 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3843 PR tree-optimization/82604
3844 * tree-loop-distribution.c (enum partition_kind): New enum item
3845 PKIND_PARTIAL_MEMSET.
3846 (partition_builtin_p): Support above new enum item.
3847 (generate_code_for_partition): Ditto.
3848 (compute_access_range): Differentiate cases that equality can be
3849 proven at all loops, the innermost loops or no loops.
3850 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3851 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3852 (finalize_partitions, distribute_loop): Don't fuse partition of
3853 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3854 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3857 2018-01-23 Martin Liska <mliska@suse.cz>
3859 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3860 order to ignore the predictor.
3861 (PRED_POLYMORPHIC_CALL): Likewise.
3862 (PRED_RECURSIVE_CALL): Likewise.
3864 2018-01-23 Martin Liska <mliska@suse.cz>
3866 * tree-profile.c (tree_profiling): Print function header to
3867 aware reader which function we are working on.
3868 * value-prof.c (gimple_find_values_to_profile): Do not print
3869 not interesting value histograms.
3871 2018-01-23 Martin Liska <mliska@suse.cz>
3873 * profile-count.h (enum profile_quality): Add
3874 profile_uninitialized as the first value. Do not number values
3875 as they are zero based.
3876 (profile_count::verify): Update sanity check.
3877 (profile_probability::verify): Likewise.
3879 2018-01-23 Nathan Sidwell <nathan@acm.org>
3881 * doc/invoke.texi (ffor-scope): Deprecate.
3883 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3885 PR tree-optimization/83510
3886 * domwalk.c (set_all_edges_as_executable): New function.
3887 (dom_walker::dom_walker): Convert bool param
3888 "skip_unreachable_blocks" to enum reachability. Move setup of
3889 edge flags to set_all_edges_as_executable and only do it when
3890 reachability is REACHABLE_BLOCKS.
3891 * domwalk.h (enum dom_walker::reachability): New enum.
3892 (dom_walker::dom_walker): Convert bool param
3893 "skip_unreachable_blocks" to enum reachability.
3894 (set_all_edges_as_executable): New decl.
3895 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3896 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3898 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3899 but converting true to REACHABLE_BLOCKS.
3900 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3902 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3903 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3904 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3906 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3907 if check_all_array_refs will be called.
3909 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3911 * tree.c (selftest::test_location_wrappers): Add more test
3914 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3916 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3917 (selftest::test_bit_in_range): Likewise.
3919 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3922 * doc/sourcebuild.texi (vect_float): Say that the selector
3923 only describes the situation when -funsafe-math-optimizations is on.
3924 (vect_float_strict): Document.
3926 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3928 PR tree-optimization/83965
3929 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3930 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3931 instead of checking only for a reduction.
3932 (vect_recog_widen_sum_pattern): Likewise.
3934 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3936 * predict.c (probably_never_executed): Only use precise profile info.
3937 (compute_function_frequency): Skip after inlining hack since we now
3938 have quality checking.
3940 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3942 * profile-count.h (profile_probability::very_unlikely,
3943 profile_probability::unlikely, profile_probability::even): Set
3944 precision to guessed.
3946 2018-01-23 Richard Biener <rguenther@suse.de>
3948 PR tree-optimization/83963
3949 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3950 Properly terminate dominator walk when crossing the exit edge not
3951 when visiting its source block.
3953 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3956 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3957 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3959 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3961 PR tree-optimization/83957
3962 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3963 semicolon after for body surrounded by braces.
3965 PR tree-optimization/83081
3966 * profile-count.h (profile_probability::split): New method.
3967 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3968 Use profile_probability::split.
3969 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3970 when splitting a single conditional jump into 2.
3972 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3974 PR tree-optimization/69452
3975 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3978 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3980 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3981 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3982 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3986 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3987 * config/rl78/rl78.md (movdi): New define_expand.
3988 * config/rl78/rl78.c (rl78_split_movdi): New function.
3990 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3993 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3995 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3996 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3997 128-bit to produce an UNSPEC move to get the double word with the
3998 signbit and then a shift directly to do signbit.
3999 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4000 implementation with a new version that just does either a direct
4001 move or a regular move. Move memory interface to separate insns.
4002 Move insns so they are next to the expander.
4003 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4004 with signbit move. Split big and little endian case.
4005 (signbit<mode>2_dm_mem_le): Likewise.
4006 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4007 (signbit<mode>2_dm2): Likewise.
4009 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4011 * config/rl78/rl78.md (anddi3): New define_expand.
4013 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4015 * config/rl78/rl78.md (umindi3): New define_expand.
4017 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4019 * config/rl78/rl78.md (smindi3): New define_expand.
4021 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4023 * config/rl78/rl78.md (smaxdi3): New define_expand.
4025 2018-01-22 Carl Love <cel@us.ibm.com>
4027 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4028 LVX_V1TI): Add macro expansion.
4029 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4030 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4031 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4032 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4033 Change check to determine if the instruction is a byte reversing
4034 entry. Fix typo in comment.
4035 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4036 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4037 Add def_builtin calls for new builtins.
4038 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4039 Add define_insn expansion.
4041 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4043 * config/rl78/rl78.md (umaxdi3): New define_expand.
4045 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4047 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4048 for non-QImode registers.
4050 2018-01-22 Richard Biener <rguenther@suse.de>
4052 PR tree-optimization/83963
4053 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4054 including the loop exit block.
4055 (scop_detection::merge_sese): Likewise.
4056 (scop_detection::add_scop): Do it here instead.
4058 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4060 * doc/sourcebuild.texi (arm_softfloat): Document.
4062 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4065 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4066 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4067 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4069 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4070 David Edelsohn <dje.gcc@gmail.com>
4073 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4074 Change "crset eq" to "crset 2".
4075 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4076 (*call_indirect_aix<mode>_nospec): Likewise.
4077 (*call_value_indirect_aix<mode>_nospec): Likewise.
4078 (*call_indirect_elfv2<mode>_nospec): Likewise.
4079 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4080 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4081 change assembly output from . to $.
4082 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4083 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4084 (*tablejump<mode>_internal1_nospec): Likewise.
4086 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4089 * config/sh/sh_optimize_sett_clrt.cc:
4090 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4092 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4094 PR tree-optimization/83940
4095 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4096 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4097 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4098 use it to pass back the definition type.
4099 (vect_check_store_rhs): Likewise rhs_dt_out.
4100 (vect_build_gather_load_calls): Add a mask_dt argument and use
4101 it instead of a call to vect_is_simple_use.
4102 (vectorizable_store): Update calls to vect_check_load_store_mask
4103 and vect_check_store_rhs. Use the dt returned by the latter instead
4104 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4105 instead of calls to vect_is_simple_use. Pass the scalar rather
4106 than the vector operand to vect_is_simple_use when handling
4107 second and subsequent copies of an rhs value.
4108 (vectorizable_load): Update calls to vect_check_load_store_mask
4109 and vect_build_gather_load_calls. Use the cached mask_dt and
4110 gs_info.offset_dt instead of calls to vect_is_simple_use.
4112 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4115 * tree-emutls.c: Include gimplify.h.
4116 (lower_emutls_2): New function.
4117 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4118 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4119 it before further processing.
4122 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4123 UINTVAL (trueop1) instead of INTVAL (op1).
4125 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4129 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4130 INCOMING_FRAME_SP_OFFSET if not defined.
4131 (scan_trace): Add ENTRY argument. If true and
4132 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4133 emit a note to adjust the CFA offset.
4134 (create_cfi_notes): Adjust scan_trace callers.
4135 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4136 INCOMING_FRAME_SP_OFFSET in the CIE.
4137 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4138 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4140 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4141 * doc/tm.texi: Regenerated.
4143 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4145 PR rtl-optimization/83147
4146 * lra-constraints.c (remove_inheritance_pseudos): Use
4147 lra_substitute_pseudo_within_insn.
4149 2018-01-19 Tom de Vries <tom@codesourcery.com>
4150 Cesar Philippidis <cesar@codesourcery.com>
4153 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4155 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4158 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4159 spaces for function labels.
4161 2018-01-19 Martin Liska <mliska@suse.cz>
4163 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4164 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4165 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4166 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4167 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4168 (PRED_CONST_RETURN): Change from 69 to 65.
4169 (PRED_NULL_RETURN): Change from 91 to 71.
4170 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4171 (PRED_LOOP_GUARD): Change from 66 to 73.
4173 2018-01-19 Martin Liska <mliska@suse.cz>
4175 * predict.c (predict_insn_def): Add new assert.
4176 (struct branch_predictor): Change type to signed integer.
4177 (test_prediction_value_range): Amend test to cover
4179 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4180 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4181 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4182 (PRED_LOOP_IV_COMPARE): Likewise.
4183 * predict.h (PROB_UNINITIALIZED): Define new constant.
4185 2018-01-19 Martin Liska <mliska@suse.cz>
4187 * predict.c (dump_prediction): Add new format for
4188 analyze_brprob.py script which is enabled with -details
4190 * profile-count.h (precise_p): New function.
4192 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4194 PR tree-optimization/83922
4195 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4196 there are no statements that need masking.
4197 (vect_active_double_reduction_p): New function.
4198 (vect_analyze_loop_operations): Use it when handling phis that
4199 are not in the loop header.
4201 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4203 PR tree-optimization/83914
4204 * tree-vect-loop.c (vectorizable_induction): Don't convert
4205 init_expr or apply the peeling adjustment for inductions
4206 that are nested within the vectorized loop.
4208 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4210 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4213 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4217 * function.h (gimplify_parameters): Add gimple_seq * argument.
4218 * function.c: Include gimple.h and options.h.
4219 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4220 for the added local temporaries if needed.
4221 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4222 if there are any parameter cleanups, wrap whole body into a
4223 try/finally with the cleanups.
4225 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4228 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4229 Use GET_MODE_CLASS for scalar floating point.
4231 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4235 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4236 Fix call of call_cgraph_insertion_hooks.
4238 2018-01-18 Martin Sebor <msebor@redhat.com>
4240 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4242 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4245 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4248 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4251 * common.opt: (-ffile-prefix-map): New option.
4252 * opts.c (common_handle_option): Defer it.
4253 * opts-global.c (handle_common_deferred_options): Handle it.
4254 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4255 * file-prefix-map.h: New file.
4256 (remap_debug_filename, add_debug_prefix_map): ...here.
4257 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4258 * final.c (debug_prefix_map, add_debug_prefix_map
4259 remap_debug_filename): Move to...
4260 * file-prefix-map.c: New file.
4261 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4262 generalize, get rid of alloca(), use strrchr() instead of strchr().
4263 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4264 Implement in terms of add_prefix_map().
4265 (remap_macro_filename, remap_debug_filename): Implement in term of
4267 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4268 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4269 * dbxout.c: Include file-prefix-map.h.
4270 * varasm.c: Likewise.
4271 * vmsdbgout.c: Likewise.
4272 * xcoffout.c: Likewise.
4273 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4274 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4275 * doc/invoke.texi (-ffile-prefix-map): Document.
4276 (-fdebug-prefix-map): Update description.
4278 2018-01-18 Martin Liska <mliska@suse.cz>
4280 * config/i386/i386.c (indirect_thunk_name): Document that also
4282 (output_indirect_thunk): Document why both instructions
4283 (pause and lfence) are generated.
4285 2018-01-18 Richard Biener <rguenther@suse.de>
4287 PR tree-optimization/83887
4288 * graphite-scop-detection.c
4289 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4290 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4291 (scop_detection::merge_sese): Re-implement with a flood-fill
4292 algorithm that properly finds a SESE region if it exists.
4294 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4297 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4298 pointer_diff optimizations use view_convert instead of convert.
4300 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4302 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4303 Generate different code for -mno-speculate-indirect-jumps.
4304 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4305 (*call_indirect_aix<mode>): Disable for
4306 -mno-speculate-indirect-jumps.
4307 (*call_indirect_aix<mode>_nospec): New define_insn.
4308 (*call_value_indirect_aix<mode>): Disable for
4309 -mno-speculate-indirect-jumps.
4310 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4311 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4312 -mno-speculate-indirect-jumps.
4313 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4315 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4317 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4318 long double type, set the flags for noting the default long double
4319 type, even if we don't pass or return a long double type.
4321 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4324 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4327 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4329 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4330 support for merge[hl].
4331 (fold_mergehl_helper): New helper function.
4332 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4333 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4334 (altivec_vmrglw_direct): Add xxmrglw insn.
4336 2018-01-17 Andrew Waterman <andrew@sifive.com>
4338 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4339 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4341 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4344 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4345 call the lto_location_cache before reading the
4346 DECL_SOURCE_LOCATION of the types.
4348 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4349 Richard Sandiford <richard.sandiford@linaro.org>
4351 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4352 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4353 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4354 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4355 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4357 * config/aarch64/constraints.md (aarch64_movti_operand):
4359 * config/aarch64/predicates.md (Uti): Add new constraint.
4361 2018-01-17 Carl Love <cel@us.ibm.com>
4363 * config/rs6000/vsx.md (define_expand xl_len_r,
4364 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4365 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4367 (define_expand, define_insn): Move the shift left from the
4368 define_insn to the define_expand for lxvl and stxvl instructions.
4369 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4370 and XL_LEN_R definitions to PURE.
4372 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4374 * config/i386/i386.c (indirect_thunk_name): Declare regno
4375 as unsigned int. Compare regno with INVALID_REGNUM.
4376 (output_indirect_thunk): Ditto.
4377 (output_indirect_thunk_function): Ditto.
4378 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4379 in the call to output_indirect_thunk_function.
4381 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4384 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4385 rather than the size of inner_type to determine the stack slot size
4386 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4388 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4391 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4394 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4396 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4397 endian Linux systems to optionally enable multilibs for selecting
4398 the long double type if the user configured an explicit type.
4399 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4400 have no long double multilibs if not defined.
4401 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4402 warn if the user used -mabi={ieee,ibm}longdouble and we built
4403 multilibs for long double.
4404 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4405 appropriate multilib option.
4406 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4408 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4409 for building long double multilibs.
4410 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4412 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4414 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4417 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4419 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4422 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4425 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4428 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4430 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4431 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4433 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4435 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4436 different rtl trees depending on TARGET_64BIT.
4437 (rs6000_gen_lvx): Likewise.
4439 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4441 * config/visium/visium.md (nop): Tweak comment.
4442 (hazard_nop): Likewise.
4444 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4446 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4447 -mspeculate-indirect-jumps.
4448 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4449 for -mno-speculate-indirect-jumps.
4450 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4451 (*call_value_indirect_elfv2<mode>): Disable for
4452 -mno-speculate-indirect-jumps.
4453 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4454 (indirect_jump): Emit different RTL for
4455 -mno-speculate-indirect-jumps.
4456 (*indirect_jump<mode>): Disable for
4457 -mno-speculate-indirect-jumps.
4458 (*indirect_jump<mode>_nospec): New define_insn.
4459 (tablejump): Emit different RTL for
4460 -mno-speculate-indirect-jumps.
4461 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4462 (tablejumpsi_nospec): New define_expand.
4463 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4464 (tablejumpdi_nospec): New define_expand.
4465 (*tablejump<mode>_internal1): Disable for
4466 -mno-speculate-indirect-jumps.
4467 (*tablejump<mode>_internal1_nospec): New define_insn.
4468 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4471 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4473 * caller-save.c (insert_save): Drop unnecessary parameter. All
4476 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4477 Richard Biener <rguenth@suse.de>
4480 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4481 return early, inline manually is_gimple_sizepos. Make sure if we
4482 call gimplify_expr we don't end up with a gimple constant.
4483 * tree.c (variably_modified_type_p): Don't return true for
4484 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4485 * gimplify.h (is_gimple_sizepos): Remove.
4487 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4489 PR tree-optimization/83857
4490 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4491 vectorizable_live_operation for pure SLP statements.
4492 (vectorizable_live_operation): Handle PHIs.
4494 2018-01-16 Richard Biener <rguenther@suse.de>
4496 PR tree-optimization/83867
4497 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4498 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4500 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4503 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4504 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4505 If off is not INTEGER_CST, issue a may not be aligned warning
4506 rather than isn't aligned. Use isn%'t rather than isn't.
4507 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4509 <case MULT_EXPR>: Improve the case when bottom and one of the
4510 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4511 operand, in that case check if the other operand is multiple of
4512 bottom divided by the INTEGER_CST operand.
4514 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4517 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4518 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4519 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4520 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4521 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4522 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4523 (pa_function_arg_size): New function.
4525 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4527 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4528 in a separate statement.
4530 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4532 PR tree-optimization/83847
4533 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4534 group gathers and scatters.
4536 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4538 PR rtl-optimization/86620
4539 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4541 PR rtl-optimization/83213
4542 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4543 to last if both are JUMP_INSNs.
4545 PR tree-optimization/83843
4546 * gimple-ssa-store-merging.c
4547 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4548 store_immediate_info for bswap/nop orig_stores.
4550 2018-01-15 Andrew Waterman <andrew@sifive.com>
4552 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4554 <UDIV>: Increase cost if !TARGET_DIV.
4556 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4558 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4559 (define_attr "cr_logical_3op"): New.
4560 (cceq_ior_compare): Adjust.
4561 (cceq_ior_compare_complement): Adjust.
4562 (*cceq_rev_compare): Adjust.
4563 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4564 (is_cracked_insn): Adjust.
4565 (insn_must_be_first_in_group): Adjust.
4566 * config/rs6000/40x.md: Adjust.
4567 * config/rs6000/440.md: Adjust.
4568 * config/rs6000/476.md: Adjust.
4569 * config/rs6000/601.md: Adjust.
4570 * config/rs6000/603.md: Adjust.
4571 * config/rs6000/6xx.md: Adjust.
4572 * config/rs6000/7450.md: Adjust.
4573 * config/rs6000/7xx.md: Adjust.
4574 * config/rs6000/8540.md: Adjust.
4575 * config/rs6000/cell.md: Adjust.
4576 * config/rs6000/e300c2c3.md: Adjust.
4577 * config/rs6000/e500mc.md: Adjust.
4578 * config/rs6000/e500mc64.md: Adjust.
4579 * config/rs6000/e5500.md: Adjust.
4580 * config/rs6000/e6500.md: Adjust.
4581 * config/rs6000/mpc.md: Adjust.
4582 * config/rs6000/power4.md: Adjust.
4583 * config/rs6000/power5.md: Adjust.
4584 * config/rs6000/power6.md: Adjust.
4585 * config/rs6000/power7.md: Adjust.
4586 * config/rs6000/power8.md: Adjust.
4587 * config/rs6000/power9.md: Adjust.
4588 * config/rs6000/rs64.md: Adjust.
4589 * config/rs6000/titan.md: Adjust.
4591 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4593 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4594 ix86_indirect_branch_register logic.
4596 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4598 * config/i386/constraints.md (Bs): Update
4599 ix86_indirect_branch_register check. Don't check
4600 ix86_indirect_branch_register with GOT_memory_operand.
4602 * config/i386/predicates.md (GOT_memory_operand): Don't check
4603 ix86_indirect_branch_register here.
