1 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8 Paul Hua <paul.hua.gm@gmail.com>
11 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
13 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
15 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19 2018-03-15 Tamar Christina <tamar.christina@arm.com>
22 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23 instead of GET_MODE_SIZE when comparing Units.
25 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
28 * varasm.c (hash_section): Return an unchangeble hash value
29 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
30 Return !aarch64_can_use_per_function_literal_pools_p ().
32 2018-03-15 Jakub Jelinek <jakub@redhat.com>
35 * optabs.c (emit_conditional_move): Pass address of cmode's copy
36 rather than address of cmode as last argument to prepare_cmp_insn.
38 2018-03-15 Julia Koval <julia.koval@intel.com>
40 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
41 F_AVX512VNNI, F_AVX512BITALG): New.
43 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
46 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
47 insn for floating-point loads and stores.
49 2018-03-14 Carl Love <cel@us.ibm.com>
51 * config/rs6000/rs6000-c.c: Add macro definitions for
52 ALTIVEC_BUILTIN_VEC_PERMXOR.
53 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
54 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
55 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
56 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
58 * config/doc/extend.texi: Add prototypes for vec_permxor.
60 2018-03-14 David Malcolm <dmalcolm@redhat.com>
63 * diagnostic-show-locus.c (class layout_point): Convert m_line
64 from int to linenum_type.
65 (line_span::comparator): Use linenum "compare" function when
66 comparing line numbers.
67 (test_line_span): New function.
68 (layout_range::contains_point): Convert param "row" from int to
70 (layout_range::intersects_line_p): Likewise.
71 (layout::will_show_line_p): Likewise.
72 (layout::print_source_line): Likewise.
73 (layout::should_print_annotation_line_p): Likewise.
74 (layout::print_annotation_line): Likewise.
75 (layout::print_leading_fixits): Likewise.
76 (layout::annotation_line_showed_range_p): Likewise.
77 (struct line_corrections): Likewise for field m_row.
78 (line_corrections::line_corrections): Likewise for param "row".
79 (layout::print_trailing_fixits): Likewise.
80 (layout::get_state_at_point): Likewise.
81 (layout::get_x_bound_for_row): Likewise.
82 (layout::print_line): Likewise.
83 (diagnostic_show_locus): Likewise for locals "last_line" and
85 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
86 * input.c (selftest::test_linenum_comparisons): New function.
87 (selftest::input_c_tests): Call it.
88 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
89 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
90 * selftest.h (ASSERT_GT): New macro.
91 (ASSERT_GT_AT): New macro.
92 (ASSERT_LT): New macro.
93 (ASSERT_LT_AT): New macro.
95 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
97 PR rtl-optimization/84780
98 * combine.c (distribute_links): Don't make a link based on pc_rtx.
100 2018-03-14 Martin Liska <mliska@suse.cz>
102 * tree.c (record_node_allocation_statistics): Use
104 (get_stats_node_kind): New function extracted from
105 record_node_allocation_statistics.
106 (free_node): Use get_stats_node_kind.
108 2018-03-14 Richard Biener <rguenther@suse.de>
110 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
111 that the value-set of ANTIC_IN doesn't grow.
114 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
116 (BB_VISITED_WITH_VISITED_SUCCS): New define.
117 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
119 2018-03-14 Julia Koval <julia.koval@intel.com>
121 * config.gcc (icelake-client, icelake-server): New.
123 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
124 (initial_ix86_arch_features): Ditto.
125 (PTA_SKYLAKE): Add SGX.
126 (PTA_ICELAKE): Remove.
127 (PTA_ICELAKE_CLIENT): New.
128 (PTA_ICELAKE_SERVER): New.
129 (ix86_option_override_internal): Split up icelake on icelake client and
131 (get_builtin_code_for_version): Ditto.
132 (fold_builtin_cpu): Ditto.
133 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
134 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
135 * config/i386/i386.h (processor_type): Ditto.
136 * doc/invoke.texi: Ditto.
138 2018-03-14 Jakub Jelinek <jakub@redhat.com>
141 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
142 INTEGER_CST offset, add it together with bitpos / 8 and
143 sign extend based on POINTER_SIZE.
147 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
150 * config/i386/constraints.md (Yc): New register constraint.
151 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
152 Use Yc constraint for alternative 2 of operand 0. Remove
153 preferred_for_speed attribute.
155 2018-03-14 Richard Biener <rguenther@suse.de>
157 PR tree-optimization/84830
158 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
159 with the old one to avoid oscillations.
161 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
164 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
166 (assign_by_spills): Return a flag of reload assignment failure.
167 Do not process the reload assignment failures. Do not spill other
168 reload pseudos if they has the same reg class. Update n if
170 (lra_assign): Add a return arg. Set up from the result of
171 assign_by_spills call.
172 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
173 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
174 usage_insns if it is not NULL.
175 (spill_hard_reg_in_range): New function.
176 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
177 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
179 (lra_assign): Change prototype.
180 * lra.c (lra): Add code to deal with fails by splitting hard reg
183 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
185 * config/riscv/riscv.opt (mrelax): New option.
186 * config/riscv/riscv.c (riscv_file_start): Emit ".option
187 "norelax" when riscv_mrelax is disabled.
188 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
190 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
193 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
194 reassociation for int modes.
196 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
198 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
199 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
201 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
202 * config/aarch64/aarch64-sve.md
203 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
204 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
205 (*extend<mode><Vwide>2): Rename to...
206 (aarch64_sve_extend<mode><Vwide>2): ...this.
207 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
208 renaming the old pattern to...
209 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
211 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
212 define_expand, renaming the old pattern to...
213 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
214 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
215 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
216 account when deciding which SVE instruction the optab should use.
217 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
219 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
221 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
222 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
223 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
224 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
225 (tlsdesc_small_<mode>): Turn a define_expand and use
226 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
227 (tlsdesc_small_advsimd_<mode>): ...this.
228 (tlsdesc_small_sve_<mode>): New pattern.
230 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
232 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
233 (UNSPEC_UMUL_HIGHPART): New constants.
234 (MUL_HIGHPART): New int iteraor.
235 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
236 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
238 (*<su>mul<mode>3_highpart): New define_insn.
240 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
243 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
246 2018-03-13 Martin Liska <mliska@suse.cz>
249 * (sem_item_optimizer::sem_item_optimizer): Initialize new
251 (sem_item_optimizer::~sem_item_optimizer): Release it.
252 (sem_item_optimizer::merge_classes): Register variable aliases.
253 (sem_item_optimizer::fixup_pt_set): New function.
254 (sem_item_optimizer::fixup_points_to_sets): Likewise.
255 * ipa-icf.h: Declare new variables and functions.
257 2018-03-13 Jakub Jelinek <jakub@redhat.com>
260 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
261 integer_pow2p@2 and test integer_pow2p in condition.
262 (A < 0 ? C : 0): Similarly for @1.
265 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
266 characters starting at p contain '\0' character, don't look beyond
270 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
271 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
274 * reg-stack.c (change_stack): Change update_end var from int to
275 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
276 also call set_block_for_insn on the newly added insns and rescan.
279 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
283 * tree.c (stabilize_reference_1): Return save_expr (e) for
284 STATEMENT_LIST even if it doesn't have side-effects.
286 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
288 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
290 2018-03-12 Renlin Li <renlin.li@arm.com>
292 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
293 aarch64_output_scalar_simd_mov_immediate.
295 2018-03-12 Martin Sebor <msebor@redhat.com>
297 PR tree-optimization/83456
298 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
299 for perfectly overlapping calls to memcpy.
300 (gimple_fold_builtin_memory_chk): Same.
301 (gimple_fold_builtin_strcpy): Handle no-warning.
302 (gimple_fold_builtin_stxcpy_chk): Same.
303 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
305 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
307 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
308 parameter. Use it for SFmode.
309 (rs6000_function_arg_advance_1): Adjust.
310 (rs6000_function_arg): Adjust.
311 (rs6000_gimplify_va_arg): Pass false for that new parameter.
313 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
315 PR rtl-optimization/84169
316 PR rtl-optimization/84780
317 * combine.c (can_combine_p): Check for a 2-insn combination whether
318 the destination register is used between the two insns, too.
320 2018-03-12 Richard Biener <rguenther@suse.de>
322 PR tree-optimization/84803
323 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
324 for refs DR analysis didn't process.
326 2018-03-12 Richard Biener <rguenther@suse.de>
328 PR tree-optimization/84777
329 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
330 force-vectorize loops ignore whether we are optimizing for size.
332 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
334 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
335 (TARGET_MD_ASM_ADJUST): Define.
337 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
338 Kito Cheng <kito.cheng@gmail.com>
339 Chung-Ju Wu <jasonwucj@gmail.com>
341 * config/nds32/nds32.c (nds32_compute_stack_frame,
342 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
343 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
344 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
345 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
346 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
347 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
348 * config/nds32/nds32.md (prologue, epilogue): Use macro
349 NDS32_V3PUSH_AVAILABLE_P to do checking.
351 2018-03-11 Jakub Jelinek <jakub@redhat.com>
354 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
355 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
356 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
357 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
358 addition of most attributes on !orig_type_die or the attribute not
359 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
361 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
362 Chung-Ju Wu <jasonwucj@gmail.com>
364 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
366 * config/nds32/nds32.opt (mvh): New option.
368 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
369 Chung-Ju Wu <jasonwucj@gmail.com>
371 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
373 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
374 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
377 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
378 Chung-Ju Wu <jasonwucj@gmail.com>
380 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
382 * config/nds32/nds32-multiple.md (strlensi): New pattern.
383 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
385 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
386 Kito Cheng <kito.cheng@gmail.com>
387 Chung-Ju Wu <jasonwucj@gmail.com>
389 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
390 UNSPEC_FFMISM and UNSPEC_FLMISM.
391 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
392 for ffb, ffmism and flmism.
393 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
394 (unspec_ffmism): Ditto.
395 (unspec_flmism): Ditto.
396 (nds32_expand_builtin_impl): Check if string extension is available.
397 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
398 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
400 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
403 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
406 * lra-assigns.c (assign_by_spills): Return a flag of reload
407 assignment failure. Do not process the reload assignment
408 failures. Do not spill other reload pseudos if they has the same
410 (lra_assign): Add a return arg. Set up from the result of
411 assign_by_spills call.
412 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
413 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
414 usage_insns if it is not NULL.
415 (spill_hard_reg_in_range): New function.
416 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
417 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
419 (lra_assign): Change prototype.
420 * lra.c (lra): Add code to deal with fails by splitting hard reg
423 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
426 * config/i386/i386.opt: Replace Enforcment with Enforcement.
428 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
431 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
432 (dw_val_node): Add val_symbolic_view.
433 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
434 (symview_upper_bound): New.
435 (new_line_info_table): Initialize symviews_since_reset.
436 (dwarf2out_source_line): Count symviews_since_reset and set
438 (dw_val_equal_p): Handle symview.
439 (add_AT_symview): New.
440 (print_dw_val): Handle symview.
441 (attr_checksum, attr_checksum_ordered): Likewise.
442 (same_dw_val_p, size_of_die): Likewise.
443 (value_format, output_die): Likewise.
444 (add_high_low_attributes): Use add_AT_symview for entry_view.
445 (dwarf2out_finish): Reset symview_upper_bound, clear
448 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
451 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
452 Add strict argument and use it.
453 (rs6000_split_multireg_move): Update for new strict argument.
454 (mem_operand_gpr): Disallow all non-offsettable addresses.
455 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
457 2018-03-09 Jakub Jelinek <jakub@redhat.com>
460 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
461 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
462 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
465 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
466 decl, use remap_type if we want to use the type.
468 2018-03-09 Martin Sebor <msebor@redhat.com>
470 PR tree-optimization/84526
471 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
473 (builtin_access::generic_overlap): Be prepared to handle non-array
476 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
478 PR rtl-optimization/84682
479 * lra-constraints.c (process_address_1): Check is_address flag
480 for address constraints.
481 (process_alt_operands): Likewise.
482 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
483 preprocess_constraints.
484 * recog.h (preprocess_constraints): Add oploc parameter.
486 * recog.c (preprocess_constraints): Test address_operand for
487 CT_ADDRESS constraints.
489 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
492 * lra-assigns.c (assign_by_spills): Return a flag of reload
493 assignment failure. Do not process the reload assignment
494 failures. Do not spill other reload pseudos if they has the same
496 (lra_assign): Add a return arg. Set up from the result of
497 assign_by_spills call.
498 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
499 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
500 usage_insns if it is not NULL.
501 (spill_hard_reg_in_range): New function.
502 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
503 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
505 (lra_assign): Change prototype.
506 * lra.c (lra): Add code to deal with fails by splitting hard reg
509 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
512 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
513 Accept complain bool parameter. Only emit errors if it is true.
514 (arm_parse_cpu_option_name): Likewise.
515 (arm_target_thumb_only): Adjust callers of the above.
516 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
517 prototype to take a default true bool parameter.
518 (arm_parse_arch_option_name): Likewise.
520 2018-03-09 David Malcolm <dmalcolm@redhat.com>
521 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
525 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
526 * configure: Regenerate.
527 * configure.ac ("linker --version-script option"): New.
528 ("linker soname option"): New.
530 2018-03-09 Richard Biener <rguenther@suse.de>
532 PR tree-optimization/84775
533 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
534 immediate uses of predicate stmts and mark them modified.
537 PR tree-optimization/84178
538 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
540 (version_loop_for_if_conversion): Delay update_ssa call.
541 (tree_if_conversion): Delay update_ssa until after predicate
544 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
547 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
548 when the function accesses prior frames.
550 2018-03-08 Jakub Jelinek <jakub@redhat.com>
553 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
554 gen_llsym, otherwise call maybe_gen_llsym.
557 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
558 has ',' character inside of it.
560 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
563 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
564 as clobbering CC_REGNUM.
566 2018-03-08 Richard Biener <rguenther@suse.de>
569 * tree-scalar-evolution.c: Include tree-into-ssa.h.
570 (follow_copies_to_constant): Do not follow SSA names registered
573 2018-03-08 Richard Biener <rguenther@suse.de>
575 PR tree-optimization/84178
576 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
578 (version_loop_for_if_conversion): Delay update_ssa call.
579 (tree_if_conversion): Delay update_ssa until after predicate
582 2018-03-08 David Malcolm <dmalcolm@redhat.com>
584 PR tree-optimization/84178
585 * tree-if-conv.c (release_bb_predicate): Remove the
586 the assertion that the stmts have NULL use_ops.
587 Discard the statements, asserting that they haven't
588 yet been added to a BB.
590 2018-03-08 Richard Biener <rguenther@suse.de>
592 PR tree-optimization/84746
593 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
594 (phi_translate): Pass in destination ANTIC_OUT set.
595 (phi_translate_1): Likewise. For a simplified result lookup
596 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
597 (phi_translate_set): Adjust.
598 (do_pre_regular_insertion): Likewise.
599 (do_pre_partial_partial_insertion): Likewise.
601 2018-03-08 Martin Liska <mliska@suse.cz>
603 PR gcov-profile/84735
604 * doc/gcov.texi: Document usage of profile files.
605 * gcov-io.h: Document changes in the format.
607 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
611 * dwarf2out.c (struct dw_line_info_table): Update comments for
613 (FORCE_RESET_NEXT_VIEW): New.
614 (FORCE_RESETTING_VIEW_P): New.
615 (RESETTING_VIEW_P): Check for -1 too.
616 (ZERO_VIEW_P): Likewise.
617 (new_line_info_table): Force-reset next view.
618 (dwarf2out_begin_function): Likewise.
619 (dwarf2out_source_line): Simplify zero_view_p initialization.
620 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
621 view directly. Omit view when omitting .loc at line 0.
623 2018-03-08 Jakub Jelinek <jakub@redhat.com>
625 PR tree-optimization/84740
626 * tree-switch-conversion.c (process_switch): Call build_constructors
627 only if info.phi_count is non-zero.
629 PR tree-optimization/84739
630 * tree-tailcall.c (find_tail_calls): Check call arguments against
631 DECL_ARGUMENTS (current_function_decl) rather than
632 DECL_ARGUMENTS (func) when checking for tail recursion.
634 2018-03-07 Jakub Jelinek <jakub@redhat.com>
636 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
637 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
638 Volker Reichelt's entry and add entries for people that perform
639 GCC fuzzy testing and report numerous bugs.
641 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
644 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
645 readonly data in sdata, if that is disabled.
646 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
647 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
648 -mreadonly-in-sdata option.
650 2018-03-07 Martin Sebor <msebor@redhat.com>
652 PR tree-optimization/84468
653 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
654 basic block when looking for nul assignment.
656 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
659 * except.h (output_function_exception_table): Adjust prototype.
660 * except.c (output_function_exception_table): Remove FNNAME parameter
661 and add SECTION parameter. Ouput one part of the table at a time.
662 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
663 the first part of the exception table and emit unwind directives.
664 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
665 (i386_pe_seh_cold_init): Likewise.
666 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
667 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
668 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
669 (ix86_output_call_insn): Emit a nop in one more case for SEH.
670 * config/i386/winnt.c: Include except.h.
671 (struct seh_frame_state): Add reg_offset, after_prologue and
672 in_cold_section fields.
673 (i386_pe_seh_end_prologue): Set seh->after_prologue.
674 (i386_pe_seh_cold_init): New function.
675 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
676 to seh->in_cold_section.
677 (seh_emit_push): Record the offset of the push.
678 (seh_emit_save): Record the offet of the save.
679 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
680 Test seh->after_prologue to disregard the epilogue.
681 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
682 (i386_pe_end_cold_function): New function.
684 2018-03-07 Jakub Jelinek <jakub@redhat.com>
687 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
688 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
691 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
693 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
694 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
697 * multiple_target.c: Include tree-inline.h and intl.h.
698 (expand_target_clones): Diagnose and fail if node->definition and
699 !tree_versionable_function_p (node->decl).
701 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
703 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
705 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
706 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
707 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
709 2018-03-06 Jakub Jelinek <jakub@redhat.com>
712 * combine.c (try_combine): Use reg_or_subregno instead of handling
713 just paradoxical SUBREGs and REGs.
715 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
717 * config/arc/arc.c (arc_finalize_pic): Remove function.
718 (arc_must_save_register): We use single base PIC register, remove
719 checks to save/restore the PIC register.
720 (arc_expand_prologue): Likewise.
721 * config/arc/arc-protos.h (arc_set_default_type_attributes):
723 (arc_verify_short): Likewise.
724 (arc_attr_type): Likewise.
725 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
726 (walk_stores): Likewise.
727 (arc_address_cost): Make it static.
728 (arc_verify_short): Likewise.
729 (branch_dest): Likewise.
730 (arc_attr_type): Likewise.
731 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
732 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
733 (arc_final_prescan_insn): Remove inserting the nops due to
734 hardware hazards. It is done in reorg step.
735 (insn_length_variant_t): Remove.
736 (insn_length_parameters_t): Likewise.
737 (arc_insn_length_parameters): Likewise.
738 (arc_get_insn_variants): Likewise.
739 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
741 2018-03-06 Jakub Jelinek <jakub@redhat.com>
744 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
747 PR tree-optimization/84687
748 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
750 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
752 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
754 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
755 Rename to ppc_speculation_barrier.
756 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
757 __builtin_ppc_speculation_barrier.
759 2018-03-05 Jakub Jelinek <jakub@redhat.com>
762 * combine.c (combine_simplify_rtx): Don't try to simplify if
763 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
766 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
768 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
769 to 32 bytes when compiling for POWER9.
771 2018-03-05 Jakub Jelinek <jakub@redhat.com>
774 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
775 regparm >= 3 with no arg reg available also for calls with
776 flag_force_indirect_call. Pass decl to ix86_function_regparm.
779 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
781 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
783 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
786 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
788 2018-03-05 Richard Biener <rguenther@suse.de>
790 PR tree-optimization/84486
791 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
792 When inserting a __builtin_assume_aligned call set the LHS
793 SSA name alignment info accordingly.
795 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
797 PR tree-optimization/84114
798 * config/aarch64/aarch64.c (aarch64_reassociation_width)
799 Avoid reassociation of FLOAT_MODE addition.
801 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
803 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
804 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
805 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
806 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
807 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
808 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
809 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
811 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
812 __builtin_ia32_wbinvd): New builtins.
813 (SPECIAL_ARGS2): New.
814 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
815 (SPECIAL_ARGS2): New.
816 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
817 (ix86_valid_target_attribute_inner_p): Ditto.
818 (ix86_init_mmx_sse_builtins): Add special_args2.
819 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
820 TARGET_WBNOINVD_P): New.
821 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
822 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
823 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
824 * config/i386/immintrin.h (_wbinvd): New intrinsic.
825 * config/i386/pconfigintrin.h: New file.
826 * config/i386/wbnoinvdintrin.h: Ditto.
827 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
828 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
830 2018-03-05 Richard Biener <rguenther@suse.de>
832 PR tree-optimization/84670
833 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
835 (BB_VISITED_WITH_VISITED_SUCCS): New define.
836 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
837 (compute_antic_aux): Only assert the number of values in ANTIC_IN
838 doesn't grow if all successors (recursively) were visited at least
841 2018-03-05 Richard Biener <rguenther@suse.de>
843 PR tree-optimization/84650
844 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
845 if executed in the loop pipeline.
847 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
849 * doc/configfiles.texi (Configuration Files): Move info about
850 conditionalizing $target-protos.h to...
851 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
852 differs from $target-protos.h.
854 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
855 Chung-Ju Wu <jasonwucj@gmail.com>
857 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
858 * config/nds32/nds32-multiple.md (setmemsi): Define.
859 * config/nds32/nds32-memory-manipulation.c
860 (nds32_gen_dup_4_byte_to_word_value): New.
861 (emit_setmem_word_loop): New.
862 (emit_setmem_byte_loop): New.
863 (nds32_expand_setmem_loop): New.
864 (nds32_expand_setmem_loop_v3m): New.
865 (nds32_expand_setmem_unroll): New.
866 (nds32_expand_setmem): New.
868 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
869 Chung-Ju Wu <jasonwucj@gmail.com>
871 * config/nds32/nds32-memory-manipulation.c
872 (nds32_emit_load_store): New.
873 (nds32_emit_post_inc_load_store): New.
874 (nds32_emit_mem_move): New.
875 (nds32_emit_mem_move_block): New.
876 (nds32_expand_movmemsi_loop_unknown_size): New.
877 (nds32_expand_movmemsi_loop_known_size): New.
878 (nds32_expand_movmemsi_loop): New.
879 (nds32_expand_movmemsi_unroll): New.
880 (nds32_expand_movmemqi): Rename ...
881 (nds32_expand_movmemsi): ... to this.
882 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
883 (movmemsi): ... to this.
884 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
885 (nds32_expand_movmemsi): ... to this.
887 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
888 Monk Chiang <sh.chiang04@gmail.com>
889 Chung-Ju Wu <jasonwucj@gmail.com>
891 * config/nds32/nds32-protos.h
892 (nds32_expand_load_multiple): New arguments.
893 (nds32_expand_store_multiple): Ditto.
894 (nds32_valid_multiple_load_store): Rename ...
895 (nds32_valid_multiple_load_store_p): ... to this.
896 * config/nds32/nds32-memory-manipulation.c
897 (nds32_expand_load_multiple): Refine implementation.
898 (nds32_expand_store_multiple): Ditto.
899 * config/nds32/nds32-multiple.md
900 (load_multiple): Update nds32_expand_load_multiple interface.
901 (store_multiple): Update nds32_expand_store_multiple interface.
902 * config/nds32/nds32-predicates.c
903 (nds32_valid_multiple_load_store): Rename ...
904 (nds32_valid_multiple_load_store_p): ... to this and refine
906 * config/nds32/predicates.md
907 (nds32_load_multiple_and_update_address_operation): New predicate.
908 (nds32_store_multiple_and_update_address_operation): New predicate.
910 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
911 Chung-Ju Wu <jasonwucj@gmail.com>
913 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
914 (combo): New attribute.
915 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
917 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
919 * config/nds32/nds32.opt: Change -mcmodel= default value.
921 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
922 Monk Chiang <sh.chiang04@gmail.com>
923 Chung-Ju Wu <jasonwucj@gmail.com>
925 * config/nds32/constants.md (unspec_element): New enum.
926 * config/nds32/constraints.md (Umw): New constraint.
927 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
928 * config/nds32/nds32-intrinsic.md: Likewise.
929 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
930 (nds32_valid_smw_lwm_base_p): New.
931 (nds32_output_smw_single_word): New.
932 (nds32_output_lmw_single_word): New.
933 (nds32_expand_unaligned_load): New.
934 (nds32_expand_unaligned_store): New.
935 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
936 (nds32_output_smw_single_word): Declare.
937 (nds32_output_lmw_single_word): Declare.
938 (nds32_expand_unaligned_load): Declare.
939 (nds32_expand_unaligned_store): Declare.
940 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
941 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
942 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
943 NDS32_BUILTIN_UASTORE_DW.
944 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
947 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
948 Kito Cheng <kito.cheng@gmail.com>
949 Chung-Ju Wu <jasonwucj@gmail.com>
951 * config/nds32/nds32-intrinsic.c
952 (nds32_expand_builtin_null_ftype_reg): Delete.
953 (nds32_expand_builtin_reg_ftype_imm): Ditto.
954 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
955 (nds32_read_argument): New.
956 (nds32_legitimize_target): Ditto.
957 (nds32_legitimize_argument): Ditto.
958 (nds32_check_constant_argument): Ditto.
959 (nds32_expand_unop_builtin): Ditto.
960 (nds32_expand_unopimm_builtin): Ditto.
961 (nds32_expand_binop_builtin): Ditto.
962 (nds32_builtin_decl_impl): Ditto.
963 (builtin_description): Ditto.
964 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
965 (nds32_init_builtins_impl): Ditto.
966 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
967 (nds32_builtin_decl): New.
968 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
969 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
971 2018-03-02 Jeff Law <law@redhat.com>
973 * reorg.c (stop_search_p): Handle DEBUG_INSN.
974 (redundant_insn, fill_simple_delay_slots): Likewise.
975 (fill_slots_from_thread): Likewise.
976 * resource.c (mark_referenced_resources): Likewise.
977 (mark_set_resources, find_dead_or_set_registers): Likewise.
979 2018-03-02 Jakub Jelinek <jakub@redhat.com>
981 * substring-locations.h (format_warning_va): Formatting fix for
983 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
985 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
986 * substring-locations.c: Include intl.h.
987 (format_warning_va): Turned into small wrapper around
988 format_warning_n_va, renamed to ...
989 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
990 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
992 (format_warning_at_substring_n): New function.
993 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
994 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
995 format_warning_at_substring with just a shorter name instead of
996 const function pointer.
997 (fmtwarn_n): New function.
998 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
999 appropriate, get rid of all the fmtstr temporaries, move conditionals
1000 with G_() wrapped string literals directly into fmtwarn arguments,
1001 cast dir.len to (int), formatting fixes.
1003 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1005 * doc/invoke.texi: Remove "Cilk Plus" references.
1007 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1008 Richard Biener <rguenther@suse.de>
1011 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1012 for error or warning attributes if CALL_FROM_THUNK_P is set.
1015 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1018 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1019 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1022 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1023 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1025 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1026 instead of -1U in last predictors element's probability member.
1028 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1031 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1032 arguments if they are comparable.
1034 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1036 PR tree-optimization/84634
1037 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1038 masks and masked_loop_p with a single loop_masks, making sure it's
1039 null for bb vectorization.
1041 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1043 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1044 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1045 loop->force_vectorize to check whether there is no alias.
1047 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1050 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1052 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1054 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1055 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1056 instead of a loop around prev_real_insn.
1057 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1061 * config/i386/i386.c (ix86_print_operand): Use conditional
1062 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1065 2018-03-02 Richard Biener <rguenther@suse.de>
1067 PR tree-optimization/84427
1068 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1069 (bitmap_set_subtract_values): Rewrite to handle multiple
1072 (prune_clobbered_mems): Likewise.
1073 (phi_translate): Take edge instead of pred/phiblock.
1074 (phi_translate_1): Likewise.
1075 (phi_translate_set): Likewise. Insert all translated
1076 exprs for a value into the set, keeping possibly multiple
1077 expressions per value.
1078 (compute_antic_aux): Adjust for phi_translate changes.
1079 When intersecting union the expressions and prune those
1080 not in the final value set, keeping possibly multiple
1081 expressions per value. Do not use value-insertion
1082 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1083 all expressions. Add verification that the value-sets
1084 only shrink during iteration.
1085 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1086 (do_pre_regular_insertion): Likewise.
1087 (do_pre_partial_partial_insertion): Likewise.
1089 2018-03-02 Richard Biener <rguenther@suse.de>
1092 * config/darwin.c (saved_debug_info_level): New static global.
1093 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1094 (darwin_asm_lto_end): Restore debug info generation settings.
1096 2018-03-01 Martin Liska <mliska@suse.cz>
1099 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1102 2018-03-01 Richard Biener <rguenther@suse.de>
1105 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1106 type annotation with LTO when debug was disabled at compile-time.
1108 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1110 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1112 (mips_final_postscan_insn): Likewise.
1114 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1116 PR rtl-optimization/84528
1117 * alias.c (init_alias_target): Add commentary.
1118 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1119 a unique base value if the frame pointer is not eliminated
1120 to the stack pointer.
1122 2018-03-01 Tom de Vries <tom@codesourcery.com>
1124 PR rtl-optimization/83327
1125 * lra-int.h (hard_regs_spilled_into): Declare.
1126 * lra.c (hard_regs_spilled_into): Define.
1127 (init_reg_info): Init hard_regs_spilled_into.
1128 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1129 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1130 (process_bb_lives): Handle hard_regs_spilled_into.
1131 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1132 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1134 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1136 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1137 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1138 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1139 * config/rs6000/aix72.h: New file.
1141 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1143 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1144 instead of warning_at with conditional singular and plural messages
1148 * stor-layout.c (update_alignment_for_field): For
1149 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1150 && !DECL_PACKED (field), do the alignment update, just use
1151 only desired_align instead of MAX (type_align, desired_align)
1153 (place_field): Don't do known_align < desired_align handling
1154 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1155 is non-NULL, instead do it after rli->prev_field handling and
1156 only if not within a bitfield word. For DECL_PACKED (field)
1157 use type_align of BITS_PER_UNIT.
1159 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1161 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1162 superfluous parentheses and trailing spaces.
1164 2018-02-28 Richard Biener <rguenther@suse.de>
1166 PR tree-optimization/84584
1167 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1168 SCoPs with fake exit edge.
