1 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5 insn for floating-point loads and stores.
7 2018-03-14 Carl Love <cel@us.ibm.com>
9 * config/rs6000/rs6000-c.c: Add macro definitions for
10 ALTIVEC_BUILTIN_VEC_PERMXOR.
11 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
13 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16 * config/doc/extend.texi: Add prototypes for vec_permxor.
18 2018-03-14 David Malcolm <dmalcolm@redhat.com>
21 * diagnostic-show-locus.c (class layout_point): Convert m_line
22 from int to linenum_type.
23 (line_span::comparator): Use linenum "compare" function when
24 comparing line numbers.
25 (test_line_span): New function.
26 (layout_range::contains_point): Convert param "row" from int to
28 (layout_range::intersects_line_p): Likewise.
29 (layout::will_show_line_p): Likewise.
30 (layout::print_source_line): Likewise.
31 (layout::should_print_annotation_line_p): Likewise.
32 (layout::print_annotation_line): Likewise.
33 (layout::print_leading_fixits): Likewise.
34 (layout::annotation_line_showed_range_p): Likewise.
35 (struct line_corrections): Likewise for field m_row.
36 (line_corrections::line_corrections): Likewise for param "row".
37 (layout::print_trailing_fixits): Likewise.
38 (layout::get_state_at_point): Likewise.
39 (layout::get_x_bound_for_row): Likewise.
40 (layout::print_line): Likewise.
41 (diagnostic_show_locus): Likewise for locals "last_line" and
43 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
44 * input.c (selftest::test_linenum_comparisons): New function.
45 (selftest::input_c_tests): Call it.
46 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
47 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
48 * selftest.h (ASSERT_GT): New macro.
49 (ASSERT_GT_AT): New macro.
50 (ASSERT_LT): New macro.
51 (ASSERT_LT_AT): New macro.
53 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
55 PR rtl-optimization/84780
56 * combine.c (distribute_links): Don't make a link based on pc_rtx.
58 2018-03-14 Martin Liska <mliska@suse.cz>
60 * tree.c (record_node_allocation_statistics): Use
62 (get_stats_node_kind): New function extracted from
63 record_node_allocation_statistics.
64 (free_node): Use get_stats_node_kind.
66 2018-03-14 Richard Biener <rguenther@suse.de>
68 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
69 that the value-set of ANTIC_IN doesn't grow.
72 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
74 (BB_VISITED_WITH_VISITED_SUCCS): New define.
75 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
77 2018-03-14 Julia Koval <julia.koval@intel.com>
79 * config.gcc (icelake-client, icelake-server): New.
81 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
82 (initial_ix86_arch_features): Ditto.
83 (PTA_SKYLAKE): Add SGX.
84 (PTA_ICELAKE): Remove.
85 (PTA_ICELAKE_CLIENT): New.
86 (PTA_ICELAKE_SERVER): New.
87 (ix86_option_override_internal): Split up icelake on icelake client and
89 (get_builtin_code_for_version): Ditto.
90 (fold_builtin_cpu): Ditto.
91 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
92 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
93 * config/i386/i386.h (processor_type): Ditto.
94 * doc/invoke.texi: Ditto.
96 2018-03-14 Jakub Jelinek <jakub@redhat.com>
99 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
100 INTEGER_CST offset, add it together with bitpos / 8 and
101 sign extend based on POINTER_SIZE.
105 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
108 * config/i386/constraints.md (Yc): New register constraint.
109 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
110 Use Yc constraint for alternative 2 of operand 0. Remove
111 preferred_for_speed attribute.
113 2018-03-14 Richard Biener <rguenther@suse.de>
115 PR tree-optimization/84830
116 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
117 with the old one to avoid oscillations.
119 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
122 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
124 (assign_by_spills): Return a flag of reload assignment failure.
125 Do not process the reload assignment failures. Do not spill other
126 reload pseudos if they has the same reg class. Update n if
128 (lra_assign): Add a return arg. Set up from the result of
129 assign_by_spills call.