4604 (GOT32_symbol_operand): Likewise.
4606 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4608 * config/i386/predicates.md (constant_call_address_operand):
4609 Rewrite ix86_indirect_branch_register logic.
4610 (sibcall_insn_operand): Likewise.
4612 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4614 * config/i386/constraints.md (Bs): Replace
4615 ix86_indirect_branch_thunk_register with
4616 ix86_indirect_branch_register.
4618 * config/i386/i386.md (indirect_jump): Likewise.
4619 (tablejump): Likewise.
4620 (*sibcall_memory): Likewise.
4621 (*sibcall_value_memory): Likewise.
4622 Peepholes of indirect call and jump via memory: Likewise.
4623 * config/i386/i386.opt: Likewise.
4624 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4625 (GOT_memory_operand): Likewise.
4626 (call_insn_operand): Likewise.
4627 (sibcall_insn_operand): Likewise.
4628 (GOT32_symbol_operand): Likewise.
4630 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4633 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4634 type rather than type addr's type points to.
4635 (expand_omp_atomic_mutex): Likewise.
4636 (expand_omp_atomic): Likewise.
4638 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4641 * config/i386/i386.c (output_indirect_thunk_function): Use
4642 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4643 for __x86_return_thunk.
4645 2018-01-15 Richard Biener <rguenther@suse.de>
4648 * expmed.c (extract_bit_field_1): Fix typo.
4650 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4653 * config/arm/iterators.md (VF): New mode iterator.
4654 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4655 Remove integer-related logic from pattern.
4656 (neon_vabd<mode>_3): Likewise.
4658 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4661 * common.opt (fstrict-overflow): No longer an alias.
4662 (fwrapv-pointer): New option.
4663 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4664 also for pointer types based on flag_wrapv_pointer.
4665 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4666 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4667 opts->x_flag_wrapv got set.
4668 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4669 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4670 POINTER_TYPE_OVERFLOW_UNDEFINED.
4671 * match.pd: Likewise in address comparison pattern.
4672 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4674 2018-01-15 Richard Biener <rguenther@suse.de>
4677 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4678 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4679 Reset type names to their identifier if their TYPE_DECL doesn't
4680 have linkage (and thus is used for ODR and devirt).
4681 (save_debug_info_for_decl): Remove.
4682 (save_debug_info_for_type): Likewise.
4683 (add_tree_to_fld_list): Adjust.
4684 * tree-pretty-print.c (dump_generic_node): Make dumping of
4685 type names more robust.
4687 2018-01-15 Richard Biener <rguenther@suse.de>
4689 * BASE-VER: Bump to 8.0.1.
4691 2018-01-14 Martin Sebor <msebor@redhat.com>
4694 * builtins.c (check_access): Avoid warning when the no-warning bit
4697 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4699 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4700 * ira-color (allocno_hard_regs_compare): Likewise.
4702 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4705 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4706 Use .pushsection/.popsection.
4708 2018-01-14 Martin Sebor <msebor@redhat.com>
4711 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4713 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4715 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4716 entry from extra_headers.
4717 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4718 extra_headers, make the list bitwise identical to the i?86-*-* one.
4720 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4722 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4723 -mcmodel=large with -mindirect-branch=thunk,
4724 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4725 -mfunction-return=thunk-extern.
4726 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4727 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4728 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4730 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4732 * config/i386/i386.c (print_reg): Print the name of the full
4733 integer register without '%'.
4734 (ix86_print_operand): Handle 'V'.
4735 * doc/extend.texi: Document 'V' modifier.
4737 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4739 * config/i386/constraints.md (Bs): Disallow memory operand for
4740 -mindirect-branch-register.
4742 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4743 (GOT_memory_operand): Likewise.
4744 (call_insn_operand): Likewise.
4745 (sibcall_insn_operand): Likewise.
4746 (GOT32_symbol_operand): Likewise.
4747 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4748 for -mindirect-branch-register.
4749 (tablejump): Likewise.
4750 (*sibcall_memory): Likewise.
4751 (*sibcall_value_memory): Likewise.
4752 Disallow peepholes of indirect call and jump via memory for
4753 -mindirect-branch-register.
4754 (*call_pop): Replace m with Bw.
4755 (*call_value_pop): Likewise.
4756 (*sibcall_pop_memory): Replace m with Bs.
4757 * config/i386/i386.opt (mindirect-branch-register): New option.
4758 * doc/invoke.texi: Document -mindirect-branch-register option.
4760 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4762 * config/i386/i386-protos.h (ix86_output_function_return): New.
4763 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4764 set function_return_type.
4765 (indirect_thunk_name): Add ret_p to indicate thunk for function
4767 (output_indirect_thunk_function): Pass false to
4768 indirect_thunk_name.
4769 (ix86_output_indirect_branch_via_reg): Likewise.
4770 (ix86_output_indirect_branch_via_push): Likewise.
4771 (output_indirect_thunk_function): Create alias for function
4772 return thunk if regno < 0.
4773 (ix86_output_function_return): New function.
4774 (ix86_handle_fndecl_attribute): Handle function_return.
4775 (ix86_attribute_table): Add function_return.
4776 * config/i386/i386.h (machine_function): Add
4777 function_return_type.
4778 * config/i386/i386.md (simple_return_internal): Use
4779 ix86_output_function_return.
4780 (simple_return_internal_long): Likewise.
4781 * config/i386/i386.opt (mfunction-return=): New option.
4782 (indirect_branch): Mention -mfunction-return=.
4783 * doc/extend.texi: Document function_return function attribute.
4784 * doc/invoke.texi: Document -mfunction-return= option.
4786 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4788 * config/i386/i386-opts.h (indirect_branch): New.
4789 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4790 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4791 with local indirect jump when converting indirect call and jump.
4792 (ix86_set_indirect_branch_type): New.
4793 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4794 (indirectlabelno): New.
4795 (indirect_thunk_needed): Likewise.
4796 (indirect_thunk_bnd_needed): Likewise.
4797 (indirect_thunks_used): Likewise.
4798 (indirect_thunks_bnd_used): Likewise.
4799 (INDIRECT_LABEL): Likewise.
4800 (indirect_thunk_name): Likewise.
4801 (output_indirect_thunk): Likewise.
4802 (output_indirect_thunk_function): Likewise.
4803 (ix86_output_indirect_branch_via_reg): Likewise.
4804 (ix86_output_indirect_branch_via_push): Likewise.
4805 (ix86_output_indirect_branch): Likewise.
4806 (ix86_output_indirect_jmp): Likewise.
4807 (ix86_code_end): Call output_indirect_thunk_function if needed.
4808 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4810 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4811 (ix86_attribute_table): Add indirect_branch.
4812 * config/i386/i386.h (machine_function): Add indirect_branch_type
4813 and has_local_indirect_jump.
4814 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4816 (tablejump): Likewise.
4817 (*indirect_jump): Use ix86_output_indirect_jmp.
4818 (*tablejump_1): Likewise.
4819 (simple_return_indirect_internal): Likewise.
4820 * config/i386/i386.opt (mindirect-branch=): New option.
4821 (indirect_branch): New.
4824 (thunk-inline): Likewise.
4825 (thunk-extern): Likewise.
4826 * doc/extend.texi: Document indirect_branch function attribute.
4827 * doc/invoke.texi: Document -mindirect-branch= option.
4829 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4832 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4834 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4836 * ipa-inline.c (want_inline_small_function_p): Return false if
4837 inlining has already failed with CIF_FINAL_ERROR.
4838 (update_caller_keys): Call want_inline_small_function_p before
4840 (update_callee_keys): Likewise.
4842 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4844 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4846 (rs6000_quadword_masked_address_p): Likewise.
4847 (quad_aligned_load_p): Likewise.
4848 (quad_aligned_store_p): Likewise.
4849 (const_load_sequence_p): Add comment to describe the outer-most loop.
4850 (mimic_memory_attributes_and_flags): New function.
4851 (rs6000_gen_stvx): Likewise.
4852 (replace_swapped_aligned_store): Likewise.
4853 (rs6000_gen_lvx): Likewise.
4854 (replace_swapped_aligned_load): Likewise.
4855 (replace_swapped_load_constant): Capitalize argument name in
4856 comment describing this function.
4857 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4858 and stores that access quad-word aligned addresses and replace
4859 with stvx or lvx instructions when appropriate.
4860 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4861 New function prototype.
4862 (rs6000_quadword_masked_address_p): Likewise.
4863 (rs6000_gen_lvx): Likewise.
4864 (rs6000_gen_stvx): Likewise.
4865 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4866 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4867 when memory address is aligned.
4868 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4869 this split to select lvx instruction when memory address is aligned.
4870 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4871 instruction when memory address is aligned.
4872 (*vsx_le_perm_load_v16qi): Likewise.
4873 (four unnamed splitters): Modify to select the stvx instruction
4874 when memory is aligned.
4876 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4878 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4879 which appears in the queue multiple times.
4881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4882 Alan Hayward <alan.hayward@arm.com>
4883 David Sherwood <david.sherwood@arm.com>
4885 * tree-vectorizer.h (vec_lower_bound): New structure.
4886 (_loop_vec_info): Add check_nonzero and lower_bounds.
4887 (LOOP_VINFO_CHECK_NONZERO): New macro.
4888 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4889 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4890 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4891 fields. Make seg_len the distance travelled, not including the
4893 (dr_direction_indicator): Declare.
4894 (dr_zero_step_indicator): Likewise.
4895 (dr_known_forward_stride_p): Likewise.
4896 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4898 (runtime_alias_check_p): Allow runtime alias checks with
4900 (operator ==): Compare access_size and align.
4901 (prune_runtime_alias_test_list): Rework for new distinction between
4902 the access_size and seg_len.
4903 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4905 (get_segment_min_max): New function.
4906 (create_intersect_range_checks): Use it.
4907 (dr_step_indicator): New function.
4908 (dr_direction_indicator): Likewise.
4909 (dr_zero_step_indicator): Likewise.
4910 (dr_known_forward_stride_p): Likewise.
4911 * tree-loop-distribution.c (data_ref_segment_size): Return
4912 DR_STEP * (niters - 1).
4913 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4915 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4916 (vect_preserves_scalar_order_p): New function, split out from...
4917 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4918 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4919 (vect_vfa_access_size): New function.
4920 (vect_vfa_align): Likewise.
4921 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4922 (dump_lower_bound): New function.
4923 (vect_check_lower_bound): Likewise.
4924 (vect_small_gap_p): Likewise.
4925 (vectorizable_with_step_bound_p): Likewise.
4926 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4927 depencies if the vectorization factor is 1. Convert the checks
4928 for nonzero steps into checks on the bounds of DR_STEP. Try using
4929 a bunds check for variable steps if the minimum required step is
4930 relatively small. Update calls to the dr_with_seg_len
4931 constructor and to vect_compile_time_alias.
4932 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4934 (vect_loop_versioning): Call it.
4935 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4937 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4940 Alan Hayward <alan.hayward@arm.com>
4941 David Sherwood <david.sherwood@arm.com>
4943 * doc/sourcebuild.texi (vect_scatter_store): Document.
4944 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4946 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4948 * genopinit.c (main): Add supports_vec_scatter_store and
4949 supports_vec_scatter_store_cached to target_optabs.
4950 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4951 IFN_MASK_SCATTER_STORE.
4952 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4954 * internal-fn.h (internal_store_fn_p): Declare.
4955 (internal_fn_stored_value_index): Likewise.
4956 * internal-fn.c (scatter_store_direct): New macro.
4957 (expand_scatter_store_optab_fn): New function.
4958 (direct_scatter_store_optab_supported_p): New macro.
4959 (internal_store_fn_p): New function.
4960 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4961 IFN_MASK_SCATTER_STORE.
4962 (internal_fn_mask_index): Likewise.
4963 (internal_fn_stored_value_index): New function.
4964 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4966 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4967 * optabs-query.c (supports_vec_scatter_store_p): New function.
4968 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4969 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4970 true for scatter stores.
4971 (vect_gather_scatter_fn_p): Handle scatter stores too.
4972 (vect_check_gather_scatter): Consider using scatter stores if
4973 supports_vec_scatter_store_p.
4974 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4976 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4977 internal_fn_stored_value_index.
4978 (check_load_store_masking): Handle scatter stores too.
4979 (vect_get_store_rhs): Make public.
4980 (vectorizable_call): Use internal_store_fn_p.
4981 (vectorizable_store): Handle scatter store internal functions.
4982 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4983 when deciding whether the end of the group has been reached.
4984 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4985 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4986 (mask_scatter_store<mode>): New insns.
4988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4989 Alan Hayward <alan.hayward@arm.com>
4990 David Sherwood <david.sherwood@arm.com>
4992 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4993 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4994 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4996 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4997 Use vect_truncate_gather_scatter_offset if we can't treat the
4998 operation as a normal gather load or scatter store.
4999 (get_group_load_store_type): Take the gather_scatter_info
5000 as argument. Try using a gather load or scatter store for
5001 single-element groups.
5002 (get_load_store_type): Update calls to get_group_load_store_type
5003 and vect_use_strided_gather_scatters_p.
5005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5006 Alan Hayward <alan.hayward@arm.com>
5007 David Sherwood <david.sherwood@arm.com>
5009 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5010 optional tree argument.
5011 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5013 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5014 but continue to use the current value as a fallback.
5015 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5016 to compare the updates.
5017 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5018 (get_load_store_type): Use it when handling a strided access.
5019 (vect_get_strided_load_store_ops): New function.
5020 (vect_get_data_ptr_increment): Likewise.
5021 (vectorizable_load): Handle strided gather loads. Always pass
5022 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5025 Alan Hayward <alan.hayward@arm.com>
5026 David Sherwood <david.sherwood@arm.com>
5028 * doc/md.texi (gather_load@var{m}): Document.
5029 (mask_gather_load@var{m}): Likewise.
5030 * genopinit.c (main): Add supports_vec_gather_load and
5031 supports_vec_gather_load_cached to target_optabs.
5032 * optabs-tree.c (init_tree_optimization_optabs): Use
5033 ggc_cleared_alloc to allocate target_optabs.
5034 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5035 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5037 * internal-fn.h (internal_load_fn_p): Declare.
5038 (internal_gather_scatter_fn_p): Likewise.
5039 (internal_fn_mask_index): Likewise.
5040 (internal_gather_scatter_fn_supported_p): Likewise.
5041 * internal-fn.c (gather_load_direct): New macro.
5042 (expand_gather_load_optab_fn): New function.
5043 (direct_gather_load_optab_supported_p): New macro.
5044 (direct_internal_fn_optab): New function.
5045 (internal_load_fn_p): Likewise.
5046 (internal_gather_scatter_fn_p): Likewise.
5047 (internal_fn_mask_index): Likewise.
5048 (internal_gather_scatter_fn_supported_p): Likewise.
5049 * optabs-query.c (supports_at_least_one_mode_p): New function.
5050 (supports_vec_gather_load_p): Likewise.
5051 * optabs-query.h (supports_vec_gather_load_p): Declare.
5052 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5053 and memory_type field.
5054 (NUM_PATTERNS): Bump to 15.
5055 * tree-vect-data-refs.c: Include internal-fn.h.
5056 (vect_gather_scatter_fn_p): New function.
5057 (vect_describe_gather_scatter_call): Likewise.
5058 (vect_check_gather_scatter): Try using internal functions for
5059 gather loads. Recognize existing calls to a gather load function.
5060 (vect_analyze_data_refs): Consider using gather loads if
5061 supports_vec_gather_load_p.
5062 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5063 (vect_get_gather_scatter_offset_type): Likewise.
5064 (vect_convert_mask_for_vectype): Likewise.
5065 (vect_add_conversion_to_patterm): Likewise.
5066 (vect_try_gather_scatter_pattern): Likewise.
5067 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5068 (vect_vect_recog_func_ptrs): Add it.
5069 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5070 internal_fn_mask_index and internal_gather_scatter_fn_p.
5071 (check_load_store_masking): Take the gather_scatter_info as an
5072 argument and handle gather loads.
5073 (vect_get_gather_scatter_ops): New function.
5074 (vectorizable_call): Check internal_load_fn_p.
5075 (vectorizable_load): Likewise. Handle gather load internal
5077 (vectorizable_store): Update call to check_load_store_masking.
5078 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5079 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5080 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5081 (aarch64_gather_scale_operand_d): New predicates.
5082 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5083 (mask_gather_load<mode>): New insns.
5085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5086 Alan Hayward <alan.hayward@arm.com>
5087 David Sherwood <david.sherwood@arm.com>
5089 * optabs.def (fold_left_plus_optab): New optab.
5090 * doc/md.texi (fold_left_plus_@var{m}): Document.
5091 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5092 * internal-fn.c (fold_left_direct): Define.
5093 (expand_fold_left_optab_fn): Likewise.
5094 (direct_fold_left_optab_supported_p): Likewise.
5095 * fold-const-call.c (fold_const_fold_left): New function.
5096 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5097 * tree-parloops.c (valid_reduction_p): New function.
5098 (gather_scalar_reductions): Use it.
5099 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5100 (vect_finish_replace_stmt): Declare.
5101 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5102 (needs_fold_left_reduction_p): New function, split out from...
5103 (vect_is_simple_reduction): ...here. Accept reductions that
5104 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5105 (vect_force_simple_reduction): Also store the reduction type in
5106 the assignment's STMT_VINFO_REDUC_TYPE.
5107 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5108 (merge_with_identity): New function.
5109 (vect_expand_fold_left): Likewise.
5110 (vectorize_fold_left_reduction): Likewise.
5111 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5112 scalar phi in place for it. Check for target support and reject
5113 cases that would reassociate the operation. Defer the transform
5114 phase to vectorize_fold_left_reduction.
5115 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5116 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5117 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5121 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5122 call to ifc_temp_var.
5124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5125 Alan Hayward <alan.hayward@arm.com>
5126 David Sherwood <david.sherwood@arm.com>
5128 * target.def (legitimize_address_displacement): Take the original
5129 offset as a poly_int.
5130 * targhooks.h (default_legitimize_address_displacement): Update
5132 * targhooks.c (default_legitimize_address_displacement): Likewise.
5133 * doc/tm.texi: Regenerate.
5134 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5135 as an argument, moving assert of ad->disp == ad->disp_term to...
5136 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5137 Try calling targetm.legitimize_address_displacement before expanding
5138 the address rather than afterwards, and adjust for the new interface.
5139 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5140 Match the new hook interface. Handle SVE addresses.
5141 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5146 * Makefile.in (OBJS): Add early-remat.o.
5147 * target.def (select_early_remat_modes): New hook.
5148 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5149 * doc/tm.texi: Regenerate.
5150 * targhooks.h (default_select_early_remat_modes): Declare.
5151 * targhooks.c (default_select_early_remat_modes): New function.
5152 * timevar.def (TV_EARLY_REMAT): New timevar.
5153 * passes.def (pass_early_remat): New pass.
5154 * tree-pass.h (make_pass_early_remat): Declare.
5155 * early-remat.c: New file.