1170 2018-02-28 Martin Liska <mliska@suse.cz>
1173 * timevar.c (timer::print): Fix format to properly print 100%
1176 2018-02-28 Richard Biener <rguenther@suse.de>
1179 * genmatch.c (capture_info::walk_match): Do not mark
1180 captured expressions without operands as expr_p given
1181 they act more like predicates and should be subject to
1182 "lost tail" side-effect preserving.
1184 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1186 PR rtl-optimization/81611
1187 * auto-inc-dec.c (attempt_change): Move dead note from
1188 mem_insn if it's the next use of regno
1189 (find_address): Take address use of reg holding
1190 non-incremented value. Add parm to limit search to the named
1192 (merge_in_block): Attempt to use a mem insn that is the next
1193 use of the original regno.
1195 2018-02-27 Martin Sebor <msebor@redhat.com>
1198 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1199 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1201 2018-02-27 Martin Sebor <msebor@redhat.com>
1203 PR translation/84207
1204 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1205 n argument to unsigned HOST_WIDE_INT.
1206 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1207 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1208 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1209 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1211 2018-02-27 Richard Biener <rguenther@suse.de>
1213 PR tree-optimization/84512
1214 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1215 Do not use the estimate returned from record_stmt_cost for
1216 the scalar iteration cost but sum properly using add_stmt_cost.
1218 2018-02-27 Richard Biener <rguenther@suse.de>
1220 PR tree-optimization/84466
1221 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1222 Adjust last change to less strictly validate use operands.
1224 2018-02-27 Martin Liska <mliska@suse.cz>
1226 PR gcov-profile/84548
1227 * gcov.c (process_file): Allow partial overlap and consider it
1228 also as group functions.
1229 (output_lines): Properly calculate range of lines for a group.
1231 2018-02-27 Martin Liska <mliska@suse.cz>
1233 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1234 'ggc' suffixes. Change first column width.
1235 (timer::print): Fix formatting of the column.
1237 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1239 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1240 preserve inline entry blocks for the sake of debug inline
1241 entry point markers alone.
1242 (remove_unused_locals): Suggest in comments a better place to
1243 force the preservation of inline entry blocks that are
1244 otherwise unused, but do not preserve them.
1246 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1248 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1250 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1253 * config/i386/constraints.md (Bs): Replace
1254 ix86_indirect_branch_register with
1255 TARGET_INDIRECT_BRANCH_REGISTER.
1257 * config/i386/i386.md (indirect_jump): Likewise.
1258 (tablejump): Likewise.
1259 (*sibcall_memory): Likewise.
1260 (*sibcall_value_memory): Likewise.
1261 Peepholes of indirect call and jump via memory: Likewise.
1262 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1263 (*sibcall_value_GOT_32): Likewise.
1264 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1265 (GOT_memory_operand): Likewise.
1266 (call_insn_operand): Likewise.
1267 (sibcall_insn_operand): Likewise.
1268 (GOT32_symbol_operand): Likewise.
1269 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1271 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1273 PR rtl-optimization/83496
1274 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1275 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1276 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1277 redundant insn, if any.
1278 (relax_delay_slots): Likewise.
1279 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1281 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1283 PR tree-optimization/83965
1284 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1285 that grouped statements are part of a reduction chain. Return
1286 true if the statement is not marked as a reduction itself but
1288 (vect_recog_dot_prod_pattern): Don't check whether the statement
1289 is part of a group here.
1290 (vect_recog_sad_pattern): Likewise.
1291 (vect_recog_widen_sum_pattern): Likewise.
1293 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1296 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1298 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1301 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1303 (ix86_output_indirect_function_return): New prototype.
1304 (ix86_split_simple_return_pop_internal): Likewise.
1305 * config/i386/i386.c (indirect_return_via_cx): New.
1306 (indirect_return_via_cx_bnd): Likewise.
1307 (indirect_thunk_name): Handle return va CX_REG.
1308 (output_indirect_thunk_function): Create alias for
1309 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1310 (ix86_output_indirect_jmp): Remove the bool argument.
1311 (ix86_output_indirect_function_return): New function.
1312 (ix86_split_simple_return_pop_internal): Likewise.
1313 * config/i386/i386.md (*indirect_jump): Don't pass false
1314 to ix86_output_indirect_jmp.
1315 (*tablejump_1): Likewise.
1316 (simple_return_pop_internal): Change it to define_insn_and_split.
1317 Call ix86_split_simple_return_pop_internal to split it for
1319 (simple_return_indirect_internal): Call
1320 ix86_output_indirect_function_return instead of
1321 ix86_output_indirect_jmp.
1323 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1326 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1327 memset and value initialization afterwards.
1329 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1331 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1333 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1336 * common/config/aarch64/aarch64-common.c
1337 (aarch_option_optimization_table[]): Switch
1338 off fomit-frame-pointer
1340 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1341 Chung-Ju Wu <jasonwucj@gmail.com>
1343 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1345 (store_multiple): Ditto.
1347 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1349 * config.gcc: Add --with-cpu support for nds32 target.
1350 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1351 * config/nds32/nds32.opt: Add -mcpu= option.
1353 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1355 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1356 isel=yes): Warn for these deprecated options.
1358 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1360 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1361 ISA_2_5_MASKS_EMBEDDED.
1363 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1365 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1366 p->max as pointers rather than using iterative_hash_expr.
1368 2018-02-23 Carl Love <cel@us.ibm.com>
1370 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1371 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1373 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1374 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1375 P8V_BUILTIN_VEC_VUNSIGNED2.
1377 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1380 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1381 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1383 * lra-constraints.c (curr_insn_transform): Set up
1384 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1386 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1387 * lra-eliminations.c (spill_pseudos): Ditto.
1388 (process_insn_for_elimination): Ditto.
1389 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1390 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1391 LRA_NON_CLOBBERED_ALT.
1393 2018-02-22 Martin Sebor <msebor@redhat.com>
1395 PR tree-optimization/84480
1396 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1397 to maybe_diag_stxncpy_trunc. Call it.
1398 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1399 from gimple_fold_builtin_strcpy. Print inlining stack.
1400 (handle_builtin_stxncpy): Print inlining stack.
1401 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1403 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1406 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1407 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1408 and -fcheck-pointer-bounds are used together.
1409 (indirect_thunk_prefix): New enum.
1410 (indirect_thunk_need_prefix): New function.
1411 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1412 "_nt" instead of "_bnd" for NOTRACK prefix.
1413 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1414 (output_indirect_thunk_function): Likewise.
1416 (ix86_code_end): Update output_indirect_thunk_function calls.
1417 (ix86_output_indirect_branch_via_reg): Replace
1418 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1419 (ix86_output_indirect_branch_via_push): Likewise.
1420 (ix86_output_function_return): Likewise.
1421 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1422 incompatible with -fcf-protection=branch and
1423 -fcheck-pointer-bounds.
1425 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1428 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1429 Change gcc_assert call to output_operand_lossage.
1431 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1433 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1435 2018-02-22 DJ Delorie <dj@redhat.com>
1436 Sebastian Perta <sebastian.perta@renesas.com>
1437 Oleg Endo <olegendo@gcc.gnu.org>
1439 * config/rx/rx.c (rx_rtx_costs): New function.
1440 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1442 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1444 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1446 2018-02-22 Martin Liska <mliska@suse.cz>
1449 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1450 Add "native" as a possible value.
1452 2018-02-22 Martin Liska <mliska@suse.cz>
1455 * config/i386/i386.c (ix86_option_override_internal):
1456 Add "native" as a possible value for -march and -mtune.
1458 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1461 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1462 to all type variants.
1464 PR tree-optimization/84503
1465 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1466 width as info->bitpos + info->bitsize - start.
1467 (merged_store_group::merge_overlapping): Simplify width computation.
1468 (check_no_overlap): New function.
1469 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1470 start + width and last_order of the group, fail if check_no_overlap
1472 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1473 to group if check_no_overlap fails.
1475 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1477 * config/rs6000/altivec.md: Delete contraint arguments to
1478 define_expand, define_split, and define_peephole2, and in
1479 define_insn_and_split if always unused.
1480 * config/rs6000/darwin.md: Ditto.
1481 * config/rs6000/dfp.md: Ditto.
1482 * config/rs6000/rs6000.md: Ditto.
1483 * config/rs6000/sync.md: Ditto.
1484 * config/rs6000/vector.md: Ditto.
1485 * config/rs6000/vsx.md: Ditto.
1487 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1489 * config/rs6000/altivec.md: Write output control strings as braced
1490 blocks instead of double-quoted strings.
1491 * config/rs6000/darwin.md: Ditto.
1492 * config/rs6000/rs6000.md: Ditto.
1493 * config/rs6000/vector.md: Ditto.
1494 * config/rs6000/vsx.md: Ditto.
1496 2018-02-21 Jason Merrill <jason@redhat.com>
1498 PR c++/84314 - ICE with templates and fastcall attribute.
1499 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1501 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1503 * ipa-cp.c (determine_versionability): Fix comment typos.
1505 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1508 * ipa-cp.c (determine_versionability): Do not version functions caling
1511 2018-02-21 Martin Liska <mliska@suse.cz>
1514 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1515 Add "native" as a possible value.
1516 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1517 the macro when native cpu detection is available.
1519 2018-02-21 Martin Liska <mliska@suse.cz>
1522 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1523 Add "native" as a possible value.
1524 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1525 when native cpu detection is available.
1527 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1528 Martin Sebor <msebor@redhat.com>
1530 PR tree-optimization/84478
1531 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1533 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1534 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1535 support which is conservatively correct, for 2 only stay conservative
1536 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1537 argument to the 2 argument get_range_strlen, adjust 6 arg
1538 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1540 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1541 (gimple_fold_builtin_strlen): Pass true as last argument to
1544 2018-02-20 Martin Sebor <msebor@redhat.com>
1547 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1548 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1549 (builtin_memref::builtin_memref): Factor out parts into
1550 set_base_and_offset and call it.
1552 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1555 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1556 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1557 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1558 search at the associated MODE_INT.
1560 2018-02-20 Jeff Law <law@redhat.com>
1563 PR tree-optimization/81592
1565 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1566 for range data rather than using global data.
1567 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1568 range data rather than using global data.
1569 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1570 pass it to children as needed.
1571 (struct directive::fmtresult): Similarly.
1572 (struct directive::set_width): Similarly.
1573 (struct directive::set_precision): Similarly.
1574 (format_integer, format_directive, parse_directive): Similarly.
1575 (format_none): Accept unnamed vr_values parameter.
1576 (format_percent, format_floating, format_character): Similarly.
1577 (format_string, format_plain): Similarly.
1578 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1579 the EVRP range analyzer for range data rather than using global data.
1580 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1581 gimple-ssa-evrp-analyze.h
1582 (class sprintf_dom_walker): Add after_dom_children member function.
1583 Add evrp_range_analyzer member.
1584 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1585 range analyzer as needed.
1586 (sprintf_dom_walker::after_dom_children): New member function.
1587 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1589 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1590 (evrp_range_analyzer::pop_to_marker): Likewise.
1592 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1594 PR tree-optimization/84419
1595 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1596 with the required type if its current type is compatible but
1599 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1602 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1603 after vectorization.
1605 2018-02-20 Martin Liska <mliska@suse.cz>
1608 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1609 possible values if we don't have a hint.
1611 2018-02-20 Martin Liska <mliska@suse.cz>
1615 * final.c (shorten_branches): Build align_tab array with one
1617 * opts.c (finish_options): Add alignment option limit check.
1618 (MAX_CODE_ALIGN): Likewise.
1619 (MAX_CODE_ALIGN_VALUE): Likewise.
1620 * doc/invoke.texi: Document maximum allowed option value for
1621 all -falign-* options.
1623 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1626 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1627 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1628 * var-tracking.c (emit_note_insn_var_location): Remove all references
1629 to NOTE_INSN_CALL_ARG_LOCATION.
1630 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1631 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1632 Use copy_rtx_if_shared.
1633 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1634 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1635 (dwarf2out_var_location): Remove handling of
1636 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1638 * final.c (final_scan_insn): Remove all references to
1639 NOTE_INSN_CALL_ARG_LOCATION.
1640 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1641 before dumping final insns.
1642 * except.c (emit_note_eh_region_end): Remove all references to
1643 NOTE_INSN_CALL_ARG_LOCATION.
1644 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1645 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1646 * config/arc/arc.c (hwloop_optimize): Likewise.
1647 * config/arm/arm.c (create_fix_barrier): Likewise.
1648 * config/s390/s390.c (s390_chunkify_start): Likewise.
1649 * config/sh/sh.c (find_barrier): Likewise.
1650 * config/i386/i386.c (rest_of_insert_endbranch,
1651 ix86_seh_fixup_eh_fallthru): Likewise.
1652 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1653 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1654 * config/frv/frv.c (frv_function_prologue): Likewise.
1655 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1657 (note_outside_basic_block_p): Remove all references to
1658 NOTE_INSN_CALL_ARG_LOCATION.
1659 * gengtype.c (adjust_field_rtx_def): Likewise.
1660 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1662 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1663 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1666 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1669 PR tree-optimization/84452
1670 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1671 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1674 2018-02-19 Martin Liska <mliska@suse.cz>
1677 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1679 2018-02-19 Martin Liska <mliska@suse.cz>
1680 Richard Sandiford <richard.sandiford@linaro.org>
1682 PR tree-optimization/82491
1683 * gimple-fold.c (get_base_constructor): Make earlier bail out
1686 2018-02-19 Carl Love <cel@us.ibm.com>
1688 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1689 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1691 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1692 P8V_BUILTIN_VEC_NEG.
1694 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1696 * config/rl78/rl78.md (movdf): New define expand.
1698 2018-02-19 Martin Liska <mliska@suse.cz>
1701 * doc/invoke.texi: Fix typo.
1702 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1704 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1706 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1707 handle rs6000_single_float and rs6000_double_float specially for
1710 2018-02-16 Jeff Law <law@redhat.com>
1712 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1715 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1718 * ipa-inline.c (inline_small_functions): Fix a typo.
1720 2018-02-16 Nathan Sidwell <nathan@acm.org>
1722 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1724 2018-02-16 Carl Love <cel@us.ibm.com>
1726 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1727 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1728 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1729 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1730 expansion to P8V_BUILTIN_VEC_FLOAT2.
1732 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1734 PR rtl-optimization/70023
1735 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1736 src_regno into account.
1738 2018-02-16 Carl Love <cel@us.ibm.com>
1740 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1741 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1742 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1743 * config/rs6000/rs6000.c: Remove case statements for
1744 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1745 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1746 and P9V_BUILTIN_VEC_VINSERT4B.
1747 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1748 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1749 * config/rs6000/vsx.md:
1750 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1753 2018-02-16 Carl Love <cel@us.ibm.com>
1755 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1757 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1759 * config/rs6000/rs6000-c.c: Add the definitions for
1760 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1761 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1762 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1763 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1764 definition for insert4b and define insn *insert3b_internal.
1765 * doc/extend.texi: Add documentation for vec_extract4b.
1767 2018-02-16 Nathan Sidwell <nathan@acm.org>
1769 * doc/extend.texi (Backwards Compatibility): Mention friend
1770 injection. Note for-scope is deprecated.
1771 * doc/invoke.texi (-ffriend-injection): Deprecate.
1773 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1775 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1776 that moved to I2, also allow destinations that are a paradoxical
1777 subreg (instead of a normal reg).
1779 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1782 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1785 2018-02-16 Richard Biener <rguenther@suse.de>
1787 PR tree-optimization/84037
1788 PR tree-optimization/84016
1790 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1791 Adjust vec_construct for the fact we need additional higher latency
1792 128bit inserts for AVX256 and AVX512 vector builds.
1793 (ix86_add_stmt_cost): Scale vector construction cost for
1796 2018-02-16 Richard Biener <rguenther@suse.de>
1798 PR tree-optimization/84417
1799 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1800 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1801 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1803 2018-02-16 Martin Liska <mliska@suse.cz>
1806 * internal-fn.def (ASAN_CHECK): Set proper flags.
1807 (ASAN_MARK): Likewise.
1809 2018-02-16 Julia Koval <julia.koval@intel.com>
1811 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1812 from PTA_CANNONLAKE.
1814 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1817 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1818 Use ++iter rather than iter++ for std::list iterators.
1819 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1820 defer deleting them until all nodes in the forest are processed. Do
1821 free even leaf nodes. Change to_process into auto_vec.
1824 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1825 * vec.h (vec_default_construct): Use memset instead of placement new
1826 if BROKEN_VALUE_INITIALIZATION is defined.
1827 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1828 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1831 PR rtl-optimization/83723
1832 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1833 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1834 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1836 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1838 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1840 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1842 PR rtl-optimization/81443
1843 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1844 from inner REGs to paradoxical SUBREGs.
1846 2018-02-16 Richard Biener <rguenther@suse.de>
1848 PR tree-optimization/84399
1849 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1850 For operands we can analyze at their definition make sure we can
1851 analyze them at each use as well.
1853 2018-02-16 Richard Biener <rguenther@suse.de>
1855 PR tree-optimization/84190
1856 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1857 volatile accesses if the decl isn't volatile.
1859 2018-02-15 Jason Merrill <jason@redhat.com>
1861 PR c++/84314 - ICE with templates and fastcall attribute.
1862 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1863 TYPE_CANONICAL on an existing type.
1865 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1867 PR tree-optimization/84383
1868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1869 dstoff nor call operand_equal_p if dstbase is NULL.
1871 PR tree-optimization/84334
1872 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1873 also a CONSTANT_CLASS_P, punt.
1875 2018-02-14 Jim Wilson <jimw@sifive.com>
1877 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1878 first SMALL_OPERAND check. New local min_second_step. Move assert
1879 to where locals are set. Add TARGET_RVC support.
1880 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1882 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1884 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1886 2018-02-14 Martin Sebor <msebor@redhat.com>
1888 PR tree-optimization/83698
1889 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1890 arrays constrain the offset range to their bounds.
1891 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1892 (builtin_access::overlap): Avoid setting the size of overlap if it's
1894 (maybe_diag_overlap): Also consider arrays when deciding what values
1895 of offsets to include in diagnostics.
1897 2018-02-14 Martin Sebor <msebor@redhat.com>
1900 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1901 that correspond to the kind of a declaration.
1903 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1906 * config/pa/pa.md: Load address of PIC label using the linkage table
1907 if the label is nonlocal.
1909 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1911 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1912 warning message if user requests -maltivec=be.
1913 * doc/invoke.texi: Document deprecation of -maltivec=be.
1915 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1918 * config/rs6000/rs6000-c.c: Update definitions for
1919 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1920 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1922 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1925 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1926 add _get_ssp intrinsics. Remove argument from
1927 __builtin_ia32_rdssp[d|q].
1928 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1929 * config/i386/i386-builtin.def: Remove argument from
1930 __builtin_ia32_rdssp[d|q].
1931 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1932 ix86_expand_special_args_builtin for _rdssp[d|q].
1933 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1934 Clear register before usage.
1935 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1936 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1938 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1940 PR tree-optimization/84357
1941 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1942 operand 1 of an ARRAY_REF too.
1944 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1947 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1948 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1950 (set_of_reg): New struct.
1951 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1952 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1953 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1955 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1956 Split into bitclr, bitset, bitinvert patterns if appropriate.
1957 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1958 use rx_fuse_in_memory_bitop.
1959 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1960 to named insn, correct maximum insn length.
1962 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1965 * machmode.def: Define a complex mode for PARTIAL_INT.
1966 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1968 * doc/rtl.texi: Document CSPImode.
1969 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1971 (msp430_hard_regno_nregs_with_padding): Likewise.
1973 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1976 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1978 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1980 PR rtl-optimization/84169
1981 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1982 we generated a parallel as new i3 and we split that to new i2 and i3
1983 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1984 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1985 those to i2, not i1. Partially rewrite this scan code.
1987 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1990 * stor-layout.c (place_field): For variable length fields, adjust
1991 offset_align afterwards not just based on the field's alignment,
1992 but also on the size.
1995 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1996 of exps and logs in the use_exp2 case.
1998 2018-02-13 Jeff Law <law@redhat.com>
2000 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2003 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2006 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2010 * common.opt (gas-loc-support, gas-locview-support): New.
2011 (ginline-points, ginternal-reset-location-views): New.
2012 * doc/invoke.texi: Document them. Use @itemx where intended.
2013 (gvariable-location-views): Adjust.
2014 * target.def (reset_location_view): New.
2015 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2016 (TARGET_RESET_LOCATION_VIEW): New.
2017 * doc/tm.texi: Rebuilt.
2018 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2019 (dwarf2out_default_as_locview_support): New.
2020 (output_asm_line_debug_info): Use option variables.
2021 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2022 (output_loc_list): Likewise.
2023 (add_high_low_attributes): Check option variables.
2024 Don't output entry view attribute in strict mode.
2025 (gen_inlined_subroutine_die): Check option variables.
2026 (dwarf2out_inline_entry): Likewise.
2027 (init_sections_and_labels): Likewise.
2028 (dwarf2out_early_finish): Likewise.
2029 (maybe_reset_location_view): New, from...
2030 (dwarf2out_var_location): ... here. Call it.
2031 * debug.h (dwarf2out_default_as_loc_support): Declare.
2032 (dwarf2out_default_as_locview_support): Declare.
2033 * hooks.c (hook_int_rtx_insn_0): New.
2034 * hooks.h (hook_int_rtx_insn_0): Declare.
2035 * toplev.c (process_options): Take -gas-loc-support and
2036 -gas-locview-support from dwarf2out. Enable
2037 -gvariable-location-views by default only with locview
2038 assembler support. Enable -ginternal-reset-location-views by
2039 default only if the target defines the corresponding hook.
2040 Enable -ginline-points by default if location views are
2041 enabled; force it disabled if statement frontiers are
2043 * tree-inline.c (expand_call_inline): Check option variables.
2044 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2046 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2048 PR tree-optimization/84321
2049 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2050 handling. Also check whether the anti-range contains any values
2051 that satisfy the mask; switch to a VR_RANGE if not.
2053 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2056 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2058 2018-02-13 Martin Jambor <mjambor@suse.cz>
2061 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2062 of call statements, also set location of a load to a temporary.
2064 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2066 * config/rl78/rl78.c (add_vector_labels): New function.
2067 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2068 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2069 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2070 which checks that no arguments are passed.
2071 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2072 * doc/extend.texi: Documentation for the new attribute.
2074 2018-02-13 Andreas Schwab <schwab@suse.de>
2076 * config/riscv/linux.h (CPP_SPEC): Define.
2078 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2081 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2082 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2083 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2084 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2085 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2086 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2087 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2088 temporarily for AES and PCLMUL builtins.
2090 PR tree-optimization/84339
2091 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2092 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2096 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2097 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2098 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2100 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2102 * omp-simd-clone.h: New file.
2103 * omp-simd-clone.c: Include omp-simd-clone.h.
2104 (expand_simd_clones): No longer static.
2105 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2106 cgraph.h and omp-simd-clone.h.
2107 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2108 (vect_recog_widen_shift_pattern): Formatting fix.
2109 (vect_pattern_recog_1): Don't check optab for calls.
2112 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2113 operands[2] into a REG before using gen_lowpart on it.
2115 2018-02-12 Jeff Law <law@redhat.com>
2118 * config/sh/sh.c (find_barrier): Consider a sibling call
2121 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2122 successfully back substituting a reg.
2124 2018-02-12 Richard Biener <rguenther@suse.de>
2126 PR tree-optimization/84037
2127 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2128 parameter, move visited init to caller.
2129 (vect_slp_analyze_operations): Separate cost from validity
2130 check, initialize visited once for all instances.
2131 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2133 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2135 (vect_model_promotion_demotion_cost): Likewise.
2136 (vectorizable_bswap): Guard cost modeling with !slp_node
2137 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2139 (vectorizable_call): Likewise.
2140 (vectorizable_conversion): Likewise.
2141 (vectorizable_assignment): Likewise.
2142 (vectorizable_shift): Likewise.
2143 (vectorizable_operation): Likewise.
2144 (vectorizable_store): Likewise.
2145 (vectorizable_load): Likewise.
2146 (vectorizable_condition): Likewise.
2147 (vectorizable_comparison): Likewise.
2149 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2152 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2153 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2154 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2156 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2159 * params.def (inline-min-speedup): Increase from 8 to 15.
2160 (max-inline-insns-auto): Decrease from 40 to 30.
2161 * ipa-split.c (consider_split): Add some buffer for function to
2162 be considered inlining candidate.
2163 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2166 2018-02-12 Richard Biener <rguenther@suse.de>
2168 PR tree-optimization/84037
2169 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2170 matched stmts if we cannot swap the non-matched ones.
2172 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2174 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2175 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2176 _mm_maskz_scalef_round_ss): New intrinsics.
2177 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2178 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2179 __builtin_ia32_scalefss_round): Remove.
2180 (__builtin_ia32_scalefsd_mask_round,
2181 __builtin_ia32_scalefss_mask_round): New intrinsics.
2182 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2183 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2184 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2185 "<round_constraint>")): Changed to ...
2186 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2187 "<round_scalar_constraint>")): ... this.
2188 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2189 %0, %1, %2<round_op3>}"): Changed to ...
2190 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2191 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2192 %2<round_scalar_mask_op3>}"): ... this.
2193 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2195 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2197 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2198 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2199 (_mm_maskz_sqrt_round_ss): New intrinsics.
2200 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2201 (__builtin_ia32_sqrtsd_mask_round)
2202 (__builtin_ia32_sqrtss_mask_round): New builtins.
2203 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2204 (__builtin_ia32_sqrtss_round): Remove.
2205 (__builtin_ia32_sqrtsd_mask_round)
2206 (__builtin_ia32_sqrtss_mask_round): New builtins.
2207 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2208 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2209 ((match_operand:VF_128 1 "vector_operand"
2210 "xBm,<round_constraint>")): Changed to ...
2211 ((match_operand:VF_128 1 "vector_operand"
2212 "xBm,<round_scalar_constraint>")): ... this.
2213 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2214 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2215 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2216 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2217 %<iptr>1<round_scalar_mask_op3>}): ... this.
2218 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2219 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2221 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2224 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2225 Cast vec_cmpeq result to correct type.
2226 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2227 Cast vec_cmpgt result to correct type.
2229 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2231 * final.c (final_scan_insn_1): Renamed from...
2232 (final_scan_insn): ... this. New wrapper, to recover
2233 seen from the outermost call in recursive ones.
2234 * config/sparc/sparc.c (output_return): Drop seen from call.
2235 (output_sibcall): Likewise.
2236 * config/visium/visium.c (output_branch): Likewise.
2238 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2240 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2243 2018-02-10 Alan Modra <amodra@gmail.com>
2246 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2247 Specify LR as an input.
2249 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2252 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2253 remove_member_access_dummy_vars): New functions.
2254 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2255 lower_omp_1, execute_lower_omp): Use them.
2257 PR rtl-optimization/84308
2258 * shrink-wrap.c (spread_components): Release todo vector.
2260 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2262 PR rtl-optimization/57193
2263 * ira-color.c (struct allocno_color_data): Add member
2264 conflict_allocno_hard_prefs.
2265 (update_conflict_allocno_hard_prefs): New.
2266 (bucket_allocno_compare_func): Add a preference based on
2267 conflict_allocno_hard_prefs.
2268 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2269 (color_allocnos): Remove a dead code. Initiate
2270 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2272 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2275 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2276 constraint from =wa to wa. Avoid a subreg on the output operand,
2277 instead use a pseudo and subreg it in a move.
2278 (p9_xxbrd_<mode>): Changed to ...
2279 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2280 (p9_xxbrd_v2df): New expander.
2281 (p9_xxbrw_<mode>): Changed to ...
2282 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2283 (p9_xxbrw_v4sf): New expander.
2285 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2287 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2289 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2292 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2293 multiply in 32-bit mode.
2294 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2295 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2298 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2300 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2301 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2302 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2303 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2305 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2308 * dwarf2out.c (is_trivial_indirect_ref): New function.
2309 (dwarf2out_late_global_decl): Do not generate a location
2310 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2311 and that are not defined in the current unit.
2313 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2315 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2316 instead of a libcall for UNORDERED.
2318 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2321 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2322 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2324 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2327 * config/s390/s390.c (s390_set_current_function): Invoke
2328 s390_indirect_branch_settings also if fndecl didn't change.
2330 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2332 * config/rs6000/rs6000.md (blockage): Set length to zero.
2334 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2336 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2338 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2341 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2342 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2346 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2347 PARALLEL incoming that failed vt_get_decl_and_offset check.
2350 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2351 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2353 (get_variable_section): For decls in named .bss* sections pass true as
2354 second argument to bss_initializer_p.
2356 2018-02-09 Marek Polacek <polacek@redhat.com>
2357 Jakub Jelinek <jakub@redhat.com>
2360 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2361 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2362 Sync some changes from cxx_fold_indirect_ref.
2364 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2366 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2368 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2369 (BLOCK_INLINE_ENTRY_LABEL): New.
2370 (dwarf2out_var_location): Disregard inline entry markers.
2371 (inline_entry_data): New struct.
2372 (inline_entry_data_hasher): New hashtable type.
2373 (inline_entry_data_hasher::hash): New.
2374 (inline_entry_data_hasher::equal): New.
2375 (inline_entry_data_table): New variable.
2376 (add_high_low_attributes): Add DW_AT_entry_pc and
2377 DW_AT_GNU_entry_view attributes if a pending entry is found
2378 in inline_entry_data_table. Add old entry_pc attribute only
2379 if debug nonbinding markers are disabled.
2380 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2381 markers are enabled.
2382 (block_within_block_p, dwarf2out_inline_entry): New.
2383 (dwarf2out_finish): Check that no entries remained in
2384 inline_entry_data_table.
2385 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2386 (final_scan_insn, notice_source_line): Likewise.
2387 (rest_of_clean_state): Skip inline entry markers.
2388 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2390 * gimple.c (gimple_build_debug_inline_entry): New.
2391 * gimple.h (enum gimple_debug_subcode): Add
2392 GIMPLE_DEBUG_INLINE_ENTRY.
2393 (gimple_build_debug_inline_entry): Declare.
2394 (gimple_debug_inline_entry_p): New.
2395 (gimple_debug_nonbind_marker_p): Adjust.
2396 * insn-notes.def (INLINE_ENTRY): New.
2397 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2398 inline entry marker notes.
2399 (print_insn): Likewise.
2400 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2401 (INSN_DEBUG_MARKER_KIND): Likewise.
2402 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2403 * tree-inline.c (expand_call_inline): Build and insert
2404 debug_inline_entry stmt.
2405 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2406 inline entry blocks early, if nonbind markers are enabled.
2407 (dump_scope_block): Dump fragment info.
2408 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2409 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2410 (gimple_build_debug_inline_entry): New.
2411 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2412 Enable/disable inline entry points too.
2413 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2414 (DEBUG_INSN): Describe inline entry markers.
2416 * common.opt (gvariable-location-views): New.