130 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
131 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
132 usage_insns if it is not NULL.
133 (spill_hard_reg_in_range): New function.
134 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
135 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
137 (lra_assign): Change prototype.
138 * lra.c (lra): Add code to deal with fails by splitting hard reg
141 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
143 * config/riscv/riscv.opt (mrelax): New option.
144 * config/riscv/riscv.c (riscv_file_start): Emit ".option
145 "norelax" when riscv_mrelax is disabled.
146 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
148 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
151 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
152 reassociation for int modes.
154 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
156 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
157 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
159 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
160 * config/aarch64/aarch64-sve.md
161 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
162 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
163 (*extend<mode><Vwide>2): Rename to...
164 (aarch64_sve_extend<mode><Vwide>2): ...this.
165 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
166 renaming the old pattern to...
167 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
169 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
170 define_expand, renaming the old pattern to...
171 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
172 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
173 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
174 account when deciding which SVE instruction the optab should use.
175 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
177 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
179 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
180 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
181 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
182 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
183 (tlsdesc_small_<mode>): Turn a define_expand and use
184 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
185 (tlsdesc_small_advsimd_<mode>): ...this.
186 (tlsdesc_small_sve_<mode>): New pattern.
188 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
190 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
191 (UNSPEC_UMUL_HIGHPART): New constants.
192 (MUL_HIGHPART): New int iteraor.
193 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
194 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
196 (*<su>mul<mode>3_highpart): New define_insn.
198 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
201 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
204 2018-03-13 Martin Liska <mliska@suse.cz>
207 * (sem_item_optimizer::sem_item_optimizer): Initialize new
209 (sem_item_optimizer::~sem_item_optimizer): Release it.
210 (sem_item_optimizer::merge_classes): Register variable aliases.
211 (sem_item_optimizer::fixup_pt_set): New function.
212 (sem_item_optimizer::fixup_points_to_sets): Likewise.
213 * ipa-icf.h: Declare new variables and functions.
215 2018-03-13 Jakub Jelinek <jakub@redhat.com>
218 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
219 integer_pow2p@2 and test integer_pow2p in condition.
220 (A < 0 ? C : 0): Similarly for @1.
223 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
224 characters starting at p contain '\0' character, don't look beyond
228 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
229 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
232 * reg-stack.c (change_stack): Change update_end var from int to
233 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
234 also call set_block_for_insn on the newly added insns and rescan.
237 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
241 * tree.c (stabilize_reference_1): Return save_expr (e) for
242 STATEMENT_LIST even if it doesn't have side-effects.
244 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
246 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
248 2018-03-12 Renlin Li <renlin.li@arm.com>
250 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
251 aarch64_output_scalar_simd_mov_immediate.
253 2018-03-12 Martin Sebor <msebor@redhat.com>
255 PR tree-optimization/83456
256 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
257 for perfectly overlapping calls to memcpy.
258 (gimple_fold_builtin_memory_chk): Same.
259 (gimple_fold_builtin_strcpy): Handle no-warning.
260 (gimple_fold_builtin_stxcpy_chk): Same.
261 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
263 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
265 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
266 parameter. Use it for SFmode.
267 (rs6000_function_arg_advance_1): Adjust.
268 (rs6000_function_arg): Adjust.
269 (rs6000_gimplify_va_arg): Pass false for that new parameter.
271 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
273 PR rtl-optimization/84169
274 PR rtl-optimization/84780
275 * combine.c (can_combine_p): Check for a 2-insn combination whether
276 the destination register is used between the two insns, too.
278 2018-03-12 Richard Biener <rguenther@suse.de>
280 PR tree-optimization/84803
281 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
282 for refs DR analysis didn't process.
284 2018-03-12 Richard Biener <rguenther@suse.de>
286 PR tree-optimization/84777
287 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
288 force-vectorize loops ignore whether we are optimizing for size.