5156 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5158 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5161 Alan Hayward <alan.hayward@arm.com>
5162 David Sherwood <david.sherwood@arm.com>
5164 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5165 vfm1 with a bound_epilog parameter.
5166 (vect_do_peeling): Update calls accordingly, and move the prologue
5167 call earlier in the function. Treat the base bound_epilog as 0 for
5168 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5169 this base when peeling for gaps.
5170 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5171 with fully-masked loops.
5172 (vect_estimate_min_profitable_iters): Handle the single peeled
5173 iteration in that case.
5175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5176 Alan Hayward <alan.hayward@arm.com>
5177 David Sherwood <david.sherwood@arm.com>
5179 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5180 single-element interleaving even if the size is not a power of 2.
5181 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5182 accesses for single-element interleaving if the group size is
5185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5186 Alan Hayward <alan.hayward@arm.com>
5187 David Sherwood <david.sherwood@arm.com>
5189 * doc/md.texi (fold_extract_last_@var{m}): Document.
5190 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5191 * optabs.def (fold_extract_last_optab): New optab.
5192 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5193 * internal-fn.c (fold_extract_direct): New macro.
5194 (expand_fold_extract_optab_fn): Likewise.
5195 (direct_fold_extract_optab_supported_p): Likewise.
5196 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5197 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5198 EXTRACT_LAST_REDUCTION.
5199 (get_initial_def_for_reduction): Do not create an initial vector
5200 for EXTRACT_LAST_REDUCTION reductions.
5201 (vectorizable_reduction): Leave the scalar phi in place for
5202 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5203 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5204 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5205 transform phase to vectorizable_condition.
5206 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5208 (vect_finish_stmt_generation): ...here.
5209 (vect_finish_replace_stmt): New function.
5210 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5211 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5213 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5216 Alan Hayward <alan.hayward@arm.com>
5217 David Sherwood <david.sherwood@arm.com>
5219 * doc/md.texi (extract_last_@var{m}): Document.
5220 * optabs.def (extract_last_optab): New optab.
5221 * internal-fn.def (EXTRACT_LAST): New internal function.
5222 * internal-fn.c (cond_unary_direct): New macro.
5223 (expand_cond_unary_optab_fn): Likewise.
5224 (direct_cond_unary_optab_supported_p): Likewise.
5225 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5226 loops using EXTRACT_LAST.
5227 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5228 (extract_last_<mode>): ...this optab.
5229 (vec_extract<mode><Vel>): Update accordingly.
5231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5232 Alan Hayward <alan.hayward@arm.com>
5233 David Sherwood <david.sherwood@arm.com>
5235 * target.def (empty_mask_is_expensive): New hook.
5236 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5237 * doc/tm.texi: Regenerate.
5238 * targhooks.h (default_empty_mask_is_expensive): Declare.
5239 * targhooks.c (default_empty_mask_is_expensive): New function.
5240 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5241 if the target says that empty masks are expensive.
5242 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5244 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5247 Alan Hayward <alan.hayward@arm.com>
5248 David Sherwood <david.sherwood@arm.com>
5250 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5251 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5252 (vect_use_loop_mask_for_alignment_p): New function.
5253 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5254 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5255 niters_skip argument. Make sure that the first niters_skip elements
5256 of the first iteration are inactive.
5257 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5258 Update call to vect_set_loop_masks_directly.
5259 (get_misalign_in_elems): New function, split out from...
5260 (vect_gen_prolog_loop_niters): ...here.
5261 (vect_update_init_of_dr): Take a code argument that specifies whether
5262 the adjustment should be added or subtracted.
5263 (vect_update_init_of_drs): Likewise.
5264 (vect_prepare_for_masked_peels): New function.
5265 (vect_do_peeling): Skip prologue peeling if we're using a mask
5266 instead. Update call to vect_update_inits_of_drs.
5267 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5269 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5270 alignment. Do not include the number of peeled iterations in
5271 the minimum threshold in that case.
5272 (vectorizable_induction): Adjust the start value down by
5273 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5274 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5275 Take the number of skipped iterations into account when calculating
5277 * tree-vect-stmts.c (vect_gen_while_not): New function.
5279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5280 Alan Hayward <alan.hayward@arm.com>
5281 David Sherwood <david.sherwood@arm.com>
5283 * doc/sourcebuild.texi (vect_fully_masked): Document.
5284 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5286 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5288 (vect_analyze_loop_2): ...here. Don't check the vectorization
5289 factor against the number of loop iterations if the loop is
5292 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5293 Alan Hayward <alan.hayward@arm.com>
5294 David Sherwood <david.sherwood@arm.com>
5296 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5297 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5298 (dump_groups): Update accordingly.
5299 (iv_use::mem_type): New member variable.
5300 (address_p): New function.
5301 (record_use): Add a mem_type argument and initialize the new
5303 (record_group_use): Add a mem_type argument. Use address_p.
5304 Remove obsolete null checks of base_object. Update call to record_use.
5305 (find_interesting_uses_op): Update call to record_group_use.
5306 (find_interesting_uses_cond): Likewise.
5307 (find_interesting_uses_address): Likewise.
5308 (get_mem_type_for_internal_fn): New function.
5309 (find_address_like_use): Likewise.
5310 (find_interesting_uses_stmt): Try find_address_like_use before
5311 calling find_interesting_uses_op.
5312 (addr_offset_valid_p): Use the iv mem_type field as the type
5313 of the addressed memory.
5314 (add_autoinc_candidates): Likewise.
5315 (get_address_cost): Likewise.
5316 (split_small_address_groups_p): Use address_p.
5317 (split_address_groups): Likewise.
5318 (add_iv_candidate_for_use): Likewise.
5319 (autoinc_possible_for_pair): Likewise.
5320 (rewrite_groups): Likewise.
5321 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5322 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5323 (get_alias_ptr_type_for_ptr_address): New function.
5324 (rewrite_use_address): Rewrite address uses in calls that were
5325 identified by find_address_like_use.
5327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5328 Alan Hayward <alan.hayward@arm.com>
5329 David Sherwood <david.sherwood@arm.com>
5331 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5333 * gimple-expr.h (is_gimple_addressable: Likewise.
5334 * gimple-expr.c (is_gimple_address): Likewise.
5335 * internal-fn.c (expand_call_mem_ref): New function.
5336 (expand_mask_load_optab_fn): Use it.
5337 (expand_mask_store_optab_fn): Likewise.
5339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5340 Alan Hayward <alan.hayward@arm.com>
5341 David Sherwood <david.sherwood@arm.com>
5343 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5344 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5345 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5346 (cond_umax@var{mode}): Document.
5347 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5348 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5349 (cond_umin_optab, cond_umax_optab): New optabs.
5350 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5351 (COND_IOR, COND_XOR): New internal functions.
5352 * internal-fn.h (get_conditional_internal_fn): Declare.
5353 * internal-fn.c (cond_binary_direct): New macro.
5354 (expand_cond_binary_optab_fn): Likewise.
5355 (direct_cond_binary_optab_supported_p): Likewise.
5356 (get_conditional_internal_fn): New function.
5357 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5358 Cope with reduction statements that are vectorized as calls rather
5360 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5361 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5362 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5363 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5364 (UNSPEC_COND_EOR): New unspecs.
5365 (optab): Add mappings for them.
5366 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5367 (sve_int_op, sve_fp_op): New int attributes.
5369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5370 Alan Hayward <alan.hayward@arm.com>
5371 David Sherwood <david.sherwood@arm.com>
5373 * optabs.def (while_ult_optab): New optab.
5374 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5375 * internal-fn.def (WHILE_ULT): New internal function.
5376 * internal-fn.h (direct_internal_fn_supported_p): New override
5377 that takes two types as argument.
5378 * internal-fn.c (while_direct): New macro.
5379 (expand_while_optab_fn): New function.
5380 (convert_optab_supported_p): Likewise.
5381 (direct_while_optab_supported_p): New macro.
5382 * wide-int.h (wi::udiv_ceil): New function.
5383 * tree-vectorizer.h (rgroup_masks): New structure.
5384 (vec_loop_masks): New typedef.
5385 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5387 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5388 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5389 (vect_max_vf): New function.
5390 (slpeel_make_loop_iterate_ntimes): Delete.
5391 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5392 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5393 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5394 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5395 internal-fn.h, stor-layout.h and optabs-query.h.
5396 (vect_set_loop_mask): New function.
5397 (add_preheader_seq): Likewise.
5398 (add_header_seq): Likewise.
5399 (interleave_supported_p): Likewise.
5400 (vect_maybe_permute_loop_masks): Likewise.
5401 (vect_set_loop_masks_directly): Likewise.
5402 (vect_set_loop_condition_masked): Likewise.
5403 (vect_set_loop_condition_unmasked): New function, split out from
5404 slpeel_make_loop_iterate_ntimes.
5405 (slpeel_make_loop_iterate_ntimes): Rename to..
5406 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5407 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5408 (vect_do_peeling): Update call accordingly.
5409 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5411 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5412 mask_compare_type, can_fully_mask_p and fully_masked_p.
5413 (release_vec_loop_masks): New function.
5414 (_loop_vec_info): Use it to free the loop masks.
5415 (can_produce_all_loop_masks_p): New function.
5416 (vect_get_max_nscalars_per_iter): Likewise.
5417 (vect_verify_full_masking): Likewise.
5418 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5419 retries, and free the mask rgroups before retrying. Check loop-wide
5420 reasons for disallowing fully-masked loops. Make the final decision
5421 about whether use a fully-masked loop or not.
5422 (vect_estimate_min_profitable_iters): Do not assume that peeling
5423 for the number of iterations will be needed for fully-masked loops.
5424 (vectorizable_reduction): Disable fully-masked loops.
5425 (vectorizable_live_operation): Likewise.
5426 (vect_halve_mask_nunits): New function.
5427 (vect_double_mask_nunits): Likewise.
5428 (vect_record_loop_mask): Likewise.
5429 (vect_get_loop_mask): Likewise.
5430 (vect_transform_loop): Handle the case in which the final loop
5431 iteration might handle a partial vector. Call vect_set_loop_condition
5432 instead of slpeel_make_loop_iterate_ntimes.
5433 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5434 (check_load_store_masking): New function.
5435 (prepare_load_store_mask): Likewise.
5436 (vectorizable_store): Handle fully-masked loops.
5437 (vectorizable_load): Likewise.
5438 (supportable_widening_operation): Use vect_halve_mask_nunits for
5440 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5441 (vect_gen_while): New function.
5442 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5443 (aarch64_uqdec<mode>): New insn.
5445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5446 Alan Hayward <alan.hayward@arm.com>
5447 David Sherwood <david.sherwood@arm.com>
5449 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5450 (reduc_xor_scal_optab): New optabs.
5451 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5452 (reduc_xor_scal_@var{m}): Document.
5453 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5454 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5456 * fold-const-call.c (fold_const_call): Handle them.
5457 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5458 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5459 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5460 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5461 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5462 (UNSPEC_XORV): New unspecs.
5463 (optab): Add entries for them.
5464 (BITWISEV): New int iterator.
5465 (bit_reduc_op): New int attributes.
5467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5468 Alan Hayward <alan.hayward@arm.com>
5469 David Sherwood <david.sherwood@arm.com>
5471 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5472 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5473 * optabs.def (vec_shl_insert_optab): New optab.
5474 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5475 (duplicate_and_interleave): Likewise.
5476 * tree-vect-loop.c: Include internal-fn.h.
5477 (neutral_op_for_slp_reduction): New function, split out from
5478 get_initial_defs_for_reduction.
5479 (get_initial_def_for_reduction): Handle option 2 for variable-length
5480 vectors by loading the neutral value into a vector and then shifting
5481 the initial value into element 0.
5482 (get_initial_defs_for_reduction): Replace the code argument with
5483 the neutral value calculated by neutral_op_for_slp_reduction.
5484 Use gimple_build_vector for constant-length vectors.
5485 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5486 but the first group_size elements have a neutral value.
5487 Use duplicate_and_interleave otherwise.
5488 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5489 Update call to get_initial_defs_for_reduction. Handle SLP
5490 reductions for variable-length vectors by creating one vector
5491 result for each scalar result, with the elements associated
5492 with other scalar results stubbed out with the neutral value.
5493 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5494 Require IFN_VEC_SHL_INSERT for double reductions on
5495 variable-length vectors, or SLP reductions that have
5496 a neutral value. Require can_duplicate_and_interleave_p
5497 support for variable-length unchained SLP reductions if there
5498 is no neutral value, such as for MIN/MAX reductions. Also require
5499 the number of vector elements to be a multiple of the number of
5500 SLP statements when doing variable-length unchained SLP reductions.
5501 Update call to vect_create_epilog_for_reduction.
5502 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5503 and remove initial values.
5504 (duplicate_and_interleave): Make public.
5505 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5506 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5509 Alan Hayward <alan.hayward@arm.com>
5510 David Sherwood <david.sherwood@arm.com>
5512 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5513 (can_duplicate_and_interleave_p): New function.
5514 (vect_get_and_check_slp_defs): Take the vector of statements
5515 rather than just the current one. Remove excess parentheses.
5516 Restriction rejectinon of vect_constant_def and vect_external_def
5517 for variable-length vectors to boolean types, or types for which
5518 can_duplicate_and_interleave_p is false.
5519 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5520 (duplicate_and_interleave): New function.
5521 (vect_get_constant_vectors): Use gimple_build_vector for
5522 constant-length vectors and suitable variable-length constant
5523 vectors. Use duplicate_and_interleave for other variable-length
5524 vectors. Don't defer the update when inserting new statements.
5526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5527 Alan Hayward <alan.hayward@arm.com>
5528 David Sherwood <david.sherwood@arm.com>
5530 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5531 min_profitable_iters doesn't go negative.
5533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5534 Alan Hayward <alan.hayward@arm.com>
5535 David Sherwood <david.sherwood@arm.com>
5537 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5538 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5539 * optabs.def (vec_mask_load_lanes_optab): New optab.
5540 (vec_mask_store_lanes_optab): Likewise.
5541 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5542 (MASK_STORE_LANES): Likewise.
5543 * internal-fn.c (mask_load_lanes_direct): New macro.
5544 (mask_store_lanes_direct): Likewise.
5545 (expand_mask_load_optab_fn): Handle masked operations.
5546 (expand_mask_load_lanes_optab_fn): New macro.
5547 (expand_mask_store_optab_fn): Handle masked operations.
5548 (expand_mask_store_lanes_optab_fn): New macro.
5549 (direct_mask_load_lanes_optab_supported_p): Likewise.
5550 (direct_mask_store_lanes_optab_supported_p): Likewise.
5551 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5553 (vect_load_lanes_supported): Likewise.
5554 * tree-vect-data-refs.c (strip_conversion): New function.
5555 (can_group_stmts_p): Likewise.
5556 (vect_analyze_data_ref_accesses): Use it instead of checking
5557 for a pair of assignments.
5558 (vect_store_lanes_supported): Take a masked_p parameter.
5559 (vect_load_lanes_supported): Likewise.
5560 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5561 vect_store_lanes_supported and vect_load_lanes_supported.
5562 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5563 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5564 parameter. Don't allow gaps for masked accesses.
5565 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5566 and vect_load_lanes_supported.
5567 (get_load_store_type): Take a masked_p parameter and update
5568 call to get_group_load_store_type.
5569 (vectorizable_store): Update call to get_load_store_type.
5570 Handle IFN_MASK_STORE_LANES.
5571 (vectorizable_load): Update call to get_load_store_type.
5572 Handle IFN_MASK_LOAD_LANES.
5574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5575 Alan Hayward <alan.hayward@arm.com>
5576 David Sherwood <david.sherwood@arm.com>
5578 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5580 * config/aarch64/aarch64-protos.h
5581 (aarch64_sve_struct_memory_operand_p): Declare.
5582 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5583 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5584 (VPRED, vpred): Handle SVE structure modes.
5585 * config/aarch64/constraints.md (Utx): New constraint.
5586 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5587 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5588 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5589 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5590 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5591 structure modes. Split into pieces after RA.
5592 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5593 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5595 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5596 SVE structure modes.
5597 (aarch64_classify_address): Likewise.
5598 (sizetochar): Move earlier in file.
5599 (aarch64_print_operand): Handle SVE register lists.
5600 (aarch64_array_mode): New function.
5601 (aarch64_sve_struct_memory_operand_p): Likewise.
5602 (TARGET_ARRAY_MODE): Redefine.
5604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5605 Alan Hayward <alan.hayward@arm.com>
5606 David Sherwood <david.sherwood@arm.com>
5608 * target.def (array_mode): New target hook.
5609 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5610 * doc/tm.texi: Regenerate.
5611 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5612 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5613 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5615 * stor-layout.c (mode_for_array): Likewise. Support polynomial
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 * fold-const.c (fold_binary_loc): Check the argument types
5623 rather than the result type when testing for a vector operation.
5625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5627 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5628 * doc/tm.texi: Regenerate.
5630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5631 Alan Hayward <alan.hayward@arm.com>
5632 David Sherwood <david.sherwood@arm.com>
5634 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5635 (sve): Document new AArch64 extension.
5636 * doc/md.texi (w): Extend the description of the AArch64
5637 constraint to include SVE vectors.
5638 (Upl, Upa): Document new AArch64 predicate constraints.
5639 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5641 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5642 (msve-vector-bits=): New option.
5643 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5644 SVE when these are disabled.
5645 (sve): New extension.
5646 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5647 modes. Adjust their number of units based on aarch64_sve_vg.
5648 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5649 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5650 aarch64_addr_query_type.
5651 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5652 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5653 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5654 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5655 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5656 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5657 (aarch64_simd_imm_zero_p): Delete.
5658 (aarch64_check_zero_based_sve_index_immediate): Declare.
5659 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5660 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5661 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5662 (aarch64_sve_float_mul_immediate_p): Likewise.
5663 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5665 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5666 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5667 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5668 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5669 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5670 (aarch64_regmode_natural_size): Likewise.
5671 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5672 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5674 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5675 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5676 for VG and the SVE predicate registers.
5677 (V_ALIASES): Add a "z"-prefixed alias.
5678 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5679 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5680 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5681 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5682 (REG_CLASS_NAMES): Add entries for them.
5683 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5684 and the predicate registers.
5685 (aarch64_sve_vg): Declare.
5686 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5687 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5688 (REGMODE_NATURAL_SIZE): Define.
5689 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5691 * config/aarch64/aarch64.c: Include cfgrtl.h.
5692 (simd_immediate_info): Add a constructor for series vectors,
5693 and an associated step field.
5694 (aarch64_sve_vg): New variable.
5695 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5696 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5697 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5698 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5699 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5700 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5701 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5702 (aarch64_get_mask_mode): New functions.
5703 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5704 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5705 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5706 predicate modes and predicate registers. Explicitly restrict
5707 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5708 to store a vector mode if it is recognized by
5709 aarch64_classify_vector_mode.
5710 (aarch64_regmode_natural_size): New function.
5711 (aarch64_hard_regno_caller_save_mode): Return the original mode
5713 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5714 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5715 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5716 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5718 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5719 does not overlap dest if the function is frame-related. Handle
5721 (aarch64_split_add_offset): New function.