2417 (gvariable-location-views=incompat5): New.
2418 * config.in: Rebuilt.
2419 * configure: Rebuilt.
2420 * configure.ac: Test assembler for view support.
2421 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2422 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2423 * dwarf2out.c (var_loc_view): New typedef.
2424 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2425 (dwarf2out_locviews_in_attribute): New.
2426 (dwarf2out_locviews_in_loclist): New.
2427 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2428 (enum dw_line_info_opcode): Add LI_adv_address.
2429 (struct dw_line_info_table): Add view.
2430 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2431 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2432 (zero_view_p): New variable.
2433 (ZERO_VIEW_P): New macro.
2434 (output_asm_line_debug_info): New.
2435 (struct var_loc_node): Add view.
2436 (add_AT_view_list, AT_loc_list): New.
2437 (add_var_loc_to_decl): Add view param. Test it against last.
2438 (new_loc_list): Add view params. Record them.
2439 (AT_loc_list_ptr): Handle loc and view lists.
2440 (view_list_to_loc_list_val_node): New.
2441 (print_dw_val): Handle dw_val_class_view_list.
2442 (size_of_die): Likewise.
2443 (value_format): Likewise.
2444 (loc_list_has_views): New.
2445 (gen_llsym): Set vl_symbol too.
2446 (maybe_gen_llsym, skip_loc_list_entry): New.
2447 (dwarf2out_maybe_output_loclist_view_pair): New.
2448 (output_loc_list): Output view list or entries too.
2449 (output_view_list_offset): New.
2450 (output_die): Handle dw_val_class_view_list.
2451 (output_dwarf_version): New.
2452 (output_compilation_unit_header): Use it.
2453 (output_skeleton_debug_sections): Likewise.
2454 (output_rnglists, output_line_info): Likewise.
2455 (output_pubnames, output_aranges): Update version comments.
2456 (output_one_line_info_table): Output view numbers in asm comments.
2457 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2458 Call maybe_gen_llsym.
2459 (loc_list_from_tree_1): Adjust.
2460 (add_AT_location_description): Create view list attribute if
2461 needed, check it's absent otherwise.
2462 (convert_cfa_to_fb_loc_list): Adjust.
2463 (maybe_emit_file): Call output_asm_line_debug_info for test.
2464 (dwarf2out_var_location): Reset views as needed. Precompute
2465 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2466 attribute. Set view.
2467 (new_line_info_table): Reset next view.
2468 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2469 (dwarf2out_source_line): Likewise. Output view resets and labels to
2470 the assembler, or select appropriate line info opcodes.
2471 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2472 (optimize_string_length): Catch it. Adjust.
2473 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2474 dw_val_class_view_list, and remove it if no longer needed.
2475 (hash_loc_list): Hash view numbers.
2476 (loc_list_hasher::equal): Compare them.
2477 (optimize_location_lists): Check whether a view list symbol is
2478 needed, and whether the locview attribute is present, and
2479 whether they match. Remove the locview attribute if no longer
2481 (index_location_lists): Call skip_loc_list_entry for test.
2482 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2483 Use output_dwarf_version.
2484 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2485 (struct dw_val_node): Add val_view_list.
2486 * final.c (SEEN_NEXT_VIEW): New.
2487 (set_next_view_needed): New.
2488 (clear_next_view_needed): New.
2489 (maybe_output_next_view): New.
2490 (final_start_function): Rename to...
2491 (final_start_function_1): ... this. Take pointer to FIRST,
2492 add SEEN parameter. Emit param bindings in the initial view.
2493 (final_start_function): Reintroduce SEEN-less interface.
2494 (final): Rename to...
2495 (final_1): ... this. Take SEEN parameter. Output final pending
2496 next view at the end.
2497 (final): Reintroduce seen-less interface.
2498 (final_scan_insn): Output pending next view before switching
2499 sections or ending a block. Mark the next view as needed when
2500 outputting variable locations. Notify debug backend of section
2501 changes, and of location view changes.
2502 (rest_of_handle_final): Adjust.
2503 * toplev.c (process_options): Autodetect value for debug variable
2504 location views option. Warn on incompat5 without -gdwarf-5.
2505 * doc/invoke.texi (gvariable-location-views): New.
2506 (gvariable-location-views=incompat5): New.
2507 (gno-variable-location-views): New.
2509 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2511 PR tree-optimization/84136
2512 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2513 that the result of find_edge is non-NULL.
2515 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2518 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2519 storing integer register in SImode. Fix cost of 256 and 512
2520 byte aligned SSE register store.
2522 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2524 * config/i386/i386.c (ix86_multiplication_cost): Fix
2525 multiplication cost for TARGET_AVX512DQ.
2527 2018-02-08 Marek Polacek <polacek@redhat.com>
2529 PR tree-optimization/84238
2530 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2533 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2535 PR tree-optimization/84265
2536 * tree-vect-stmts.c (vectorizable_store): Don't treat
2537 VMAT_CONTIGUOUS accesses as grouped.
2538 (vectorizable_load): Likewise.
2540 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2542 PR tree-optimization/81635
2543 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2544 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2545 (test_round_for_mask): New functions.
2546 (wide_int_cc_tests): Call test_round_for_mask.
2547 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2548 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2549 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2550 range returned by get_range_info.
2552 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2555 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2556 * symtab.c: Include builtins.h
2557 (symtab_node::output_to_lto_symbol_table_p): Move here
2558 from lto-streamer-out.c:output_symbol_p.
2559 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2560 (output_symbol_p): Move all logic to symtab.c
2561 (produce_symtab): Update.
2563 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2565 * config/s390/s390-opts.h (enum indirect_branch): Define.
2566 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2567 (s390_indirect_branch_via_thunk)
2568 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2569 (enum s390_indirect_branch_type): Define.
2570 * config/s390/s390.c (struct s390_frame_layout, struct
2571 machine_function): Remove.
2572 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2573 (indirect_branch_table_label_no, indirect_branch_table_name):
2575 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2576 (enum s390_indirect_branch_option): Define.
2577 (s390_return_addr_from_memory): New function.
2578 (s390_handle_string_attribute): New function.
2579 (s390_attribute_table): Add new attribute handler.
2580 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2581 (s390_indirect_branch_via_thunk): New function.
2582 (s390_indirect_branch_via_inline_thunk): New function.
2583 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2584 sibling call optimization for non z10 compiles.
2585 (s390_emit_call): Force indirect branch target to be a single
2586 register. Add r1 clobber for non-z10 compiles.
2587 (s390_emit_epilogue): Emit return jump via return_use expander.
2588 (s390_reorg): Handle JUMP_INSNs as execute targets.
2589 (s390_option_override_internal): Perform validity checks for the
2590 new command line options.
2591 (s390_indirect_branch_attrvalue): New function.
2592 (s390_indirect_branch_settings): New function.
2593 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2594 (s390_output_indirect_thunk_function): New function.
2595 (s390_code_end): Implement target hook.
2596 (s390_case_values_threshold): Implement target hook.
2597 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2599 * config/s390/s390.h (struct s390_frame_layout)
2600 (struct machine_function): Move here from s390.c.
2601 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2602 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2603 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2604 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2605 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2606 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2607 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2608 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2609 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2610 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2611 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2612 (mnemonic attribute): Add values which aren't recognized
2614 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2615 pattern for branch conversion. Fix mnemonic attribute.
2616 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2617 indirect branch via thunk if requested.
2618 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2619 ("*indirect_jump"): Disable for branch conversion using out of
2621 ("indirect_jump_via_thunk<mode>_z10")
2622 ("indirect_jump_via_thunk<mode>")
2623 ("indirect_jump_via_inlinethunk<mode>_z10")
2624 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2625 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2626 ("casesi_jump_via_inlinethunk<mode>_z10")
2627 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2628 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2629 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2630 ("*indirect2_jump"): Disable for branch conversion.
2631 ("casesi_jump"): Turn into expander and expand patterns for branch
2633 ("return_use"): New expander.
2634 ("*return"): Emit return via thunk and rename it to ...
2635 ("*return<mode>"): ... this one.
2636 * config/s390/s390.opt: Add new options and and enum for the
2639 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2641 * lra-constraints.c (match_reload): Unconditionally use
2642 gen_lowpart_SUBREG, rather than selecting between that
2643 and equivalent gen_rtx_SUBREG code.
2645 2018-02-08 Richard Biener <rguenther@suse.de>
2647 PR tree-optimization/84233
2648 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2649 changed flag instead of boguously re-using phi_inserted.
2651 2018-02-08 Martin Jambor <mjambor@suse.cz>
2653 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2654 static local variables.
2656 2018-02-08 Richard Biener <rguenther@suse.de>
2658 PR tree-optimization/84278
2659 * tree-vect-stmts.c (vectorizable_store): When looking for
2660 smaller vector types to perform grouped strided loads/stores
2661 make sure the mode is supported by the target.
2662 (vectorizable_load): Likewise.
2664 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2666 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2667 Increase LDP/STP opportunities by adding adjacent callee-saves.
2669 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2671 PR rtl-optimization/84068
2672 PR rtl-optimization/83459
2673 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2675 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2677 PR tree-optimization/84224
2678 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2679 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2682 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2685 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2686 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2687 count, remove one USE.
2689 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2691 * doc/install.texi (Configuration): Document the
2692 --with-long-double-format={ibm,ieee} PowerPC configuration
2696 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2697 Convert from define_expand to be define_insn_and_split. Rework
2698 float/double/_Float128 conversions to QI/HI/SImode to work with
2699 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2700 conversions to QI/HImode types did a store and then a load to
2701 truncate the value. For conversions to VSX registers, don't split
2702 the insn, instead emit the code directly. Use the code iterator
2703 any_fix to combine signed and unsigned conversions.
2704 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2705 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2706 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2707 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2708 (fix_<mode>di2_hw): Likewise.
2709 (fixuns_<mode>di2_hw): Likewise.
2710 (fix_<mode>si2_hw): Likewise.
2711 (fixuns_<mode>si2_hw): Likewise.
2712 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2713 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2714 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2715 fix<uns>_trunc<SFDF:mode>si2_p8.
2716 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2718 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2719 (fix<uns>_<mode>_mem): Likewise.
2720 (fctiw<u>z_<mode>_mem): Likewise.
2721 (fix<uns>_<mode>_mem): Likewise.
2722 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2723 the register allocator from doing a direct move to the GPRs to do
2724 a store, and instead use the ISA 3.0 store byte/half-word from
2725 vector register instruction. For IEEE 128-bit floating point,
2726 also optimize stores of 32-bit ints.
2727 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2729 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2731 * genextract.c (push_pathstr_operand): New function to support
2733 (walk_rtx): Call push_pathstr_operand.
2734 (print_path): Support [a-zA-Z].
2736 2018-02-07 Richard Biener <rguenther@suse.de>
2738 PR tree-optimization/84037
2739 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2740 (cse_and_gimplify_to_preheader): Declare.
2741 (vect_get_place_in_interleaving_chain): Likewise.
2742 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2744 (_loop_vec_info::~_loop_vec_info): Delete it.
2745 (cse_and_gimplify_to_preheader): New function.
2746 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2747 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2748 (vectorizable_load): Likewise. For grouped stores always base
2749 the IV on the first element.
2750 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2751 condition before gimplifying.
2753 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2755 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2756 *DIV_EXPR and *MOD_EXPR.
2758 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2761 * config/i386/i386.c (ix86_option_override_internal): Mask out
2762 the CF_SET bit when checking -fcf-protection.
2764 2018-02-07 Tom de Vries <tom@codesourcery.com>
2767 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2770 2018-02-07 Richard Biener <rguenther@suse.de>
2772 PR tree-optimization/84204
2773 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2776 PR tree-optimization/84205
2777 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2778 special-case isl_ast_op_zdiv_r.
2780 PR tree-optimization/84223
2781 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2782 Only add conditions from within the region.
2783 (gather_bbs::after_dom_children): Adjust.
2785 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2788 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2789 * config/avr/avr.md: Only post-reload split REG-REG moves if
2790 either register is GENERAL_REG_P.
2792 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2794 PR tree-optimization/84235
2795 * tree-ssa-scopedtables.c
2796 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2797 if the subtraction is performed in floating point type where NaNs are
2798 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2799 build 1. Formatting fix.
2801 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2804 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2805 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2806 and skip it regardless of bb boundaries. Use CALL_P macro,
2807 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2808 unnecessarily, formatting fix.
2810 2018-02-06 Michael Collison <michael.collison@arm.com>
2812 * config/arm/thumb2.md:
2813 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2814 (*thumb_mov_notscc): Ditto.
2816 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2819 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2820 unsigned_fix, not "s".
2822 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2824 * configure.ac (gcc_fn_eh_frame_ro): New function.
2825 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2826 correct .eh_frame permissions.
2827 * configure: Regenerate.
2829 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2831 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2834 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2836 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2837 Display warning message for -mno-speculate-indirect-jumps.
2839 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2841 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2843 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2845 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2847 PR tree-optimization/84225
2848 * tree-eh.c (find_trapping_overflow): Only call
2849 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2851 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2854 * config/i386/i386.c: Reimplement the check of possible options
2855 -mibt/-mshstk conbination. Change error messages.
2856 * doc/invoke.texi: Fix a typo: remove extra '='.
2858 2018-02-06 Marek Polacek <polacek@redhat.com>
2860 PR tree-optimization/84228
2861 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2863 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2866 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2867 emitted arch directives.
2868 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2869 __ARM_FEATURE_COPROC before changing architectures.
2871 2018-02-06 Richard Biener <rguenther@suse.de>
2873 * config/i386/i386.c (print_reg): Fix typo.
2874 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2876 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2878 * configure: Regenerate.
2880 2018-02-05 Martin Sebor <msebor@redhat.com>
2882 PR tree-optimization/83369
2883 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2886 2018-02-05 Martin Liska <mliska@suse.cz>
2888 * doc/invoke.texi: Cherry-pick upstream r323995.
2890 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2892 * ira.c (ira_init_register_move_cost): Adjust comment.
2894 2018-02-05 Martin Liska <mliska@suse.cz>
2896 PR gcov-profile/84137
2897 * doc/gcov.texi: Fix typo in documentation.
2899 2018-02-05 Martin Liska <mliska@suse.cz>
2901 PR gcov-profile/83879
2902 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2903 using dlopen functionality.
2905 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2907 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2908 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2909 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2910 _mm_maskz_range_round_ss): New intrinsics.
2911 (__builtin_ia32_rangesd128_round)
2912 (__builtin_ia32_rangess128_round): Remove.
2913 (__builtin_ia32_rangesd128_mask_round,
2914 __builtin_ia32_rangess128_mask_round): New builtins.
2915 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2916 __builtin_ia32_rangess128_round): Remove.
2917 (__builtin_ia32_rangesd128_mask_round,
2918 __builtin_ia32_rangess128_mask_round): New builtins.
2919 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2920 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2921 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2922 "<round_saeonly_constraint>")): Changed to ...
2923 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2924 "<round_saeonly_scalar_constraint>")): ... this.
2925 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2926 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2927 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2928 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2929 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2931 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2933 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2935 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2936 Remove all values except native, 8540 and 8548.
2938 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2940 * config/i386/i386.c (ix86_output_function_return): Pass
2941 INVALID_REGNUM, instead of -1, as invalid register number to
2942 indirect_thunk_name and output_indirect_thunk.
2944 2018-02-02 Julia Koval <julia.koval@intel.com>
2946 * config.gcc: Add -march=icelake.
2947 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2948 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2949 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2950 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2951 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2952 (processor_target_table): Add icelake.
2953 (ix86_option_override_internal): Handle new PTAs.
2954 (get_builtin_code_for_version): Handle icelake.
2955 (M_INTEL_COREI7_ICELAKE): New.
2956 (fold_builtin_cpu): Handle icelake.
2957 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2958 * doc/invoke.texi: Add -march=icelake.
2960 2018-02-02 Julia Koval <julia.koval@intel.com>
2962 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2963 to wide_int_bitmask.
2964 * wide-int-bitmask.h: New.
2966 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2969 * config/i386/i386.md: Replace Pmode with word_mode in
2970 builtin_setjmp_setup and builtin_longjmp to support x32.
2972 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2976 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2978 (rs6000_supported_cpu_names): New static variable.
2979 (linux_cpu_translation_table): Likewise.
2980 (elf_platform) <cpu>: Define new static variable and use it.
2981 Translate kernel AT_PLATFORM name to canonical name if needed.
2982 Error if platform name is unknown.
2984 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2987 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2989 2018-02-01 Jeff Law <law@redhat.com>
2992 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2993 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2994 the scratch if RELEASE_VIA_POP is false.
2995 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2996 If we have to save a temporary register, decrement SIZE appropriately.
2997 Pass new arguments to release_scratch_register_on_entry.
2998 (ix86_adjust_stack_and_probe): Likewise.
2999 (ix86_emit_probe_stack_range): Pass new arguments to
3000 release_scratch_register_on_entry.
3002 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3004 PR rtl-optimization/84157
3005 * combine.c (change_zero_ext): Use REG_P predicate in
3006 front of HARD_REGISTER_P predicate.
3008 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3010 * config/avr/avr.c (avr_option_override): Move disabling of
3011 -fdelete-null-pointer-checks to...
3012 * common/config/avr/avr-common.c (avr_option_optimization_table):
3015 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3017 PR tree-optimization/81635
3018 * tree-data-ref.c (split_constant_offset_1): For types that
3019 wrap on overflow, try to use range info to prove that wrapping
3022 2018-02-01 Renlin Li <renlin.li@arm.com>
3025 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3027 (aarch64_register_move_cost): Likewise.
3028 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3030 (REG_CLASS_NAMES): Likewise.
3031 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3032 TAILCALL_ADDR_REGS. Remove IP registers.
3033 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3035 2018-02-01 Richard Biener <rguenther@suse.de>
3037 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3038 for specifying RPO order and allow NULL for that.
3039 * domwalk.c (dom_walker::dom_walker): Likewise.
3040 (dom_walker::walk): Handle NULL RPO order.
3041 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3043 (rewrite_update_dom_walker): Likewise.
3044 (mark_def_dom_walker): Likewise.
3046 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3048 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3049 (aarch64_maybe_expand_sve_subreg_move): Declare.
3050 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3051 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3053 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3054 that are semantically a reverse operation.
3055 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3056 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3057 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3059 (aarch64_can_change_mode_class): For big-endian, forbid changes
3060 between two SVE modes if they have different element sizes.
3062 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3064 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3065 the TImode handling for big-endian targets.
3067 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3069 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3070 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3072 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3073 Remove BSWAP handing for big-endian targets and use the form of
3074 LD1RQ appropariate for the mode.
3076 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3078 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3079 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3082 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3085 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3086 check for operands that need to go through aarch64_sve_reload_be.
3088 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3090 PR tree-optimization/81661
3091 PR tree-optimization/84117
3092 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3093 * tree-eh.c: Include gimplify.h.
3094 (find_trapping_overflow, replace_trapping_overflow,
3095 rewrite_to_non_trapping_overflow): New functions.
3096 * tree-vect-loop.c: Include tree-eh.h.
3097 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3098 * tree-data-ref.c: Include tree-eh.h.
3099 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3101 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3103 PR rtl-optimization/84123
3104 * combine.c (change_zero_ext): Check if hard register satisfies
3105 can_change_dest_mode before calling gen_lowpart_SUBREG.
3107 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3110 * ira.c (ira_init_register_move_cost): Remove assert.
3112 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3114 PR rtl-optimization/84071
3115 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3116 * doc/tm.texi: Regenerate.
3118 2018-01-31 Richard Biener <rguenther@suse.de>
3120 PR tree-optimization/84132
3121 * tree-data-ref.c (analyze_miv_subscript): Properly
3122 check whether evolution_function_is_affine_multivariate_p
3123 before calling gcd_of_steps_may_divide_p.
3125 2018-01-31 Julia Koval <julia.koval@intel.com>
3128 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3129 * config/i386/i386.md (rdpid_rex64) New.
3130 (rdpid): Make 32bit only.
3132 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3135 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3136 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3138 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3141 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3143 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3145 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3147 PR rtl-optimization/84071
3148 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3149 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3151 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3153 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3154 (arc_attribute_table): Add 'aux' attribute.
3155 (arc_in_small_data_p): Consider aux like variables.
3156 (arc_is_aux_reg_p): New function.
3157 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3158 (arc_get_aux_arg): New function.
3159 (prepare_move_operands): Handle aux-register access.
3160 (arc_handle_aux_attribute): New function.
3161 * doc/extend.texi (ARC Variable attributes): Add subsection.
3163 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3165 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3166 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3167 (arc_attribute_table): Add 'uncached' attribute.
3168 (arc_print_operand): Print '.di' flag for uncached memory
3170 (arc_in_small_data_p): Do not consider for small data the uncached
3172 (arc_is_uncached_mem_p): New function.
3173 * config/arc/predicates.md (compact_store_memory_operand): Check
3174 for uncached memory accesses.
3175 (nonvol_nonimm_operand): Likewise.
3176 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3178 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3181 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3182 falign-loops=): Add Optimization flag.
3184 2018-01-30 Jeff Law <law@redhat.com>
3187 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3188 INT_REGISTERS_SAVED. Check it prior to calling
3189 get_scratch_register_on_entry.
3190 (ix86_adjust_stack_and_probe): Similarly.
3191 (ix86_emit_probe_stack_range): Similarly.
3192 (ix86_expand_prologue): Corresponding changes.
3194 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3197 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3198 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3200 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3203 * lra-constraints.c (curr_insn_transform): Process AND in the
3206 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3208 PR rtl-optimization/83986
3209 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3210 dependence against last_pending_memory_flush in addition to
3213 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3215 PR tree-optimization/81611
3216 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3219 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3222 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3225 2018-01-30 Richard Biener <rguenther@suse.de>
3226 Jakub Jelinek <jakub@redhat.com>
3228 PR tree-optimization/84111
3229 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3230 inner loops added during recursion, as they don't have up-to-date
3233 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3236 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3237 (can_inline_edge_by_limits_p): ... here.
3238 (can_early_inline_edge_p, check_callers,
3239 update_caller_keys, update_callee_keys, recursive_inlining,
3240 add_new_edges_to_heap, speculation_useful_p,
3241 inline_small_functions,
3242 inline_small_functions, flatten_function,
3243 inline_to_all_callers_1): Update.
3245 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3247 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3250 2018-01-30 Richard Biener <rguenther@suse.de>
3252 PR tree-optimization/83008
3253 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3254 invariant and constant vector uses in stmts when they need
3257 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3260 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3261 * configure: Regenerate.
3263 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3265 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3267 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3268 Use gen_rtx_REG rather than gen_lowpart.
3270 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3272 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3273 rather than 0 when creating partial subregs.
3275 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3277 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3280 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3283 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3284 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3285 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3286 flags. This restores the settings used before the 2017-07-24.
3287 Turning off pre increment/decrement/modify allows IVOPTS to
3288 optimize DF/SF loops where the index is an int.
3290 2018-01-29 Richard Biener <rguenther@suse.de>
3291 Kelvin Nilsen <kelvin@gcc.gnu.org>
3294 * tree-vect-stmts.c (vectorizable_call): Don't call
3295 targetm.vectorize_builtin_md_vectorized_function if callee is
3298 2018-01-22 Carl Love <cel@us.ibm.com>
3300 * doc/extend.tex: Fix typo in second arg in
3301 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3303 2018-01-29 Richard Biener <rguenther@suse.de>
3305 PR tree-optimization/84086
3306 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3307 (flush_ssaname_freelist): When SSA names were released reset
3308 the SCEV hash table.
3310 2018-01-29 Richard Biener <rguenther@suse.de>
3312 PR tree-optimization/84057
3313 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3314 removed paths when removing edges.
3316 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3318 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3319 -mfunction-return=@var{choice}.
3321 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3324 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3326 (layout::print_source_line): Likewise.
3327 (test_get_line_width_without_trailing_whitespace): Add test cases.
3329 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3332 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3335 2018-01-26 Jim Wilson <jimw@sifive.com>
3337 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3339 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3342 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3344 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3345 and CMP + SUB-immediate -> SUBS.
3347 2018-01-26 Martin Sebor <msebor@redhat.com>
3349 PR tree-optimization/83896
3350 * tree-ssa-strlen.c (get_string_len): Rename...
3351 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3352 Avoid assuming length is constant.
3353 (handle_char_store): Use HOST_WIDE_INT for string length.
3355 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3358 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3359 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3361 2018-01-26 Richard Biener <rguenther@suse.de>
3363 PR rtl-optimization/84003
3364 * dse.c (record_store): Only record redundant stores when
3365 the earlier store aliases at least all accesses the later one does.
3367 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3369 PR rtl-optimization/83985
3370 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3371 REG_CFA_RESTORE insns.
3372 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3373 REG_CFA_RESTORE insns here.
3376 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3377 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3379 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3381 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3382 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3383 (arc_init): Likewise.
3384 (arc_override_options): Likewise.
3385 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3387 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3389 * config/arc/arc.h (TARGET_DBNZ): Define.
3390 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3391 properly set the tune attribute.
3392 (dbnz): Use TARGET_DBNZ guard.
3393 * config/arc/arc.opt (mtune): Add core3 option.
3395 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3397 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3398 recognize new pic like addresses.
3399 (arc_delegitimize_address): Clean up.
3401 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3403 * config/arc/arc-arches.def: Option mrf16 valid for all
3405 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3406 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3407 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3408 * config/arc/arc-tables.opt: Regenerate.
3409 * config/arc/arc.c (arc_conditional_register_usage): Handle
3410 reduced register file case.
3411 (arc_file_start): Set must have build attributes.
3412 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3414 * config/arc/arc.opt (mrf16): Add new option.
3415 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3416 * config/arc/genmultilib.awk: Handle new mrf16 option.
3417 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3418 * config/arc/t-multilib: Regenerate.
3419 * doc/invoke.texi (ARC Options): Document mrf16 option.
3421 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3423 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3424 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3425 (arc_attribute_table): Add 'secure_call' attribute.
3426 (arc_print_operand): Print secure call operand.
3427 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3429 (arc_is_secure_call_p): New function. * config/arc/arc.md
3430 (call_i): Add support for sjli instruction.
3431 (call_value_i): Likewise.
3432 * config/arc/constraints.md (Csc): New constraint.
3434 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3435 John Eric Martin <John.Martin@emmicro-us.com>
3437 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3438 * config/arc/arc.c (_arc_jli_section): New struct.
3439 (arc_jli_section): New type.
3440 (rc_jli_sections): New static variable.
3441 (arc_handle_jli_attribute): New function.
3442 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3443 (arc_file_end): New function.
3444 (TARGET_ASM_FILE_END): Define.
3445 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3446 (arc_add_jli_section): New function.
3447 (jli_call_scan): Likewise.
3448 (arc_reorg): Call jli_call_scan.
3449 (arc_output_addsi): Remove 'S' from printing asm operand.
3450 (arc_is_jli_call_p): New function.
3451 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3453 (movhi_insn): Likewise.
3454 (movsi_insn): Likewise.
3455 (movsi_set_cc_insn): Likewise.
3456 (loadqi_update): Likewise.
3457 (load_zeroextendqisi_update): Likewise.
3458 (load_signextendqisi_update): Likewise.
3459 (loadhi_update): Likewise.
3460 (load_zeroextendhisi_update): Likewise.
3461 (load_signextendhisi_update): Likewise.
3462 (loadsi_update): Likewise.
3463 (loadsf_update): Likewise.
3464 (movsicc_insn): Likewise.
3465 (bset_insn): Likewise.
3466 (bxor_insn): Likewise.
3467 (bclr_insn): Likewise.
3468 (bmsk_insn): Likewise.
3469 (bicsi3_insn): Likewise.
3470 (cmpsi_cc_c_insn): Likewise.
3471 (movsi_ne): Likewise.
3472 (movsi_cond_exec): Likewise.
3473 (clrsbsi2): Likewise.
3484 (call_i): Remove 'S' asm letter, add jli instruction.
3485 (call_value_i): Likewise.
3486 * config/arc/arc.op (mjli-always): New option.
3487 * config/arc/constraints.md (Cji): New constraint.
3488 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3490 (subsf3_fpx): Likewise.
3491 (mulsf3_fpx): Likewise.
3492 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3494 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3496 * doc/invoke.texi (ARC): Document mjli-always option.
3498 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3500 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3501 avoid addition with 0 and use incw and decw where possible.
3503 2018-01-26 Richard Biener <rguenther@suse.de>
3505 PR tree-optimization/81082
3506 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3507 association if it requires casting to unsigned.
3508 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3509 from fold_plusminus_mult_expr to catch important cases late when
3510 range info is available.
3512 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3514 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3515 * configure.ac (hidden_linkonce): New test.
3516 * configure: Regenerate.
3517 * config.in: Regenerate.
3519 2018-01-26 Julia Koval <julia.koval@intel.com>
3521 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3522 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3523 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3524 _mm_mask_bitshuffle_epi64_mask): Fix type.
3525 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3526 USI_FTYPE_V4DI_V4DI_USI): Remove.
3527 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3528 __builtin_ia32_vpshufbitqmb256_mask,
3529 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3530 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3531 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3533 2018-01-26 Alan Modra <amodra@gmail.com>
3536 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3537 UNSPEC_VBPERMQ. Sort other unspecs.
3539 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3541 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3543 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3546 * predict.c (drop_profile): Do not push/pop cfun; update also
3548 (handle_missing_profiles): Fix logic looking for zero profiles.
3550 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3553 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3554 on functions with #pragma omp declare simd or functions with simd
3556 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3557 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3558 Remove trailing \n from warning_at calls.
3560 2018-01-25 Tom de Vries <tom@codesourcery.com>
3563 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3564 for neutered workers.
3566 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3569 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3570 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3572 2018-01-24 Jeff Law <law@redhat.com>
3575 * i386.c (get_probe_interval): Move to earlier point.
3576 (ix86_compute_frame_layout): If -fstack-clash-protection and
3577 the frame is larger than the probe interval, then use pushes
3578 to save registers rather than reg->mem moves.
3579 (ix86_expand_prologue): Remove conditional for int_registers_saved
3582 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3585 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3586 min/max for never referenced object.
3588 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3591 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3593 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3594 attributes from DECL_ATTRIBUTES (decl) without affecting
3595 DECL_ATTRIBUTES (current_function_decl).
3596 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3597 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3599 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3601 PR tree-optimization/83979
3602 * fold-const.c (fold_comparison): Use constant_boolean_node
3603 instead of boolean_{true,false}_node.
3605 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3607 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3610 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3612 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3613 Simplify the clause that sets the length attribute.
3614 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3615 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3616 clause that sets the length attribute.
3617 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3619 2018-01-24 Tom de Vries <tom@codesourcery.com>
3622 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3623 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3624 Add strict parameter.