290 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
292 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
293 (TARGET_MD_ASM_ADJUST): Define.
295 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
296 Kito Cheng <kito.cheng@gmail.com>
297 Chung-Ju Wu <jasonwucj@gmail.com>
299 * config/nds32/nds32.c (nds32_compute_stack_frame,
300 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
301 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
302 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
303 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
304 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
305 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
306 * config/nds32/nds32.md (prologue, epilogue): Use macro
307 NDS32_V3PUSH_AVAILABLE_P to do checking.
309 2018-03-11 Jakub Jelinek <jakub@redhat.com>
312 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
313 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
314 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
315 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
316 addition of most attributes on !orig_type_die or the attribute not
317 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
319 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
320 Chung-Ju Wu <jasonwucj@gmail.com>
322 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
324 * config/nds32/nds32.opt (mvh): New option.
326 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
327 Chung-Ju Wu <jasonwucj@gmail.com>
329 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
331 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
332 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
335 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
336 Chung-Ju Wu <jasonwucj@gmail.com>
338 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
340 * config/nds32/nds32-multiple.md (strlensi): New pattern.
341 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
343 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
344 Kito Cheng <kito.cheng@gmail.com>
345 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
348 UNSPEC_FFMISM and UNSPEC_FLMISM.
349 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
350 for ffb, ffmism and flmism.
351 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
352 (unspec_ffmism): Ditto.
353 (unspec_flmism): Ditto.
354 (nds32_expand_builtin_impl): Check if string extension is available.
355 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
356 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
358 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
361 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
364 * lra-assigns.c (assign_by_spills): Return a flag of reload
365 assignment failure. Do not process the reload assignment
366 failures. Do not spill other reload pseudos if they has the same
368 (lra_assign): Add a return arg. Set up from the result of
369 assign_by_spills call.
370 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
371 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
372 usage_insns if it is not NULL.
373 (spill_hard_reg_in_range): New function.
374 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
375 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
377 (lra_assign): Change prototype.
378 * lra.c (lra): Add code to deal with fails by splitting hard reg
381 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
384 * config/i386/i386.opt: Replace Enforcment with Enforcement.
386 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
389 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
390 (dw_val_node): Add val_symbolic_view.
391 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
392 (symview_upper_bound): New.
393 (new_line_info_table): Initialize symviews_since_reset.
394 (dwarf2out_source_line): Count symviews_since_reset and set
396 (dw_val_equal_p): Handle symview.
397 (add_AT_symview): New.
398 (print_dw_val): Handle symview.
399 (attr_checksum, attr_checksum_ordered): Likewise.
400 (same_dw_val_p, size_of_die): Likewise.
401 (value_format, output_die): Likewise.
402 (add_high_low_attributes): Use add_AT_symview for entry_view.
403 (dwarf2out_finish): Reset symview_upper_bound, clear
406 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
409 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
410 Add strict argument and use it.
411 (rs6000_split_multireg_move): Update for new strict argument.
412 (mem_operand_gpr): Disallow all non-offsettable addresses.
413 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
415 2018-03-09 Jakub Jelinek <jakub@redhat.com>
418 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
419 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
420 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
423 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
424 decl, use remap_type if we want to use the type.
426 2018-03-09 Martin Sebor <msebor@redhat.com>
428 PR tree-optimization/84526
429 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
431 (builtin_access::generic_overlap): Be prepared to handle non-array
434 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
436 PR rtl-optimization/84682
437 * lra-constraints.c (process_address_1): Check is_address flag
438 for address constraints.
439 (process_alt_operands): Likewise.
440 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
441 preprocess_constraints.
442 * recog.h (preprocess_constraints): Add oploc parameter.
444 * recog.c (preprocess_constraints): Test address_operand for
445 CT_ADDRESS constraints.
447 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
450 * lra-assigns.c (assign_by_spills): Return a flag of reload
451 assignment failure. Do not process the reload assignment
452 failures. Do not spill other reload pseudos if they has the same
454 (lra_assign): Add a return arg. Set up from the result of
455 assign_by_spills call.