5722 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5723 them aarch64_add_offset.
5724 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5725 and update call to aarch64_sub_sp.
5726 (aarch64_add_cfa_expression): New function.
5727 (aarch64_expand_prologue): Pass extra temporary registers to the
5728 functions above. Handle the case in which we need to emit new
5729 DW_CFA_expressions for registers that were originally saved
5730 relative to the stack pointer, but now have to be expressed
5731 relative to the frame pointer.
5732 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5734 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5735 IP0 and IP1 values for SVE frames.
5736 (aarch64_expand_vec_series): New function.
5737 (aarch64_expand_sve_widened_duplicate): Likewise.
5738 (aarch64_expand_sve_const_vector): Likewise.
5739 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5740 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5741 into the register, rather than emitting a SET directly.
5742 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5743 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5744 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5745 (offset_9bit_signed_scaled_p): New functions.
5746 (aarch64_replicate_bitmask_imm): New function.
5747 (aarch64_bitmask_imm): Use it.
5748 (aarch64_cannot_force_const_mem): Reject expressions involving
5749 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5750 (aarch64_classify_index): Handle SVE indices, by requiring
5751 a plain register index with a scale that matches the element size.
5752 (aarch64_classify_address): Handle SVE addresses. Assert that
5753 the mode of the address is VOIDmode or an integer mode.
5754 Update call to aarch64_classify_symbol.
5755 (aarch64_classify_symbolic_expression): Update call to
5756 aarch64_classify_symbol.
5757 (aarch64_const_vec_all_in_range_p): New function.
5758 (aarch64_print_vector_float_operand): Likewise.
5759 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5760 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5761 and the FP immediates 1.0 and 0.5.
5762 (aarch64_print_address_internal): Handle SVE addresses.
5763 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5764 (aarch64_regno_regclass): Handle predicate registers.
5765 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5767 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5768 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5769 (aarch64_convert_sve_vector_bits): New function.
5770 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5771 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5773 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5774 Handle SVE vector and predicate modes. Accept VL-based constants
5775 that need only one temporary register, and VL offsets that require
5776 no temporary registers.
5777 (aarch64_conditional_register_usage): Mark the predicate registers
5778 as fixed if SVE isn't available.
5779 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5780 Return true for SVE vector and predicate modes.
5781 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5782 rather than an unsigned int. Handle SVE modes.
5783 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5785 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5787 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5788 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5789 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5790 (aarch64_sve_float_mul_immediate_p): New functions.
5791 (aarch64_sve_valid_immediate): New function.
5792 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5793 Explicitly reject structure modes. Check for INDEX constants.
5794 Handle PTRUE and PFALSE constants.
5795 (aarch64_check_zero_based_sve_index_immediate): New function.
5796 (aarch64_simd_imm_zero_p): Delete.
5797 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5798 vector modes. Accept constants in the range of CNT[BHWD].
5799 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5800 ask for an Advanced SIMD mode.
5801 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5802 (aarch64_simd_vector_alignment): Handle SVE predicates.
5803 (aarch64_vectorize_preferred_vector_alignment): New function.
5804 (aarch64_simd_vector_alignment_reachable): Use it instead of
5806 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5807 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5809 (MAX_VECT_LEN): Delete.
5810 (expand_vec_perm_d): Add a vec_flags field.
5811 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5812 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5813 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5815 (aarch64_evpc_rev): Rename to...
5816 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5817 (aarch64_evpc_rev_global): New function.
5818 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5819 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5821 (aarch64_evpc_sve_tbl): New function.
5822 (aarch64_expand_vec_perm_const_1): Update after rename of
5823 aarch64_evpc_rev. Handle SVE permutes too, trying
5824 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5825 than aarch64_evpc_tbl.
5826 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5827 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5828 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5829 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5830 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5831 (aarch64_expand_sve_vcond): New functions.
5832 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5833 of aarch64_vector_mode_p.
5834 (aarch64_dwarf_poly_indeterminate_value): New function.
5835 (aarch64_compute_pressure_classes): Likewise.
5836 (aarch64_can_change_mode_class): Likewise.
5837 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5838 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5839 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5840 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5841 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5842 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5843 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5844 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5846 (Dn, Dl, Dr): Accept const as well as const_vector.
5847 (Dz): Likewise. Compare against CONST0_RTX.
5848 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5849 of "vector" where appropriate.
5850 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5851 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5852 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5853 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5854 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5855 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5856 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5857 (v_int_equiv): Extend to SVE modes.
5858 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5860 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5861 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5862 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5863 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5864 (SVE_COND_FP_CMP): New int iterators.
5865 (perm_hilo): Handle the new unpack unspecs.
5866 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5868 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5869 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5870 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5871 (aarch64_equality_operator, aarch64_constant_vector_operand)
5872 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5873 (aarch64_sve_nonimmediate_operand): Likewise.
5874 (aarch64_sve_general_operand): Likewise.
5875 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5876 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5877 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5878 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5879 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5880 (aarch64_sve_float_arith_immediate): Likewise.
5881 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5882 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5883 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5884 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5885 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5886 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5887 (aarch64_sve_float_arith_operand): Likewise.
5888 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5889 (aarch64_sve_float_mul_operand): Likewise.
5890 (aarch64_sve_vec_perm_operand): Likewise.
5891 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5892 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5893 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5894 as well as const_vector.
5895 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5896 in file. Use CONST0_RTX and CONSTM1_RTX.
5897 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5898 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5899 Use aarch64_simd_imm_zero.
5900 * config/aarch64/aarch64-sve.md: New file.
5901 * config/aarch64/aarch64.md: Include it.
5902 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5903 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5904 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5905 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5906 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5907 (sve): New attribute.
5908 (enabled): Disable instructions with the sve attribute unless
5910 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5911 aarch64_expand_mov_immediate.
5912 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5913 CNT[BHSD] immediates.
5914 (movti): Split CONST_POLY_INT moves into two halves.
5915 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5916 Split additions that need a temporary here if the destination
5917 is the stack pointer.
5918 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5919 (*add<mode>3_poly_1): New instruction.
5920 (set_clobber_cc): New expander.
5922 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5924 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5925 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5926 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5927 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5928 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5929 Change innermode from fixed_mode_size to machine_mode.
5930 (simplify_subreg): Update call accordingly. Handle a constant-sized
5931 subreg of a variable-length CONST_VECTOR.
5933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5934 Alan Hayward <alan.hayward@arm.com>
5935 David Sherwood <david.sherwood@arm.com>
5937 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5938 (add_offset_to_base): New function, split out from...
5939 (create_mem_ref): ...here. When handling a scale other than 1,
5940 check first whether the address is valid without the offset.
5941 Add it into the base if so, leaving the index and scale as-is.
5943 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5946 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5947 fold_for_warn before checking if arg2 is INTEGER_CST.
5949 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5951 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5952 (store_multiple_operation): Delete.
5953 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5954 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5955 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5956 guarded by TARGET_STRING.
5957 (rs6000_output_load_multiple): Delete.
5958 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5959 OPTION_MASK_STRING / TARGET_STRING handling.
5960 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5961 (const rs6000_opt_masks) <"string">: Change mask to 0.
5962 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5963 (MASK_STRING): Delete.
5964 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5966 (load_multiple): Delete.
5973 (store_multiple): Delete.
5980 (movmemsi_8reg): Delete.
5981 (corresponding unnamed define_insn): Delete.
5982 (movmemsi_6reg): Delete.
5983 (corresponding unnamed define_insn): Delete.
5984 (movmemsi_4reg): Delete.
5985 (corresponding unnamed define_insn): Delete.
5986 (movmemsi_2reg): Delete.
5987 (corresponding unnamed define_insn): Delete.
5988 (movmemsi_1reg): Delete.
5989 (corresponding unnamed define_insn): Delete.
5990 * config/rs6000/rs6000.opt (mno-string): New.
5991 (mstring): Replace by deprecation warning stub.
5992 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5994 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5996 * regrename.c (regrename_do_replace): If replacing the same
5997 reg multiple times, try to reuse last created gen_raw_REG.
6000 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6001 main to workaround a bug in GDB.
6003 2018-01-12 Tom de Vries <tom@codesourcery.com>
6006 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6008 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6010 PR rtl-optimization/80481
6011 * ira-color.c (get_cap_member): New function.
6012 (allocnos_conflict_by_live_ranges_p): Use it.
6013 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6014 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6016 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6019 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6020 (*saddl_se_1): Ditto.
6022 (*ssubl_se_1): Ditto.
6024 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6026 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6027 rather than wi::to_widest for DR_INITs.
6028 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6029 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6030 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6032 (vect_analyze_group_access_1): Note that here.
6034 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6036 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6037 polynomial type sizes.
6039 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6041 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6042 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6043 (gimple_add_tmp_var): Likewise.
6045 2018-01-12 Martin Liska <mliska@suse.cz>
6047 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6048 (gimple_alloc_sizes): Likewise.
6049 (dump_gimple_statistics): Use PRIu64 in printf format.
6050 * gimple.h: Change uint64_t to int.
6052 2018-01-12 Martin Liska <mliska@suse.cz>
6054 * tree-core.h: Use uint64_t instead of int.
6055 * tree.c (tree_node_counts): Likewise.
6056 (tree_node_sizes): Likewise.
6057 (dump_tree_statistics): Use PRIu64 in printf format.
6059 2018-01-12 Martin Liska <mliska@suse.cz>
6061 * Makefile.in: As qsort_chk is implemented in vec.c, add
6062 vec.o to linkage of gencfn-macros.
6063 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6064 passing the info to record_node_allocation_statistics.
6065 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6067 * ggc-common.c (struct ggc_usage): Add operator== and use
6068 it in operator< and compare function.
6069 * mem-stats.h (struct mem_usage): Likewise.
6070 * vec.c (struct vec_usage): Remove operator< and compare
6071 function. Can be simply inherited.
6073 2018-01-12 Martin Jambor <mjambor@suse.cz>
6076 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6077 * tree-ssa-math-opts.c: Include domwalk.h.
6078 (convert_mult_to_fma_1): New function.
6079 (fma_transformation_info): New type.
6080 (fma_deferring_state): Likewise.
6081 (cancel_fma_deferring): New function.
6082 (result_of_phi): Likewise.
6083 (last_fma_candidate_feeds_initial_phi): Likewise.
6084 (convert_mult_to_fma): Added deferring logic, split actual
6085 transformation to convert_mult_to_fma_1.
6086 (math_opts_dom_walker): New type.
6087 (math_opts_dom_walker::after_dom_children): New method, body moved
6088 here from pass_optimize_widening_mul::execute, added deferring logic
6090 (pass_optimize_widening_mul::execute): Moved most of code to
6091 math_opts_dom_walker::after_dom_children.
6092 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6093 * config/i386/i386.c (ix86_option_override_internal): Added
6094 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6096 2018-01-12 Richard Biener <rguenther@suse.de>
6099 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6100 inline instance vars.
6102 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6105 * config/rx/rx.c (rx_is_restricted_memory_address):
6108 2018-01-12 Richard Biener <rguenther@suse.de>
6110 PR tree-optimization/80846
6111 * target.def (split_reduction): New target hook.
6112 * targhooks.c (default_split_reduction): New function.
6113 * targhooks.h (default_split_reduction): Declare.
6114 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6115 target requests first reduce vectors by combining low and high
6117 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6118 (get_vectype_for_scalar_type_and_size): Export.
6119 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6120 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6121 * doc/tm.texi: Regenerate.
6122 * config/i386/i386.c (ix86_split_reduction): Implement
6123 TARGET_VECTORIZE_SPLIT_REDUCTION.
6125 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6128 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6129 in PIC mode except for TARGET_VXWORKS_RTP.
6130 * config/sparc/sparc.c: Include cfgrtl.h.
6131 (TARGET_INIT_PIC_REG): Define.
6132 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6133 (sparc_pic_register_p): New predicate.
6134 (sparc_legitimate_address_p): Use it.
6135 (sparc_legitimize_pic_address): Likewise.
6136 (sparc_delegitimize_address): Likewise.
6137 (sparc_mode_dependent_address_p): Likewise.
6138 (gen_load_pcrel_sym): Remove 4th parameter.
6139 (load_got_register): Adjust call to above. Remove obsolete stuff.
6140 (sparc_expand_prologue): Do not call load_got_register here.
6141 (sparc_flat_expand_prologue): Likewise.
6142 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6143 (sparc_use_pseudo_pic_reg): New function.
6144 (sparc_init_pic_reg): Likewise.
6145 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6146 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6148 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6150 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6151 Add item for branch_cost.
6153 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6155 PR rtl-optimization/83565
6156 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6157 not extend the result to a larger mode for rotate operations.
6158 (num_sign_bit_copies1): Likewise.
6160 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6163 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6165 Use values-Xc.o for -pedantic.
6166 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6168 2018-01-12 Martin Liska <mliska@suse.cz>
6171 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6173 (possible_polymorphic_call_targets): Use it.
6174 (ipa_devirt): Likewise.
6176 2018-01-12 Martin Liska <mliska@suse.cz>
6178 * profile-count.h (enum profile_quality): Use 0 as invalid
6179 enum value of profile_quality.
6181 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6183 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6184 -mext-string options.
6186 2018-01-12 Richard Biener <rguenther@suse.de>
6188 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6189 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6190 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6192 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6194 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6196 * configure.ac (--with-long-double-format): Add support for the
6197 configuration option to change the default long double format on
6199 * config.gcc (powerpc*-linux*-*): Likewise.
6200 * configure: Regenerate.
6201 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6202 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6203 used without modification.
6205 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6207 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6208 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6209 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6210 MISC_BUILTIN_SPEC_BARRIER.
6211 (rs6000_init_builtins): Likewise.
6212 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6214 (speculation_barrier): New define_insn.
6215 * doc/extend.texi: Document __builtin_speculation_barrier.
6217 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6220 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6221 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6222 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6224 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6225 integral modes instead of "ss" and "sd".
6226 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6227 vectors with 32-bit and 64-bit elements.
6228 (vecdupssescalarmodesuffix): New mode attribute.
6229 (vec_dup<mode>): Use it.
6231 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6234 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6235 frame if argument is passed on stack.
6237 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6240 * ree.c (combine_reaching_defs): Optimize also
6241 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6242 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6244 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6247 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6249 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6252 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6253 after they are computed.
6255 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6257 PR tree-optimization/83695
6258 * gimple-loop-linterchange.cc
6259 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6260 reset cached scev information after interchange.
6261 (pass_linterchange::execute): Remove call to scev_reset_htab.
6263 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6265 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6266 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6267 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6268 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6269 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6270 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6271 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6272 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6273 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6274 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6275 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6276 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6277 (V_lane_reg): Likewise.
6278 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6280 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6281 (vfmal_lane_low<mode>_intrinsic,
6282 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6283 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6284 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6285 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6286 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6287 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6289 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6291 * config/arm/arm-cpus.in (fp16fml): New feature.
6292 (ALL_SIMD): Add fp16fml.
6293 (armv8.2-a): Add fp16fml as an option.
6294 (armv8.3-a): Likewise.
6295 (armv8.4-a): Add fp16fml as part of fp16.
6296 * config/arm/arm.h (TARGET_FP16FML): Define.
6297 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6299 * config/arm/arm-modes.def (V2HF): Define.
6300 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6301 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6302 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6303 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6304 vfmsl_low, vfmsl_high): New set of builtins.
6305 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6306 (vfml_op): New code attribute.
6307 (VFMLHALVES): New int iterator.
6308 (VFML, VFMLSEL): New mode attributes.
6309 (V_reg): Define mapping for V2HF.
6310 (V_hi, V_lo): New mode attributes.
6311 (VF_constraint): Likewise.
6312 (vfml_half, vfml_half_selector): New int attributes.
6313 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6315 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6316 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6318 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6319 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6320 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6321 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6323 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6324 Document new effective target and option set.
6326 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6328 * config/arm/arm-cpus.in (armv8_4): New feature.
6329 (ARMv8_4a): New fgroup.
6330 (armv8.4-a): New arch.
6331 * config/arm/arm-tables.opt: Regenerate.
6332 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6333 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6334 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6335 Add matching rules for -march=armv8.4-a and extensions.
6336 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6338 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6341 * config/rx/rx.md (BW): New mode attribute.
6342 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6344 2018-01-11 Richard Biener <rguenther@suse.de>
6346 PR tree-optimization/83435
6347 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6348 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6349 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6351 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6352 Alan Hayward <alan.hayward@arm.com>
6353 David Sherwood <david.sherwood@arm.com>
6355 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6357 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6358 (aarch64_print_address_internal): Use it to check for a zero offset.
6360 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6361 Alan Hayward <alan.hayward@arm.com>
6362 David Sherwood <david.sherwood@arm.com>
6364 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6365 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6366 Return a poly_int64 rather than a HOST_WIDE_INT.
6367 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6368 rather than a HOST_WIDE_INT.
6369 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6370 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6371 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6372 final_offset from HOST_WIDE_INT to poly_int64.
6373 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6374 to_constant when getting the number of units in an Advanced SIMD
6376 (aarch64_builtin_vectorized_function): Check for a constant number
6378 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6380 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6381 attribute instead of GET_MODE_NUNITS.
6382 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6383 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6384 GET_MODE_SIZE for fixed-size registers.
6385 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6386 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6387 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6388 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6389 (aarch64_print_operand, aarch64_print_address_internal)
6390 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6391 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6392 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6393 Handle polynomial GET_MODE_SIZE.
6394 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6395 wider than SImode without modification.
6396 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6397 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6398 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6399 passing and returning SVE modes.
6400 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6401 rather than GEN_INT.
6402 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6403 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6404 (aarch64_allocate_and_probe_stack_space): Likewise.
6405 (aarch64_layout_frame): Cope with polynomial offsets.
6406 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6407 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6409 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6410 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6411 poly_int64 rather than a HOST_WIDE_INT.
6412 (aarch64_get_separate_components, aarch64_process_components)
6413 (aarch64_expand_prologue, aarch64_expand_epilogue)
6414 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6415 (aarch64_anchor_offset): New function, split out from...
6416 (aarch64_legitimize_address): ...here.
6417 (aarch64_builtin_vectorization_cost): Handle polynomial
6418 TYPE_VECTOR_SUBPARTS.
6419 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6421 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6422 number of elements from the PARALLEL rather than the mode.
6423 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6424 rather than GET_MODE_BITSIZE.
6425 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6426 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6427 (aarch64_expand_vec_perm_const_1): Handle polynomial
6428 d->perm.length () and d->perm elements.
6429 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6430 Apply to_constant to d->perm elements.
6431 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6432 polynomial CONST_VECTOR_NUNITS.
6433 (aarch64_move_pointer): Take amount as a poly_int64 rather
6435 (aarch64_progress_pointer): Avoid temporary variable.
6436 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6437 the mode attribute instead of GET_MODE.
6439 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6440 Alan Hayward <alan.hayward@arm.com>
6441 David Sherwood <david.sherwood@arm.com>
6443 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6444 x exists before using it.