3625 (prevent_branch_around_nothing): Insert dummy insn between branch to
3626 label and label with no ptx insn inbetween.
3627 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3629 2018-01-24 Tom de Vries <tom@codesourcery.com>
3632 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3633 for neutered threads in warp.
3634 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3636 2018-01-24 Richard Biener <rguenther@suse.de>
3638 PR tree-optimization/83176
3639 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3642 2018-01-24 Richard Biener <rguenther@suse.de>
3644 PR tree-optimization/82819
3645 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3646 code generating pluses that are no-ops in the target precision.
3648 2018-01-24 Richard Biener <rguenther@suse.de>
3651 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3653 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3655 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3656 to merge probabilities.
3657 * predict.c (probably_never_executed): Also mark as cold functions
3658 with global 0 profile and guessed local profile.
3659 * profile-count.c (profile_probability::combine_with_count): New
3661 * profile-count.h (profile_probability::operator*,
3662 profile_probability::operator*=, profile_probability::operator/,
3663 profile_probability::operator/=): Reduce precision to adjusted
3664 and set value to guessed on contradictory divisions.
3665 (profile_probability::combine_with_freq): Remove.
3666 (profile_probability::combine_wiht_count): Declare.
3667 (profile_count::force_nonzero):: Set to adjusted.
3668 (profile_count::probability_in):: Set quality to adjusted.
3669 * tree-ssa-tail-merge.c (replace_block_by): Use
3672 2018-01-23 Andrew Waterman <andrew@sifive.com>
3673 Jim Wilson <jimw@sifive.com>
3675 * config/riscv/riscv.c (riscv_stack_boundary): New.
3676 (riscv_option_override): Set riscv_stack_boundary. Handle
3677 riscv_preferred_stack_boundary_arg.
3678 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3679 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3680 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3681 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3682 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3683 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3685 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3688 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3689 of struct ix86_frame.
3690 (ix86_expand_epilogue): Likewise. Add a local variable for
3691 the reg_save_offset field in struct ix86_frame.
3693 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3695 PR tree-optimization/82604
3696 * tree-loop-distribution.c (enum partition_kind): New enum item
3697 PKIND_PARTIAL_MEMSET.
3698 (partition_builtin_p): Support above new enum item.
3699 (generate_code_for_partition): Ditto.
3700 (compute_access_range): Differentiate cases that equality can be
3701 proven at all loops, the innermost loops or no loops.
3702 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3703 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3704 (finalize_partitions, distribute_loop): Don't fuse partition of
3705 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3706 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3709 2018-01-23 Martin Liska <mliska@suse.cz>
3711 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3712 order to ignore the predictor.
3713 (PRED_POLYMORPHIC_CALL): Likewise.
3714 (PRED_RECURSIVE_CALL): Likewise.
3716 2018-01-23 Martin Liska <mliska@suse.cz>
3718 * tree-profile.c (tree_profiling): Print function header to
3719 aware reader which function we are working on.
3720 * value-prof.c (gimple_find_values_to_profile): Do not print
3721 not interesting value histograms.
3723 2018-01-23 Martin Liska <mliska@suse.cz>
3725 * profile-count.h (enum profile_quality): Add
3726 profile_uninitialized as the first value. Do not number values
3727 as they are zero based.
3728 (profile_count::verify): Update sanity check.
3729 (profile_probability::verify): Likewise.
3731 2018-01-23 Nathan Sidwell <nathan@acm.org>
3733 * doc/invoke.texi (ffor-scope): Deprecate.
3735 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3737 PR tree-optimization/83510
3738 * domwalk.c (set_all_edges_as_executable): New function.
3739 (dom_walker::dom_walker): Convert bool param
3740 "skip_unreachable_blocks" to enum reachability. Move setup of
3741 edge flags to set_all_edges_as_executable and only do it when
3742 reachability is REACHABLE_BLOCKS.
3743 * domwalk.h (enum dom_walker::reachability): New enum.
3744 (dom_walker::dom_walker): Convert bool param
3745 "skip_unreachable_blocks" to enum reachability.
3746 (set_all_edges_as_executable): New decl.
3747 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3748 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3750 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3751 but converting true to REACHABLE_BLOCKS.
3752 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3754 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3755 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3756 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3758 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3759 if check_all_array_refs will be called.
3761 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3763 * tree.c (selftest::test_location_wrappers): Add more test
3766 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3768 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3769 (selftest::test_bit_in_range): Likewise.
3771 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3774 * doc/sourcebuild.texi (vect_float): Say that the selector
3775 only describes the situation when -funsafe-math-optimizations is on.
3776 (vect_float_strict): Document.
3778 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3780 PR tree-optimization/83965
3781 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3782 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3783 instead of checking only for a reduction.
3784 (vect_recog_widen_sum_pattern): Likewise.
3786 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3788 * predict.c (probably_never_executed): Only use precise profile info.
3789 (compute_function_frequency): Skip after inlining hack since we now
3790 have quality checking.
3792 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3794 * profile-count.h (profile_probability::very_unlikely,
3795 profile_probability::unlikely, profile_probability::even): Set
3796 precision to guessed.
3798 2018-01-23 Richard Biener <rguenther@suse.de>
3800 PR tree-optimization/83963
3801 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3802 Properly terminate dominator walk when crossing the exit edge not
3803 when visiting its source block.
3805 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3808 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3809 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3811 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3813 PR tree-optimization/83957
3814 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3815 semicolon after for body surrounded by braces.
3817 PR tree-optimization/83081
3818 * profile-count.h (profile_probability::split): New method.
3819 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3820 Use profile_probability::split.
3821 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3822 when splitting a single conditional jump into 2.
3824 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3826 PR tree-optimization/69452
3827 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3830 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3832 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3833 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3834 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3836 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3838 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3839 * config/rl78/rl78.md (movdi): New define_expand.
3840 * config/rl78/rl78.c (rl78_split_movdi): New function.
3842 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3845 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3847 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3848 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3849 128-bit to produce an UNSPEC move to get the double word with the
3850 signbit and then a shift directly to do signbit.
3851 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3852 implementation with a new version that just does either a direct
3853 move or a regular move. Move memory interface to separate insns.
3854 Move insns so they are next to the expander.
3855 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3856 with signbit move. Split big and little endian case.
3857 (signbit<mode>2_dm_mem_le): Likewise.
3858 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3859 (signbit<mode>2_dm2): Likewise.
3861 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3863 * config/rl78/rl78.md (anddi3): New define_expand.
3865 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3867 * config/rl78/rl78.md (umindi3): New define_expand.
3869 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3871 * config/rl78/rl78.md (smindi3): New define_expand.
3873 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3875 * config/rl78/rl78.md (smaxdi3): New define_expand.
3877 2018-01-22 Carl Love <cel@us.ibm.com>
3879 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3880 LVX_V1TI): Add macro expansion.
3881 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3882 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3883 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3884 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3885 Change check to determine if the instruction is a byte reversing
3886 entry. Fix typo in comment.
3887 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3888 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3889 Add def_builtin calls for new builtins.
3890 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3891 Add define_insn expansion.
3893 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3895 * config/rl78/rl78.md (umaxdi3): New define_expand.
3897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3899 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3900 for non-QImode registers.
3902 2018-01-22 Richard Biener <rguenther@suse.de>
3904 PR tree-optimization/83963
3905 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3906 including the loop exit block.
3907 (scop_detection::merge_sese): Likewise.
3908 (scop_detection::add_scop): Do it here instead.
3910 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3912 * doc/sourcebuild.texi (arm_softfloat): Document.
3914 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3917 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3918 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3919 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3921 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3922 David Edelsohn <dje.gcc@gmail.com>
3925 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3926 Change "crset eq" to "crset 2".
3927 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3928 (*call_indirect_aix<mode>_nospec): Likewise.
3929 (*call_value_indirect_aix<mode>_nospec): Likewise.
3930 (*call_indirect_elfv2<mode>_nospec): Likewise.
3931 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3932 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3933 change assembly output from . to $.
3934 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3935 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3936 (*tablejump<mode>_internal1_nospec): Likewise.
3938 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3941 * config/sh/sh_optimize_sett_clrt.cc:
3942 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3944 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3946 PR tree-optimization/83940
3947 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3948 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3949 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3950 use it to pass back the definition type.
3951 (vect_check_store_rhs): Likewise rhs_dt_out.
3952 (vect_build_gather_load_calls): Add a mask_dt argument and use
3953 it instead of a call to vect_is_simple_use.
3954 (vectorizable_store): Update calls to vect_check_load_store_mask
3955 and vect_check_store_rhs. Use the dt returned by the latter instead
3956 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3957 instead of calls to vect_is_simple_use. Pass the scalar rather
3958 than the vector operand to vect_is_simple_use when handling
3959 second and subsequent copies of an rhs value.
3960 (vectorizable_load): Update calls to vect_check_load_store_mask
3961 and vect_build_gather_load_calls. Use the cached mask_dt and
3962 gs_info.offset_dt instead of calls to vect_is_simple_use.
3964 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3967 * tree-emutls.c: Include gimplify.h.
3968 (lower_emutls_2): New function.
3969 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3970 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3971 it before further processing.
3974 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3975 UINTVAL (trueop1) instead of INTVAL (op1).
3977 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3981 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3982 INCOMING_FRAME_SP_OFFSET if not defined.
3983 (scan_trace): Add ENTRY argument. If true and
3984 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3985 emit a note to adjust the CFA offset.
3986 (create_cfi_notes): Adjust scan_trace callers.
3987 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3988 INCOMING_FRAME_SP_OFFSET in the CIE.
3989 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3990 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3992 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3993 * doc/tm.texi: Regenerated.
3995 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3997 PR rtl-optimization/83147
3998 * lra-constraints.c (remove_inheritance_pseudos): Use
3999 lra_substitute_pseudo_within_insn.
4001 2018-01-19 Tom de Vries <tom@codesourcery.com>
4002 Cesar Philippidis <cesar@codesourcery.com>
4005 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4007 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4010 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4011 spaces for function labels.
4013 2018-01-19 Martin Liska <mliska@suse.cz>
4015 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4016 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4017 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4018 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4019 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4020 (PRED_CONST_RETURN): Change from 69 to 65.
4021 (PRED_NULL_RETURN): Change from 91 to 71.
4022 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4023 (PRED_LOOP_GUARD): Change from 66 to 73.
4025 2018-01-19 Martin Liska <mliska@suse.cz>
4027 * predict.c (predict_insn_def): Add new assert.
4028 (struct branch_predictor): Change type to signed integer.
4029 (test_prediction_value_range): Amend test to cover
4031 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4032 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4033 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4034 (PRED_LOOP_IV_COMPARE): Likewise.
4035 * predict.h (PROB_UNINITIALIZED): Define new constant.
4037 2018-01-19 Martin Liska <mliska@suse.cz>
4039 * predict.c (dump_prediction): Add new format for
4040 analyze_brprob.py script which is enabled with -details
4042 * profile-count.h (precise_p): New function.
4044 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4046 PR tree-optimization/83922
4047 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4048 there are no statements that need masking.
4049 (vect_active_double_reduction_p): New function.
4050 (vect_analyze_loop_operations): Use it when handling phis that
4051 are not in the loop header.
4053 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4055 PR tree-optimization/83914
4056 * tree-vect-loop.c (vectorizable_induction): Don't convert
4057 init_expr or apply the peeling adjustment for inductions
4058 that are nested within the vectorized loop.
4060 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4062 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4065 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4069 * function.h (gimplify_parameters): Add gimple_seq * argument.
4070 * function.c: Include gimple.h and options.h.
4071 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4072 for the added local temporaries if needed.
4073 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4074 if there are any parameter cleanups, wrap whole body into a
4075 try/finally with the cleanups.
4077 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4080 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4081 Use GET_MODE_CLASS for scalar floating point.
4083 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4087 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4088 Fix call of call_cgraph_insertion_hooks.
4090 2018-01-18 Martin Sebor <msebor@redhat.com>
4092 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4094 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4097 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4100 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4103 * common.opt: (-ffile-prefix-map): New option.
4104 * opts.c (common_handle_option): Defer it.
4105 * opts-global.c (handle_common_deferred_options): Handle it.
4106 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4107 * file-prefix-map.h: New file.
4108 (remap_debug_filename, add_debug_prefix_map): ...here.
4109 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4110 * final.c (debug_prefix_map, add_debug_prefix_map
4111 remap_debug_filename): Move to...
4112 * file-prefix-map.c: New file.
4113 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4114 generalize, get rid of alloca(), use strrchr() instead of strchr().
4115 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4116 Implement in terms of add_prefix_map().
4117 (remap_macro_filename, remap_debug_filename): Implement in term of
4119 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4120 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4121 * dbxout.c: Include file-prefix-map.h.
4122 * varasm.c: Likewise.
4123 * vmsdbgout.c: Likewise.
4124 * xcoffout.c: Likewise.
4125 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4126 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4127 * doc/invoke.texi (-ffile-prefix-map): Document.
4128 (-fdebug-prefix-map): Update description.
4130 2018-01-18 Martin Liska <mliska@suse.cz>
4132 * config/i386/i386.c (indirect_thunk_name): Document that also
4134 (output_indirect_thunk): Document why both instructions
4135 (pause and lfence) are generated.
4137 2018-01-18 Richard Biener <rguenther@suse.de>
4139 PR tree-optimization/83887
4140 * graphite-scop-detection.c
4141 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4142 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4143 (scop_detection::merge_sese): Re-implement with a flood-fill
4144 algorithm that properly finds a SESE region if it exists.
4146 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4149 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4150 pointer_diff optimizations use view_convert instead of convert.
4152 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4154 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4155 Generate different code for -mno-speculate-indirect-jumps.
4156 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4157 (*call_indirect_aix<mode>): Disable for
4158 -mno-speculate-indirect-jumps.
4159 (*call_indirect_aix<mode>_nospec): New define_insn.
4160 (*call_value_indirect_aix<mode>): Disable for
4161 -mno-speculate-indirect-jumps.
4162 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4163 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4164 -mno-speculate-indirect-jumps.
4165 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4167 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4169 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4170 long double type, set the flags for noting the default long double
4171 type, even if we don't pass or return a long double type.
4173 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4176 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4179 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4181 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4182 support for merge[hl].
4183 (fold_mergehl_helper): New helper function.
4184 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4185 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4186 (altivec_vmrglw_direct): Add xxmrglw insn.
4188 2018-01-17 Andrew Waterman <andrew@sifive.com>
4190 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4191 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4193 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4196 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4197 call the lto_location_cache before reading the
4198 DECL_SOURCE_LOCATION of the types.
4200 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4201 Richard Sandiford <richard.sandiford@linaro.org>
4203 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4204 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4205 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4206 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4207 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4209 * config/aarch64/constraints.md (aarch64_movti_operand):
4211 * config/aarch64/predicates.md (Uti): Add new constraint.
4213 2018-01-17 Carl Love <cel@us.ibm.com>
4215 * config/rs6000/vsx.md (define_expand xl_len_r,
4216 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4217 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4219 (define_expand, define_insn): Move the shift left from the
4220 define_insn to the define_expand for lxvl and stxvl instructions.
4221 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4222 and XL_LEN_R definitions to PURE.
4224 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4226 * config/i386/i386.c (indirect_thunk_name): Declare regno
4227 as unsigned int. Compare regno with INVALID_REGNUM.
4228 (output_indirect_thunk): Ditto.
4229 (output_indirect_thunk_function): Ditto.
4230 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4231 in the call to output_indirect_thunk_function.
4233 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4236 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4237 rather than the size of inner_type to determine the stack slot size
4238 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4240 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4243 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4246 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4248 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4249 endian Linux systems to optionally enable multilibs for selecting
4250 the long double type if the user configured an explicit type.
4251 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4252 have no long double multilibs if not defined.
4253 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4254 warn if the user used -mabi={ieee,ibm}longdouble and we built
4255 multilibs for long double.
4256 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4257 appropriate multilib option.
4258 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4260 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4261 for building long double multilibs.
4262 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4264 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4266 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4269 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4271 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4274 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4277 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4280 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4282 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4283 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4285 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4287 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4288 different rtl trees depending on TARGET_64BIT.
4289 (rs6000_gen_lvx): Likewise.
4291 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4293 * config/visium/visium.md (nop): Tweak comment.
4294 (hazard_nop): Likewise.
4296 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4298 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4299 -mspeculate-indirect-jumps.
4300 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4301 for -mno-speculate-indirect-jumps.
4302 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4303 (*call_value_indirect_elfv2<mode>): Disable for
4304 -mno-speculate-indirect-jumps.
4305 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4306 (indirect_jump): Emit different RTL for
4307 -mno-speculate-indirect-jumps.
4308 (*indirect_jump<mode>): Disable for
4309 -mno-speculate-indirect-jumps.
4310 (*indirect_jump<mode>_nospec): New define_insn.
4311 (tablejump): Emit different RTL for
4312 -mno-speculate-indirect-jumps.
4313 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4314 (tablejumpsi_nospec): New define_expand.
4315 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4316 (tablejumpdi_nospec): New define_expand.
4317 (*tablejump<mode>_internal1): Disable for
4318 -mno-speculate-indirect-jumps.
4319 (*tablejump<mode>_internal1_nospec): New define_insn.
4320 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4323 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4325 * caller-save.c (insert_save): Drop unnecessary parameter. All
4328 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4329 Richard Biener <rguenth@suse.de>
4332 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4333 return early, inline manually is_gimple_sizepos. Make sure if we
4334 call gimplify_expr we don't end up with a gimple constant.
4335 * tree.c (variably_modified_type_p): Don't return true for
4336 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4337 * gimplify.h (is_gimple_sizepos): Remove.
4339 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4341 PR tree-optimization/83857
4342 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4343 vectorizable_live_operation for pure SLP statements.
4344 (vectorizable_live_operation): Handle PHIs.
4346 2018-01-16 Richard Biener <rguenther@suse.de>
4348 PR tree-optimization/83867
4349 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4350 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4352 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4355 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4356 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4357 If off is not INTEGER_CST, issue a may not be aligned warning
4358 rather than isn't aligned. Use isn%'t rather than isn't.
4359 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4361 <case MULT_EXPR>: Improve the case when bottom and one of the
4362 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4363 operand, in that case check if the other operand is multiple of
4364 bottom divided by the INTEGER_CST operand.
4366 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4369 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4370 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4371 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4372 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4373 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4374 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4375 (pa_function_arg_size): New function.
4377 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4379 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4380 in a separate statement.
4382 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4384 PR tree-optimization/83847
4385 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4386 group gathers and scatters.
4388 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4390 PR rtl-optimization/86620
4391 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4393 PR rtl-optimization/83213
4394 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4395 to last if both are JUMP_INSNs.
4397 PR tree-optimization/83843
4398 * gimple-ssa-store-merging.c
4399 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4400 store_immediate_info for bswap/nop orig_stores.
4402 2018-01-15 Andrew Waterman <andrew@sifive.com>
4404 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4406 <UDIV>: Increase cost if !TARGET_DIV.
4408 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4410 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4411 (define_attr "cr_logical_3op"): New.
4412 (cceq_ior_compare): Adjust.
4413 (cceq_ior_compare_complement): Adjust.
4414 (*cceq_rev_compare): Adjust.
4415 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4416 (is_cracked_insn): Adjust.
4417 (insn_must_be_first_in_group): Adjust.
4418 * config/rs6000/40x.md: Adjust.
4419 * config/rs6000/440.md: Adjust.
4420 * config/rs6000/476.md: Adjust.
4421 * config/rs6000/601.md: Adjust.
4422 * config/rs6000/603.md: Adjust.
4423 * config/rs6000/6xx.md: Adjust.
4424 * config/rs6000/7450.md: Adjust.
4425 * config/rs6000/7xx.md: Adjust.
4426 * config/rs6000/8540.md: Adjust.
4427 * config/rs6000/cell.md: Adjust.
4428 * config/rs6000/e300c2c3.md: Adjust.
4429 * config/rs6000/e500mc.md: Adjust.
4430 * config/rs6000/e500mc64.md: Adjust.
4431 * config/rs6000/e5500.md: Adjust.
4432 * config/rs6000/e6500.md: Adjust.
4433 * config/rs6000/mpc.md: Adjust.
4434 * config/rs6000/power4.md: Adjust.
4435 * config/rs6000/power5.md: Adjust.
4436 * config/rs6000/power6.md: Adjust.
4437 * config/rs6000/power7.md: Adjust.
4438 * config/rs6000/power8.md: Adjust.
4439 * config/rs6000/power9.md: Adjust.
4440 * config/rs6000/rs64.md: Adjust.
4441 * config/rs6000/titan.md: Adjust.
4443 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4445 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4446 ix86_indirect_branch_register logic.
4448 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4450 * config/i386/constraints.md (Bs): Update
4451 ix86_indirect_branch_register check. Don't check
4452 ix86_indirect_branch_register with GOT_memory_operand.
4454 * config/i386/predicates.md (GOT_memory_operand): Don't check
4455 ix86_indirect_branch_register here.
4456 (GOT32_symbol_operand): Likewise.
4458 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4460 * config/i386/predicates.md (constant_call_address_operand):
4461 Rewrite ix86_indirect_branch_register logic.
4462 (sibcall_insn_operand): Likewise.
4464 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4466 * config/i386/constraints.md (Bs): Replace
4467 ix86_indirect_branch_thunk_register with
4468 ix86_indirect_branch_register.
4470 * config/i386/i386.md (indirect_jump): Likewise.
4471 (tablejump): Likewise.
4472 (*sibcall_memory): Likewise.
4473 (*sibcall_value_memory): Likewise.
4474 Peepholes of indirect call and jump via memory: Likewise.
4475 * config/i386/i386.opt: Likewise.
4476 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4477 (GOT_memory_operand): Likewise.
4478 (call_insn_operand): Likewise.
4479 (sibcall_insn_operand): Likewise.
4480 (GOT32_symbol_operand): Likewise.
4482 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4485 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4486 type rather than type addr's type points to.
4487 (expand_omp_atomic_mutex): Likewise.
4488 (expand_omp_atomic): Likewise.
4490 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4493 * config/i386/i386.c (output_indirect_thunk_function): Use
4494 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4495 for __x86_return_thunk.
4497 2018-01-15 Richard Biener <rguenther@suse.de>
4500 * expmed.c (extract_bit_field_1): Fix typo.
4502 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4505 * config/arm/iterators.md (VF): New mode iterator.
4506 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4507 Remove integer-related logic from pattern.
4508 (neon_vabd<mode>_3): Likewise.
4510 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4513 * common.opt (fstrict-overflow): No longer an alias.
4514 (fwrapv-pointer): New option.
4515 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4516 also for pointer types based on flag_wrapv_pointer.
4517 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4518 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4519 opts->x_flag_wrapv got set.
4520 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4521 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4522 POINTER_TYPE_OVERFLOW_UNDEFINED.
4523 * match.pd: Likewise in address comparison pattern.
4524 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4526 2018-01-15 Richard Biener <rguenther@suse.de>
4529 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4530 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4531 Reset type names to their identifier if their TYPE_DECL doesn't
4532 have linkage (and thus is used for ODR and devirt).
4533 (save_debug_info_for_decl): Remove.
4534 (save_debug_info_for_type): Likewise.
4535 (add_tree_to_fld_list): Adjust.
4536 * tree-pretty-print.c (dump_generic_node): Make dumping of
4537 type names more robust.
4539 2018-01-15 Richard Biener <rguenther@suse.de>
4541 * BASE-VER: Bump to 8.0.1.
4543 2018-01-14 Martin Sebor <msebor@redhat.com>
4546 * builtins.c (check_access): Avoid warning when the no-warning bit
4549 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4551 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4552 * ira-color (allocno_hard_regs_compare): Likewise.
4554 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4557 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4558 Use .pushsection/.popsection.
4560 2018-01-14 Martin Sebor <msebor@redhat.com>
4563 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4565 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4567 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4568 entry from extra_headers.
4569 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4570 extra_headers, make the list bitwise identical to the i?86-*-* one.
4572 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4574 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4575 -mcmodel=large with -mindirect-branch=thunk,
4576 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4577 -mfunction-return=thunk-extern.
4578 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4579 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4580 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4582 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4584 * config/i386/i386.c (print_reg): Print the name of the full
4585 integer register without '%'.
4586 (ix86_print_operand): Handle 'V'.
4587 * doc/extend.texi: Document 'V' modifier.
4589 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4591 * config/i386/constraints.md (Bs): Disallow memory operand for
4592 -mindirect-branch-register.
4594 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4595 (GOT_memory_operand): Likewise.
4596 (call_insn_operand): Likewise.
4597 (sibcall_insn_operand): Likewise.
4598 (GOT32_symbol_operand): Likewise.
4599 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4600 for -mindirect-branch-register.
4601 (tablejump): Likewise.
4602 (*sibcall_memory): Likewise.
4603 (*sibcall_value_memory): Likewise.
4604 Disallow peepholes of indirect call and jump via memory for
4605 -mindirect-branch-register.
4606 (*call_pop): Replace m with Bw.
4607 (*call_value_pop): Likewise.
4608 (*sibcall_pop_memory): Replace m with Bs.
4609 * config/i386/i386.opt (mindirect-branch-register): New option.
4610 * doc/invoke.texi: Document -mindirect-branch-register option.
4612 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4614 * config/i386/i386-protos.h (ix86_output_function_return): New.
4615 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4616 set function_return_type.
4617 (indirect_thunk_name): Add ret_p to indicate thunk for function
4619 (output_indirect_thunk_function): Pass false to
4620 indirect_thunk_name.
4621 (ix86_output_indirect_branch_via_reg): Likewise.
4622 (ix86_output_indirect_branch_via_push): Likewise.
4623 (output_indirect_thunk_function): Create alias for function
4624 return thunk if regno < 0.
4625 (ix86_output_function_return): New function.
4626 (ix86_handle_fndecl_attribute): Handle function_return.
4627 (ix86_attribute_table): Add function_return.
4628 * config/i386/i386.h (machine_function): Add
4629 function_return_type.
4630 * config/i386/i386.md (simple_return_internal): Use
4631 ix86_output_function_return.
4632 (simple_return_internal_long): Likewise.
4633 * config/i386/i386.opt (mfunction-return=): New option.
4634 (indirect_branch): Mention -mfunction-return=.
4635 * doc/extend.texi: Document function_return function attribute.
4636 * doc/invoke.texi: Document -mfunction-return= option.
4638 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4640 * config/i386/i386-opts.h (indirect_branch): New.
4641 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4642 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4643 with local indirect jump when converting indirect call and jump.
4644 (ix86_set_indirect_branch_type): New.
4645 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4646 (indirectlabelno): New.
4647 (indirect_thunk_needed): Likewise.
4648 (indirect_thunk_bnd_needed): Likewise.
4649 (indirect_thunks_used): Likewise.
4650 (indirect_thunks_bnd_used): Likewise.
4651 (INDIRECT_LABEL): Likewise.
4652 (indirect_thunk_name): Likewise.
4653 (output_indirect_thunk): Likewise.
4654 (output_indirect_thunk_function): Likewise.
4655 (ix86_output_indirect_branch_via_reg): Likewise.
4656 (ix86_output_indirect_branch_via_push): Likewise.
4657 (ix86_output_indirect_branch): Likewise.
4658 (ix86_output_indirect_jmp): Likewise.
4659 (ix86_code_end): Call output_indirect_thunk_function if needed.
4660 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4662 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4663 (ix86_attribute_table): Add indirect_branch.
4664 * config/i386/i386.h (machine_function): Add indirect_branch_type
4665 and has_local_indirect_jump.
4666 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4668 (tablejump): Likewise.
4669 (*indirect_jump): Use ix86_output_indirect_jmp.
4670 (*tablejump_1): Likewise.
4671 (simple_return_indirect_internal): Likewise.
4672 * config/i386/i386.opt (mindirect-branch=): New option.
4673 (indirect_branch): New.
4676 (thunk-inline): Likewise.
4677 (thunk-extern): Likewise.
4678 * doc/extend.texi: Document indirect_branch function attribute.
4679 * doc/invoke.texi: Document -mindirect-branch= option.
4681 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4684 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4686 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4688 * ipa-inline.c (want_inline_small_function_p): Return false if
4689 inlining has already failed with CIF_FINAL_ERROR.
4690 (update_caller_keys): Call want_inline_small_function_p before
4692 (update_callee_keys): Likewise.
4694 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4696 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4698 (rs6000_quadword_masked_address_p): Likewise.
4699 (quad_aligned_load_p): Likewise.
4700 (quad_aligned_store_p): Likewise.
4701 (const_load_sequence_p): Add comment to describe the outer-most loop.
4702 (mimic_memory_attributes_and_flags): New function.
4703 (rs6000_gen_stvx): Likewise.
4704 (replace_swapped_aligned_store): Likewise.
4705 (rs6000_gen_lvx): Likewise.
4706 (replace_swapped_aligned_load): Likewise.
4707 (replace_swapped_load_constant): Capitalize argument name in
4708 comment describing this function.
4709 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4710 and stores that access quad-word aligned addresses and replace
4711 with stvx or lvx instructions when appropriate.
4712 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4713 New function prototype.
4714 (rs6000_quadword_masked_address_p): Likewise.
4715 (rs6000_gen_lvx): Likewise.
4716 (rs6000_gen_stvx): Likewise.
4717 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4718 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4719 when memory address is aligned.
4720 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4721 this split to select lvx instruction when memory address is aligned.
4722 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4723 instruction when memory address is aligned.
4724 (*vsx_le_perm_load_v16qi): Likewise.
4725 (four unnamed splitters): Modify to select the stvx instruction
4726 when memory is aligned.
4728 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4730 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4731 which appears in the queue multiple times.
4733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4734 Alan Hayward <alan.hayward@arm.com>
4735 David Sherwood <david.sherwood@arm.com>
4737 * tree-vectorizer.h (vec_lower_bound): New structure.
4738 (_loop_vec_info): Add check_nonzero and lower_bounds.
4739 (LOOP_VINFO_CHECK_NONZERO): New macro.
4740 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4741 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4742 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4743 fields. Make seg_len the distance travelled, not including the
4745 (dr_direction_indicator): Declare.
4746 (dr_zero_step_indicator): Likewise.
4747 (dr_known_forward_stride_p): Likewise.
4748 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4750 (runtime_alias_check_p): Allow runtime alias checks with
4752 (operator ==): Compare access_size and align.
4753 (prune_runtime_alias_test_list): Rework for new distinction between
4754 the access_size and seg_len.
4755 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4757 (get_segment_min_max): New function.
4758 (create_intersect_range_checks): Use it.
4759 (dr_step_indicator): New function.
4760 (dr_direction_indicator): Likewise.
4761 (dr_zero_step_indicator): Likewise.