456 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
457 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
458 usage_insns if it is not NULL.
459 (spill_hard_reg_in_range): New function.
460 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
461 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
463 (lra_assign): Change prototype.
464 * lra.c (lra): Add code to deal with fails by splitting hard reg
467 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
470 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
471 Accept complain bool parameter. Only emit errors if it is true.
472 (arm_parse_cpu_option_name): Likewise.
473 (arm_target_thumb_only): Adjust callers of the above.
474 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
475 prototype to take a default true bool parameter.
476 (arm_parse_arch_option_name): Likewise.
478 2018-03-09 David Malcolm <dmalcolm@redhat.com>
479 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
483 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
484 * configure: Regenerate.
485 * configure.ac ("linker --version-script option"): New.
486 ("linker soname option"): New.
488 2018-03-09 Richard Biener <rguenther@suse.de>
490 PR tree-optimization/84775
491 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
492 immediate uses of predicate stmts and mark them modified.
495 PR tree-optimization/84178
496 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
498 (version_loop_for_if_conversion): Delay update_ssa call.
499 (tree_if_conversion): Delay update_ssa until after predicate
502 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
505 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
506 when the function accesses prior frames.
508 2018-03-08 Jakub Jelinek <jakub@redhat.com>
511 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
512 gen_llsym, otherwise call maybe_gen_llsym.
515 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
516 has ',' character inside of it.
518 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
521 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
522 as clobbering CC_REGNUM.
524 2018-03-08 Richard Biener <rguenther@suse.de>
527 * tree-scalar-evolution.c: Include tree-into-ssa.h.
528 (follow_copies_to_constant): Do not follow SSA names registered
531 2018-03-08 Richard Biener <rguenther@suse.de>
533 PR tree-optimization/84178
534 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
536 (version_loop_for_if_conversion): Delay update_ssa call.
537 (tree_if_conversion): Delay update_ssa until after predicate
540 2018-03-08 David Malcolm <dmalcolm@redhat.com>
542 PR tree-optimization/84178
543 * tree-if-conv.c (release_bb_predicate): Remove the
544 the assertion that the stmts have NULL use_ops.
545 Discard the statements, asserting that they haven't
546 yet been added to a BB.
548 2018-03-08 Richard Biener <rguenther@suse.de>
550 PR tree-optimization/84746
551 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
552 (phi_translate): Pass in destination ANTIC_OUT set.
553 (phi_translate_1): Likewise. For a simplified result lookup
554 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
555 (phi_translate_set): Adjust.
556 (do_pre_regular_insertion): Likewise.
557 (do_pre_partial_partial_insertion): Likewise.
559 2018-03-08 Martin Liska <mliska@suse.cz>
561 PR gcov-profile/84735
562 * doc/gcov.texi: Document usage of profile files.
563 * gcov-io.h: Document changes in the format.
565 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
569 * dwarf2out.c (struct dw_line_info_table): Update comments for
571 (FORCE_RESET_NEXT_VIEW): New.
572 (FORCE_RESETTING_VIEW_P): New.
573 (RESETTING_VIEW_P): Check for -1 too.
574 (ZERO_VIEW_P): Likewise.
575 (new_line_info_table): Force-reset next view.
576 (dwarf2out_begin_function): Likewise.
577 (dwarf2out_source_line): Simplify zero_view_p initialization.
578 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
579 view directly. Omit view when omitting .loc at line 0.
581 2018-03-08 Jakub Jelinek <jakub@redhat.com>
583 PR tree-optimization/84740
584 * tree-switch-conversion.c (process_switch): Call build_constructors
585 only if info.phi_count is non-zero.