6445 (aarch64_add_constant_internal): Rename to...
6446 (aarch64_add_offset_1): ...this. Replace regnum with separate
6447 src and dest rtxes. Handle the case in which they're different,
6448 including when the offset is zero. Replace scratchreg with an rtx.
6449 Use 2 additions if there is no spare register into which we can
6450 move a 16-bit constant.
6451 (aarch64_add_constant): Delete.
6452 (aarch64_add_offset): Replace reg with separate src and dest
6453 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6454 Use aarch64_add_offset_1.
6455 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6456 an rtx rather than an int. Take the delta as a poly_int64
6457 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6458 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6459 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6460 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6461 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6463 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6464 aarch64_add_constant.
6466 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6468 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6469 Use scalar_float_mode.
6471 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6473 * config/aarch64/aarch64-simd.md
6474 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6475 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6476 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6477 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6478 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6479 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6480 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6481 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6482 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6483 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6485 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6488 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6489 targ_options->x_arm_arch_string is non NULL.
6491 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6493 * config/aarch64/aarch64.h
6494 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6496 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6499 * expmed.c (emit_store_flag_force): Swap if const op0
6500 and change VOIDmode to mode of op0.
6502 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6504 PR rtl-optimization/83761
6505 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6506 than bytes to mode_for_size.
6508 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6511 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6512 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6515 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6518 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6519 when in layout mode.
6520 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6521 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6524 2018-01-10 Michael Collison <michael.collison@arm.com>
6526 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6527 * config/aarch64/aarch64-option-extension.def: Add
6528 AARCH64_OPT_EXTENSION of 'fp16fml'.
6529 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6530 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6531 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6532 * config/aarch64/constraints.md (Ui7): New constraint.
6533 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6534 (VFMLA_SEL_W): Ditto.
6537 (VFMLA16_LOW): New int iterator.
6538 (VFMLA16_HIGH): Ditto.
6539 (UNSPEC_FMLAL): New unspec.
6540 (UNSPEC_FMLSL): Ditto.
6541 (UNSPEC_FMLAL2): Ditto.
6542 (UNSPEC_FMLSL2): Ditto.
6543 (f16mac): New code attribute.
6544 * config/aarch64/aarch64-simd-builtins.def
6545 (aarch64_fmlal_lowv2sf): Ditto.
6546 (aarch64_fmlsl_lowv2sf): Ditto.
6547 (aarch64_fmlalq_lowv4sf): Ditto.
6548 (aarch64_fmlslq_lowv4sf): Ditto.
6549 (aarch64_fmlal_highv2sf): Ditto.
6550 (aarch64_fmlsl_highv2sf): Ditto.
6551 (aarch64_fmlalq_highv4sf): Ditto.
6552 (aarch64_fmlslq_highv4sf): Ditto.
6553 (aarch64_fmlal_lane_lowv2sf): Ditto.
6554 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6555 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6556 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6557 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6558 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6559 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6560 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6561 (aarch64_fmlal_lane_highv2sf): Ditto.
6562 (aarch64_fmlsl_lane_highv2sf): Ditto.
6563 (aarch64_fmlal_laneq_highv2sf): Ditto.
6564 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6565 (aarch64_fmlalq_lane_highv4sf): Ditto.
6566 (aarch64_fmlsl_lane_highv4sf): Ditto.
6567 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6568 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6569 * config/aarch64/aarch64-simd.md:
6570 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6571 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6572 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6573 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6574 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6575 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6576 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6577 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6578 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6579 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6580 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6581 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6582 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6583 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6584 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6585 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6586 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6587 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6588 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6589 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6590 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6591 (vfmlsl_low_u32): Ditto.
6592 (vfmlalq_low_u32): Ditto.
6593 (vfmlslq_low_u32): Ditto.
6594 (vfmlal_high_u32): Ditto.
6595 (vfmlsl_high_u32): Ditto.
6596 (vfmlalq_high_u32): Ditto.
6597 (vfmlslq_high_u32): Ditto.
6598 (vfmlal_lane_low_u32): Ditto.
6599 (vfmlsl_lane_low_u32): Ditto.
6600 (vfmlal_laneq_low_u32): Ditto.
6601 (vfmlsl_laneq_low_u32): Ditto.
6602 (vfmlalq_lane_low_u32): Ditto.
6603 (vfmlslq_lane_low_u32): Ditto.
6604 (vfmlalq_laneq_low_u32): Ditto.
6605 (vfmlslq_laneq_low_u32): Ditto.
6606 (vfmlal_lane_high_u32): Ditto.
6607 (vfmlsl_lane_high_u32): Ditto.
6608 (vfmlal_laneq_high_u32): Ditto.
6609 (vfmlsl_laneq_high_u32): Ditto.
6610 (vfmlalq_lane_high_u32): Ditto.
6611 (vfmlslq_lane_high_u32): Ditto.
6612 (vfmlalq_laneq_high_u32): Ditto.
6613 (vfmlslq_laneq_high_u32): Ditto.
6614 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6615 (AARCH64_FL_FOR_ARCH8_4): New.
6616 (AARCH64_ISA_F16FML): New ISA flag.
6617 (TARGET_F16FML): New feature flag for fp16fml.
6618 (doc/invoke.texi): Document new fp16fml option.
6620 2018-01-10 Michael Collison <michael.collison@arm.com>
6622 * config/aarch64/aarch64-builtins.c:
6623 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6624 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6625 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6626 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6627 (AARCH64_ISA_SHA3): New ISA flag.
6628 (TARGET_SHA3): New feature flag for sha3.
6629 * config/aarch64/iterators.md (sha512_op): New int attribute.
6630 (CRYPTO_SHA512): New int iterator.
6631 (UNSPEC_SHA512H): New unspec.
6632 (UNSPEC_SHA512H2): Ditto.
6633 (UNSPEC_SHA512SU0): Ditto.
6634 (UNSPEC_SHA512SU1): Ditto.
6635 * config/aarch64/aarch64-simd-builtins.def
6636 (aarch64_crypto_sha512hqv2di): New builtin.
6637 (aarch64_crypto_sha512h2qv2di): Ditto.
6638 (aarch64_crypto_sha512su0qv2di): Ditto.
6639 (aarch64_crypto_sha512su1qv2di): Ditto.
6640 (aarch64_eor3qv8hi): Ditto.
6641 (aarch64_rax1qv2di): Ditto.
6642 (aarch64_xarqv2di): Ditto.
6643 (aarch64_bcaxqv8hi): Ditto.
6644 * config/aarch64/aarch64-simd.md:
6645 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6646 (aarch64_crypto_sha512su0qv2di): Ditto.
6647 (aarch64_crypto_sha512su1qv2di): Ditto.
6648 (aarch64_eor3qv8hi): Ditto.
6649 (aarch64_rax1qv2di): Ditto.
6650 (aarch64_xarqv2di): Ditto.
6651 (aarch64_bcaxqv8hi): Ditto.
6652 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6653 (vsha512h2q_u64): Ditto.
6654 (vsha512su0q_u64): Ditto.
6655 (vsha512su1q_u64): Ditto.
6656 (veor3q_u16): Ditto.
6657 (vrax1q_u64): Ditto.
6659 (vbcaxq_u16): Ditto.
6660 * config/arm/types.md (crypto_sha512): New type attribute.
6661 (crypto_sha3): Ditto.
6662 (doc/invoke.texi): Document new sha3 option.
6664 2018-01-10 Michael Collison <michael.collison@arm.com>
6666 * config/aarch64/aarch64-builtins.c:
6667 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6668 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6669 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6670 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6671 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6672 (AARCH64_ISA_SM4): New ISA flag.
6673 (TARGET_SM4): New feature flag for sm4.
6674 * config/aarch64/aarch64-simd-builtins.def
6675 (aarch64_sm3ss1qv4si): Ditto.
6676 (aarch64_sm3tt1aq4si): Ditto.
6677 (aarch64_sm3tt1bq4si): Ditto.
6678 (aarch64_sm3tt2aq4si): Ditto.
6679 (aarch64_sm3tt2bq4si): Ditto.
6680 (aarch64_sm3partw1qv4si): Ditto.
6681 (aarch64_sm3partw2qv4si): Ditto.
6682 (aarch64_sm4eqv4si): Ditto.
6683 (aarch64_sm4ekeyqv4si): Ditto.
6684 * config/aarch64/aarch64-simd.md:
6685 (aarch64_sm3ss1qv4si): Ditto.
6686 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6687 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6688 (aarch64_sm4eqv4si): Ditto.
6689 (aarch64_sm4ekeyqv4si): Ditto.
6690 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6691 (sm3part_op): Ditto.
6692 (CRYPTO_SM3TT): Ditto.
6693 (CRYPTO_SM3PART): Ditto.
6694 (UNSPEC_SM3SS1): New unspec.
6695 (UNSPEC_SM3TT1A): Ditto.
6696 (UNSPEC_SM3TT1B): Ditto.
6697 (UNSPEC_SM3TT2A): Ditto.
6698 (UNSPEC_SM3TT2B): Ditto.
6699 (UNSPEC_SM3PARTW1): Ditto.
6700 (UNSPEC_SM3PARTW2): Ditto.
6701 (UNSPEC_SM4E): Ditto.
6702 (UNSPEC_SM4EKEY): Ditto.
6703 * config/aarch64/constraints.md (Ui2): New constraint.
6704 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6705 * config/arm/types.md (crypto_sm3): New type attribute.
6706 (crypto_sm4): Ditto.
6707 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6708 (vsm3tt1aq_u32): Ditto.
6709 (vsm3tt1bq_u32): Ditto.
6710 (vsm3tt2aq_u32): Ditto.
6711 (vsm3tt2bq_u32): Ditto.
6712 (vsm3partw1q_u32): Ditto.
6713 (vsm3partw2q_u32): Ditto.
6714 (vsm4eq_u32): Ditto.
6715 (vsm4ekeyq_u32): Ditto.
6716 (doc/invoke.texi): Document new sm4 option.
6718 2018-01-10 Michael Collison <michael.collison@arm.com>
6720 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6721 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6722 (AARCH64_FL_FOR_ARCH8_4): New.
6723 (AARCH64_FL_V8_4): New flag.
6724 (doc/invoke.texi): Document new armv8.4-a option.
6726 2018-01-10 Michael Collison <michael.collison@arm.com>
6728 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6729 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6730 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6731 * config/aarch64/aarch64-option-extension.def: Add
6732 AARCH64_OPT_EXTENSION of 'sha2'.
6733 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6734 (crypto): Disable sha2 and aes if crypto disabled.
6735 (crypto): Enable aes and sha2 if enabled.
6736 (simd): Disable sha2 and aes if simd disabled.
6737 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6739 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6740 (TARGET_SHA2): New feature flag for sha2.
6741 (TARGET_AES): New feature flag for aes.
6742 * config/aarch64/aarch64-simd.md:
6743 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6744 conditional on TARGET_AES.
6745 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6746 (aarch64_crypto_sha1hsi): Make pattern conditional
6748 (aarch64_crypto_sha1hv4si): Ditto.
6749 (aarch64_be_crypto_sha1hv4si): Ditto.
6750 (aarch64_crypto_sha1su1v4si): Ditto.
6751 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6752 (aarch64_crypto_sha1su0v4si): Ditto.
6753 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6754 (aarch64_crypto_sha256su0v4si): Ditto.
6755 (aarch64_crypto_sha256su1v4si): Ditto.
6756 (doc/invoke.texi): Document new aes and sha2 options.
6758 2018-01-10 Martin Sebor <msebor@redhat.com>
6760 PR tree-optimization/83781
6761 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6764 2018-01-11 Martin Sebor <msebor@gmail.com>
6765 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6767 PR tree-optimization/83501
6768 PR tree-optimization/81703
6770 * tree-ssa-strlen.c (get_string_cst): Rename...
6771 (get_string_len): ...to this. Handle global constants.
6772 (handle_char_store): Adjust.
6774 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6775 Jim Wilson <jimw@sifive.com>
6777 * config/riscv/riscv-protos.h (riscv_output_return): New.
6778 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6779 (riscv_attribute_table, riscv_output_return),
6780 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6781 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6782 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6784 (riscv_expand_prologue): Add early return for naked function.
6785 (riscv_expand_epilogue): Likewise.
6786 (riscv_function_ok_for_sibcall): Return false for naked function.
6787 (riscv_set_current_function): New.
6788 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6789 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6790 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6791 * doc/extend.texi (RISC-V Function Attributes): New.
6793 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6795 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6796 check for 128-bit long double before checking TCmode.
6797 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6798 128-bit long doubles before checking TFmode or TCmode.
6799 (FLOAT128_IBM_P): Likewise.
6801 2018-01-10 Martin Sebor <msebor@redhat.com>
6803 PR tree-optimization/83671
6804 * builtins.c (c_strlen): Unconditionally return zero for the empty
6806 Use -Warray-bounds for warnings.
6807 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6808 for non-constant array indices with COMPONENT_REF, arrays of
6809 arrays, and pointers to arrays.
6810 (gimple_fold_builtin_strlen): Determine and set length range for
6811 non-constant character arrays.
6813 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6816 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6819 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6821 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6823 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6826 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6827 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6828 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6829 indexed_or_indirect_operand predicate.
6830 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6831 (*vsx_le_perm_load_v8hi): Likewise.
6832 (*vsx_le_perm_load_v16qi): Likewise.
6833 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6834 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6835 (*vsx_le_perm_store_v8hi): Likewise.
6836 (*vsx_le_perm_store_v16qi): Likewise.
6837 (eight unnamed splitters): Likewise.
6839 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6841 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6842 * config/rs6000/emmintrin.h: Likewise.
6843 * config/rs6000/mmintrin.h: Likewise.
6844 * config/rs6000/xmmintrin.h: Likewise.
6846 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6849 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6851 * tree.c (tree_nop_conversion): Return true for location wrapper
6853 (maybe_wrap_with_location): New function.
6854 (selftest::check_strip_nops): New function.
6855 (selftest::test_location_wrappers): New function.
6856 (selftest::tree_c_tests): Call it.
6857 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6858 (maybe_wrap_with_location): New decl.
6859 (EXPR_LOCATION_WRAPPER_P): New macro.
6860 (location_wrapper_p): New inline function.
6861 (tree_strip_any_location_wrapper): New inline function.
6863 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6866 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6867 stack_realign_offset for the largest alignment of stack slot
6869 (ix86_find_max_used_stack_alignment): New function.
6870 (ix86_finalize_stack_frame_flags): Use it. Set
6871 max_used_stack_alignment if we don't realign stack.
6872 * config/i386/i386.h (machine_function): Add
6873 max_used_stack_alignment.
6875 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6877 * config/arm/arm.opt (-mbranch-cost): New option.
6878 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6881 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6884 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6885 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6887 2018-01-10 Richard Biener <rguenther@suse.de>
6890 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6891 early out so it also covers the case where we have a non-NULL
6894 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6896 PR tree-optimization/83753
6897 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6898 for non-strided grouped accesses if the number of elements is 1.
6900 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6903 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6904 * i386.h (TARGET_USE_GATHER): Define.
6905 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6907 2018-01-10 Martin Liska <mliska@suse.cz>
6910 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6911 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6913 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6914 CLEANUP_NO_PARTITIONING is not set.
6916 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6918 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6919 for vectors, as a partial revert of r254296.
6920 * rtl.h (const_vec_p): Delete.
6921 (const_vec_duplicate_p): Don't test for vector CONSTs.
6922 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6923 * expmed.c (make_tree): Likewise.
6926 * common.md (E, F): Use CONSTANT_P instead of checking for
6928 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6929 checking for CONST_VECTOR.
6931 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6934 * predict.c (force_edge_cold): Handle in more sane way edges
6937 2018-01-09 Carl Love <cel@us.ibm.com>
6939 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6941 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6942 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6943 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6944 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6945 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6946 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6947 * config/rs6000/rs6000-protos.h: Add extern defition for
6948 rs6000_generate_float2_double_code.
6949 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6951 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6952 (float2_v2df): Add define_expand.
6954 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6957 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6958 op_mode in the force_to_mode call.
6960 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6962 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6963 instead of checking each element individually.
6964 (aarch64_evpc_uzp): Likewise.
6965 (aarch64_evpc_zip): Likewise.
6966 (aarch64_evpc_ext): Likewise.
6967 (aarch64_evpc_rev): Likewise.
6968 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6969 instead of checking each element individually. Return true without
6971 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6972 whether all selected elements come from the same input, instead of
6973 checking each element individually. Remove calls to gen_rtx_REG,
6974 start_sequence and end_sequence and instead assert that no rtl is
6977 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6979 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6980 order of HIGH and CONST checks.
6982 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6984 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6985 if the destination isn't an SSA_NAME.
6987 2018-01-09 Richard Biener <rguenther@suse.de>
6989 PR tree-optimization/83668
6990 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6992 (canonicalize_loop_form): ... here, renamed from ...
6993 (canonicalize_loop_closed_ssa_form): ... this and amended to
6994 swap successor edges for loop exit blocks to make us use
6995 the RPO order we need for initial schedule generation.
6997 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6999 PR tree-optimization/64811
7000 * match.pd: When optimizing comparisons with Inf, avoid
7001 introducing or losing exceptions from comparisons with NaN.
7003 2018-01-09 Martin Liska <mliska@suse.cz>
7006 * asan.c (shadow_mem_size): Add gcc_assert.
7008 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7010 Don't save registers in main().
7013 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7014 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7015 * config/avr/avr.c (avr_set_current_function): Don't error if
7016 naked, OS_task or OS_main are specified at the same time.
7017 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7019 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7021 * common/config/avr/avr-common.c (avr_option_optimization_table):
7022 Switch on -mmain-is-OS_task for optimizing compilations.
7024 2018-01-09 Richard Biener <rguenther@suse.de>
7026 PR tree-optimization/83572
7027 * graphite.c: Include cfganal.h.
7028 (graphite_transform_loops): Connect infinite loops to exit
7029 and remove fake edges at the end.
7031 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7033 * ipa-inline.c (edge_badness): Revert accidental checkin.
7035 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7038 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7039 symbols; not inline clones.
7041 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7044 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7045 hard registers. Formatting fixes.
7047 PR preprocessor/83722
7048 * gcc.c (try_generate_repro): Pass
7049 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7050 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7053 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7054 Kito Cheng <kito.cheng@gmail.com>
7056 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7057 (riscv_leaf_function_p): Delete.
7058 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7060 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7062 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7064 (do_ifelse): New function.
7065 (do_isel): New function.
7066 (do_sub3): New function.
7067 (do_add3): New function.
7068 (do_load_mask_compare): New function.
7069 (do_overlap_load_compare): New function.
7070 (expand_compare_loop): New function.
7071 (expand_block_compare): Call expand_compare_loop() when appropriate.
7072 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7074 (-mblock-compare-inline-loop-limit): New option.
7076 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7079 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7080 Reverse order of second and third operands in first alternative.
7081 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7082 of first and second elements in UNSPEC_VPERMR vector.
7083 (altivec_expand_vec_perm_le): Likewise.