4762 (dr_known_forward_stride_p): Likewise.
4763 * tree-loop-distribution.c (data_ref_segment_size): Return
4764 DR_STEP * (niters - 1).
4765 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4767 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4768 (vect_preserves_scalar_order_p): New function, split out from...
4769 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4770 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4771 (vect_vfa_access_size): New function.
4772 (vect_vfa_align): Likewise.
4773 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4774 (dump_lower_bound): New function.
4775 (vect_check_lower_bound): Likewise.
4776 (vect_small_gap_p): Likewise.
4777 (vectorizable_with_step_bound_p): Likewise.
4778 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4779 depencies if the vectorization factor is 1. Convert the checks
4780 for nonzero steps into checks on the bounds of DR_STEP. Try using
4781 a bunds check for variable steps if the minimum required step is
4782 relatively small. Update calls to the dr_with_seg_len
4783 constructor and to vect_compile_time_alias.
4784 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4786 (vect_loop_versioning): Call it.
4787 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4789 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4792 Alan Hayward <alan.hayward@arm.com>
4793 David Sherwood <david.sherwood@arm.com>
4795 * doc/sourcebuild.texi (vect_scatter_store): Document.
4796 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4798 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4800 * genopinit.c (main): Add supports_vec_scatter_store and
4801 supports_vec_scatter_store_cached to target_optabs.
4802 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4803 IFN_MASK_SCATTER_STORE.
4804 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4806 * internal-fn.h (internal_store_fn_p): Declare.
4807 (internal_fn_stored_value_index): Likewise.
4808 * internal-fn.c (scatter_store_direct): New macro.
4809 (expand_scatter_store_optab_fn): New function.
4810 (direct_scatter_store_optab_supported_p): New macro.
4811 (internal_store_fn_p): New function.
4812 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4813 IFN_MASK_SCATTER_STORE.
4814 (internal_fn_mask_index): Likewise.
4815 (internal_fn_stored_value_index): New function.
4816 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4818 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4819 * optabs-query.c (supports_vec_scatter_store_p): New function.
4820 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4821 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4822 true for scatter stores.
4823 (vect_gather_scatter_fn_p): Handle scatter stores too.
4824 (vect_check_gather_scatter): Consider using scatter stores if
4825 supports_vec_scatter_store_p.
4826 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4828 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4829 internal_fn_stored_value_index.
4830 (check_load_store_masking): Handle scatter stores too.
4831 (vect_get_store_rhs): Make public.
4832 (vectorizable_call): Use internal_store_fn_p.
4833 (vectorizable_store): Handle scatter store internal functions.
4834 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4835 when deciding whether the end of the group has been reached.
4836 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4837 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4838 (mask_scatter_store<mode>): New insns.
4840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4841 Alan Hayward <alan.hayward@arm.com>
4842 David Sherwood <david.sherwood@arm.com>
4844 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4845 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4846 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4848 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4849 Use vect_truncate_gather_scatter_offset if we can't treat the
4850 operation as a normal gather load or scatter store.
4851 (get_group_load_store_type): Take the gather_scatter_info
4852 as argument. Try using a gather load or scatter store for
4853 single-element groups.
4854 (get_load_store_type): Update calls to get_group_load_store_type
4855 and vect_use_strided_gather_scatters_p.
4857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4858 Alan Hayward <alan.hayward@arm.com>
4859 David Sherwood <david.sherwood@arm.com>
4861 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4862 optional tree argument.
4863 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4865 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4866 but continue to use the current value as a fallback.
4867 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4868 to compare the updates.
4869 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4870 (get_load_store_type): Use it when handling a strided access.
4871 (vect_get_strided_load_store_ops): New function.
4872 (vect_get_data_ptr_increment): Likewise.
4873 (vectorizable_load): Handle strided gather loads. Always pass
4874 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4877 Alan Hayward <alan.hayward@arm.com>
4878 David Sherwood <david.sherwood@arm.com>
4880 * doc/md.texi (gather_load@var{m}): Document.
4881 (mask_gather_load@var{m}): Likewise.
4882 * genopinit.c (main): Add supports_vec_gather_load and
4883 supports_vec_gather_load_cached to target_optabs.
4884 * optabs-tree.c (init_tree_optimization_optabs): Use
4885 ggc_cleared_alloc to allocate target_optabs.
4886 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4887 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4889 * internal-fn.h (internal_load_fn_p): Declare.
4890 (internal_gather_scatter_fn_p): Likewise.
4891 (internal_fn_mask_index): Likewise.
4892 (internal_gather_scatter_fn_supported_p): Likewise.
4893 * internal-fn.c (gather_load_direct): New macro.
4894 (expand_gather_load_optab_fn): New function.
4895 (direct_gather_load_optab_supported_p): New macro.
4896 (direct_internal_fn_optab): New function.
4897 (internal_load_fn_p): Likewise.
4898 (internal_gather_scatter_fn_p): Likewise.
4899 (internal_fn_mask_index): Likewise.
4900 (internal_gather_scatter_fn_supported_p): Likewise.
4901 * optabs-query.c (supports_at_least_one_mode_p): New function.
4902 (supports_vec_gather_load_p): Likewise.
4903 * optabs-query.h (supports_vec_gather_load_p): Declare.
4904 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4905 and memory_type field.
4906 (NUM_PATTERNS): Bump to 15.
4907 * tree-vect-data-refs.c: Include internal-fn.h.
4908 (vect_gather_scatter_fn_p): New function.
4909 (vect_describe_gather_scatter_call): Likewise.
4910 (vect_check_gather_scatter): Try using internal functions for
4911 gather loads. Recognize existing calls to a gather load function.
4912 (vect_analyze_data_refs): Consider using gather loads if
4913 supports_vec_gather_load_p.
4914 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4915 (vect_get_gather_scatter_offset_type): Likewise.
4916 (vect_convert_mask_for_vectype): Likewise.
4917 (vect_add_conversion_to_patterm): Likewise.
4918 (vect_try_gather_scatter_pattern): Likewise.
4919 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4920 (vect_vect_recog_func_ptrs): Add it.
4921 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4922 internal_fn_mask_index and internal_gather_scatter_fn_p.
4923 (check_load_store_masking): Take the gather_scatter_info as an
4924 argument and handle gather loads.
4925 (vect_get_gather_scatter_ops): New function.
4926 (vectorizable_call): Check internal_load_fn_p.
4927 (vectorizable_load): Likewise. Handle gather load internal
4929 (vectorizable_store): Update call to check_load_store_masking.
4930 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4931 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4932 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4933 (aarch64_gather_scale_operand_d): New predicates.
4934 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4935 (mask_gather_load<mode>): New insns.
4937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4938 Alan Hayward <alan.hayward@arm.com>
4939 David Sherwood <david.sherwood@arm.com>
4941 * optabs.def (fold_left_plus_optab): New optab.
4942 * doc/md.texi (fold_left_plus_@var{m}): Document.
4943 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4944 * internal-fn.c (fold_left_direct): Define.
4945 (expand_fold_left_optab_fn): Likewise.
4946 (direct_fold_left_optab_supported_p): Likewise.
4947 * fold-const-call.c (fold_const_fold_left): New function.
4948 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4949 * tree-parloops.c (valid_reduction_p): New function.
4950 (gather_scalar_reductions): Use it.
4951 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4952 (vect_finish_replace_stmt): Declare.
4953 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4954 (needs_fold_left_reduction_p): New function, split out from...
4955 (vect_is_simple_reduction): ...here. Accept reductions that
4956 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4957 (vect_force_simple_reduction): Also store the reduction type in
4958 the assignment's STMT_VINFO_REDUC_TYPE.
4959 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4960 (merge_with_identity): New function.
4961 (vect_expand_fold_left): Likewise.
4962 (vectorize_fold_left_reduction): Likewise.
4963 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4964 scalar phi in place for it. Check for target support and reject
4965 cases that would reassociate the operation. Defer the transform
4966 phase to vectorize_fold_left_reduction.
4967 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4968 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4969 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4973 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4974 call to ifc_temp_var.
4976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4977 Alan Hayward <alan.hayward@arm.com>
4978 David Sherwood <david.sherwood@arm.com>
4980 * target.def (legitimize_address_displacement): Take the original
4981 offset as a poly_int.
4982 * targhooks.h (default_legitimize_address_displacement): Update
4984 * targhooks.c (default_legitimize_address_displacement): Likewise.
4985 * doc/tm.texi: Regenerate.
4986 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4987 as an argument, moving assert of ad->disp == ad->disp_term to...
4988 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4989 Try calling targetm.legitimize_address_displacement before expanding
4990 the address rather than afterwards, and adjust for the new interface.
4991 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4992 Match the new hook interface. Handle SVE addresses.
4993 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4998 * Makefile.in (OBJS): Add early-remat.o.
4999 * target.def (select_early_remat_modes): New hook.
5000 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5001 * doc/tm.texi: Regenerate.
5002 * targhooks.h (default_select_early_remat_modes): Declare.
5003 * targhooks.c (default_select_early_remat_modes): New function.
5004 * timevar.def (TV_EARLY_REMAT): New timevar.
5005 * passes.def (pass_early_remat): New pass.
5006 * tree-pass.h (make_pass_early_remat): Declare.
5007 * early-remat.c: New file.
5008 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5010 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5013 Alan Hayward <alan.hayward@arm.com>
5014 David Sherwood <david.sherwood@arm.com>
5016 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5017 vfm1 with a bound_epilog parameter.
5018 (vect_do_peeling): Update calls accordingly, and move the prologue
5019 call earlier in the function. Treat the base bound_epilog as 0 for
5020 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5021 this base when peeling for gaps.
5022 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5023 with fully-masked loops.
5024 (vect_estimate_min_profitable_iters): Handle the single peeled
5025 iteration in that case.
5027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5028 Alan Hayward <alan.hayward@arm.com>
5029 David Sherwood <david.sherwood@arm.com>
5031 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5032 single-element interleaving even if the size is not a power of 2.
5033 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5034 accesses for single-element interleaving if the group size is
5037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5038 Alan Hayward <alan.hayward@arm.com>
5039 David Sherwood <david.sherwood@arm.com>
5041 * doc/md.texi (fold_extract_last_@var{m}): Document.
5042 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5043 * optabs.def (fold_extract_last_optab): New optab.
5044 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5045 * internal-fn.c (fold_extract_direct): New macro.
5046 (expand_fold_extract_optab_fn): Likewise.
5047 (direct_fold_extract_optab_supported_p): Likewise.
5048 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5049 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5050 EXTRACT_LAST_REDUCTION.
5051 (get_initial_def_for_reduction): Do not create an initial vector
5052 for EXTRACT_LAST_REDUCTION reductions.
5053 (vectorizable_reduction): Leave the scalar phi in place for
5054 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5055 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5056 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5057 transform phase to vectorizable_condition.
5058 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5060 (vect_finish_stmt_generation): ...here.
5061 (vect_finish_replace_stmt): New function.
5062 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5063 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5065 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5068 Alan Hayward <alan.hayward@arm.com>
5069 David Sherwood <david.sherwood@arm.com>
5071 * doc/md.texi (extract_last_@var{m}): Document.
5072 * optabs.def (extract_last_optab): New optab.
5073 * internal-fn.def (EXTRACT_LAST): New internal function.
5074 * internal-fn.c (cond_unary_direct): New macro.
5075 (expand_cond_unary_optab_fn): Likewise.
5076 (direct_cond_unary_optab_supported_p): Likewise.
5077 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5078 loops using EXTRACT_LAST.
5079 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5080 (extract_last_<mode>): ...this optab.
5081 (vec_extract<mode><Vel>): Update accordingly.
5083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5084 Alan Hayward <alan.hayward@arm.com>
5085 David Sherwood <david.sherwood@arm.com>
5087 * target.def (empty_mask_is_expensive): New hook.
5088 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5089 * doc/tm.texi: Regenerate.
5090 * targhooks.h (default_empty_mask_is_expensive): Declare.
5091 * targhooks.c (default_empty_mask_is_expensive): New function.
5092 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5093 if the target says that empty masks are expensive.
5094 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5096 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5099 Alan Hayward <alan.hayward@arm.com>
5100 David Sherwood <david.sherwood@arm.com>
5102 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5103 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5104 (vect_use_loop_mask_for_alignment_p): New function.
5105 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5106 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5107 niters_skip argument. Make sure that the first niters_skip elements
5108 of the first iteration are inactive.
5109 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5110 Update call to vect_set_loop_masks_directly.
5111 (get_misalign_in_elems): New function, split out from...
5112 (vect_gen_prolog_loop_niters): ...here.
5113 (vect_update_init_of_dr): Take a code argument that specifies whether
5114 the adjustment should be added or subtracted.
5115 (vect_update_init_of_drs): Likewise.
5116 (vect_prepare_for_masked_peels): New function.
5117 (vect_do_peeling): Skip prologue peeling if we're using a mask
5118 instead. Update call to vect_update_inits_of_drs.
5119 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5121 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5122 alignment. Do not include the number of peeled iterations in
5123 the minimum threshold in that case.
5124 (vectorizable_induction): Adjust the start value down by
5125 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5126 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5127 Take the number of skipped iterations into account when calculating
5129 * tree-vect-stmts.c (vect_gen_while_not): New function.
5131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5132 Alan Hayward <alan.hayward@arm.com>
5133 David Sherwood <david.sherwood@arm.com>
5135 * doc/sourcebuild.texi (vect_fully_masked): Document.
5136 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5138 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5140 (vect_analyze_loop_2): ...here. Don't check the vectorization
5141 factor against the number of loop iterations if the loop is
5144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5145 Alan Hayward <alan.hayward@arm.com>
5146 David Sherwood <david.sherwood@arm.com>
5148 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5149 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5150 (dump_groups): Update accordingly.
5151 (iv_use::mem_type): New member variable.
5152 (address_p): New function.
5153 (record_use): Add a mem_type argument and initialize the new
5155 (record_group_use): Add a mem_type argument. Use address_p.
5156 Remove obsolete null checks of base_object. Update call to record_use.
5157 (find_interesting_uses_op): Update call to record_group_use.
5158 (find_interesting_uses_cond): Likewise.
5159 (find_interesting_uses_address): Likewise.
5160 (get_mem_type_for_internal_fn): New function.
5161 (find_address_like_use): Likewise.
5162 (find_interesting_uses_stmt): Try find_address_like_use before
5163 calling find_interesting_uses_op.
5164 (addr_offset_valid_p): Use the iv mem_type field as the type
5165 of the addressed memory.
5166 (add_autoinc_candidates): Likewise.
5167 (get_address_cost): Likewise.
5168 (split_small_address_groups_p): Use address_p.
5169 (split_address_groups): Likewise.
5170 (add_iv_candidate_for_use): Likewise.
5171 (autoinc_possible_for_pair): Likewise.
5172 (rewrite_groups): Likewise.
5173 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5174 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5175 (get_alias_ptr_type_for_ptr_address): New function.
5176 (rewrite_use_address): Rewrite address uses in calls that were
5177 identified by find_address_like_use.
5179 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5180 Alan Hayward <alan.hayward@arm.com>
5181 David Sherwood <david.sherwood@arm.com>
5183 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5185 * gimple-expr.h (is_gimple_addressable: Likewise.
5186 * gimple-expr.c (is_gimple_address): Likewise.
5187 * internal-fn.c (expand_call_mem_ref): New function.
5188 (expand_mask_load_optab_fn): Use it.
5189 (expand_mask_store_optab_fn): Likewise.
5191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5192 Alan Hayward <alan.hayward@arm.com>
5193 David Sherwood <david.sherwood@arm.com>
5195 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5196 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5197 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5198 (cond_umax@var{mode}): Document.
5199 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5200 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5201 (cond_umin_optab, cond_umax_optab): New optabs.
5202 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5203 (COND_IOR, COND_XOR): New internal functions.
5204 * internal-fn.h (get_conditional_internal_fn): Declare.
5205 * internal-fn.c (cond_binary_direct): New macro.
5206 (expand_cond_binary_optab_fn): Likewise.
5207 (direct_cond_binary_optab_supported_p): Likewise.
5208 (get_conditional_internal_fn): New function.
5209 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5210 Cope with reduction statements that are vectorized as calls rather
5212 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5213 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5214 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5215 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5216 (UNSPEC_COND_EOR): New unspecs.
5217 (optab): Add mappings for them.
5218 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5219 (sve_int_op, sve_fp_op): New int attributes.
5221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5222 Alan Hayward <alan.hayward@arm.com>
5223 David Sherwood <david.sherwood@arm.com>
5225 * optabs.def (while_ult_optab): New optab.
5226 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5227 * internal-fn.def (WHILE_ULT): New internal function.
5228 * internal-fn.h (direct_internal_fn_supported_p): New override
5229 that takes two types as argument.
5230 * internal-fn.c (while_direct): New macro.
5231 (expand_while_optab_fn): New function.
5232 (convert_optab_supported_p): Likewise.
5233 (direct_while_optab_supported_p): New macro.
5234 * wide-int.h (wi::udiv_ceil): New function.
5235 * tree-vectorizer.h (rgroup_masks): New structure.
5236 (vec_loop_masks): New typedef.
5237 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5239 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5240 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5241 (vect_max_vf): New function.
5242 (slpeel_make_loop_iterate_ntimes): Delete.
5243 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5244 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5245 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5246 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5247 internal-fn.h, stor-layout.h and optabs-query.h.
5248 (vect_set_loop_mask): New function.
5249 (add_preheader_seq): Likewise.
5250 (add_header_seq): Likewise.
5251 (interleave_supported_p): Likewise.
5252 (vect_maybe_permute_loop_masks): Likewise.
5253 (vect_set_loop_masks_directly): Likewise.
5254 (vect_set_loop_condition_masked): Likewise.
5255 (vect_set_loop_condition_unmasked): New function, split out from
5256 slpeel_make_loop_iterate_ntimes.
5257 (slpeel_make_loop_iterate_ntimes): Rename to..
5258 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5259 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5260 (vect_do_peeling): Update call accordingly.
5261 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5263 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5264 mask_compare_type, can_fully_mask_p and fully_masked_p.
5265 (release_vec_loop_masks): New function.
5266 (_loop_vec_info): Use it to free the loop masks.
5267 (can_produce_all_loop_masks_p): New function.
5268 (vect_get_max_nscalars_per_iter): Likewise.
5269 (vect_verify_full_masking): Likewise.
5270 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5271 retries, and free the mask rgroups before retrying. Check loop-wide
5272 reasons for disallowing fully-masked loops. Make the final decision
5273 about whether use a fully-masked loop or not.
5274 (vect_estimate_min_profitable_iters): Do not assume that peeling
5275 for the number of iterations will be needed for fully-masked loops.
5276 (vectorizable_reduction): Disable fully-masked loops.
5277 (vectorizable_live_operation): Likewise.
5278 (vect_halve_mask_nunits): New function.
5279 (vect_double_mask_nunits): Likewise.
5280 (vect_record_loop_mask): Likewise.
5281 (vect_get_loop_mask): Likewise.
5282 (vect_transform_loop): Handle the case in which the final loop
5283 iteration might handle a partial vector. Call vect_set_loop_condition
5284 instead of slpeel_make_loop_iterate_ntimes.
5285 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5286 (check_load_store_masking): New function.
5287 (prepare_load_store_mask): Likewise.
5288 (vectorizable_store): Handle fully-masked loops.
5289 (vectorizable_load): Likewise.
5290 (supportable_widening_operation): Use vect_halve_mask_nunits for
5292 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5293 (vect_gen_while): New function.
5294 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5295 (aarch64_uqdec<mode>): New insn.
5297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5298 Alan Hayward <alan.hayward@arm.com>
5299 David Sherwood <david.sherwood@arm.com>
5301 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5302 (reduc_xor_scal_optab): New optabs.
5303 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5304 (reduc_xor_scal_@var{m}): Document.
5305 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5306 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5308 * fold-const-call.c (fold_const_call): Handle them.
5309 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5310 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5311 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5312 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5313 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5314 (UNSPEC_XORV): New unspecs.
5315 (optab): Add entries for them.
5316 (BITWISEV): New int iterator.
5317 (bit_reduc_op): New int attributes.
5319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5320 Alan Hayward <alan.hayward@arm.com>
5321 David Sherwood <david.sherwood@arm.com>
5323 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5324 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5325 * optabs.def (vec_shl_insert_optab): New optab.
5326 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5327 (duplicate_and_interleave): Likewise.
5328 * tree-vect-loop.c: Include internal-fn.h.
5329 (neutral_op_for_slp_reduction): New function, split out from
5330 get_initial_defs_for_reduction.
5331 (get_initial_def_for_reduction): Handle option 2 for variable-length
5332 vectors by loading the neutral value into a vector and then shifting
5333 the initial value into element 0.
5334 (get_initial_defs_for_reduction): Replace the code argument with
5335 the neutral value calculated by neutral_op_for_slp_reduction.
5336 Use gimple_build_vector for constant-length vectors.
5337 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5338 but the first group_size elements have a neutral value.
5339 Use duplicate_and_interleave otherwise.
5340 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5341 Update call to get_initial_defs_for_reduction. Handle SLP
5342 reductions for variable-length vectors by creating one vector
5343 result for each scalar result, with the elements associated
5344 with other scalar results stubbed out with the neutral value.
5345 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5346 Require IFN_VEC_SHL_INSERT for double reductions on
5347 variable-length vectors, or SLP reductions that have
5348 a neutral value. Require can_duplicate_and_interleave_p
5349 support for variable-length unchained SLP reductions if there
5350 is no neutral value, such as for MIN/MAX reductions. Also require
5351 the number of vector elements to be a multiple of the number of
5352 SLP statements when doing variable-length unchained SLP reductions.
5353 Update call to vect_create_epilog_for_reduction.
5354 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5355 and remove initial values.
5356 (duplicate_and_interleave): Make public.
5357 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5358 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5361 Alan Hayward <alan.hayward@arm.com>
5362 David Sherwood <david.sherwood@arm.com>
5364 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5365 (can_duplicate_and_interleave_p): New function.
5366 (vect_get_and_check_slp_defs): Take the vector of statements
5367 rather than just the current one. Remove excess parentheses.
5368 Restriction rejectinon of vect_constant_def and vect_external_def
5369 for variable-length vectors to boolean types, or types for which
5370 can_duplicate_and_interleave_p is false.
5371 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5372 (duplicate_and_interleave): New function.
5373 (vect_get_constant_vectors): Use gimple_build_vector for
5374 constant-length vectors and suitable variable-length constant
5375 vectors. Use duplicate_and_interleave for other variable-length
5376 vectors. Don't defer the update when inserting new statements.
5378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5379 Alan Hayward <alan.hayward@arm.com>
5380 David Sherwood <david.sherwood@arm.com>
5382 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5383 min_profitable_iters doesn't go negative.
5385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5386 Alan Hayward <alan.hayward@arm.com>
5387 David Sherwood <david.sherwood@arm.com>
5389 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5390 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5391 * optabs.def (vec_mask_load_lanes_optab): New optab.
5392 (vec_mask_store_lanes_optab): Likewise.
5393 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5394 (MASK_STORE_LANES): Likewise.
5395 * internal-fn.c (mask_load_lanes_direct): New macro.
5396 (mask_store_lanes_direct): Likewise.
5397 (expand_mask_load_optab_fn): Handle masked operations.
5398 (expand_mask_load_lanes_optab_fn): New macro.
5399 (expand_mask_store_optab_fn): Handle masked operations.
5400 (expand_mask_store_lanes_optab_fn): New macro.
5401 (direct_mask_load_lanes_optab_supported_p): Likewise.
5402 (direct_mask_store_lanes_optab_supported_p): Likewise.
5403 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5405 (vect_load_lanes_supported): Likewise.
5406 * tree-vect-data-refs.c (strip_conversion): New function.
5407 (can_group_stmts_p): Likewise.
5408 (vect_analyze_data_ref_accesses): Use it instead of checking
5409 for a pair of assignments.
5410 (vect_store_lanes_supported): Take a masked_p parameter.
5411 (vect_load_lanes_supported): Likewise.
5412 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5413 vect_store_lanes_supported and vect_load_lanes_supported.
5414 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5415 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5416 parameter. Don't allow gaps for masked accesses.
5417 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5418 and vect_load_lanes_supported.
5419 (get_load_store_type): Take a masked_p parameter and update
5420 call to get_group_load_store_type.
5421 (vectorizable_store): Update call to get_load_store_type.
5422 Handle IFN_MASK_STORE_LANES.
5423 (vectorizable_load): Update call to get_load_store_type.
5424 Handle IFN_MASK_LOAD_LANES.
5426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5427 Alan Hayward <alan.hayward@arm.com>
5428 David Sherwood <david.sherwood@arm.com>
5430 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5432 * config/aarch64/aarch64-protos.h
5433 (aarch64_sve_struct_memory_operand_p): Declare.
5434 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5435 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5436 (VPRED, vpred): Handle SVE structure modes.
5437 * config/aarch64/constraints.md (Utx): New constraint.
5438 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5439 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5440 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5441 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5442 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5443 structure modes. Split into pieces after RA.
5444 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5445 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5447 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5448 SVE structure modes.
5449 (aarch64_classify_address): Likewise.
5450 (sizetochar): Move earlier in file.
5451 (aarch64_print_operand): Handle SVE register lists.
5452 (aarch64_array_mode): New function.
5453 (aarch64_sve_struct_memory_operand_p): Likewise.
5454 (TARGET_ARRAY_MODE): Redefine.
5456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5457 Alan Hayward <alan.hayward@arm.com>
5458 David Sherwood <david.sherwood@arm.com>
5460 * target.def (array_mode): New target hook.
5461 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5462 * doc/tm.texi: Regenerate.
5463 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5464 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5465 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5467 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5471 Alan Hayward <alan.hayward@arm.com>
5472 David Sherwood <david.sherwood@arm.com>
5474 * fold-const.c (fold_binary_loc): Check the argument types
5475 rather than the result type when testing for a vector operation.
5477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5479 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5480 * doc/tm.texi: Regenerate.
5482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5483 Alan Hayward <alan.hayward@arm.com>
5484 David Sherwood <david.sherwood@arm.com>
5486 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5487 (sve): Document new AArch64 extension.
5488 * doc/md.texi (w): Extend the description of the AArch64
5489 constraint to include SVE vectors.
5490 (Upl, Upa): Document new AArch64 predicate constraints.
5491 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5493 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5494 (msve-vector-bits=): New option.
5495 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5496 SVE when these are disabled.
5497 (sve): New extension.
5498 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5499 modes. Adjust their number of units based on aarch64_sve_vg.
5500 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5501 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5502 aarch64_addr_query_type.
5503 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5504 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5505 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5506 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5507 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5508 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5509 (aarch64_simd_imm_zero_p): Delete.
5510 (aarch64_check_zero_based_sve_index_immediate): Declare.
5511 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5512 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5513 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5514 (aarch64_sve_float_mul_immediate_p): Likewise.
5515 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5517 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5518 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5519 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5520 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5521 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5522 (aarch64_regmode_natural_size): Likewise.
5523 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5524 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5526 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5527 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5528 for VG and the SVE predicate registers.
5529 (V_ALIASES): Add a "z"-prefixed alias.
5530 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5531 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5532 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5533 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5534 (REG_CLASS_NAMES): Add entries for them.
5535 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5536 and the predicate registers.
5537 (aarch64_sve_vg): Declare.
5538 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5539 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5540 (REGMODE_NATURAL_SIZE): Define.
5541 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5543 * config/aarch64/aarch64.c: Include cfgrtl.h.
5544 (simd_immediate_info): Add a constructor for series vectors,
5545 and an associated step field.
5546 (aarch64_sve_vg): New variable.
5547 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5548 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5549 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5550 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5551 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5552 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5553 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5554 (aarch64_get_mask_mode): New functions.
5555 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5556 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5557 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5558 predicate modes and predicate registers. Explicitly restrict
5559 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5560 to store a vector mode if it is recognized by
5561 aarch64_classify_vector_mode.
5562 (aarch64_regmode_natural_size): New function.
5563 (aarch64_hard_regno_caller_save_mode): Return the original mode
5565 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5566 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5567 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5568 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5570 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5571 does not overlap dest if the function is frame-related. Handle
5573 (aarch64_split_add_offset): New function.
5574 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5575 them aarch64_add_offset.
5576 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5577 and update call to aarch64_sub_sp.
5578 (aarch64_add_cfa_expression): New function.
5579 (aarch64_expand_prologue): Pass extra temporary registers to the
5580 functions above. Handle the case in which we need to emit new
5581 DW_CFA_expressions for registers that were originally saved
5582 relative to the stack pointer, but now have to be expressed
5583 relative to the frame pointer.
5584 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5586 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5587 IP0 and IP1 values for SVE frames.
5588 (aarch64_expand_vec_series): New function.
5589 (aarch64_expand_sve_widened_duplicate): Likewise.
5590 (aarch64_expand_sve_const_vector): Likewise.
5591 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5592 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5593 into the register, rather than emitting a SET directly.
5594 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5595 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5596 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5597 (offset_9bit_signed_scaled_p): New functions.
5598 (aarch64_replicate_bitmask_imm): New function.
5599 (aarch64_bitmask_imm): Use it.
5600 (aarch64_cannot_force_const_mem): Reject expressions involving
5601 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5602 (aarch64_classify_index): Handle SVE indices, by requiring
5603 a plain register index with a scale that matches the element size.
5604 (aarch64_classify_address): Handle SVE addresses. Assert that
5605 the mode of the address is VOIDmode or an integer mode.
5606 Update call to aarch64_classify_symbol.
5607 (aarch64_classify_symbolic_expression): Update call to
5608 aarch64_classify_symbol.
5609 (aarch64_const_vec_all_in_range_p): New function.
5610 (aarch64_print_vector_float_operand): Likewise.
5611 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5612 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5613 and the FP immediates 1.0 and 0.5.
5614 (aarch64_print_address_internal): Handle SVE addresses.
5615 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5616 (aarch64_regno_regclass): Handle predicate registers.
5617 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5619 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5620 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5621 (aarch64_convert_sve_vector_bits): New function.
5622 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5623 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5625 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5626 Handle SVE vector and predicate modes. Accept VL-based constants
5627 that need only one temporary register, and VL offsets that require
5628 no temporary registers.
5629 (aarch64_conditional_register_usage): Mark the predicate registers
5630 as fixed if SVE isn't available.
5631 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5632 Return true for SVE vector and predicate modes.
5633 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5634 rather than an unsigned int. Handle SVE modes.
5635 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5637 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5639 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5640 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5641 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5642 (aarch64_sve_float_mul_immediate_p): New functions.
5643 (aarch64_sve_valid_immediate): New function.
5644 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5645 Explicitly reject structure modes. Check for INDEX constants.
5646 Handle PTRUE and PFALSE constants.
5647 (aarch64_check_zero_based_sve_index_immediate): New function.
5648 (aarch64_simd_imm_zero_p): Delete.
5649 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5650 vector modes. Accept constants in the range of CNT[BHWD].
5651 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5652 ask for an Advanced SIMD mode.
5653 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5654 (aarch64_simd_vector_alignment): Handle SVE predicates.
5655 (aarch64_vectorize_preferred_vector_alignment): New function.
5656 (aarch64_simd_vector_alignment_reachable): Use it instead of
5658 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5659 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5661 (MAX_VECT_LEN): Delete.
5662 (expand_vec_perm_d): Add a vec_flags field.
5663 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5664 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5665 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5667 (aarch64_evpc_rev): Rename to...
5668 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5669 (aarch64_evpc_rev_global): New function.
5670 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5671 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5673 (aarch64_evpc_sve_tbl): New function.
5674 (aarch64_expand_vec_perm_const_1): Update after rename of
5675 aarch64_evpc_rev. Handle SVE permutes too, trying
5676 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5677 than aarch64_evpc_tbl.
5678 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5679 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5680 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5681 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5682 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5683 (aarch64_expand_sve_vcond): New functions.
5684 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5685 of aarch64_vector_mode_p.
5686 (aarch64_dwarf_poly_indeterminate_value): New function.
5687 (aarch64_compute_pressure_classes): Likewise.
5688 (aarch64_can_change_mode_class): Likewise.
5689 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5690 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5691 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5692 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5693 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5694 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5695 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5696 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5698 (Dn, Dl, Dr): Accept const as well as const_vector.
5699 (Dz): Likewise. Compare against CONST0_RTX.
5700 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5701 of "vector" where appropriate.
5702 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5703 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5704 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5705 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5706 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5707 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5708 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5709 (v_int_equiv): Extend to SVE modes.
5710 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5712 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5713 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5714 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5715 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5716 (SVE_COND_FP_CMP): New int iterators.
5717 (perm_hilo): Handle the new unpack unspecs.
5718 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5720 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5721 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5722 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5723 (aarch64_equality_operator, aarch64_constant_vector_operand)
5724 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5725 (aarch64_sve_nonimmediate_operand): Likewise.
5726 (aarch64_sve_general_operand): Likewise.
5727 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5728 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5729 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5730 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5731 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5732 (aarch64_sve_float_arith_immediate): Likewise.
5733 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5734 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5735 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5736 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5737 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5738 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5739 (aarch64_sve_float_arith_operand): Likewise.
5740 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5741 (aarch64_sve_float_mul_operand): Likewise.
5742 (aarch64_sve_vec_perm_operand): Likewise.
5743 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5744 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5745 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5746 as well as const_vector.
5747 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5748 in file. Use CONST0_RTX and CONSTM1_RTX.
5749 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5750 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5751 Use aarch64_simd_imm_zero.
5752 * config/aarch64/aarch64-sve.md: New file.
5753 * config/aarch64/aarch64.md: Include it.
5754 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5755 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5756 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5757 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5758 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5759 (sve): New attribute.
5760 (enabled): Disable instructions with the sve attribute unless
5762 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5763 aarch64_expand_mov_immediate.
5764 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5765 CNT[BHSD] immediates.
5766 (movti): Split CONST_POLY_INT moves into two halves.
5767 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5768 Split additions that need a temporary here if the destination
5769 is the stack pointer.
5770 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5771 (*add<mode>3_poly_1): New instruction.
5772 (set_clobber_cc): New expander.
5774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5776 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5777 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5778 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5779 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5780 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5781 Change innermode from fixed_mode_size to machine_mode.
5782 (simplify_subreg): Update call accordingly. Handle a constant-sized
5783 subreg of a variable-length CONST_VECTOR.
5785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5786 Alan Hayward <alan.hayward@arm.com>
5787 David Sherwood <david.sherwood@arm.com>
5789 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5790 (add_offset_to_base): New function, split out from...
5791 (create_mem_ref): ...here. When handling a scale other than 1,
5792 check first whether the address is valid without the offset.
5793 Add it into the base if so, leaving the index and scale as-is.
5795 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5798 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5799 fold_for_warn before checking if arg2 is INTEGER_CST.
5801 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5803 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5804 (store_multiple_operation): Delete.
5805 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5806 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5807 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5808 guarded by TARGET_STRING.
5809 (rs6000_output_load_multiple): Delete.
5810 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5811 OPTION_MASK_STRING / TARGET_STRING handling.
5812 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5813 (const rs6000_opt_masks) <"string">: Change mask to 0.
5814 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5815 (MASK_STRING): Delete.
5816 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5818 (load_multiple): Delete.
5825 (store_multiple): Delete.
5832 (movmemsi_8reg): Delete.
5833 (corresponding unnamed define_insn): Delete.
5834 (movmemsi_6reg): Delete.
5835 (corresponding unnamed define_insn): Delete.
5836 (movmemsi_4reg): Delete.
5837 (corresponding unnamed define_insn): Delete.
5838 (movmemsi_2reg): Delete.
5839 (corresponding unnamed define_insn): Delete.
5840 (movmemsi_1reg): Delete.
5841 (corresponding unnamed define_insn): Delete.
5842 * config/rs6000/rs6000.opt (mno-string): New.
5843 (mstring): Replace by deprecation warning stub.
5844 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5846 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5848 * regrename.c (regrename_do_replace): If replacing the same
5849 reg multiple times, try to reuse last created gen_raw_REG.
5852 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5853 main to workaround a bug in GDB.
5855 2018-01-12 Tom de Vries <tom@codesourcery.com>
5858 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5860 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5862 PR rtl-optimization/80481
5863 * ira-color.c (get_cap_member): New function.
5864 (allocnos_conflict_by_live_ranges_p): Use it.
5865 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5866 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5868 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5871 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5872 (*saddl_se_1): Ditto.
5874 (*ssubl_se_1): Ditto.
5876 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5878 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5879 rather than wi::to_widest for DR_INITs.
5880 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5881 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5882 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5884 (vect_analyze_group_access_1): Note that here.
5886 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5888 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5889 polynomial type sizes.
5891 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5893 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5894 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5895 (gimple_add_tmp_var): Likewise.
5897 2018-01-12 Martin Liska <mliska@suse.cz>
5899 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5900 (gimple_alloc_sizes): Likewise.
5901 (dump_gimple_statistics): Use PRIu64 in printf format.
5902 * gimple.h: Change uint64_t to int.
5904 2018-01-12 Martin Liska <mliska@suse.cz>
5906 * tree-core.h: Use uint64_t instead of int.
5907 * tree.c (tree_node_counts): Likewise.
5908 (tree_node_sizes): Likewise.
5909 (dump_tree_statistics): Use PRIu64 in printf format.
5911 2018-01-12 Martin Liska <mliska@suse.cz>
5913 * Makefile.in: As qsort_chk is implemented in vec.c, add
5914 vec.o to linkage of gencfn-macros.
5915 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5916 passing the info to record_node_allocation_statistics.
5917 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5919 * ggc-common.c (struct ggc_usage): Add operator== and use
5920 it in operator< and compare function.
5921 * mem-stats.h (struct mem_usage): Likewise.
5922 * vec.c (struct vec_usage): Remove operator< and compare
5923 function. Can be simply inherited.
5925 2018-01-12 Martin Jambor <mjambor@suse.cz>
5928 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5929 * tree-ssa-math-opts.c: Include domwalk.h.
5930 (convert_mult_to_fma_1): New function.
5931 (fma_transformation_info): New type.
5932 (fma_deferring_state): Likewise.
5933 (cancel_fma_deferring): New function.
5934 (result_of_phi): Likewise.
5935 (last_fma_candidate_feeds_initial_phi): Likewise.
5936 (convert_mult_to_fma): Added deferring logic, split actual
5937 transformation to convert_mult_to_fma_1.
5938 (math_opts_dom_walker): New type.
5939 (math_opts_dom_walker::after_dom_children): New method, body moved
5940 here from pass_optimize_widening_mul::execute, added deferring logic
5942 (pass_optimize_widening_mul::execute): Moved most of code to
5943 math_opts_dom_walker::after_dom_children.
5944 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5945 * config/i386/i386.c (ix86_option_override_internal): Added
5946 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5948 2018-01-12 Richard Biener <rguenther@suse.de>
5951 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5952 inline instance vars.
5954 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5957 * config/rx/rx.c (rx_is_restricted_memory_address):
5960 2018-01-12 Richard Biener <rguenther@suse.de>
5962 PR tree-optimization/80846
5963 * target.def (split_reduction): New target hook.
5964 * targhooks.c (default_split_reduction): New function.
5965 * targhooks.h (default_split_reduction): Declare.
5966 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5967 target requests first reduce vectors by combining low and high
5969 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5970 (get_vectype_for_scalar_type_and_size): Export.
5971 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5972 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5973 * doc/tm.texi: Regenerate.
5974 * config/i386/i386.c (ix86_split_reduction): Implement
5975 TARGET_VECTORIZE_SPLIT_REDUCTION.
5977 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5980 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5981 in PIC mode except for TARGET_VXWORKS_RTP.
5982 * config/sparc/sparc.c: Include cfgrtl.h.
5983 (TARGET_INIT_PIC_REG): Define.
5984 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5985 (sparc_pic_register_p): New predicate.
5986 (sparc_legitimate_address_p): Use it.
5987 (sparc_legitimize_pic_address): Likewise.
5988 (sparc_delegitimize_address): Likewise.
5989 (sparc_mode_dependent_address_p): Likewise.
5990 (gen_load_pcrel_sym): Remove 4th parameter.
5991 (load_got_register): Adjust call to above. Remove obsolete stuff.
5992 (sparc_expand_prologue): Do not call load_got_register here.
5993 (sparc_flat_expand_prologue): Likewise.
5994 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5995 (sparc_use_pseudo_pic_reg): New function.
5996 (sparc_init_pic_reg): Likewise.
5997 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5998 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6000 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6002 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6003 Add item for branch_cost.
6005 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6007 PR rtl-optimization/83565
6008 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6009 not extend the result to a larger mode for rotate operations.
6010 (num_sign_bit_copies1): Likewise.
6012 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6015 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6017 Use values-Xc.o for -pedantic.
6018 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6020 2018-01-12 Martin Liska <mliska@suse.cz>
6023 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6025 (possible_polymorphic_call_targets): Use it.
6026 (ipa_devirt): Likewise.
6028 2018-01-12 Martin Liska <mliska@suse.cz>
6030 * profile-count.h (enum profile_quality): Use 0 as invalid
6031 enum value of profile_quality.
6033 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6035 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6036 -mext-string options.
6038 2018-01-12 Richard Biener <rguenther@suse.de>
6040 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6041 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6042 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6044 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6046 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6048 * configure.ac (--with-long-double-format): Add support for the
6049 configuration option to change the default long double format on
6051 * config.gcc (powerpc*-linux*-*): Likewise.
6052 * configure: Regenerate.
6053 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6054 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6055 used without modification.
6057 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6059 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6060 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6061 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6062 MISC_BUILTIN_SPEC_BARRIER.
6063 (rs6000_init_builtins): Likewise.
6064 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6066 (speculation_barrier): New define_insn.
6067 * doc/extend.texi: Document __builtin_speculation_barrier.
6069 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6072 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6073 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6074 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6076 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6077 integral modes instead of "ss" and "sd".
6078 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6079 vectors with 32-bit and 64-bit elements.
6080 (vecdupssescalarmodesuffix): New mode attribute.
6081 (vec_dup<mode>): Use it.
6083 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6086 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6087 frame if argument is passed on stack.
6089 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6092 * ree.c (combine_reaching_defs): Optimize also
6093 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6094 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6096 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6099 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6101 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6104 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6105 after they are computed.
6107 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6109 PR tree-optimization/83695
6110 * gimple-loop-linterchange.cc
6111 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6112 reset cached scev information after interchange.
6113 (pass_linterchange::execute): Remove call to scev_reset_htab.
6115 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6117 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6118 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6119 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6120 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6121 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6122 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6123 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6124 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6125 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6126 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6127 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6128 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6129 (V_lane_reg): Likewise.
6130 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6132 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6133 (vfmal_lane_low<mode>_intrinsic,
6134 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6135 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6136 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6137 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6138 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6139 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6141 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6143 * config/arm/arm-cpus.in (fp16fml): New feature.
6144 (ALL_SIMD): Add fp16fml.
6145 (armv8.2-a): Add fp16fml as an option.
6146 (armv8.3-a): Likewise.
6147 (armv8.4-a): Add fp16fml as part of fp16.
6148 * config/arm/arm.h (TARGET_FP16FML): Define.
6149 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6151 * config/arm/arm-modes.def (V2HF): Define.
6152 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6153 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6154 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6155 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6156 vfmsl_low, vfmsl_high): New set of builtins.
6157 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6158 (vfml_op): New code attribute.
6159 (VFMLHALVES): New int iterator.
6160 (VFML, VFMLSEL): New mode attributes.
6161 (V_reg): Define mapping for V2HF.
6162 (V_hi, V_lo): New mode attributes.
6163 (VF_constraint): Likewise.
6164 (vfml_half, vfml_half_selector): New int attributes.
6165 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6167 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6168 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6170 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6171 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6172 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6173 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6175 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6176 Document new effective target and option set.
6178 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6180 * config/arm/arm-cpus.in (armv8_4): New feature.
6181 (ARMv8_4a): New fgroup.
6182 (armv8.4-a): New arch.
6183 * config/arm/arm-tables.opt: Regenerate.
6184 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6185 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6186 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6187 Add matching rules for -march=armv8.4-a and extensions.
6188 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6190 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6193 * config/rx/rx.md (BW): New mode attribute.
6194 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6196 2018-01-11 Richard Biener <rguenther@suse.de>
6198 PR tree-optimization/83435
6199 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6200 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6201 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6203 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6204 Alan Hayward <alan.hayward@arm.com>
6205 David Sherwood <david.sherwood@arm.com>
6207 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6209 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6210 (aarch64_print_address_internal): Use it to check for a zero offset.
6212 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6213 Alan Hayward <alan.hayward@arm.com>
6214 David Sherwood <david.sherwood@arm.com>
6216 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6217 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6218 Return a poly_int64 rather than a HOST_WIDE_INT.
6219 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6220 rather than a HOST_WIDE_INT.
6221 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6222 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6223 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6224 final_offset from HOST_WIDE_INT to poly_int64.
6225 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6226 to_constant when getting the number of units in an Advanced SIMD
6228 (aarch64_builtin_vectorized_function): Check for a constant number
6230 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6232 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6233 attribute instead of GET_MODE_NUNITS.
6234 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6235 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6236 GET_MODE_SIZE for fixed-size registers.
6237 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6238 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6239 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6240 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6241 (aarch64_print_operand, aarch64_print_address_internal)
6242 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6243 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6244 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6245 Handle polynomial GET_MODE_SIZE.
6246 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6247 wider than SImode without modification.
6248 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6249 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6250 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6251 passing and returning SVE modes.
6252 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6253 rather than GEN_INT.
6254 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6255 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6256 (aarch64_allocate_and_probe_stack_space): Likewise.
6257 (aarch64_layout_frame): Cope with polynomial offsets.
6258 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6259 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6261 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6262 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6263 poly_int64 rather than a HOST_WIDE_INT.
6264 (aarch64_get_separate_components, aarch64_process_components)
6265 (aarch64_expand_prologue, aarch64_expand_epilogue)
6266 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6267 (aarch64_anchor_offset): New function, split out from...
6268 (aarch64_legitimize_address): ...here.
6269 (aarch64_builtin_vectorization_cost): Handle polynomial
6270 TYPE_VECTOR_SUBPARTS.
6271 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6273 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6274 number of elements from the PARALLEL rather than the mode.
6275 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6276 rather than GET_MODE_BITSIZE.
6277 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6278 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6279 (aarch64_expand_vec_perm_const_1): Handle polynomial
6280 d->perm.length () and d->perm elements.
6281 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6282 Apply to_constant to d->perm elements.
6283 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6284 polynomial CONST_VECTOR_NUNITS.
6285 (aarch64_move_pointer): Take amount as a poly_int64 rather
6287 (aarch64_progress_pointer): Avoid temporary variable.
6288 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6289 the mode attribute instead of GET_MODE.
6291 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6292 Alan Hayward <alan.hayward@arm.com>
6293 David Sherwood <david.sherwood@arm.com>
6295 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6296 x exists before using it.
6297 (aarch64_add_constant_internal): Rename to...
6298 (aarch64_add_offset_1): ...this. Replace regnum with separate
6299 src and dest rtxes. Handle the case in which they're different,
6300 including when the offset is zero. Replace scratchreg with an rtx.
6301 Use 2 additions if there is no spare register into which we can
6302 move a 16-bit constant.
6303 (aarch64_add_constant): Delete.
6304 (aarch64_add_offset): Replace reg with separate src and dest
6305 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6306 Use aarch64_add_offset_1.
6307 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6308 an rtx rather than an int. Take the delta as a poly_int64
6309 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6310 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6311 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6312 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6313 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6315 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6316 aarch64_add_constant.
6318 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6320 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6321 Use scalar_float_mode.
6323 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6325 * config/aarch64/aarch64-simd.md
6326 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6327 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6328 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6329 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6330 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6331 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6332 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6333 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6334 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6335 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6337 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6340 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6341 targ_options->x_arm_arch_string is non NULL.
6343 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6345 * config/aarch64/aarch64.h
6346 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6348 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6351 * expmed.c (emit_store_flag_force): Swap if const op0
6352 and change VOIDmode to mode of op0.
6354 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6356 PR rtl-optimization/83761
6357 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6358 than bytes to mode_for_size.
6360 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6363 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6364 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6367 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6370 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6371 when in layout mode.
6372 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6373 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6376 2018-01-10 Michael Collison <michael.collison@arm.com>
6378 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6379 * config/aarch64/aarch64-option-extension.def: Add
6380 AARCH64_OPT_EXTENSION of 'fp16fml'.
6381 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6382 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6383 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6384 * config/aarch64/constraints.md (Ui7): New constraint.
6385 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6386 (VFMLA_SEL_W): Ditto.
6389 (VFMLA16_LOW): New int iterator.
6390 (VFMLA16_HIGH): Ditto.
6391 (UNSPEC_FMLAL): New unspec.
6392 (UNSPEC_FMLSL): Ditto.
6393 (UNSPEC_FMLAL2): Ditto.
6394 (UNSPEC_FMLSL2): Ditto.
6395 (f16mac): New code attribute.
6396 * config/aarch64/aarch64-simd-builtins.def
6397 (aarch64_fmlal_lowv2sf): Ditto.
6398 (aarch64_fmlsl_lowv2sf): Ditto.
6399 (aarch64_fmlalq_lowv4sf): Ditto.
6400 (aarch64_fmlslq_lowv4sf): Ditto.
6401 (aarch64_fmlal_highv2sf): Ditto.
6402 (aarch64_fmlsl_highv2sf): Ditto.
6403 (aarch64_fmlalq_highv4sf): Ditto.
6404 (aarch64_fmlslq_highv4sf): Ditto.
6405 (aarch64_fmlal_lane_lowv2sf): Ditto.
6406 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6407 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6408 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6409 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6410 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6411 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6412 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6413 (aarch64_fmlal_lane_highv2sf): Ditto.
6414 (aarch64_fmlsl_lane_highv2sf): Ditto.
6415 (aarch64_fmlal_laneq_highv2sf): Ditto.
6416 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6417 (aarch64_fmlalq_lane_highv4sf): Ditto.
6418 (aarch64_fmlsl_lane_highv4sf): Ditto.
6419 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6420 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6421 * config/aarch64/aarch64-simd.md:
6422 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6423 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6424 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6425 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6426 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6427 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6428 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6429 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6430 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6431 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6432 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6433 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6434 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6435 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6436 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6437 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6438 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6439 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6440 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6441 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6442 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6443 (vfmlsl_low_u32): Ditto.
6444 (vfmlalq_low_u32): Ditto.
6445 (vfmlslq_low_u32): Ditto.
6446 (vfmlal_high_u32): Ditto.
6447 (vfmlsl_high_u32): Ditto.
6448 (vfmlalq_high_u32): Ditto.
6449 (vfmlslq_high_u32): Ditto.
6450 (vfmlal_lane_low_u32): Ditto.
6451 (vfmlsl_lane_low_u32): Ditto.
6452 (vfmlal_laneq_low_u32): Ditto.
6453 (vfmlsl_laneq_low_u32): Ditto.
6454 (vfmlalq_lane_low_u32): Ditto.
6455 (vfmlslq_lane_low_u32): Ditto.
6456 (vfmlalq_laneq_low_u32): Ditto.
6457 (vfmlslq_laneq_low_u32): Ditto.
6458 (vfmlal_lane_high_u32): Ditto.
6459 (vfmlsl_lane_high_u32): Ditto.
6460 (vfmlal_laneq_high_u32): Ditto.
6461 (vfmlsl_laneq_high_u32): Ditto.
6462 (vfmlalq_lane_high_u32): Ditto.
6463 (vfmlslq_lane_high_u32): Ditto.
6464 (vfmlalq_laneq_high_u32): Ditto.
6465 (vfmlslq_laneq_high_u32): Ditto.
6466 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6467 (AARCH64_FL_FOR_ARCH8_4): New.
6468 (AARCH64_ISA_F16FML): New ISA flag.
6469 (TARGET_F16FML): New feature flag for fp16fml.
6470 (doc/invoke.texi): Document new fp16fml option.
6472 2018-01-10 Michael Collison <michael.collison@arm.com>
6474 * config/aarch64/aarch64-builtins.c:
6475 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6476 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6477 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6478 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6479 (AARCH64_ISA_SHA3): New ISA flag.
6480 (TARGET_SHA3): New feature flag for sha3.
6481 * config/aarch64/iterators.md (sha512_op): New int attribute.
6482 (CRYPTO_SHA512): New int iterator.
6483 (UNSPEC_SHA512H): New unspec.
6484 (UNSPEC_SHA512H2): Ditto.
6485 (UNSPEC_SHA512SU0): Ditto.
6486 (UNSPEC_SHA512SU1): Ditto.
6487 * config/aarch64/aarch64-simd-builtins.def
6488 (aarch64_crypto_sha512hqv2di): New builtin.
6489 (aarch64_crypto_sha512h2qv2di): Ditto.
6490 (aarch64_crypto_sha512su0qv2di): Ditto.
6491 (aarch64_crypto_sha512su1qv2di): Ditto.
6492 (aarch64_eor3qv8hi): Ditto.
6493 (aarch64_rax1qv2di): Ditto.
6494 (aarch64_xarqv2di): Ditto.
6495 (aarch64_bcaxqv8hi): Ditto.
6496 * config/aarch64/aarch64-simd.md:
6497 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6498 (aarch64_crypto_sha512su0qv2di): Ditto.
6499 (aarch64_crypto_sha512su1qv2di): Ditto.
6500 (aarch64_eor3qv8hi): Ditto.
6501 (aarch64_rax1qv2di): Ditto.
6502 (aarch64_xarqv2di): Ditto.
6503 (aarch64_bcaxqv8hi): Ditto.
6504 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6505 (vsha512h2q_u64): Ditto.
6506 (vsha512su0q_u64): Ditto.
6507 (vsha512su1q_u64): Ditto.
6508 (veor3q_u16): Ditto.
6509 (vrax1q_u64): Ditto.
6511 (vbcaxq_u16): Ditto.
6512 * config/arm/types.md (crypto_sha512): New type attribute.
6513 (crypto_sha3): Ditto.
6514 (doc/invoke.texi): Document new sha3 option.
6516 2018-01-10 Michael Collison <michael.collison@arm.com>
6518 * config/aarch64/aarch64-builtins.c:
6519 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6520 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6521 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6522 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6523 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6524 (AARCH64_ISA_SM4): New ISA flag.
6525 (TARGET_SM4): New feature flag for sm4.
6526 * config/aarch64/aarch64-simd-builtins.def
6527 (aarch64_sm3ss1qv4si): Ditto.
6528 (aarch64_sm3tt1aq4si): Ditto.
6529 (aarch64_sm3tt1bq4si): Ditto.
6530 (aarch64_sm3tt2aq4si): Ditto.
6531 (aarch64_sm3tt2bq4si): Ditto.
6532 (aarch64_sm3partw1qv4si): Ditto.
6533 (aarch64_sm3partw2qv4si): Ditto.
6534 (aarch64_sm4eqv4si): Ditto.
6535 (aarch64_sm4ekeyqv4si): Ditto.
6536 * config/aarch64/aarch64-simd.md:
6537 (aarch64_sm3ss1qv4si): Ditto.
6538 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6539 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6540 (aarch64_sm4eqv4si): Ditto.
6541 (aarch64_sm4ekeyqv4si): Ditto.
6542 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6543 (sm3part_op): Ditto.
6544 (CRYPTO_SM3TT): Ditto.
6545 (CRYPTO_SM3PART): Ditto.
6546 (UNSPEC_SM3SS1): New unspec.
6547 (UNSPEC_SM3TT1A): Ditto.
6548 (UNSPEC_SM3TT1B): Ditto.
6549 (UNSPEC_SM3TT2A): Ditto.
6550 (UNSPEC_SM3TT2B): Ditto.
6551 (UNSPEC_SM3PARTW1): Ditto.
6552 (UNSPEC_SM3PARTW2): Ditto.
6553 (UNSPEC_SM4E): Ditto.
6554 (UNSPEC_SM4EKEY): Ditto.
6555 * config/aarch64/constraints.md (Ui2): New constraint.
6556 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6557 * config/arm/types.md (crypto_sm3): New type attribute.
6558 (crypto_sm4): Ditto.
6559 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6560 (vsm3tt1aq_u32): Ditto.
6561 (vsm3tt1bq_u32): Ditto.
6562 (vsm3tt2aq_u32): Ditto.
6563 (vsm3tt2bq_u32): Ditto.
6564 (vsm3partw1q_u32): Ditto.
6565 (vsm3partw2q_u32): Ditto.
6566 (vsm4eq_u32): Ditto.
6567 (vsm4ekeyq_u32): Ditto.
6568 (doc/invoke.texi): Document new sm4 option.
6570 2018-01-10 Michael Collison <michael.collison@arm.com>
6572 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6573 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6574 (AARCH64_FL_FOR_ARCH8_4): New.
6575 (AARCH64_FL_V8_4): New flag.
6576 (doc/invoke.texi): Document new armv8.4-a option.
6578 2018-01-10 Michael Collison <michael.collison@arm.com>
6580 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6581 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6582 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6583 * config/aarch64/aarch64-option-extension.def: Add
6584 AARCH64_OPT_EXTENSION of 'sha2'.
6585 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6586 (crypto): Disable sha2 and aes if crypto disabled.
6587 (crypto): Enable aes and sha2 if enabled.
6588 (simd): Disable sha2 and aes if simd disabled.
6589 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6591 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6592 (TARGET_SHA2): New feature flag for sha2.
6593 (TARGET_AES): New feature flag for aes.
6594 * config/aarch64/aarch64-simd.md:
6595 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6596 conditional on TARGET_AES.
6597 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6598 (aarch64_crypto_sha1hsi): Make pattern conditional
6600 (aarch64_crypto_sha1hv4si): Ditto.
6601 (aarch64_be_crypto_sha1hv4si): Ditto.
6602 (aarch64_crypto_sha1su1v4si): Ditto.
6603 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6604 (aarch64_crypto_sha1su0v4si): Ditto.
6605 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6606 (aarch64_crypto_sha256su0v4si): Ditto.
6607 (aarch64_crypto_sha256su1v4si): Ditto.
6608 (doc/invoke.texi): Document new aes and sha2 options.
6610 2018-01-10 Martin Sebor <msebor@redhat.com>
6612 PR tree-optimization/83781
6613 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6616 2018-01-11 Martin Sebor <msebor@gmail.com>
6617 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6619 PR tree-optimization/83501
6620 PR tree-optimization/81703
6622 * tree-ssa-strlen.c (get_string_cst): Rename...
6623 (get_string_len): ...to this. Handle global constants.
6624 (handle_char_store): Adjust.
6626 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6627 Jim Wilson <jimw@sifive.com>
6629 * config/riscv/riscv-protos.h (riscv_output_return): New.
6630 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6631 (riscv_attribute_table, riscv_output_return),
6632 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6633 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6634 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6636 (riscv_expand_prologue): Add early return for naked function.
6637 (riscv_expand_epilogue): Likewise.
6638 (riscv_function_ok_for_sibcall): Return false for naked function.
6639 (riscv_set_current_function): New.
6640 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6641 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6642 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6643 * doc/extend.texi (RISC-V Function Attributes): New.
6645 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6647 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6648 check for 128-bit long double before checking TCmode.
6649 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6650 128-bit long doubles before checking TFmode or TCmode.
6651 (FLOAT128_IBM_P): Likewise.
6653 2018-01-10 Martin Sebor <msebor@redhat.com>
6655 PR tree-optimization/83671
6656 * builtins.c (c_strlen): Unconditionally return zero for the empty
6658 Use -Warray-bounds for warnings.
6659 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6660 for non-constant array indices with COMPONENT_REF, arrays of
6661 arrays, and pointers to arrays.
6662 (gimple_fold_builtin_strlen): Determine and set length range for
6663 non-constant character arrays.
6665 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6668 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6671 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6673 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6675 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6678 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6679 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6680 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6681 indexed_or_indirect_operand predicate.
6682 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6683 (*vsx_le_perm_load_v8hi): Likewise.
6684 (*vsx_le_perm_load_v16qi): Likewise.
6685 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6686 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6687 (*vsx_le_perm_store_v8hi): Likewise.
6688 (*vsx_le_perm_store_v16qi): Likewise.
6689 (eight unnamed splitters): Likewise.
6691 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6693 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6694 * config/rs6000/emmintrin.h: Likewise.
6695 * config/rs6000/mmintrin.h: Likewise.
6696 * config/rs6000/xmmintrin.h: Likewise.
6698 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6701 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6703 * tree.c (tree_nop_conversion): Return true for location wrapper
6705 (maybe_wrap_with_location): New function.
6706 (selftest::check_strip_nops): New function.
6707 (selftest::test_location_wrappers): New function.
6708 (selftest::tree_c_tests): Call it.
6709 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6710 (maybe_wrap_with_location): New decl.
6711 (EXPR_LOCATION_WRAPPER_P): New macro.
6712 (location_wrapper_p): New inline function.
6713 (tree_strip_any_location_wrapper): New inline function.
6715 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6718 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6719 stack_realign_offset for the largest alignment of stack slot
6721 (ix86_find_max_used_stack_alignment): New function.
6722 (ix86_finalize_stack_frame_flags): Use it. Set
6723 max_used_stack_alignment if we don't realign stack.