7085 2018-01-08 Jeff Law <law@redhat.com>
7087 PR rtl-optimizatin/81308
7088 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7089 (process_switch): If group_case_labels makes a change, then set
7091 (pass_convert_switch::execute): If a switch is converted, then
7092 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7094 PR rtl-optimization/81308
7095 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7098 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7100 PR target/83663 - Revert r255946
7101 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7102 generation for cases where splatting a value is not useful.
7103 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7104 across a vec_duplicate and a paradoxical subreg forming a vector
7105 mode to a vec_concat.
7107 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7109 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7110 -march=armv8.3-a variants.
7111 * config/arm/t-multilib: Likewise.
7112 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7114 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7116 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7118 (cceq_ior_compare_complement): Give it a name so I can use it, and
7119 change boolean_or_operator predicate to boolean_operator so it can
7120 be used to generate a crand.
7121 (eqne): New code iterator.
7122 (bd/bd_neg): New code_attrs.
7123 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7124 a single define_insn.
7125 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7126 decrement (bdnzt/bdnzf/bdzt/bdzf).
7127 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7128 with the new names of the branch decrement patterns, and added the
7129 names of the branch decrement conditional patterns.
7131 2018-01-08 Richard Biener <rguenther@suse.de>
7133 PR tree-optimization/83563
7134 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7137 2018-01-08 Richard Biener <rguenther@suse.de>
7140 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7142 2018-01-08 Richard Biener <rguenther@suse.de>
7144 PR tree-optimization/83685
7145 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7146 references to abnormals.
7148 2018-01-08 Richard Biener <rguenther@suse.de>
7151 * dwarf2out.c (output_indirect_strings): Handle empty
7152 skeleton_debug_str_hash.
7153 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7155 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7157 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7158 (emit_store_direct): Likewise.
7159 (arc_trampoline_adjust_address): Likewise.
7160 (arc_asm_trampoline_template): New function.
7161 (arc_initialize_trampoline): Use asm_trampoline_template.
7162 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7163 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7164 * config/arc/arc.md (flush_icache): Delete pattern.
7166 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7168 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7169 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7172 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7175 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7176 by not USED_FOR_TARGET.
7177 (make_pass_resolve_sw_modes): Likewise.
7179 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7181 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7184 2018-01-08 Richard Biener <rguenther@suse.de>
7187 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7189 2018-01-08 Richard Biener <rguenther@suse.de>
7192 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7194 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7197 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7198 basic blocks with a small number of successors.
7199 (convert_control_dep_chain_into_preds): Improve handling of
7201 (dump_predicates): Split apart into...
7202 (dump_pred_chain): ...here...
7203 (dump_pred_info): ...and here.
7204 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7205 (can_chain_union_be_invalidated_p): Improve check for invalidation
7207 (uninit_uses_cannot_happen): Avoid unnecessary if
7208 convert_control_dep_chain_into_preds yielded nothing.
7210 2018-01-06 Martin Sebor <msebor@redhat.com>
7212 PR tree-optimization/83640
7213 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7214 subtracting negative offset from size.
7215 (builtin_access::overlap): Adjust offset bounds of the access to fall
7216 within the size of the object if possible.
7218 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7220 PR rtl-optimization/83699
7221 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7222 extract_bit_field_as_subreg to cases in which the extracted
7223 value is also a vector.
7225 * lra-constraints.c (process_alt_operands): Test for the equivalence
7226 substitutions when detecting a possible reload cycle.
7228 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7231 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7232 by default if flag_selective_schedling{,2}. Formatting fixes.
7234 PR rtl-optimization/83682
7235 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7236 if it has non-VECTOR_MODE element mode.
7237 (vec_duplicate_p): Likewise.
7240 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7241 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7243 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7246 * config/i386/i386-builtin.def
7247 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7248 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7249 Require also OPTION_MASK_ISA_AVX512F in addition to
7250 OPTION_MASK_ISA_GFNI.
7251 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7252 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7253 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7254 to OPTION_MASK_ISA_GFNI.
7255 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7256 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7257 OPTION_MASK_ISA_AVX512BW.
7258 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7259 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7260 addition to OPTION_MASK_ISA_GFNI.
7261 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7262 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7263 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7264 to OPTION_MASK_ISA_GFNI.
7265 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7266 a requirement for all ISAs rather than any of them with a few
7268 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7270 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7271 bitmasks to be enabled with 3 exceptions, instead of requiring any
7272 enabled ISA with lots of exceptions.
7273 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7274 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7275 Change avx512bw in isa attribute to avx512f.
7276 * config/i386/sgxintrin.h: Add license boilerplate.
7277 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7278 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7279 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7280 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7282 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7283 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7284 temporarily sse2 rather than sse if not enabled already.
7287 * config/i386/sse.md (VI248_VLBW): Rename to ...
7288 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7289 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7290 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7291 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7292 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7293 mode iterator instead of VI248_VLBW.
7295 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7297 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7298 (record_modified): Skip clobbers; add debug output.
7299 (param_change_prob): Use sreal frequencies.
7301 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7303 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7304 punt for user-aligned variables.
7306 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7308 * tree-chrec.c (chrec_contains_symbols): Return true for
7311 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7314 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7315 of (x|y) == x for BICS pattern.
7317 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7319 PR tree-optimization/83605
7320 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7321 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7324 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7326 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7327 * config/epiphany/rtems.h: New file.
7329 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7330 Uros Bizjak <ubizjak@gmail.com>
7333 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7334 QIreg_operand instead of register_operand predicate.
7335 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7336 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7337 comments instead of -fmitigate[-_]rop.
7339 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7342 * cgraphunit.c (symbol_table::compile): Switch to text_section
7343 before calling assembly_start debug hook.
7344 * run-rtl-passes.c (run_rtl_passes): Likewise.
7347 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7349 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7350 range_int_cst_p rather than !symbolic_range_p before calling
7351 extract_range_from_multiplicative_op_1.
7353 2018-01-04 Jeff Law <law@redhat.com>
7355 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7356 redundant test in assertion.
7358 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7360 * doc/rtl.texi: Document machine_mode wrapper classes.
7362 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7364 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7367 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7369 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7370 the VEC_PERM_EXPR fold to fail.
7372 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7375 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7376 to switched_sections.
7378 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7381 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7384 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7387 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7388 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7390 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7393 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7394 is BLKmode and bitpos not zero or mode change is needed.
7396 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7399 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7402 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7405 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7406 instead of MULT rtx. Update all corresponding splitters.
7408 (*ssub<modesuffix>): Ditto.
7410 (*cmp_sadd_di): Update split patterns.
7411 (*cmp_sadd_si): Ditto.
7412 (*cmp_sadd_sidi): Ditto.
7413 (*cmp_ssub_di): Ditto.
7414 (*cmp_ssub_si): Ditto.
7415 (*cmp_ssub_sidi): Ditto.
7416 * config/alpha/predicates.md (const23_operand): New predicate.
7417 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7418 Look for ASHIFT, not MULT inner operand.
7419 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7421 2018-01-04 Martin Liska <mliska@suse.cz>
7423 PR gcov-profile/83669
7424 * gcov.c (output_intermediate_file): Add version to intermediate
7426 * doc/gcov.texi: Document new field 'version' in intermediate
7427 file format. Fix location of '-k' option of gcov command.
7429 2018-01-04 Martin Liska <mliska@suse.cz>
7432 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7434 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7436 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7438 2018-01-03 Martin Sebor <msebor@redhat.com>
7440 PR tree-optimization/83655
7441 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7442 checking calls with invalid arguments.
7444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7446 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7447 (vectorizable_mask_load_store): Delete.
7448 (vectorizable_call): Return false for masked loads and stores.
7449 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7450 instead of gimple_assign_rhs1.
7451 (vectorizable_load): Handle IFN_MASK_LOAD.
7452 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7456 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7458 (vectorizable_mask_load_store): ...here.
7459 (vectorizable_load): ...and here.
7461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7463 * tree-vect-stmts.c (vect_build_all_ones_mask)
7464 (vect_build_zero_merge_argument): New functions, split out from...
7465 (vectorizable_load): ...here.
7467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7469 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7471 (vectorizable_mask_load_store): ...here.
7472 (vectorizable_store): ...and here.
7474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7476 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7478 (vectorizable_mask_load_store): ...here.
7480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7482 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7483 (vect_model_store_cost): Take a vec_load_store_type instead of a
7485 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7486 (vect_model_store_cost): Take a vec_load_store_type instead of a
7488 (vectorizable_mask_load_store): Update accordingly.
7489 (vectorizable_store): Likewise.
7490 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7495 IFN_MASK_LOAD calls here rather than...
7496 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7499 Alan Hayward <alan.hayward@arm.com>
7500 David Sherwood <david.sherwood@arm.com>
7502 * expmed.c (extract_bit_field_1): For vector extracts,
7503 fall back to extract_bit_field_as_subreg if vec_extract
7506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7507 Alan Hayward <alan.hayward@arm.com>
7508 David Sherwood <david.sherwood@arm.com>
7510 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7511 they are variable or constant sized.
7512 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7513 slots for constant-sized data.
7515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7516 Alan Hayward <alan.hayward@arm.com>
7517 David Sherwood <david.sherwood@arm.com>
7519 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7520 handling COND_EXPRs with boolean comparisons, try to find a better
7521 basis for the mask type than the boolean itself.
7523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7525 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7526 is calculated and how it can be overridden.
7527 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7528 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7530 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7534 Alan Hayward <alan.hayward@arm.com>
7535 David Sherwood <david.sherwood@arm.com>
7537 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7538 Remove the mode argument.
7539 (aarch64_simd_valid_immediate): Remove the mode and inverse
7541 * config/aarch64/iterators.md (bitsize): New iterator.
7542 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7543 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7544 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7545 aarch64_simd_valid_immediate.
7546 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7547 (aarch64_reg_or_bic_imm): Likewise.
7548 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7549 with an insn_type enum and msl with a modifier_type enum.
7550 Replace element_width with a scalar_mode. Change the shift
7551 to unsigned int. Add constructors for scalar_float_mode and
7552 scalar_int_mode elements.
7553 (aarch64_vect_float_const_representable_p): Delete.
7554 (aarch64_can_const_movi_rtx_p)
7555 (aarch64_simd_scalar_immediate_valid_for_move)
7556 (aarch64_simd_make_constant): Update call to
7557 aarch64_simd_valid_immediate.
7558 (aarch64_advsimd_valid_immediate_hs): New function.
7559 (aarch64_advsimd_valid_immediate): Likewise.
7560 (aarch64_simd_valid_immediate): Remove mode and inverse
7561 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7562 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7563 and aarch64_float_const_representable_p on the result.
7564 (aarch64_output_simd_mov_immediate): Remove mode argument.
7565 Update call to aarch64_simd_valid_immediate and use of
7566 simd_immediate_info.
7567 (aarch64_output_scalar_simd_mov_immediate): Update call
7570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7571 Alan Hayward <alan.hayward@arm.com>
7572 David Sherwood <david.sherwood@arm.com>
7574 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7575 (mode_nunits): Likewise CONST_MODE_NUNITS.
7576 * machmode.def (ADJUST_NUNITS): Document.
7577 * genmodes.c (mode_data::need_nunits_adj): New field.
7578 (blank_mode): Update accordingly.
7579 (adj_nunits): New variable.
7580 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7582 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7583 listed in adj_nunits.
7584 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7585 listed in adj_nunits. Don't emit case statements for such modes.
7586 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7587 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7588 nothing if adj_nunits is nonnull.
7589 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7590 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7591 (emit_mode_fbit): Update use of print_maybe_const_decl.
7592 (emit_move_size): Likewise. Treat the array as non-const
7594 (emit_mode_adjustments): Handle adj_nunits.
7596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7598 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7599 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7600 (VECTOR_MODES): Use it.
7601 (make_vector_modes): Take the prefix as an argument.
7603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7604 Alan Hayward <alan.hayward@arm.com>
7605 David Sherwood <david.sherwood@arm.com>
7607 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7608 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7609 for MODE_VECTOR_BOOL.
7610 * machmode.def (VECTOR_BOOL_MODE): Document.
7611 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7612 (make_vector_bool_mode): New function.
7613 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7615 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7616 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7618 * stor-layout.c (int_mode_for_mode): Likewise.
7619 * tree.c (build_vector_type_for_mode): Likewise.
7620 * varasm.c (output_constant_pool_2): Likewise.
7621 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7622 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7623 for MODE_VECTOR_BOOL.
7624 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7625 of mode class checks.
7626 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7627 instead of a list of mode class checks.
7628 (expand_vector_scalar_condition): Likewise.
7629 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7632 Alan Hayward <alan.hayward@arm.com>
7633 David Sherwood <david.sherwood@arm.com>
7635 * machmode.h (mode_size): Change from unsigned short to
7637 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7638 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7639 or if measurement_type is not polynomial.
7640 (fixed_size_mode::includes_p): Check for constant-sized modes.
7641 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7642 return a poly_uint16 rather than an unsigned short.
7643 (emit_mode_size): Change the type of mode_size from unsigned short
7644 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7645 (emit_mode_adjustments): Cope with polynomial vector sizes.
7646 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7648 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7650 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7651 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7652 * caller-save.c (setup_save_areas): Likewise.
7653 (replace_reg_with_saved_mem): Likewise.
7654 * calls.c (emit_library_call_value_1): Likewise.
7655 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7656 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7657 (gen_lowpart_for_combine): Likewise.
7658 * convert.c (convert_to_integer_1): Likewise.
7659 * cse.c (equiv_constant, cse_insn): Likewise.
7660 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7661 (cselib_subst_to_values): Likewise.
7662 * dce.c (word_dce_process_block): Likewise.
7663 * df-problems.c (df_word_lr_mark_ref): Likewise.
7664 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7665 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7666 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7667 (rtl_for_decl_location): Likewise.
7668 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7669 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7670 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7671 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7672 (expand_expr_real_1): Likewise.
7673 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7674 (pad_below): Likewise.
7675 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7676 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7677 * ira.c (get_subreg_tracking_sizes): Likewise.
7678 * ira-build.c (ira_create_allocno_objects): Likewise.
7679 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7680 (ira_sort_regnos_for_alter_reg): Likewise.
7681 * ira-costs.c (record_operand_costs): Likewise.
7682 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7683 (resolve_simple_move): Likewise.
7684 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7685 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7686 (lra_constraints): Likewise.
7687 (CONST_POOL_OK_P): Reject variable-sized modes.
7688 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7689 (add_pseudo_to_slot, lra_spill): Likewise.
7690 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7691 * optabs-query.c (get_best_extraction_insn): Likewise.
7692 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7693 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7694 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7695 * recog.c (offsettable_address_addr_space_p): Likewise.
7696 * regcprop.c (maybe_mode_change): Likewise.
7697 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7698 * regrename.c (build_def_use): Likewise.
7699 * regstat.c (dump_reg_info): Likewise.
7700 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7701 (find_reloads, find_reloads_subreg_address): Likewise.
7702 * reload1.c (eliminate_regs_1): Likewise.
7703 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7704 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7705 (simplify_binary_operation_1, simplify_subreg): Likewise.
7706 * targhooks.c (default_function_arg_padding): Likewise.
7707 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7708 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7709 (verify_gimple_assign_ternary): Likewise.
7710 * tree-inline.c (estimate_move_cost): Likewise.
7711 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7712 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7713 (get_address_cost_ainc): Likewise.
7714 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7715 (vect_supportable_dr_alignment): Likewise.
7716 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7717 (vectorizable_reduction): Likewise.
7718 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7719 (vectorizable_operation, vectorizable_load): Likewise.
7720 * tree.c (build_same_sized_truth_vector_type): Likewise.
7721 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7722 * var-tracking.c (emit_note_insn_var_location): Likewise.
7723 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7724 (ADDR_VEC_ALIGN): Likewise.
7726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7727 Alan Hayward <alan.hayward@arm.com>
7728 David Sherwood <david.sherwood@arm.com>
7730 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7732 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7733 or if measurement_type is polynomial.
7734 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7735 * combine.c (make_extraction): Likewise.
7736 * dse.c (find_shift_sequence): Likewise.
7737 * dwarf2out.c (mem_loc_descriptor): Likewise.
7738 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7739 (extract_bit_field, extract_low_bits): Likewise.
7740 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7741 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7742 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7743 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7744 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7745 * reload.c (find_reloads): Likewise.
7746 * reload1.c (alter_reg): Likewise.
7747 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7748 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7749 * tree-if-conv.c (predicate_mem_writes): Likewise.
7750 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7751 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7752 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7753 * valtrack.c (dead_debug_insert_temp): Likewise.
7754 * varasm.c (mergeable_constant_section): Likewise.
7755 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7758 Alan Hayward <alan.hayward@arm.com>
7759 David Sherwood <david.sherwood@arm.com>
7761 * expr.c (expand_assignment): Cope with polynomial mode sizes
7762 when assigning to a CONCAT.
7764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7765 Alan Hayward <alan.hayward@arm.com>
7766 David Sherwood <david.sherwood@arm.com>
7768 * machmode.h (mode_precision): Change from unsigned short to
7770 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7772 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7773 or if measurement_type is not polynomial.
7774 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7775 in which the mode is already known to be a scalar_int_mode.
7776 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7777 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7779 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7780 for GET_MODE_PRECISION.
7781 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7782 for GET_MODE_PRECISION.
7783 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7785 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7786 (expand_field_assignment, make_extraction): Likewise.
7787 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7788 (get_last_value): Likewise.
7789 * convert.c (convert_to_integer_1): Likewise.
7790 * cse.c (cse_insn): Likewise.
7791 * expr.c (expand_expr_real_1): Likewise.
7792 * lra-constraints.c (simplify_operand_subreg): Likewise.
7793 * optabs-query.c (can_atomic_load_p): Likewise.
7794 * optabs.c (expand_atomic_load): Likewise.
7795 (expand_atomic_store): Likewise.
7796 * ree.c (combine_reaching_defs): Likewise.
7797 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7798 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7799 * tree.h (type_has_mode_precision_p): Likewise.
7800 * ubsan.c (instrument_si_overflow): Likewise.
7802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7803 Alan Hayward <alan.hayward@arm.com>
7804 David Sherwood <david.sherwood@arm.com>
7806 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7807 polynomial numbers of units.
7808 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7809 (valid_vector_subparts_p): New function.
7810 (build_vector_type): Remove temporary shim and take the number
7811 of units as a poly_uint64 rather than an int.
7812 (build_opaque_vector_type): Take the number of units as a
7813 poly_uint64 rather than an int.
7814 * tree.c (build_vector_from_ctor): Handle polynomial
7815 TYPE_VECTOR_SUBPARTS.
7816 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7817 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7818 (build_vector_from_val): If the number of units is variable,
7819 use build_vec_duplicate_cst for constant operands and
7820 VEC_DUPLICATE_EXPR otherwise.
7821 (make_vector_type): Remove temporary is_constant ().
7822 (build_vector_type, build_opaque_vector_type): Take the number of
7823 units as a poly_uint64 rather than an int.