6724 * config/i386/i386.h (machine_function): Add
6725 max_used_stack_alignment.
6727 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6729 * config/arm/arm.opt (-mbranch-cost): New option.
6730 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6733 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6736 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6737 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6739 2018-01-10 Richard Biener <rguenther@suse.de>
6742 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6743 early out so it also covers the case where we have a non-NULL
6746 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6748 PR tree-optimization/83753
6749 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6750 for non-strided grouped accesses if the number of elements is 1.
6752 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6755 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6756 * i386.h (TARGET_USE_GATHER): Define.
6757 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6759 2018-01-10 Martin Liska <mliska@suse.cz>
6762 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6763 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6765 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6766 CLEANUP_NO_PARTITIONING is not set.
6768 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6770 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6771 for vectors, as a partial revert of r254296.
6772 * rtl.h (const_vec_p): Delete.
6773 (const_vec_duplicate_p): Don't test for vector CONSTs.
6774 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6775 * expmed.c (make_tree): Likewise.
6778 * common.md (E, F): Use CONSTANT_P instead of checking for
6780 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6781 checking for CONST_VECTOR.
6783 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6786 * predict.c (force_edge_cold): Handle in more sane way edges
6789 2018-01-09 Carl Love <cel@us.ibm.com>
6791 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6793 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6794 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6795 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6796 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6797 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6798 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6799 * config/rs6000/rs6000-protos.h: Add extern defition for
6800 rs6000_generate_float2_double_code.
6801 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6803 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6804 (float2_v2df): Add define_expand.
6806 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6809 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6810 op_mode in the force_to_mode call.
6812 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6814 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6815 instead of checking each element individually.
6816 (aarch64_evpc_uzp): Likewise.
6817 (aarch64_evpc_zip): Likewise.
6818 (aarch64_evpc_ext): Likewise.
6819 (aarch64_evpc_rev): Likewise.
6820 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6821 instead of checking each element individually. Return true without
6823 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6824 whether all selected elements come from the same input, instead of
6825 checking each element individually. Remove calls to gen_rtx_REG,
6826 start_sequence and end_sequence and instead assert that no rtl is
6829 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6831 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6832 order of HIGH and CONST checks.
6834 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6836 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6837 if the destination isn't an SSA_NAME.
6839 2018-01-09 Richard Biener <rguenther@suse.de>
6841 PR tree-optimization/83668
6842 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6844 (canonicalize_loop_form): ... here, renamed from ...
6845 (canonicalize_loop_closed_ssa_form): ... this and amended to
6846 swap successor edges for loop exit blocks to make us use
6847 the RPO order we need for initial schedule generation.
6849 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6851 PR tree-optimization/64811
6852 * match.pd: When optimizing comparisons with Inf, avoid
6853 introducing or losing exceptions from comparisons with NaN.
6855 2018-01-09 Martin Liska <mliska@suse.cz>
6858 * asan.c (shadow_mem_size): Add gcc_assert.
6860 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6862 Don't save registers in main().
6865 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6866 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6867 * config/avr/avr.c (avr_set_current_function): Don't error if
6868 naked, OS_task or OS_main are specified at the same time.
6869 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6871 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6873 * common/config/avr/avr-common.c (avr_option_optimization_table):
6874 Switch on -mmain-is-OS_task for optimizing compilations.
6876 2018-01-09 Richard Biener <rguenther@suse.de>
6878 PR tree-optimization/83572
6879 * graphite.c: Include cfganal.h.
6880 (graphite_transform_loops): Connect infinite loops to exit
6881 and remove fake edges at the end.
6883 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6885 * ipa-inline.c (edge_badness): Revert accidental checkin.
6887 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6890 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6891 symbols; not inline clones.
6893 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6896 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6897 hard registers. Formatting fixes.
6899 PR preprocessor/83722
6900 * gcc.c (try_generate_repro): Pass
6901 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6902 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6905 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6906 Kito Cheng <kito.cheng@gmail.com>
6908 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6909 (riscv_leaf_function_p): Delete.
6910 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6912 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6914 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6916 (do_ifelse): New function.
6917 (do_isel): New function.
6918 (do_sub3): New function.
6919 (do_add3): New function.
6920 (do_load_mask_compare): New function.
6921 (do_overlap_load_compare): New function.
6922 (expand_compare_loop): New function.
6923 (expand_block_compare): Call expand_compare_loop() when appropriate.
6924 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6926 (-mblock-compare-inline-loop-limit): New option.
6928 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6931 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6932 Reverse order of second and third operands in first alternative.
6933 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6934 of first and second elements in UNSPEC_VPERMR vector.
6935 (altivec_expand_vec_perm_le): Likewise.
6937 2018-01-08 Jeff Law <law@redhat.com>
6939 PR rtl-optimizatin/81308
6940 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6941 (process_switch): If group_case_labels makes a change, then set
6943 (pass_convert_switch::execute): If a switch is converted, then
6944 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6946 PR rtl-optimization/81308
6947 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6950 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6952 PR target/83663 - Revert r255946
6953 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6954 generation for cases where splatting a value is not useful.
6955 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6956 across a vec_duplicate and a paradoxical subreg forming a vector
6957 mode to a vec_concat.
6959 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6961 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6962 -march=armv8.3-a variants.
6963 * config/arm/t-multilib: Likewise.
6964 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6966 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6968 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6970 (cceq_ior_compare_complement): Give it a name so I can use it, and
6971 change boolean_or_operator predicate to boolean_operator so it can
6972 be used to generate a crand.
6973 (eqne): New code iterator.
6974 (bd/bd_neg): New code_attrs.
6975 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6976 a single define_insn.
6977 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6978 decrement (bdnzt/bdnzf/bdzt/bdzf).
6979 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6980 with the new names of the branch decrement patterns, and added the
6981 names of the branch decrement conditional patterns.
6983 2018-01-08 Richard Biener <rguenther@suse.de>
6985 PR tree-optimization/83563
6986 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6989 2018-01-08 Richard Biener <rguenther@suse.de>
6992 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6994 2018-01-08 Richard Biener <rguenther@suse.de>
6996 PR tree-optimization/83685
6997 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6998 references to abnormals.
7000 2018-01-08 Richard Biener <rguenther@suse.de>
7003 * dwarf2out.c (output_indirect_strings): Handle empty
7004 skeleton_debug_str_hash.
7005 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7007 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7009 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7010 (emit_store_direct): Likewise.
7011 (arc_trampoline_adjust_address): Likewise.
7012 (arc_asm_trampoline_template): New function.
7013 (arc_initialize_trampoline): Use asm_trampoline_template.
7014 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7015 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7016 * config/arc/arc.md (flush_icache): Delete pattern.
7018 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7020 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7021 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7024 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7027 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7028 by not USED_FOR_TARGET.
7029 (make_pass_resolve_sw_modes): Likewise.
7031 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7033 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7036 2018-01-08 Richard Biener <rguenther@suse.de>
7039 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7041 2018-01-08 Richard Biener <rguenther@suse.de>
7044 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7046 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7049 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7050 basic blocks with a small number of successors.
7051 (convert_control_dep_chain_into_preds): Improve handling of
7053 (dump_predicates): Split apart into...
7054 (dump_pred_chain): ...here...
7055 (dump_pred_info): ...and here.
7056 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7057 (can_chain_union_be_invalidated_p): Improve check for invalidation
7059 (uninit_uses_cannot_happen): Avoid unnecessary if
7060 convert_control_dep_chain_into_preds yielded nothing.
7062 2018-01-06 Martin Sebor <msebor@redhat.com>
7064 PR tree-optimization/83640
7065 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7066 subtracting negative offset from size.
7067 (builtin_access::overlap): Adjust offset bounds of the access to fall
7068 within the size of the object if possible.
7070 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7072 PR rtl-optimization/83699
7073 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7074 extract_bit_field_as_subreg to cases in which the extracted
7075 value is also a vector.
7077 * lra-constraints.c (process_alt_operands): Test for the equivalence
7078 substitutions when detecting a possible reload cycle.
7080 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7083 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7084 by default if flag_selective_schedling{,2}. Formatting fixes.
7086 PR rtl-optimization/83682
7087 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7088 if it has non-VECTOR_MODE element mode.
7089 (vec_duplicate_p): Likewise.
7092 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7093 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7095 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7098 * config/i386/i386-builtin.def
7099 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7100 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7101 Require also OPTION_MASK_ISA_AVX512F in addition to
7102 OPTION_MASK_ISA_GFNI.
7103 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7104 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7105 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7106 to OPTION_MASK_ISA_GFNI.
7107 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7108 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7109 OPTION_MASK_ISA_AVX512BW.
7110 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7111 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7112 addition to OPTION_MASK_ISA_GFNI.
7113 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7114 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7115 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7116 to OPTION_MASK_ISA_GFNI.
7117 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7118 a requirement for all ISAs rather than any of them with a few
7120 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7122 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7123 bitmasks to be enabled with 3 exceptions, instead of requiring any
7124 enabled ISA with lots of exceptions.
7125 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7126 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7127 Change avx512bw in isa attribute to avx512f.
7128 * config/i386/sgxintrin.h: Add license boilerplate.
7129 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7130 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7131 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7132 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7134 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7135 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7136 temporarily sse2 rather than sse if not enabled already.
7139 * config/i386/sse.md (VI248_VLBW): Rename to ...
7140 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7141 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7142 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7143 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7144 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7145 mode iterator instead of VI248_VLBW.
7147 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7149 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7150 (record_modified): Skip clobbers; add debug output.
7151 (param_change_prob): Use sreal frequencies.
7153 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7155 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7156 punt for user-aligned variables.
7158 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7160 * tree-chrec.c (chrec_contains_symbols): Return true for
7163 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7166 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7167 of (x|y) == x for BICS pattern.
7169 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7171 PR tree-optimization/83605
7172 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7173 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7176 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7178 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7179 * config/epiphany/rtems.h: New file.
7181 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7182 Uros Bizjak <ubizjak@gmail.com>
7185 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7186 QIreg_operand instead of register_operand predicate.
7187 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7188 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7189 comments instead of -fmitigate[-_]rop.
7191 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7194 * cgraphunit.c (symbol_table::compile): Switch to text_section
7195 before calling assembly_start debug hook.
7196 * run-rtl-passes.c (run_rtl_passes): Likewise.
7199 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7201 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7202 range_int_cst_p rather than !symbolic_range_p before calling
7203 extract_range_from_multiplicative_op_1.
7205 2018-01-04 Jeff Law <law@redhat.com>
7207 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7208 redundant test in assertion.
7210 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7212 * doc/rtl.texi: Document machine_mode wrapper classes.
7214 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7216 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7219 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7221 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7222 the VEC_PERM_EXPR fold to fail.
7224 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7227 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7228 to switched_sections.
7230 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7233 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7236 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7239 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7240 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7242 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7245 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7246 is BLKmode and bitpos not zero or mode change is needed.
7248 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7251 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7254 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7257 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7258 instead of MULT rtx. Update all corresponding splitters.
7260 (*ssub<modesuffix>): Ditto.
7262 (*cmp_sadd_di): Update split patterns.
7263 (*cmp_sadd_si): Ditto.
7264 (*cmp_sadd_sidi): Ditto.
7265 (*cmp_ssub_di): Ditto.
7266 (*cmp_ssub_si): Ditto.
7267 (*cmp_ssub_sidi): Ditto.
7268 * config/alpha/predicates.md (const23_operand): New predicate.
7269 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7270 Look for ASHIFT, not MULT inner operand.
7271 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7273 2018-01-04 Martin Liska <mliska@suse.cz>
7275 PR gcov-profile/83669
7276 * gcov.c (output_intermediate_file): Add version to intermediate
7278 * doc/gcov.texi: Document new field 'version' in intermediate
7279 file format. Fix location of '-k' option of gcov command.
7281 2018-01-04 Martin Liska <mliska@suse.cz>
7284 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7286 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7288 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7290 2018-01-03 Martin Sebor <msebor@redhat.com>
7292 PR tree-optimization/83655
7293 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7294 checking calls with invalid arguments.
7296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7298 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7299 (vectorizable_mask_load_store): Delete.
7300 (vectorizable_call): Return false for masked loads and stores.
7301 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7302 instead of gimple_assign_rhs1.
7303 (vectorizable_load): Handle IFN_MASK_LOAD.
7304 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7308 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7310 (vectorizable_mask_load_store): ...here.
7311 (vectorizable_load): ...and here.
7313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7315 * tree-vect-stmts.c (vect_build_all_ones_mask)
7316 (vect_build_zero_merge_argument): New functions, split out from...
7317 (vectorizable_load): ...here.
7319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7321 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7323 (vectorizable_mask_load_store): ...here.
7324 (vectorizable_store): ...and here.
7326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7328 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7330 (vectorizable_mask_load_store): ...here.
7332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7334 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7335 (vect_model_store_cost): Take a vec_load_store_type instead of a
7337 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7338 (vect_model_store_cost): Take a vec_load_store_type instead of a
7340 (vectorizable_mask_load_store): Update accordingly.
7341 (vectorizable_store): Likewise.
7342 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7346 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7347 IFN_MASK_LOAD calls here rather than...
7348 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7351 Alan Hayward <alan.hayward@arm.com>
7352 David Sherwood <david.sherwood@arm.com>
7354 * expmed.c (extract_bit_field_1): For vector extracts,
7355 fall back to extract_bit_field_as_subreg if vec_extract
7358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7359 Alan Hayward <alan.hayward@arm.com>
7360 David Sherwood <david.sherwood@arm.com>
7362 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7363 they are variable or constant sized.
7364 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7365 slots for constant-sized data.
7367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7368 Alan Hayward <alan.hayward@arm.com>
7369 David Sherwood <david.sherwood@arm.com>
7371 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7372 handling COND_EXPRs with boolean comparisons, try to find a better
7373 basis for the mask type than the boolean itself.
7375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7377 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7378 is calculated and how it can be overridden.
7379 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7380 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7382 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7386 Alan Hayward <alan.hayward@arm.com>
7387 David Sherwood <david.sherwood@arm.com>
7389 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7390 Remove the mode argument.
7391 (aarch64_simd_valid_immediate): Remove the mode and inverse
7393 * config/aarch64/iterators.md (bitsize): New iterator.
7394 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7395 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7396 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7397 aarch64_simd_valid_immediate.
7398 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7399 (aarch64_reg_or_bic_imm): Likewise.
7400 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7401 with an insn_type enum and msl with a modifier_type enum.
7402 Replace element_width with a scalar_mode. Change the shift
7403 to unsigned int. Add constructors for scalar_float_mode and
7404 scalar_int_mode elements.
7405 (aarch64_vect_float_const_representable_p): Delete.
7406 (aarch64_can_const_movi_rtx_p)
7407 (aarch64_simd_scalar_immediate_valid_for_move)
7408 (aarch64_simd_make_constant): Update call to
7409 aarch64_simd_valid_immediate.
7410 (aarch64_advsimd_valid_immediate_hs): New function.
7411 (aarch64_advsimd_valid_immediate): Likewise.
7412 (aarch64_simd_valid_immediate): Remove mode and inverse
7413 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7414 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7415 and aarch64_float_const_representable_p on the result.
7416 (aarch64_output_simd_mov_immediate): Remove mode argument.
7417 Update call to aarch64_simd_valid_immediate and use of
7418 simd_immediate_info.
7419 (aarch64_output_scalar_simd_mov_immediate): Update call
7422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7423 Alan Hayward <alan.hayward@arm.com>
7424 David Sherwood <david.sherwood@arm.com>
7426 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7427 (mode_nunits): Likewise CONST_MODE_NUNITS.
7428 * machmode.def (ADJUST_NUNITS): Document.
7429 * genmodes.c (mode_data::need_nunits_adj): New field.
7430 (blank_mode): Update accordingly.
7431 (adj_nunits): New variable.
7432 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7434 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7435 listed in adj_nunits.
7436 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7437 listed in adj_nunits. Don't emit case statements for such modes.
7438 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7439 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7440 nothing if adj_nunits is nonnull.
7441 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7442 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7443 (emit_mode_fbit): Update use of print_maybe_const_decl.
7444 (emit_move_size): Likewise. Treat the array as non-const
7446 (emit_mode_adjustments): Handle adj_nunits.
7448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7450 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7451 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7452 (VECTOR_MODES): Use it.
7453 (make_vector_modes): Take the prefix as an argument.
7455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7456 Alan Hayward <alan.hayward@arm.com>
7457 David Sherwood <david.sherwood@arm.com>
7459 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7460 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7461 for MODE_VECTOR_BOOL.
7462 * machmode.def (VECTOR_BOOL_MODE): Document.
7463 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7464 (make_vector_bool_mode): New function.
7465 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7467 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7468 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7470 * stor-layout.c (int_mode_for_mode): Likewise.
7471 * tree.c (build_vector_type_for_mode): Likewise.
7472 * varasm.c (output_constant_pool_2): Likewise.
7473 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7474 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7475 for MODE_VECTOR_BOOL.
7476 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7477 of mode class checks.
7478 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7479 instead of a list of mode class checks.
7480 (expand_vector_scalar_condition): Likewise.
7481 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7484 Alan Hayward <alan.hayward@arm.com>
7485 David Sherwood <david.sherwood@arm.com>
7487 * machmode.h (mode_size): Change from unsigned short to
7489 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7490 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7491 or if measurement_type is not polynomial.
7492 (fixed_size_mode::includes_p): Check for constant-sized modes.
7493 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7494 return a poly_uint16 rather than an unsigned short.
7495 (emit_mode_size): Change the type of mode_size from unsigned short
7496 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7497 (emit_mode_adjustments): Cope with polynomial vector sizes.
7498 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7500 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7502 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7503 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7504 * caller-save.c (setup_save_areas): Likewise.
7505 (replace_reg_with_saved_mem): Likewise.
7506 * calls.c (emit_library_call_value_1): Likewise.
7507 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7508 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7509 (gen_lowpart_for_combine): Likewise.
7510 * convert.c (convert_to_integer_1): Likewise.
7511 * cse.c (equiv_constant, cse_insn): Likewise.
7512 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7513 (cselib_subst_to_values): Likewise.
7514 * dce.c (word_dce_process_block): Likewise.
7515 * df-problems.c (df_word_lr_mark_ref): Likewise.
7516 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7517 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7518 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7519 (rtl_for_decl_location): Likewise.
7520 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7521 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7522 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7523 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7524 (expand_expr_real_1): Likewise.
7525 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7526 (pad_below): Likewise.
7527 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7528 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7529 * ira.c (get_subreg_tracking_sizes): Likewise.
7530 * ira-build.c (ira_create_allocno_objects): Likewise.
7531 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7532 (ira_sort_regnos_for_alter_reg): Likewise.
7533 * ira-costs.c (record_operand_costs): Likewise.
7534 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7535 (resolve_simple_move): Likewise.
7536 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7537 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7538 (lra_constraints): Likewise.
7539 (CONST_POOL_OK_P): Reject variable-sized modes.
7540 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7541 (add_pseudo_to_slot, lra_spill): Likewise.
7542 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7543 * optabs-query.c (get_best_extraction_insn): Likewise.
7544 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7545 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7546 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7547 * recog.c (offsettable_address_addr_space_p): Likewise.
7548 * regcprop.c (maybe_mode_change): Likewise.
7549 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7550 * regrename.c (build_def_use): Likewise.
7551 * regstat.c (dump_reg_info): Likewise.
7552 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7553 (find_reloads, find_reloads_subreg_address): Likewise.
7554 * reload1.c (eliminate_regs_1): Likewise.
7555 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7556 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7557 (simplify_binary_operation_1, simplify_subreg): Likewise.
7558 * targhooks.c (default_function_arg_padding): Likewise.
7559 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7560 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7561 (verify_gimple_assign_ternary): Likewise.
7562 * tree-inline.c (estimate_move_cost): Likewise.
7563 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7564 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7565 (get_address_cost_ainc): Likewise.
7566 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7567 (vect_supportable_dr_alignment): Likewise.
7568 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7569 (vectorizable_reduction): Likewise.
7570 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7571 (vectorizable_operation, vectorizable_load): Likewise.
7572 * tree.c (build_same_sized_truth_vector_type): Likewise.
7573 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7574 * var-tracking.c (emit_note_insn_var_location): Likewise.
7575 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7576 (ADDR_VEC_ALIGN): Likewise.
7578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7579 Alan Hayward <alan.hayward@arm.com>
7580 David Sherwood <david.sherwood@arm.com>
7582 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7584 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7585 or if measurement_type is polynomial.
7586 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7587 * combine.c (make_extraction): Likewise.
7588 * dse.c (find_shift_sequence): Likewise.
7589 * dwarf2out.c (mem_loc_descriptor): Likewise.
7590 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7591 (extract_bit_field, extract_low_bits): Likewise.
7592 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7593 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7594 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7595 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7596 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7597 * reload.c (find_reloads): Likewise.
7598 * reload1.c (alter_reg): Likewise.
7599 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7600 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7601 * tree-if-conv.c (predicate_mem_writes): Likewise.
7602 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7603 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7604 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7605 * valtrack.c (dead_debug_insert_temp): Likewise.
7606 * varasm.c (mergeable_constant_section): Likewise.
7607 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7610 Alan Hayward <alan.hayward@arm.com>
7611 David Sherwood <david.sherwood@arm.com>
7613 * expr.c (expand_assignment): Cope with polynomial mode sizes
7614 when assigning to a CONCAT.
7616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7617 Alan Hayward <alan.hayward@arm.com>
7618 David Sherwood <david.sherwood@arm.com>
7620 * machmode.h (mode_precision): Change from unsigned short to
7622 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7624 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7625 or if measurement_type is not polynomial.
7626 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7627 in which the mode is already known to be a scalar_int_mode.
7628 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7629 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7631 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7632 for GET_MODE_PRECISION.
7633 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7634 for GET_MODE_PRECISION.
7635 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7637 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7638 (expand_field_assignment, make_extraction): Likewise.
7639 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7640 (get_last_value): Likewise.
7641 * convert.c (convert_to_integer_1): Likewise.
7642 * cse.c (cse_insn): Likewise.
7643 * expr.c (expand_expr_real_1): Likewise.
7644 * lra-constraints.c (simplify_operand_subreg): Likewise.
7645 * optabs-query.c (can_atomic_load_p): Likewise.
7646 * optabs.c (expand_atomic_load): Likewise.
7647 (expand_atomic_store): Likewise.
7648 * ree.c (combine_reaching_defs): Likewise.
7649 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7650 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7651 * tree.h (type_has_mode_precision_p): Likewise.
7652 * ubsan.c (instrument_si_overflow): Likewise.
7654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7655 Alan Hayward <alan.hayward@arm.com>
7656 David Sherwood <david.sherwood@arm.com>
7658 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7659 polynomial numbers of units.
7660 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7661 (valid_vector_subparts_p): New function.
7662 (build_vector_type): Remove temporary shim and take the number
7663 of units as a poly_uint64 rather than an int.
7664 (build_opaque_vector_type): Take the number of units as a
7665 poly_uint64 rather than an int.
7666 * tree.c (build_vector_from_ctor): Handle polynomial
7667 TYPE_VECTOR_SUBPARTS.
7668 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7669 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7670 (build_vector_from_val): If the number of units is variable,
7671 use build_vec_duplicate_cst for constant operands and
7672 VEC_DUPLICATE_EXPR otherwise.
7673 (make_vector_type): Remove temporary is_constant ().
7674 (build_vector_type, build_opaque_vector_type): Take the number of
7675 units as a poly_uint64 rather than an int.
7676 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7678 * cfgexpand.c (expand_debug_expr): Likewise.
7679 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7680 (store_constructor, expand_expr_real_1): Likewise.
7681 (const_scalar_mask_from_tree): Likewise.
7682 * fold-const-call.c (fold_const_reduction): Likewise.
7683 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7684 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7685 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7686 (fold_relational_const): Likewise.
7687 (native_interpret_vector): Likewise. Change the size from an
7688 int to an unsigned int.
7689 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7690 TYPE_VECTOR_SUBPARTS.
7691 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7692 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7693 duplicating a non-constant operand into a variable-length vector.
7694 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7695 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7696 * ipa-icf.c (sem_variable::equals): Likewise.
7697 * match.pd: Likewise.
7698 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7699 * print-tree.c (print_node): Likewise.
7700 * stor-layout.c (layout_type): Likewise.
7701 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7702 * tree-cfg.c (verify_gimple_comparison): Likewise.
7703 (verify_gimple_assign_binary): Likewise.
7704 (verify_gimple_assign_ternary): Likewise.
7705 (verify_gimple_assign_single): Likewise.
7706 * tree-pretty-print.c (dump_generic_node): Likewise.
7707 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7708 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7709 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7710 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7711 (vect_shift_permute_load_chain): Likewise.
7712 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7713 (expand_vector_condition, optimize_vector_constructor): Likewise.
7714 (lower_vec_perm, get_compute_type): Likewise.
7715 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7716 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7717 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7718 (vect_recog_mask_conversion_pattern): Likewise.
7719 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7720 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7721 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7722 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7723 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7724 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7725 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7726 (supportable_widening_operation): Likewise.
7727 (supportable_narrowing_operation): Likewise.
7728 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7730 * varasm.c (output_constant): Likewise.
7732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7733 Alan Hayward <alan.hayward@arm.com>
7734 David Sherwood <david.sherwood@arm.com>
7736 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7737 so that both the length == 3 and length != 3 cases set up their
7738 own permute vectors. Add comments explaining why we know the
7739 number of elements is constant.
7740 (vect_permute_load_chain): Likewise.
7742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7743 Alan Hayward <alan.hayward@arm.com>
7744 David Sherwood <david.sherwood@arm.com>
7746 * machmode.h (mode_nunits): Change from unsigned char to
7748 (ONLY_FIXED_SIZE_MODES): New macro.
7749 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7750 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7751 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7753 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7754 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7755 or if measurement_type is not polynomial.
7756 * genmodes.c (ZERO_COEFFS): New macro.
7757 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7759 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7760 Use ZERO_COEFFS when emitting initializers.
7761 * data-streamer.h (bp_pack_poly_value): New function.
7762 (bp_unpack_poly_value): Likewise.
7763 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7764 for GET_MODE_NUNITS.
7765 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7766 for GET_MODE_NUNITS.
7767 * tree.c (make_vector_type): Remove temporary shim and make
7768 the real function take the number of units as a poly_uint64
7770 (build_vector_type_for_mode): Handle polynomial nunits.
7771 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7772 * emit-rtl.c (const_vec_series_p_1): Likewise.
7773 (gen_rtx_CONST_VECTOR): Likewise.
7774 * fold-const.c (test_vec_duplicate_folding): Likewise.
7775 * genrecog.c (validate_pattern): Likewise.
7776 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7777 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7778 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7779 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7780 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7781 * rtlanal.c (subreg_get_info): Likewise.
7782 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7783 (vect_grouped_load_supported): Likewise.
7784 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7785 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7786 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7787 (simplify_const_unary_operation, simplify_binary_operation_1)
7788 (simplify_const_binary_operation, simplify_ternary_operation)
7789 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7790 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7791 instead of CONST_VECTOR_NUNITS.
7792 * varasm.c (output_constant_pool_2): Likewise.
7793 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7794 explicit-encoded elements in the XVEC for variable-length vectors.
7796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7798 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7801 Alan Hayward <alan.hayward@arm.com>
7802 David Sherwood <david.sherwood@arm.com>
7804 * coretypes.h (fixed_size_mode): Declare.
7805 (fixed_size_mode_pod): New typedef.
7806 * builtins.h (target_builtins::x_apply_args_mode)
7807 (target_builtins::x_apply_result_mode): Change type to
7808 fixed_size_mode_pod.
7809 * builtins.c (apply_args_size, apply_result_size, result_vector)
7810 (expand_builtin_apply_args_1, expand_builtin_apply)
7811 (expand_builtin_return): Update accordingly.
7813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7815 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7816 * cselib.c (cselib_hash_rtx): Likewise.
7817 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7818 CONST_VECTOR encoding.
7820 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7821 Jeff Law <law@redhat.com>
7824 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7825 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7826 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7827 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7830 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7831 explicitly probe *sp in a noreturn function if there were any callee
7832 register saves or frame pointer is needed.
7834 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7837 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7838 BLKmode for ternary, binary or unary expressions.
7841 * var-tracking.c (delete_vta_debug_insn): New inline function.
7842 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7843 insns from get_insns () to NULL instead of each bb separately.
7844 Use delete_vta_debug_insn. No longer static.
7845 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7846 delete_vta_debug_insns callers.
7847 * rtl.h (delete_vta_debug_insns): Declare.
7848 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7849 instead of variable_tracking_main.
7851 2018-01-03 Martin Sebor <msebor@redhat.com>
7853 PR tree-optimization/83603
7854 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7855 arguments past the endof the argument list in functions declared
7856 without a prototype.
7857 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7858 Avoid checking when arguments are null.
7860 2018-01-03 Martin Sebor <msebor@redhat.com>
7863 * doc/extend.texi (attribute const): Fix a typo.
7864 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7865 issuing -Wsuggest-attribute for void functions.
7867 2018-01-03 Martin Sebor <msebor@redhat.com>
7869 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7870 offset_int::from instead of wide_int::to_shwi.
7871 (maybe_diag_overlap): Remove assertion.
7872 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7873 * gimple-ssa-sprintf.c (format_directive): Same.
7874 (parse_directive): Same.
7875 (sprintf_dom_walker::compute_format_length): Same.
7876 (try_substitute_return_value): Same.
7878 2018-01-03 Jeff Law <law@redhat.com>
7881 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7882 non-constant residual for zero at runtime and avoid probing in
7883 that case. Reorganize code for trailing problem to mirror handling
7886 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7888 PR tree-optimization/83501
7889 * tree-ssa-strlen.c (get_string_cst): New.
7890 (handle_char_store): Call get_string_cst.
7892 2018-01-03 Martin Liska <mliska@suse.cz>
7894 PR tree-optimization/83593
7895 * tree-ssa-strlen.c: Include tree-cfg.h.
7896 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7897 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7898 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7900 (strlen_dom_walker::before_dom_children): Call
7901 gimple_purge_dead_eh_edges. Dump tranformation with details
7903 (strlen_dom_walker::before_dom_children): Update call by adding
7904 new argument cleanup_eh.
7905 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7907 2018-01-03 Martin Liska <mliska@suse.cz>
7910 * cif-code.def (VARIADIC_THUNK): New enum value.
7911 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7914 2018-01-03 Jan Beulich <jbeulich@suse.com>
7916 * sse.md (mov<mode>_internal): Tighten condition for when to use
7917 vmovdqu<ssescalarsize> for TI and OI modes.
7919 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7921 Update copyright years.
7923 2018-01-03 Martin Liska <mliska@suse.cz>
7926 * ipa-visibility.c (function_and_variable_visibility): Skip
7927 functions with noipa attribure.
7929 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7931 * gcc.c (process_command): Update copyright notice dates.
7932 * gcov-dump.c (print_version): Ditto.
7933 * gcov.c (print_version): Ditto.
7934 * gcov-tool.c (print_version): Ditto.
7935 * gengtype.c (create_file): Ditto.
7936 * doc/cpp.texi: Bump @copying's copyright year.
7937 * doc/cppinternals.texi: Ditto.
7938 * doc/gcc.texi: Ditto.
7939 * doc/gccint.texi: Ditto.
7940 * doc/gcov.texi: Ditto.
7941 * doc/install.texi: Ditto.
7942 * doc/invoke.texi: Ditto.
7944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7946 * vector-builder.h (vector_builder::m_full_nelts): Change from
7947 unsigned int to poly_uint64.
7948 (vector_builder::full_nelts): Update prototype accordingly.
7949 (vector_builder::new_vector): Likewise.
7950 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7951 (vector_builder::operator ==): Likewise.
7952 (vector_builder::finalize): Likewise.
7953 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7954 Take the number of elements as a poly_uint64 rather than an
7956 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7957 from unsigned int to poly_uint64.
7958 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7959 (vec_perm_indices::new_vector): Likewise.
7960 (vec_perm_indices::length): Likewise.
7961 (vec_perm_indices::nelts_per_input): Likewise.
7962 (vec_perm_indices::input_nelts): Likewise.
7963 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7964 number of elements per input as a poly_uint64 rather than an
7965 unsigned int. Use the original encoding for variable-length
7966 vectors, rather than clamping each individual element.
7967 For the second and subsequent elements in each pattern,
7968 clamp the step and base before clamping their sum.
7969 (vec_perm_indices::series_p): Handle polynomial element counts.
7970 (vec_perm_indices::all_in_range_p): Likewise.
7971 (vec_perm_indices_to_tree): Likewise.
7972 (vec_perm_indices_to_rtx): Likewise.
7973 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7974 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7975 (tree_vector_builder::new_binary_operation): Handle polynomial
7976 element counts. Return false if we need to know the number
7977 of elements at compile time.
7978 * fold-const.c (fold_vec_perm): Punt if the number of elements
7979 isn't known at compile time.
7981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7983 * vec-perm-indices.h (vec_perm_builder): Change element type
7984 from HOST_WIDE_INT to poly_int64.
7985 (vec_perm_indices::element_type): Update accordingly.
7986 (vec_perm_indices::clamp): Handle polynomial element_types.
7987 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7988 (vec_perm_indices::all_in_range_p): Likewise.
7989 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7991 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7992 polynomial vec_perm_indices element types.
7993 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7994 * fold-const.c (fold_vec_perm): Likewise.
7995 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7996 * tree-vect-generic.c (lower_vec_perm): Likewise.
7997 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7998 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7999 element type to HOST_WIDE_INT.
8001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8002 Alan Hayward <alan.hayward@arm.com>
8003 David Sherwood <david.sherwood@arm.com>
8005 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8006 rather than an int. Use plus_constant.
8007 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8008 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8011 Alan Hayward <alan.hayward@arm.com>
8012 David Sherwood <david.sherwood@arm.com>
8014 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8015 a HOST_WIDE_INT to a poly_int64.
8017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8018 Alan Hayward <alan.hayward@arm.com>
8019 David Sherwood <david.sherwood@arm.com>
8021 * calls.c (load_register_parameters): Cope with polynomial
8022 mode sizes. Require a constant size for BLKmode parameters
8023 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8024 forces a parameter to be padded at the lsb end in order to
8025 fill a complete number of words, require the parameter size
8026 to be ordered wrt UNITS_PER_WORD.
8028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8029 Alan Hayward <alan.hayward@arm.com>
8030 David Sherwood <david.sherwood@arm.com>
8032 * reload1.c (spill_stack_slot_width): Change element type
8033 from unsigned int to poly_uint64_pod.
8034 (alter_reg): Treat mode sizes as polynomial.
8036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8037 Alan Hayward <alan.hayward@arm.com>
8038 David Sherwood <david.sherwood@arm.com>
8040 * reload.c (complex_word_subreg_p): New function.
8041 (reload_inner_reg_of_subreg, push_reload): Use it.
8043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8044 Alan Hayward <alan.hayward@arm.com>
8045 David Sherwood <david.sherwood@arm.com>
8047 * lra-constraints.c (process_alt_operands): Reject matched
8048 operands whose sizes aren't ordered.
8049 (match_reload): Refer to this check here.
8051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8052 Alan Hayward <alan.hayward@arm.com>
8053 David Sherwood <david.sherwood@arm.com>
8055 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8056 that the mode size is in the set {1, 2, 4, 8, 16}.
8058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8059 Alan Hayward <alan.hayward@arm.com>
8060 David Sherwood <david.sherwood@arm.com>
8062 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8063 Use plus_constant instead of gen_rtx_PLUS.
8065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8066 Alan Hayward <alan.hayward@arm.com>
8067 David Sherwood <david.sherwood@arm.com>
8069 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8070 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8071 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8072 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8073 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8074 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8075 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8076 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8077 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8078 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8080 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8081 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8082 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8083 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8084 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8085 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8086 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8087 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8088 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8089 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8091 * expr.c (emit_move_resolve_push): Treat the input and result
8092 of PUSH_ROUNDING as a poly_int64.
8093 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8094 (emit_push_insn): Likewise.
8095 * lra-eliminations.c (mark_not_eliminable): Likewise.
8096 * recog.c (push_operand): Likewise.
8097 * reload1.c (elimination_effects): Likewise.
8098 * rtlanal.c (nonzero_bits1): Likewise.
8099 * calls.c (store_one_arg): Likewise. Require the padding to be
8100 known at compile time.
8102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8103 Alan Hayward <alan.hayward@arm.com>
8104 David Sherwood <david.sherwood@arm.com>
8106 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8107 Use plus_constant instead of gen_rtx_PLUS.
8109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8110 Alan Hayward <alan.hayward@arm.com>
8111 David Sherwood <david.sherwood@arm.com>
8113 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8117 Alan Hayward <alan.hayward@arm.com>
8118 David Sherwood <david.sherwood@arm.com>
8120 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8121 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8122 via stack temporaries. Treat the mode size as polynomial too.
8124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8125 Alan Hayward <alan.hayward@arm.com>
8126 David Sherwood <david.sherwood@arm.com>
8128 * expr.c (expand_expr_real_2): When handling conversions involving
8129 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8130 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8131 as a poly_uint64 too.
8133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8134 Alan Hayward <alan.hayward@arm.com>
8135 David Sherwood <david.sherwood@arm.com>
8137 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8140 Alan Hayward <alan.hayward@arm.com>
8141 David Sherwood <david.sherwood@arm.com>
8143 * combine.c (can_change_dest_mode): Handle polynomial
8144 REGMODE_NATURAL_SIZE.
8145 * expmed.c (store_bit_field_1): Likewise.
8146 * expr.c (store_constructor): Likewise.
8147 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8148 and polynomial REGMODE_NATURAL_SIZE.
8149 (gen_lowpart_common): Likewise.
8150 * reginfo.c (record_subregs_of_mode): Likewise.
8151 * rtlanal.c (read_modify_subreg_p): Likewise.
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8154 Alan Hayward <alan.hayward@arm.com>
8155 David Sherwood <david.sherwood@arm.com>
8157 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8158 numbers of elements.
8160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8161 Alan Hayward <alan.hayward@arm.com>
8162 David Sherwood <david.sherwood@arm.com>
8164 * match.pd: Cope with polynomial numbers of vector elements.
8166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8167 Alan Hayward <alan.hayward@arm.com>
8168 David Sherwood <david.sherwood@arm.com>
8170 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8171 in a POINTER_PLUS_EXPR.
8173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8174 Alan Hayward <alan.hayward@arm.com>
8175 David Sherwood <david.sherwood@arm.com>
8177 * omp-simd-clone.c (simd_clone_subparts): New function.
8178 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8179 (ipa_simd_modify_function_body): Likewise.
8181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8182 Alan Hayward <alan.hayward@arm.com>
8183 David Sherwood <david.sherwood@arm.com>
8185 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8186 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8187 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8188 (expand_vector_condition, vector_element): Likewise.
8189 (subparts_gt): New function.
8190 (get_compute_type): Use subparts_gt.
8191 (count_type_subparts): Delete.
8192 (expand_vector_operations_1): Use subparts_gt instead of
8193 count_type_subparts.
8195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8196 Alan Hayward <alan.hayward@arm.com>
8197 David Sherwood <david.sherwood@arm.com>
8199 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8200 (vect_compile_time_alias): ...this new function. Do the calculation
8201 on poly_ints rather than trees.
8202 (vect_prune_runtime_alias_test_list): Update call accordingly.
8204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8205 Alan Hayward <alan.hayward@arm.com>
8206 David Sherwood <david.sherwood@arm.com>
8208 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8210 (vect_schedule_slp_instance): Likewise.
8212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8213 Alan Hayward <alan.hayward@arm.com>
8214 David Sherwood <david.sherwood@arm.com>
8216 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8217 constant and extern definitions for variable-length vectors.
8218 (vect_get_constant_vectors): Note that the number of units
8219 is known to be constant.
8221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8222 Alan Hayward <alan.hayward@arm.com>
8223 David Sherwood <david.sherwood@arm.com>
8225 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8226 of units as polynomial. Choose between WIDE and NARROW based
8229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8230 Alan Hayward <alan.hayward@arm.com>
8231 David Sherwood <david.sherwood@arm.com>
8233 * tree-vect-stmts.c (simd_clone_subparts): New function.
8234 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8237 Alan Hayward <alan.hayward@arm.com>
8238 David Sherwood <david.sherwood@arm.com>
8240 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8241 vectors as polynomial. Use build_index_vector for
8244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8245 Alan Hayward <alan.hayward@arm.com>
8246 David Sherwood <david.sherwood@arm.com>
8248 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8249 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8250 for variable-length vectors.
8251 (vectorizable_mask_load_store): Treat the number of units as
8252 polynomial, asserting that it is constant if the condition has
8253 already been enforced.
8254 (vectorizable_store, vectorizable_load): Likewise.
8256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8257 Alan Hayward <alan.hayward@arm.com>
8258 David Sherwood <david.sherwood@arm.com>
8260 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8261 of units as polynomial. Punt if we can't tell at compile time
8262 which vector contains the final result.
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 * tree-vect-loop.c (vectorizable_induction): Treat the number
8269 of units as polynomial. Punt on SLP inductions. Use an integer
8270 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8271 cast of such a series for variable-length floating-point
8274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8275 Alan Hayward <alan.hayward@arm.com>
8276 David Sherwood <david.sherwood@arm.com>
8278 * tree.h (build_index_vector): Declare.
8279 * tree.c (build_index_vector): New function.
8280 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8281 of units as polynomial, forcibly converting it to a constant if
8282 vectorizable_reduction has already enforced the condition.
8283 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8284 to create a {1,2,3,...} vector.
8285 (vectorizable_reduction): Treat the number of units as polynomial.
8286 Choose vectype_in based on the largest scalar element size rather
8287 than the smallest number of units. Enforce the restrictions
8290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8291 Alan Hayward <alan.hayward@arm.com>
8292 David Sherwood <david.sherwood@arm.com>
8294 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8295 number of units as polynomial.
8297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8298 Alan Hayward <alan.hayward@arm.com>
8299 David Sherwood <david.sherwood@arm.com>
8301 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8302 * target.def (autovectorize_vector_sizes): Return the vector sizes
8303 by pointer, using vector_sizes rather than a bitmask.
8304 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8305 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8306 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8308 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8309 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8310 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8311 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8312 * omp-general.c (omp_max_vf): Likewise.
8313 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8314 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8315 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8316 * tree-vect-slp.c (vect_slp_bb): Likewise.
8317 * doc/tm.texi: Regenerate.
8318 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8320 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8321 the vector size as a poly_uint64 rather than an unsigned int.
8322 (current_vector_size): Change from an unsigned int to a poly_uint64.
8323 (get_vectype_for_scalar_type): Update accordingly.
8324 * tree.h (build_truth_vector_type): Take the size and number of
8325 units as a poly_uint64 rather than an unsigned int.
8326 (build_vector_type): Add a temporary overload that takes
8327 the number of units as a poly_uint64 rather than an unsigned int.
8328 * tree.c (make_vector_type): Likewise.
8329 (build_truth_vector_type): Take the number of units as a poly_uint64
8330 rather than an unsigned int.
8332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8333 Alan Hayward <alan.hayward@arm.com>
8334 David Sherwood <david.sherwood@arm.com>
8336 * target.def (get_mask_mode): Take the number of units and length
8337 as poly_uint64s rather than unsigned ints.
8338 * targhooks.h (default_get_mask_mode): Update accordingly.
8339 * targhooks.c (default_get_mask_mode): Likewise.
8340 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8341 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8348 * omp-general.c (omp_max_vf): Likewise.
8349 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8350 (expand_omp_simd): Handle polynomial safelen.
8351 * omp-low.c (omplow_simd_context): Add a default constructor.
8352 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8353 (lower_rec_simd_input_clauses): Update accordingly.
8354 (lower_rec_input_clauses): Likewise.
8356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8357 Alan Hayward <alan.hayward@arm.com>
8358 David Sherwood <david.sherwood@arm.com>
8360 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8361 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8362 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8363 (vect_analyze_slp_cost): Likewise.
8364 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8365 (vect_model_load_cost): Likewise.
8367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8368 Alan Hayward <alan.hayward@arm.com>
8369 David Sherwood <david.sherwood@arm.com>
8371 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8372 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8373 from an unsigned int * to a poly_uint64_pod *.
8374 (calculate_unrolling_factor): New function.
8375 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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-vectorizer.h (_slp_instance::unrolling_factor): Change
8382 from an unsigned int to a poly_uint64.
8383 (_loop_vec_info::slp_unrolling_factor): Likewise.
8384 (_loop_vec_info::vectorization_factor): Change from an int
8386 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8387 (vect_get_num_vectors): New function.
8388 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8389 (vect_get_num_copies): Use vect_get_num_vectors.
8390 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8391 to an unsigned int *.
8392 (vect_analyze_data_refs): Change min_vf from an int * to a
8394 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8395 than an unsigned HOST_WIDE_INT.
8396 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8397 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8398 to an unsigned int *.
8399 (vect_analyze_data_ref_dependences): Likewise.
8400 (vect_compute_data_ref_alignment): Handle polynomial vf.
8401 (vect_enhance_data_refs_alignment): Likewise.
8402 (vect_prune_runtime_alias_test_list): Likewise.
8403 (vect_shift_permute_load_chain): Likewise.
8404 (vect_supportable_dr_alignment): Likewise.
8405 (dependence_distance_ge_vf): Take the vectorization factor as a
8406 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8407 (vect_analyze_data_refs): Change min_vf from an int * to a
8409 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8410 vfm1 as a poly_uint64 rather than an int. Make the same change
8411 for the returned bound_scalar.
8412 (vect_gen_vector_loop_niters): Handle polynomial vf.
8413 (vect_do_peeling): Likewise. Update call to
8414 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8415 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8417 * tree-vect-loop.c (vect_determine_vectorization_factor)
8418 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8419 (vect_get_known_peeling_cost): Likewise.
8420 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8421 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8422 (vect_transform_loop): Likewise. Use the lowest possible VF when
8423 updating the upper bounds of the loop.
8424 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8426 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8427 polynomial unroll factors.
8428 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8429 (vect_make_slp_decision): Likewise.
8430 (vect_supported_load_permutation_p): Likewise, and polynomial
8432 (vect_analyze_slp_cost): Handle polynomial vf.
8433 (vect_slp_analyze_node_operations): Likewise.
8434 (vect_slp_analyze_bb_1): Likewise.
8435 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8436 than an unsigned HOST_WIDE_INT.
8437 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8438 (vectorizable_load): Handle polynomial vf.
8439 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8441 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8444 Alan Hayward <alan.hayward@arm.com>
8445 David Sherwood <david.sherwood@arm.com>
8447 * match.pd: Handle bit operations involving three constants
8448 and try to fold one pair.
8450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8452 * tree-vect-loop-manip.c: Include gimple-fold.h.
8453 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8454 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8455 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8456 Add a path that uses a step of VF instead of 1, but disable it
8458 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8459 and niters_no_overflow parameters. Update calls to
8460 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8461 Create a new SSA name if the latter choses to use a ste other
8462 than zero, and return it via niters_vector_mult_vf_var.
8463 * tree-vect-loop.c (vect_transform_loop): Update calls to
8464 vect_do_peeling, vect_gen_vector_loop_niters and
8465 slpeel_make_loop_iterate_ntimes.
8466 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8467 (vect_gen_vector_loop_niters): Update declarations after above changes.
8469 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8471 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8472 128-bit round to integer instructions.
8473 (ceil<mode>2): Likewise.
8474 (btrunc<mode>2): Likewise.
8475 (round<mode>2): Likewise.
8477 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8479 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8480 unaligned VSX load/store on P8/P9.
8481 (expand_block_clear): Allow the use of unaligned VSX
8482 load/store on P8/P9.
8484 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8486 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8488 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8489 swap associated with both a load and a store.
8491 2018-01-02 Andrew Waterman <andrew@sifive.com>
8493 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8494 * config/riscv/riscv.md (clear_cache): Use it.
8496 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8498 * web.c: Remove out-of-date comment.
8500 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8502 * expr.c (fixup_args_size_notes): Check that any existing
8503 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8504 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8505 (emit_single_push_insn): ...here.
8507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8509 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8510 (const_vector_encoded_nelts): New function.
8511 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8512 (const_vector_int_elt, const_vector_elt): Declare.
8513 * emit-rtl.c (const_vector_int_elt_1): New function.
8514 (const_vector_elt): Likewise.
8515 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8516 of CONST_VECTOR_ELT.
8518 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8520 * expr.c: Include rtx-vector-builder.h.
8521 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8522 directly on the tree encoding.
8523 (const_vector_from_tree): Likewise.
8524 * optabs.c: Include rtx-vector-builder.h.
8525 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8526 sequence of "u" values.
8527 * vec-perm-indices.c: Include rtx-vector-builder.h.
8528 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8529 directly on the vec_perm_indices encoding.
8531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8533 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8534 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8535 * rtx-vector-builder.h: New file.
8536 * rtx-vector-builder.c: Likewise.
8537 * rtl.h (rtx_def::u2): Add a const_vector field.
8538 (CONST_VECTOR_NPATTERNS): New macro.
8539 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8540 (CONST_VECTOR_DUPLICATE_P): Likewise.
8541 (CONST_VECTOR_STEPPED_P): Likewise.
8542 (CONST_VECTOR_ENCODED_ELT): Likewise.
8543 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8544 (unwrap_const_vec_duplicate): Likewise.
8545 (const_vec_series_p): Check for a non-duplicated vector encoding.
8546 Say that the function only returns true for integer vectors.
8547 * emit-rtl.c: Include rtx-vector-builder.h.
8548 (gen_const_vec_duplicate_1): Delete.
8549 (gen_const_vector): Call gen_const_vec_duplicate instead of
8550 gen_const_vec_duplicate_1.
8551 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8552 (gen_const_vec_duplicate): Use rtx_vector_builder.
8553 (gen_const_vec_series): Likewise.
8554 (gen_rtx_CONST_VECTOR): Likewise.
8555 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8556 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8557 Build a new vector rather than modifying a CONST_VECTOR in-place.
8558 (handle_special_swappables): Update call accordingly.
8559 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8560 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8561 Build a new vector rather than modifying a CONST_VECTOR in-place.
8562 (handle_special_swappables): Update call accordingly.
8564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8566 * simplify-rtx.c (simplify_const_binary_operation): Use
8567 CONST_VECTOR_ELT instead of XVECEXP.
8569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8571 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8572 the selector elements to be different from the data elements
8573 if the selector is a VECTOR_CST.
8574 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8575 ssizetype for the selector.
8577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8579 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8580 before testing each element individually.
8581 * tree-vect-generic.c (lower_vec_perm): Likewise.
8583 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8585 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8586 * selftest-run-tests.c (selftest::run_tests): Call it.
8587 * vector-builder.h (vector_builder::operator ==): New function.
8588 (vector_builder::operator !=): Likewise.
8589 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8590 (vec_perm_indices::all_from_input_p): New function.
8591 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8592 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8593 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8594 instead of reading the VECTOR_CST directly. Detect whether both
8595 vector inputs are the same before constructing the vec_perm_indices,
8596 and update the number of inputs argument accordingly. Use the
8597 utility functions added above. Only construct sel2 if we need to.
8599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8601 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8602 the broadcast of the low byte.
8603 (expand_mult_highpart): Use an explicit encoding for the permutes.
8604 * optabs-query.c (can_mult_highpart_p): Likewise.
8605 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8606 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8607 (vectorizable_bswap): Likewise.
8608 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8609 explicit encoding for the power-of-2 permutes.
8610 (vect_permute_store_chain): Likewise.
8611 (vect_grouped_load_supported): Likewise.
8612 (vect_permute_load_chain): Likewise.
8614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8616 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8617 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8618 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8619 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8620 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8621 (vect_gen_perm_mask_any): Likewise.
8623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8625 * int-vector-builder.h: New file.
8626 * vec-perm-indices.h: Include int-vector-builder.h.
8627 (vec_perm_indices): Redefine as an int_vector_builder.
8628 (auto_vec_perm_indices): Delete.
8629 (vec_perm_builder): Redefine as a stand-alone class.
8630 (vec_perm_indices::vec_perm_indices): New function.
8631 (vec_perm_indices::clamp): Likewise.
8632 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8633 (vec_perm_indices::new_vector): New function.
8634 (vec_perm_indices::new_expanded_vector): Update for new
8635 vec_perm_indices class.
8636 (vec_perm_indices::rotate_inputs): New function.
8637 (vec_perm_indices::all_in_range_p): Operate directly on the
8638 encoded form, without computing elided elements.
8639 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8640 encoding. Update for new vec_perm_indices class.
8641 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8642 the given vec_perm_builder.
8643 (expand_vec_perm_var): Update vec_perm_builder constructor.
8644 (expand_mult_highpart): Use vec_perm_builder instead of
8645 auto_vec_perm_indices.
8646 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8647 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8648 or double series encoding as appropriate.
8649 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8650 vec_perm_indices instead of auto_vec_perm_indices.
8651 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8652 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8653 (vect_permute_store_chain): Likewise.
8654 (vect_grouped_load_supported): Likewise.
8655 (vect_permute_load_chain): Likewise.
8656 (vect_shift_permute_load_chain): Likewise.
8657 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8658 (vect_transform_slp_perm_load): Likewise.
8659 (vect_schedule_slp_instance): Likewise.
8660 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8661 (vectorizable_mask_load_store): Likewise.
8662 (vectorizable_bswap): Likewise.
8663 (vectorizable_store): Likewise.
8664 (vectorizable_load): Likewise.
8665 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8666 vec_perm_indices instead of auto_vec_perm_indices. Use
8667 tree_to_vec_perm_builder to read the vector from a tree.
8668 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8669 vec_perm_builder instead of a vec_perm_indices.
8670 (have_whole_vector_shift): Use vec_perm_builder and
8671 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8672 truncation to calc_vec_perm_mask_for_shift.
8673 (vect_create_epilog_for_reduction): Likewise.
8674 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8675 from auto_vec_perm_indices to vec_perm_indices.
8676 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8677 instead of changing individual elements.
8678 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8679 the vector in d.perm.
8680 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8681 from auto_vec_perm_indices to vec_perm_indices.
8682 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8683 instead of changing individual elements.
8684 (arm_vectorize_vec_perm_const): Use new_vector to install
8685 the vector in d.perm.
8686 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8687 Update vec_perm_builder constructor.
8688 (rs6000_expand_interleave): Likewise.
8689 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8690 (rs6000_expand_interleave): Likewise.
8692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8694 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8695 to qimode could truncate the indices.
8696 * optabs.c (expand_vec_perm_var): Likewise.
8698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8700 * Makefile.in (OBJS): Add vec-perm-indices.o.
8701 * vec-perm-indices.h: New file.
8702 * vec-perm-indices.c: Likewise.
8703 * target.h (vec_perm_indices): Replace with a forward class
8705 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8706 * optabs.h: Include vec-perm-indices.h.
8707 (expand_vec_perm): Delete.
8708 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8709 (expand_vec_perm_const): Declare.
8710 * target.def (vec_perm_const_ok): Replace with...
8711 (vec_perm_const): ...this new hook.
8712 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8713 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8714 * doc/tm.texi: Regenerate.
8715 * optabs.def (vec_perm_const): Delete.
8716 * doc/md.texi (vec_perm_const): Likewise.
8717 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8718 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8719 expand_vec_perm for constant permutation vectors. Assert that
8720 the mode of variable permutation vectors is the integer equivalent
8721 of the mode that is being permuted.
8722 * optabs-query.h (selector_fits_mode_p): Declare.
8723 * optabs-query.c: Include vec-perm-indices.h.
8724 (selector_fits_mode_p): New function.
8725 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8726 is defined, instead of checking whether the vec_perm_const_optab
8727 exists. Use targetm.vectorize.vec_perm_const instead of
8728 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8729 fit in the vector mode before using a variable permute.
8730 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8731 vec_perm_indices instead of an rtx.
8732 (expand_vec_perm): Replace with...
8733 (expand_vec_perm_const): ...this new function. Take the selector
8734 as a vec_perm_indices rather than an rtx. Also take the mode of
8735 the selector. Update call to shift_amt_for_vec_perm_mask.
8736 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8737 Use vec_perm_indices::new_expanded_vector to expand the original
8738 selector into bytes. Check whether the indices fit in the vector
8739 mode before using a variable permute.
8740 (expand_vec_perm_var): Make global.
8741 (expand_mult_highpart): Use expand_vec_perm_const.
8742 * fold-const.c: Includes vec-perm-indices.h.
8743 * tree-ssa-forwprop.c: Likewise.
8744 * tree-vect-data-refs.c: Likewise.
8745 * tree-vect-generic.c: Likewise.
8746 * tree-vect-loop.c: Likewise.
8747 * tree-vect-slp.c: Likewise.
8748 * tree-vect-stmts.c: Likewise.
8749 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8751 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8752 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8753 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8754 (aarch64_vectorize_vec_perm_const): ...this new function.
8755 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8756 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8757 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8758 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8759 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8760 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8761 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8763 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8764 check for NEON modes.
8765 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8766 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8767 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8768 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8770 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8771 the old VEC_PERM_CONST conditions.
8772 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8773 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8774 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8775 (ia64_vectorize_vec_perm_const_ok): Merge into...
8776 (ia64_vectorize_vec_perm_const): ...this new function.
8777 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8778 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8779 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8780 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8781 * config/mips/mips.c (mips_expand_vec_perm_const)
8782 (mips_vectorize_vec_perm_const_ok): Merge into...
8783 (mips_vectorize_vec_perm_const): ...this new function.
8784 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8785 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8786 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8787 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8788 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8789 (rs6000_expand_vec_perm_const): Delete.
8790 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8792 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8793 (altivec_expand_vec_perm_const_le): Take each operand individually.
8794 Operate on constant selectors rather than rtxes.
8795 (altivec_expand_vec_perm_const): Likewise. Update call to
8796 altivec_expand_vec_perm_const_le.
8797 (rs6000_expand_vec_perm_const): Delete.
8798 (rs6000_vectorize_vec_perm_const_ok): Delete.
8799 (rs6000_vectorize_vec_perm_const): New function.
8800 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8801 an element count and rtx array.
8802 (rs6000_expand_extract_even): Update call accordingly.
8803 (rs6000_expand_interleave): Likewise.
8804 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8805 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8806 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8807 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8808 (rs6000_expand_vec_perm_const): Delete.
8809 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8810 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8811 (altivec_expand_vec_perm_const_le): Take each operand individually.
8812 Operate on constant selectors rather than rtxes.
8813 (altivec_expand_vec_perm_const): Likewise. Update call to
8814 altivec_expand_vec_perm_const_le.
8815 (rs6000_expand_vec_perm_const): Delete.
8816 (rs6000_vectorize_vec_perm_const_ok): Delete.
8817 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8818 reference to the SPE evmerge intructions.
8819 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8820 an element count and rtx array.
8821 (rs6000_expand_extract_even): Update call accordingly.
8822 (rs6000_expand_interleave): Likewise.
8823 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8824 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8826 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8830 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8831 vector mode and that that mode matches the mode of the data
8833 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8834 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8835 directly using expand_vec_perm_1 when forcing selectors into
8837 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8841 * optabs-query.h (can_vec_perm_p): Delete.
8842 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8843 * optabs-query.c (can_vec_perm_p): Split into...
8844 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8845 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8846 particular selector is valid.
8847 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8848 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8849 (vect_grouped_load_supported): Likewise.
8850 (vect_shift_permute_load_chain): Likewise.
8851 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8852 (vect_transform_slp_perm_load): Likewise.
8853 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8854 (vectorizable_bswap): Likewise.
8855 (vect_gen_perm_mask_checked): Likewise.
8856 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8857 implementations of variable permutation vectors into account
8858 when deciding which selector to use.
8859 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8860 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8861 with a false third argument.
8862 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8863 to test whether the constant selector is valid and can_vec_perm_var_p
8864 to test whether a variable selector is valid.
8866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8868 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8869 * optabs-query.c (can_vec_perm_p): Likewise.
8870 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8871 instead of vec_perm_indices.
8872 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8873 (vect_gen_perm_mask_checked): Likewise,
8874 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8875 (vect_gen_perm_mask_checked): Likewise,
8877 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8879 * optabs-query.h (qimode_for_vec_perm): Declare.
8880 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8881 (qimode_for_vec_perm): ...this new function.
8882 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8884 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8886 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8887 does not have a conditional at the top.
8889 2018-01-02 Richard Biener <rguenther@suse.de>
8891 * ipa-inline.c (big_speedup_p): Fix expression.
8893 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8896 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8899 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8903 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8904 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8905 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8906 cond_taken_branch_cost 3->4.
8908 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8910 PR tree-optimization/83581
8911 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8912 TODO_cleanup_cfg if any changes have been made.
8915 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8916 convert_modes if target mode has the right side, but different mode
8920 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8921 last argument when extracting from CONCAT. If either from_real or
8922 from_imag is NULL, use expansion through memory. If result is not
8923 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8924 the parts directly to inner mode, if even that fails, use expansion
8928 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8929 check for bswap in mode rather than HImode and use that in expand_unop
8932 Copyright (C) 2018 Free Software Foundation, Inc.
8934 Copying and distribution of this file, with or without modification,
8935 are permitted in any medium without royalty provided the copyright
8936 notice and this notice are preserved.