7824 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7826 * cfgexpand.c (expand_debug_expr): Likewise.
7827 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7828 (store_constructor, expand_expr_real_1): Likewise.
7829 (const_scalar_mask_from_tree): Likewise.
7830 * fold-const-call.c (fold_const_reduction): Likewise.
7831 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7832 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7833 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7834 (fold_relational_const): Likewise.
7835 (native_interpret_vector): Likewise. Change the size from an
7836 int to an unsigned int.
7837 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7838 TYPE_VECTOR_SUBPARTS.
7839 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7840 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7841 duplicating a non-constant operand into a variable-length vector.
7842 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7843 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7844 * ipa-icf.c (sem_variable::equals): Likewise.
7845 * match.pd: Likewise.
7846 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7847 * print-tree.c (print_node): Likewise.
7848 * stor-layout.c (layout_type): Likewise.
7849 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7850 * tree-cfg.c (verify_gimple_comparison): Likewise.
7851 (verify_gimple_assign_binary): Likewise.
7852 (verify_gimple_assign_ternary): Likewise.
7853 (verify_gimple_assign_single): Likewise.
7854 * tree-pretty-print.c (dump_generic_node): Likewise.
7855 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7856 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7857 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7858 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7859 (vect_shift_permute_load_chain): Likewise.
7860 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7861 (expand_vector_condition, optimize_vector_constructor): Likewise.
7862 (lower_vec_perm, get_compute_type): Likewise.
7863 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7864 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7865 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7866 (vect_recog_mask_conversion_pattern): Likewise.
7867 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7868 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7869 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7870 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7871 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7872 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7873 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7874 (supportable_widening_operation): Likewise.
7875 (supportable_narrowing_operation): Likewise.
7876 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7878 * varasm.c (output_constant): Likewise.
7880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7881 Alan Hayward <alan.hayward@arm.com>
7882 David Sherwood <david.sherwood@arm.com>
7884 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7885 so that both the length == 3 and length != 3 cases set up their
7886 own permute vectors. Add comments explaining why we know the
7887 number of elements is constant.
7888 (vect_permute_load_chain): Likewise.
7890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7891 Alan Hayward <alan.hayward@arm.com>
7892 David Sherwood <david.sherwood@arm.com>
7894 * machmode.h (mode_nunits): Change from unsigned char to
7896 (ONLY_FIXED_SIZE_MODES): New macro.
7897 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7898 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7899 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7901 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7902 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7903 or if measurement_type is not polynomial.
7904 * genmodes.c (ZERO_COEFFS): New macro.
7905 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7907 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7908 Use ZERO_COEFFS when emitting initializers.
7909 * data-streamer.h (bp_pack_poly_value): New function.
7910 (bp_unpack_poly_value): Likewise.
7911 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7912 for GET_MODE_NUNITS.
7913 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7914 for GET_MODE_NUNITS.
7915 * tree.c (make_vector_type): Remove temporary shim and make
7916 the real function take the number of units as a poly_uint64
7918 (build_vector_type_for_mode): Handle polynomial nunits.
7919 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7920 * emit-rtl.c (const_vec_series_p_1): Likewise.
7921 (gen_rtx_CONST_VECTOR): Likewise.
7922 * fold-const.c (test_vec_duplicate_folding): Likewise.
7923 * genrecog.c (validate_pattern): Likewise.
7924 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7925 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7926 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7927 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7928 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7929 * rtlanal.c (subreg_get_info): Likewise.
7930 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7931 (vect_grouped_load_supported): Likewise.
7932 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7933 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7934 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7935 (simplify_const_unary_operation, simplify_binary_operation_1)
7936 (simplify_const_binary_operation, simplify_ternary_operation)
7937 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7938 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7939 instead of CONST_VECTOR_NUNITS.
7940 * varasm.c (output_constant_pool_2): Likewise.
7941 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7942 explicit-encoded elements in the XVEC for variable-length vectors.
7944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7946 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7949 Alan Hayward <alan.hayward@arm.com>
7950 David Sherwood <david.sherwood@arm.com>
7952 * coretypes.h (fixed_size_mode): Declare.
7953 (fixed_size_mode_pod): New typedef.
7954 * builtins.h (target_builtins::x_apply_args_mode)
7955 (target_builtins::x_apply_result_mode): Change type to
7956 fixed_size_mode_pod.
7957 * builtins.c (apply_args_size, apply_result_size, result_vector)
7958 (expand_builtin_apply_args_1, expand_builtin_apply)
7959 (expand_builtin_return): Update accordingly.
7961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7963 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7964 * cselib.c (cselib_hash_rtx): Likewise.
7965 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7966 CONST_VECTOR encoding.
7968 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7969 Jeff Law <law@redhat.com>
7972 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7973 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7974 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7975 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7978 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7979 explicitly probe *sp in a noreturn function if there were any callee
7980 register saves or frame pointer is needed.
7982 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7985 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7986 BLKmode for ternary, binary or unary expressions.
7989 * var-tracking.c (delete_vta_debug_insn): New inline function.
7990 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7991 insns from get_insns () to NULL instead of each bb separately.
7992 Use delete_vta_debug_insn. No longer static.
7993 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7994 delete_vta_debug_insns callers.
7995 * rtl.h (delete_vta_debug_insns): Declare.
7996 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7997 instead of variable_tracking_main.
7999 2018-01-03 Martin Sebor <msebor@redhat.com>
8001 PR tree-optimization/83603
8002 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8003 arguments past the endof the argument list in functions declared
8004 without a prototype.
8005 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8006 Avoid checking when arguments are null.
8008 2018-01-03 Martin Sebor <msebor@redhat.com>
8011 * doc/extend.texi (attribute const): Fix a typo.
8012 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8013 issuing -Wsuggest-attribute for void functions.
8015 2018-01-03 Martin Sebor <msebor@redhat.com>
8017 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8018 offset_int::from instead of wide_int::to_shwi.
8019 (maybe_diag_overlap): Remove assertion.
8020 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8021 * gimple-ssa-sprintf.c (format_directive): Same.
8022 (parse_directive): Same.
8023 (sprintf_dom_walker::compute_format_length): Same.
8024 (try_substitute_return_value): Same.
8026 2018-01-03 Jeff Law <law@redhat.com>
8029 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8030 non-constant residual for zero at runtime and avoid probing in
8031 that case. Reorganize code for trailing problem to mirror handling
8034 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8036 PR tree-optimization/83501
8037 * tree-ssa-strlen.c (get_string_cst): New.
8038 (handle_char_store): Call get_string_cst.
8040 2018-01-03 Martin Liska <mliska@suse.cz>
8042 PR tree-optimization/83593
8043 * tree-ssa-strlen.c: Include tree-cfg.h.
8044 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8045 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8046 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8048 (strlen_dom_walker::before_dom_children): Call
8049 gimple_purge_dead_eh_edges. Dump tranformation with details
8051 (strlen_dom_walker::before_dom_children): Update call by adding
8052 new argument cleanup_eh.
8053 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8055 2018-01-03 Martin Liska <mliska@suse.cz>
8058 * cif-code.def (VARIADIC_THUNK): New enum value.
8059 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8062 2018-01-03 Jan Beulich <jbeulich@suse.com>
8064 * sse.md (mov<mode>_internal): Tighten condition for when to use
8065 vmovdqu<ssescalarsize> for TI and OI modes.
8067 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8069 Update copyright years.
8071 2018-01-03 Martin Liska <mliska@suse.cz>
8074 * ipa-visibility.c (function_and_variable_visibility): Skip
8075 functions with noipa attribure.
8077 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8079 * gcc.c (process_command): Update copyright notice dates.
8080 * gcov-dump.c (print_version): Ditto.
8081 * gcov.c (print_version): Ditto.
8082 * gcov-tool.c (print_version): Ditto.
8083 * gengtype.c (create_file): Ditto.
8084 * doc/cpp.texi: Bump @copying's copyright year.
8085 * doc/cppinternals.texi: Ditto.
8086 * doc/gcc.texi: Ditto.
8087 * doc/gccint.texi: Ditto.
8088 * doc/gcov.texi: Ditto.
8089 * doc/install.texi: Ditto.
8090 * doc/invoke.texi: Ditto.
8092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8094 * vector-builder.h (vector_builder::m_full_nelts): Change from
8095 unsigned int to poly_uint64.
8096 (vector_builder::full_nelts): Update prototype accordingly.
8097 (vector_builder::new_vector): Likewise.
8098 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8099 (vector_builder::operator ==): Likewise.
8100 (vector_builder::finalize): Likewise.
8101 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8102 Take the number of elements as a poly_uint64 rather than an
8104 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8105 from unsigned int to poly_uint64.
8106 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8107 (vec_perm_indices::new_vector): Likewise.
8108 (vec_perm_indices::length): Likewise.
8109 (vec_perm_indices::nelts_per_input): Likewise.
8110 (vec_perm_indices::input_nelts): Likewise.
8111 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8112 number of elements per input as a poly_uint64 rather than an
8113 unsigned int. Use the original encoding for variable-length
8114 vectors, rather than clamping each individual element.
8115 For the second and subsequent elements in each pattern,
8116 clamp the step and base before clamping their sum.
8117 (vec_perm_indices::series_p): Handle polynomial element counts.
8118 (vec_perm_indices::all_in_range_p): Likewise.
8119 (vec_perm_indices_to_tree): Likewise.
8120 (vec_perm_indices_to_rtx): Likewise.
8121 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8122 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8123 (tree_vector_builder::new_binary_operation): Handle polynomial
8124 element counts. Return false if we need to know the number
8125 of elements at compile time.
8126 * fold-const.c (fold_vec_perm): Punt if the number of elements
8127 isn't known at compile time.
8129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8131 * vec-perm-indices.h (vec_perm_builder): Change element type
8132 from HOST_WIDE_INT to poly_int64.
8133 (vec_perm_indices::element_type): Update accordingly.
8134 (vec_perm_indices::clamp): Handle polynomial element_types.
8135 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8136 (vec_perm_indices::all_in_range_p): Likewise.
8137 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8139 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8140 polynomial vec_perm_indices element types.
8141 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8142 * fold-const.c (fold_vec_perm): Likewise.
8143 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8144 * tree-vect-generic.c (lower_vec_perm): Likewise.
8145 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8146 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8147 element type to HOST_WIDE_INT.
8149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8150 Alan Hayward <alan.hayward@arm.com>
8151 David Sherwood <david.sherwood@arm.com>
8153 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8154 rather than an int. Use plus_constant.
8155 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8156 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8159 Alan Hayward <alan.hayward@arm.com>
8160 David Sherwood <david.sherwood@arm.com>
8162 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8163 a HOST_WIDE_INT to a poly_int64.
8165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8166 Alan Hayward <alan.hayward@arm.com>
8167 David Sherwood <david.sherwood@arm.com>
8169 * calls.c (load_register_parameters): Cope with polynomial
8170 mode sizes. Require a constant size for BLKmode parameters
8171 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8172 forces a parameter to be padded at the lsb end in order to
8173 fill a complete number of words, require the parameter size
8174 to be ordered wrt UNITS_PER_WORD.
8176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8177 Alan Hayward <alan.hayward@arm.com>
8178 David Sherwood <david.sherwood@arm.com>
8180 * reload1.c (spill_stack_slot_width): Change element type
8181 from unsigned int to poly_uint64_pod.
8182 (alter_reg): Treat mode sizes as polynomial.
8184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8185 Alan Hayward <alan.hayward@arm.com>
8186 David Sherwood <david.sherwood@arm.com>
8188 * reload.c (complex_word_subreg_p): New function.
8189 (reload_inner_reg_of_subreg, push_reload): Use it.
8191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8192 Alan Hayward <alan.hayward@arm.com>
8193 David Sherwood <david.sherwood@arm.com>
8195 * lra-constraints.c (process_alt_operands): Reject matched
8196 operands whose sizes aren't ordered.
8197 (match_reload): Refer to this check here.
8199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8200 Alan Hayward <alan.hayward@arm.com>
8201 David Sherwood <david.sherwood@arm.com>
8203 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8204 that the mode size is in the set {1, 2, 4, 8, 16}.
8206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8207 Alan Hayward <alan.hayward@arm.com>
8208 David Sherwood <david.sherwood@arm.com>
8210 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8211 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8218 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8219 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8220 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8221 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8222 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8223 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8224 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8225 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8226 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8228 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8229 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8230 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8231 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8232 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8233 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8234 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8235 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8236 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8237 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8239 * expr.c (emit_move_resolve_push): Treat the input and result
8240 of PUSH_ROUNDING as a poly_int64.
8241 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8242 (emit_push_insn): Likewise.
8243 * lra-eliminations.c (mark_not_eliminable): Likewise.
8244 * recog.c (push_operand): Likewise.
8245 * reload1.c (elimination_effects): Likewise.
8246 * rtlanal.c (nonzero_bits1): Likewise.
8247 * calls.c (store_one_arg): Likewise. Require the padding to be
8248 known at compile time.
8250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8251 Alan Hayward <alan.hayward@arm.com>
8252 David Sherwood <david.sherwood@arm.com>
8254 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8255 Use plus_constant instead of gen_rtx_PLUS.
8257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8258 Alan Hayward <alan.hayward@arm.com>
8259 David Sherwood <david.sherwood@arm.com>
8261 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
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 (expand_expr_real_1): Use tree_to_poly_uint64
8269 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8270 via stack temporaries. Treat the mode size as polynomial too.
8272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8273 Alan Hayward <alan.hayward@arm.com>
8274 David Sherwood <david.sherwood@arm.com>
8276 * expr.c (expand_expr_real_2): When handling conversions involving
8277 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8278 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8279 as a poly_uint64 too.
8281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8282 Alan Hayward <alan.hayward@arm.com>
8283 David Sherwood <david.sherwood@arm.com>
8285 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8288 Alan Hayward <alan.hayward@arm.com>
8289 David Sherwood <david.sherwood@arm.com>
8291 * combine.c (can_change_dest_mode): Handle polynomial
8292 REGMODE_NATURAL_SIZE.
8293 * expmed.c (store_bit_field_1): Likewise.
8294 * expr.c (store_constructor): Likewise.
8295 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8296 and polynomial REGMODE_NATURAL_SIZE.
8297 (gen_lowpart_common): Likewise.
8298 * reginfo.c (record_subregs_of_mode): Likewise.
8299 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8306 numbers of elements.
8308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8309 Alan Hayward <alan.hayward@arm.com>
8310 David Sherwood <david.sherwood@arm.com>
8312 * match.pd: Cope with polynomial numbers of vector elements.
8314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8315 Alan Hayward <alan.hayward@arm.com>
8316 David Sherwood <david.sherwood@arm.com>
8318 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8319 in a POINTER_PLUS_EXPR.
8321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8322 Alan Hayward <alan.hayward@arm.com>
8323 David Sherwood <david.sherwood@arm.com>
8325 * omp-simd-clone.c (simd_clone_subparts): New function.
8326 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8327 (ipa_simd_modify_function_body): Likewise.
8329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8330 Alan Hayward <alan.hayward@arm.com>
8331 David Sherwood <david.sherwood@arm.com>
8333 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8334 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8335 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8336 (expand_vector_condition, vector_element): Likewise.
8337 (subparts_gt): New function.
8338 (get_compute_type): Use subparts_gt.
8339 (count_type_subparts): Delete.
8340 (expand_vector_operations_1): Use subparts_gt instead of
8341 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
8348 (vect_compile_time_alias): ...this new function. Do the calculation
8349 on poly_ints rather than trees.
8350 (vect_prune_runtime_alias_test_list): Update call accordingly.
8352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8353 Alan Hayward <alan.hayward@arm.com>
8354 David Sherwood <david.sherwood@arm.com>
8356 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8358 (vect_schedule_slp_instance): Likewise.
8360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8361 Alan Hayward <alan.hayward@arm.com>
8362 David Sherwood <david.sherwood@arm.com>
8364 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8365 constant and extern definitions for variable-length vectors.
8366 (vect_get_constant_vectors): Note that the number of units
8367 is known to be constant.
8369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8370 Alan Hayward <alan.hayward@arm.com>
8371 David Sherwood <david.sherwood@arm.com>
8373 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8374 of units as polynomial. Choose between WIDE and NARROW based
8377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8378 Alan Hayward <alan.hayward@arm.com>
8379 David Sherwood <david.sherwood@arm.com>
8381 * tree-vect-stmts.c (simd_clone_subparts): New function.
8382 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8385 Alan Hayward <alan.hayward@arm.com>
8386 David Sherwood <david.sherwood@arm.com>
8388 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8389 vectors as polynomial. Use build_index_vector for
8392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8393 Alan Hayward <alan.hayward@arm.com>
8394 David Sherwood <david.sherwood@arm.com>
8396 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8397 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8398 for variable-length vectors.
8399 (vectorizable_mask_load_store): Treat the number of units as
8400 polynomial, asserting that it is constant if the condition has
8401 already been enforced.
8402 (vectorizable_store, vectorizable_load): Likewise.
8404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8405 Alan Hayward <alan.hayward@arm.com>
8406 David Sherwood <david.sherwood@arm.com>
8408 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8409 of units as polynomial. Punt if we can't tell at compile time
8410 which vector contains the final result.
8412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8413 Alan Hayward <alan.hayward@arm.com>
8414 David Sherwood <david.sherwood@arm.com>
8416 * tree-vect-loop.c (vectorizable_induction): Treat the number
8417 of units as polynomial. Punt on SLP inductions. Use an integer
8418 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8419 cast of such a series for variable-length floating-point
8422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8423 Alan Hayward <alan.hayward@arm.com>
8424 David Sherwood <david.sherwood@arm.com>
8426 * tree.h (build_index_vector): Declare.
8427 * tree.c (build_index_vector): New function.
8428 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8429 of units as polynomial, forcibly converting it to a constant if
8430 vectorizable_reduction has already enforced the condition.
8431 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8432 to create a {1,2,3,...} vector.
8433 (vectorizable_reduction): Treat the number of units as polynomial.
8434 Choose vectype_in based on the largest scalar element size rather
8435 than the smallest number of units. Enforce the restrictions
8438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8439 Alan Hayward <alan.hayward@arm.com>
8440 David Sherwood <david.sherwood@arm.com>
8442 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8443 number of units as polynomial.
8445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8446 Alan Hayward <alan.hayward@arm.com>
8447 David Sherwood <david.sherwood@arm.com>
8449 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8450 * target.def (autovectorize_vector_sizes): Return the vector sizes
8451 by pointer, using vector_sizes rather than a bitmask.
8452 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8453 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8454 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8456 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8457 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8458 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8459 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8460 * omp-general.c (omp_max_vf): Likewise.
8461 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8462 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8463 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8464 * tree-vect-slp.c (vect_slp_bb): Likewise.
8465 * doc/tm.texi: Regenerate.
8466 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8468 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8469 the vector size as a poly_uint64 rather than an unsigned int.
8470 (current_vector_size): Change from an unsigned int to a poly_uint64.
8471 (get_vectype_for_scalar_type): Update accordingly.
8472 * tree.h (build_truth_vector_type): Take the size and number of
8473 units as a poly_uint64 rather than an unsigned int.
8474 (build_vector_type): Add a temporary overload that takes
8475 the number of units as a poly_uint64 rather than an unsigned int.
8476 * tree.c (make_vector_type): Likewise.
8477 (build_truth_vector_type): Take the number of units as a poly_uint64
8478 rather than an unsigned int.
8480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8481 Alan Hayward <alan.hayward@arm.com>
8482 David Sherwood <david.sherwood@arm.com>
8484 * target.def (get_mask_mode): Take the number of units and length
8485 as poly_uint64s rather than unsigned ints.
8486 * targhooks.h (default_get_mask_mode): Update accordingly.
8487 * targhooks.c (default_get_mask_mode): Likewise.
8488 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8489 * doc/tm.texi: Regenerate.
8491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8492 Alan Hayward <alan.hayward@arm.com>
8493 David Sherwood <david.sherwood@arm.com>
8495 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8496 * omp-general.c (omp_max_vf): Likewise.
8497 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8498 (expand_omp_simd): Handle polynomial safelen.
8499 * omp-low.c (omplow_simd_context): Add a default constructor.
8500 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8501 (lower_rec_simd_input_clauses): Update accordingly.
8502 (lower_rec_input_clauses): Likewise.
8504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8505 Alan Hayward <alan.hayward@arm.com>
8506 David Sherwood <david.sherwood@arm.com>
8508 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8509 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8510 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8511 (vect_analyze_slp_cost): Likewise.
8512 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8513 (vect_model_load_cost): Likewise.
8515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8516 Alan Hayward <alan.hayward@arm.com>
8517 David Sherwood <david.sherwood@arm.com>
8519 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8520 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8521 from an unsigned int * to a poly_uint64_pod *.
8522 (calculate_unrolling_factor): New function.
8523 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8526 Alan Hayward <alan.hayward@arm.com>
8527 David Sherwood <david.sherwood@arm.com>
8529 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8530 from an unsigned int to a poly_uint64.
8531 (_loop_vec_info::slp_unrolling_factor): Likewise.
8532 (_loop_vec_info::vectorization_factor): Change from an int
8534 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8535 (vect_get_num_vectors): New function.
8536 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8537 (vect_get_num_copies): Use vect_get_num_vectors.
8538 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8539 to an unsigned int *.
8540 (vect_analyze_data_refs): Change min_vf from an int * to a
8542 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8543 than an unsigned HOST_WIDE_INT.
8544 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8545 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8546 to an unsigned int *.
8547 (vect_analyze_data_ref_dependences): Likewise.
8548 (vect_compute_data_ref_alignment): Handle polynomial vf.
8549 (vect_enhance_data_refs_alignment): Likewise.
8550 (vect_prune_runtime_alias_test_list): Likewise.
8551 (vect_shift_permute_load_chain): Likewise.
8552 (vect_supportable_dr_alignment): Likewise.
8553 (dependence_distance_ge_vf): Take the vectorization factor as a
8554 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8555 (vect_analyze_data_refs): Change min_vf from an int * to a
8557 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8558 vfm1 as a poly_uint64 rather than an int. Make the same change
8559 for the returned bound_scalar.
8560 (vect_gen_vector_loop_niters): Handle polynomial vf.
8561 (vect_do_peeling): Likewise. Update call to
8562 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8563 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8565 * tree-vect-loop.c (vect_determine_vectorization_factor)
8566 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8567 (vect_get_known_peeling_cost): Likewise.
8568 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8569 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8570 (vect_transform_loop): Likewise. Use the lowest possible VF when
8571 updating the upper bounds of the loop.
8572 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8574 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8575 polynomial unroll factors.
8576 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8577 (vect_make_slp_decision): Likewise.
8578 (vect_supported_load_permutation_p): Likewise, and polynomial
8580 (vect_analyze_slp_cost): Handle polynomial vf.
8581 (vect_slp_analyze_node_operations): Likewise.
8582 (vect_slp_analyze_bb_1): Likewise.
8583 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8584 than an unsigned HOST_WIDE_INT.
8585 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8586 (vectorizable_load): Handle polynomial vf.
8587 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8589 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8592 Alan Hayward <alan.hayward@arm.com>
8593 David Sherwood <david.sherwood@arm.com>
8595 * match.pd: Handle bit operations involving three constants
8596 and try to fold one pair.
8598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8600 * tree-vect-loop-manip.c: Include gimple-fold.h.
8601 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8602 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8603 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8604 Add a path that uses a step of VF instead of 1, but disable it
8606 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8607 and niters_no_overflow parameters. Update calls to
8608 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8609 Create a new SSA name if the latter choses to use a ste other
8610 than zero, and return it via niters_vector_mult_vf_var.
8611 * tree-vect-loop.c (vect_transform_loop): Update calls to
8612 vect_do_peeling, vect_gen_vector_loop_niters and
8613 slpeel_make_loop_iterate_ntimes.
8614 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8615 (vect_gen_vector_loop_niters): Update declarations after above changes.
8617 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8619 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8620 128-bit round to integer instructions.
8621 (ceil<mode>2): Likewise.
8622 (btrunc<mode>2): Likewise.
8623 (round<mode>2): Likewise.
8625 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8627 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8628 unaligned VSX load/store on P8/P9.
8629 (expand_block_clear): Allow the use of unaligned VSX
8630 load/store on P8/P9.
8632 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8634 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8636 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8637 swap associated with both a load and a store.
8639 2018-01-02 Andrew Waterman <andrew@sifive.com>
8641 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8642 * config/riscv/riscv.md (clear_cache): Use it.
8644 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8646 * web.c: Remove out-of-date comment.
8648 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8650 * expr.c (fixup_args_size_notes): Check that any existing
8651 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8652 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8653 (emit_single_push_insn): ...here.
8655 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8657 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8658 (const_vector_encoded_nelts): New function.
8659 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8660 (const_vector_int_elt, const_vector_elt): Declare.
8661 * emit-rtl.c (const_vector_int_elt_1): New function.
8662 (const_vector_elt): Likewise.
8663 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8664 of CONST_VECTOR_ELT.
8666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8668 * expr.c: Include rtx-vector-builder.h.
8669 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8670 directly on the tree encoding.
8671 (const_vector_from_tree): Likewise.
8672 * optabs.c: Include rtx-vector-builder.h.
8673 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8674 sequence of "u" values.
8675 * vec-perm-indices.c: Include rtx-vector-builder.h.
8676 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8677 directly on the vec_perm_indices encoding.
8679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8681 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8682 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8683 * rtx-vector-builder.h: New file.
8684 * rtx-vector-builder.c: Likewise.
8685 * rtl.h (rtx_def::u2): Add a const_vector field.
8686 (CONST_VECTOR_NPATTERNS): New macro.
8687 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8688 (CONST_VECTOR_DUPLICATE_P): Likewise.
8689 (CONST_VECTOR_STEPPED_P): Likewise.
8690 (CONST_VECTOR_ENCODED_ELT): Likewise.
8691 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8692 (unwrap_const_vec_duplicate): Likewise.
8693 (const_vec_series_p): Check for a non-duplicated vector encoding.
8694 Say that the function only returns true for integer vectors.
8695 * emit-rtl.c: Include rtx-vector-builder.h.
8696 (gen_const_vec_duplicate_1): Delete.
8697 (gen_const_vector): Call gen_const_vec_duplicate instead of
8698 gen_const_vec_duplicate_1.
8699 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8700 (gen_const_vec_duplicate): Use rtx_vector_builder.
8701 (gen_const_vec_series): Likewise.
8702 (gen_rtx_CONST_VECTOR): Likewise.
8703 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8704 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8705 Build a new vector rather than modifying a CONST_VECTOR in-place.
8706 (handle_special_swappables): Update call accordingly.
8707 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8708 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8709 Build a new vector rather than modifying a CONST_VECTOR in-place.
8710 (handle_special_swappables): Update call accordingly.
8712 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8714 * simplify-rtx.c (simplify_const_binary_operation): Use
8715 CONST_VECTOR_ELT instead of XVECEXP.
8717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8719 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8720 the selector elements to be different from the data elements
8721 if the selector is a VECTOR_CST.
8722 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8723 ssizetype for the selector.
8725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8727 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8728 before testing each element individually.
8729 * tree-vect-generic.c (lower_vec_perm): Likewise.
8731 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8733 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8734 * selftest-run-tests.c (selftest::run_tests): Call it.
8735 * vector-builder.h (vector_builder::operator ==): New function.
8736 (vector_builder::operator !=): Likewise.
8737 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8738 (vec_perm_indices::all_from_input_p): New function.
8739 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8740 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8741 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8742 instead of reading the VECTOR_CST directly. Detect whether both
8743 vector inputs are the same before constructing the vec_perm_indices,
8744 and update the number of inputs argument accordingly. Use the
8745 utility functions added above. Only construct sel2 if we need to.
8747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8749 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8750 the broadcast of the low byte.
8751 (expand_mult_highpart): Use an explicit encoding for the permutes.
8752 * optabs-query.c (can_mult_highpart_p): Likewise.
8753 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8754 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8755 (vectorizable_bswap): Likewise.
8756 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8757 explicit encoding for the power-of-2 permutes.
8758 (vect_permute_store_chain): Likewise.
8759 (vect_grouped_load_supported): Likewise.
8760 (vect_permute_load_chain): Likewise.
8762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8764 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8765 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8766 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8767 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8768 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8769 (vect_gen_perm_mask_any): Likewise.
8771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8773 * int-vector-builder.h: New file.
8774 * vec-perm-indices.h: Include int-vector-builder.h.
8775 (vec_perm_indices): Redefine as an int_vector_builder.
8776 (auto_vec_perm_indices): Delete.
8777 (vec_perm_builder): Redefine as a stand-alone class.
8778 (vec_perm_indices::vec_perm_indices): New function.
8779 (vec_perm_indices::clamp): Likewise.
8780 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8781 (vec_perm_indices::new_vector): New function.
8782 (vec_perm_indices::new_expanded_vector): Update for new
8783 vec_perm_indices class.
8784 (vec_perm_indices::rotate_inputs): New function.
8785 (vec_perm_indices::all_in_range_p): Operate directly on the
8786 encoded form, without computing elided elements.
8787 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8788 encoding. Update for new vec_perm_indices class.
8789 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8790 the given vec_perm_builder.
8791 (expand_vec_perm_var): Update vec_perm_builder constructor.
8792 (expand_mult_highpart): Use vec_perm_builder instead of
8793 auto_vec_perm_indices.
8794 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8795 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8796 or double series encoding as appropriate.
8797 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8798 vec_perm_indices instead of auto_vec_perm_indices.
8799 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8800 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8801 (vect_permute_store_chain): Likewise.
8802 (vect_grouped_load_supported): Likewise.
8803 (vect_permute_load_chain): Likewise.
8804 (vect_shift_permute_load_chain): Likewise.
8805 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8806 (vect_transform_slp_perm_load): Likewise.
8807 (vect_schedule_slp_instance): Likewise.
8808 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8809 (vectorizable_mask_load_store): Likewise.
8810 (vectorizable_bswap): Likewise.
8811 (vectorizable_store): Likewise.
8812 (vectorizable_load): Likewise.
8813 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8814 vec_perm_indices instead of auto_vec_perm_indices. Use
8815 tree_to_vec_perm_builder to read the vector from a tree.
8816 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8817 vec_perm_builder instead of a vec_perm_indices.
8818 (have_whole_vector_shift): Use vec_perm_builder and
8819 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8820 truncation to calc_vec_perm_mask_for_shift.
8821 (vect_create_epilog_for_reduction): Likewise.
8822 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8823 from auto_vec_perm_indices to vec_perm_indices.
8824 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8825 instead of changing individual elements.
8826 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8827 the vector in d.perm.
8828 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8829 from auto_vec_perm_indices to vec_perm_indices.
8830 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8831 instead of changing individual elements.
8832 (arm_vectorize_vec_perm_const): Use new_vector to install
8833 the vector in d.perm.
8834 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8835 Update vec_perm_builder constructor.
8836 (rs6000_expand_interleave): Likewise.
8837 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8838 (rs6000_expand_interleave): Likewise.
8840 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8842 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8843 to qimode could truncate the indices.
8844 * optabs.c (expand_vec_perm_var): Likewise.
8846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8848 * Makefile.in (OBJS): Add vec-perm-indices.o.
8849 * vec-perm-indices.h: New file.
8850 * vec-perm-indices.c: Likewise.
8851 * target.h (vec_perm_indices): Replace with a forward class
8853 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8854 * optabs.h: Include vec-perm-indices.h.
8855 (expand_vec_perm): Delete.
8856 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8857 (expand_vec_perm_const): Declare.
8858 * target.def (vec_perm_const_ok): Replace with...
8859 (vec_perm_const): ...this new hook.
8860 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8861 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8862 * doc/tm.texi: Regenerate.
8863 * optabs.def (vec_perm_const): Delete.
8864 * doc/md.texi (vec_perm_const): Likewise.
8865 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8866 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8867 expand_vec_perm for constant permutation vectors. Assert that
8868 the mode of variable permutation vectors is the integer equivalent
8869 of the mode that is being permuted.
8870 * optabs-query.h (selector_fits_mode_p): Declare.
8871 * optabs-query.c: Include vec-perm-indices.h.
8872 (selector_fits_mode_p): New function.
8873 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8874 is defined, instead of checking whether the vec_perm_const_optab
8875 exists. Use targetm.vectorize.vec_perm_const instead of
8876 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8877 fit in the vector mode before using a variable permute.
8878 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8879 vec_perm_indices instead of an rtx.
8880 (expand_vec_perm): Replace with...
8881 (expand_vec_perm_const): ...this new function. Take the selector
8882 as a vec_perm_indices rather than an rtx. Also take the mode of
8883 the selector. Update call to shift_amt_for_vec_perm_mask.
8884 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8885 Use vec_perm_indices::new_expanded_vector to expand the original
8886 selector into bytes. Check whether the indices fit in the vector
8887 mode before using a variable permute.
8888 (expand_vec_perm_var): Make global.
8889 (expand_mult_highpart): Use expand_vec_perm_const.
8890 * fold-const.c: Includes vec-perm-indices.h.
8891 * tree-ssa-forwprop.c: Likewise.
8892 * tree-vect-data-refs.c: Likewise.
8893 * tree-vect-generic.c: Likewise.
8894 * tree-vect-loop.c: Likewise.
8895 * tree-vect-slp.c: Likewise.
8896 * tree-vect-stmts.c: Likewise.
8897 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8899 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8900 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8901 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8902 (aarch64_vectorize_vec_perm_const): ...this new function.
8903 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8904 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8905 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8906 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8907 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8908 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8909 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8911 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8912 check for NEON modes.
8913 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8914 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8915 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8916 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8918 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8919 the old VEC_PERM_CONST conditions.
8920 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8921 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8922 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8923 (ia64_vectorize_vec_perm_const_ok): Merge into...
8924 (ia64_vectorize_vec_perm_const): ...this new function.
8925 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8926 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8927 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8928 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8929 * config/mips/mips.c (mips_expand_vec_perm_const)
8930 (mips_vectorize_vec_perm_const_ok): Merge into...
8931 (mips_vectorize_vec_perm_const): ...this new function.
8932 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8933 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8934 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8935 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8936 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8937 (rs6000_expand_vec_perm_const): Delete.
8938 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8940 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8941 (altivec_expand_vec_perm_const_le): Take each operand individually.
8942 Operate on constant selectors rather than rtxes.
8943 (altivec_expand_vec_perm_const): Likewise. Update call to
8944 altivec_expand_vec_perm_const_le.
8945 (rs6000_expand_vec_perm_const): Delete.
8946 (rs6000_vectorize_vec_perm_const_ok): Delete.
8947 (rs6000_vectorize_vec_perm_const): New function.
8948 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8949 an element count and rtx array.
8950 (rs6000_expand_extract_even): Update call accordingly.
8951 (rs6000_expand_interleave): Likewise.
8952 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8953 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8954 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8955 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8956 (rs6000_expand_vec_perm_const): Delete.
8957 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8958 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8959 (altivec_expand_vec_perm_const_le): Take each operand individually.
8960 Operate on constant selectors rather than rtxes.
8961 (altivec_expand_vec_perm_const): Likewise. Update call to
8962 altivec_expand_vec_perm_const_le.
8963 (rs6000_expand_vec_perm_const): Delete.
8964 (rs6000_vectorize_vec_perm_const_ok): Delete.
8965 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8966 reference to the SPE evmerge intructions.
8967 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8968 an element count and rtx array.
8969 (rs6000_expand_extract_even): Update call accordingly.
8970 (rs6000_expand_interleave): Likewise.
8971 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8972 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8974 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8976 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8978 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8979 vector mode and that that mode matches the mode of the data
8981 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8982 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8983 directly using expand_vec_perm_1 when forcing selectors into
8985 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8987 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8989 * optabs-query.h (can_vec_perm_p): Delete.
8990 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8991 * optabs-query.c (can_vec_perm_p): Split into...
8992 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8993 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8994 particular selector is valid.
8995 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8996 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8997 (vect_grouped_load_supported): Likewise.
8998 (vect_shift_permute_load_chain): Likewise.
8999 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9000 (vect_transform_slp_perm_load): Likewise.
9001 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9002 (vectorizable_bswap): Likewise.
9003 (vect_gen_perm_mask_checked): Likewise.
9004 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9005 implementations of variable permutation vectors into account
9006 when deciding which selector to use.
9007 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9008 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9009 with a false third argument.
9010 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9011 to test whether the constant selector is valid and can_vec_perm_var_p
9012 to test whether a variable selector is valid.
9014 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9016 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9017 * optabs-query.c (can_vec_perm_p): Likewise.
9018 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9019 instead of vec_perm_indices.
9020 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9021 (vect_gen_perm_mask_checked): Likewise,
9022 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9023 (vect_gen_perm_mask_checked): Likewise,
9025 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9027 * optabs-query.h (qimode_for_vec_perm): Declare.
9028 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9029 (qimode_for_vec_perm): ...this new function.
9030 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9032 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9034 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9035 does not have a conditional at the top.
9037 2018-01-02 Richard Biener <rguenther@suse.de>
9039 * ipa-inline.c (big_speedup_p): Fix expression.
9041 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9044 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9047 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9051 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9052 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9053 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9054 cond_taken_branch_cost 3->4.
9056 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9058 PR tree-optimization/83581
9059 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9060 TODO_cleanup_cfg if any changes have been made.
9063 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9064 convert_modes if target mode has the right side, but different mode
9068 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9069 last argument when extracting from CONCAT. If either from_real or
9070 from_imag is NULL, use expansion through memory. If result is not
9071 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9072 the parts directly to inner mode, if even that fails, use expansion
9076 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9077 check for bswap in mode rather than HImode and use that in expand_unop
9080 Copyright (C) 2018 Free Software Foundation, Inc.
9082 Copying and distribution of this file, with or without modification,
9083 are permitted in any medium without royalty provided the copyright
9084 notice and this notice are preserved.