587 PR tree-optimization/84739
588 * tree-tailcall.c (find_tail_calls): Check call arguments against
589 DECL_ARGUMENTS (current_function_decl) rather than
590 DECL_ARGUMENTS (func) when checking for tail recursion.
592 2018-03-07 Jakub Jelinek <jakub@redhat.com>
594 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
595 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
596 Volker Reichelt's entry and add entries for people that perform
597 GCC fuzzy testing and report numerous bugs.
599 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
602 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
603 readonly data in sdata, if that is disabled.
604 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
605 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
606 -mreadonly-in-sdata option.
608 2018-03-07 Martin Sebor <msebor@redhat.com>
610 PR tree-optimization/84468
611 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
612 basic block when looking for nul assignment.
614 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
617 * except.h (output_function_exception_table): Adjust prototype.
618 * except.c (output_function_exception_table): Remove FNNAME parameter
619 and add SECTION parameter. Ouput one part of the table at a time.
620 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
621 the first part of the exception table and emit unwind directives.
622 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
623 (i386_pe_seh_cold_init): Likewise.
624 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
625 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
626 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
627 (ix86_output_call_insn): Emit a nop in one more case for SEH.
628 * config/i386/winnt.c: Include except.h.
629 (struct seh_frame_state): Add reg_offset, after_prologue and
630 in_cold_section fields.
631 (i386_pe_seh_end_prologue): Set seh->after_prologue.
632 (i386_pe_seh_cold_init): New function.
633 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
634 to seh->in_cold_section.
635 (seh_emit_push): Record the offset of the push.
636 (seh_emit_save): Record the offet of the save.
637 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
638 Test seh->after_prologue to disregard the epilogue.
639 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
640 (i386_pe_end_cold_function): New function.
642 2018-03-07 Jakub Jelinek <jakub@redhat.com>
645 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
646 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
649 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
651 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
652 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
655 * multiple_target.c: Include tree-inline.h and intl.h.
656 (expand_target_clones): Diagnose and fail if node->definition and
657 !tree_versionable_function_p (node->decl).
659 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
661 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
663 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
664 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
665 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
667 2018-03-06 Jakub Jelinek <jakub@redhat.com>
670 * combine.c (try_combine): Use reg_or_subregno instead of handling
671 just paradoxical SUBREGs and REGs.
673 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
675 * config/arc/arc.c (arc_finalize_pic): Remove function.
676 (arc_must_save_register): We use single base PIC register, remove
677 checks to save/restore the PIC register.
678 (arc_expand_prologue): Likewise.
679 * config/arc/arc-protos.h (arc_set_default_type_attributes):
681 (arc_verify_short): Likewise.
682 (arc_attr_type): Likewise.
683 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
684 (walk_stores): Likewise.
685 (arc_address_cost): Make it static.
686 (arc_verify_short): Likewise.
687 (branch_dest): Likewise.
688 (arc_attr_type): Likewise.
689 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
690 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
691 (arc_final_prescan_insn): Remove inserting the nops due to
692 hardware hazards. It is done in reorg step.
693 (insn_length_variant_t): Remove.
694 (insn_length_parameters_t): Likewise.
695 (arc_insn_length_parameters): Likewise.
696 (arc_get_insn_variants): Likewise.
697 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
699 2018-03-06 Jakub Jelinek <jakub@redhat.com>
702 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
705 PR tree-optimization/84687
706 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
708 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
710 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
712 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
713 Rename to ppc_speculation_barrier.
714 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
715 __builtin_ppc_speculation_barrier.
717 2018-03-05 Jakub Jelinek <jakub@redhat.com>
720 * combine.c (combine_simplify_rtx): Don't try to simplify if
721 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
724 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
726 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
727 to 32 bytes when compiling for POWER9.
729 2018-03-05 Jakub Jelinek <jakub@redhat.com>
732 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
733 regparm >= 3 with no arg reg available also for calls with
734 flag_force_indirect_call. Pass decl to ix86_function_regparm.
737 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
739 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
741 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
744 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
746 2018-03-05 Richard Biener <rguenther@suse.de>
748 PR tree-optimization/84486
749 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
750 When inserting a __builtin_assume_aligned call set the LHS
751 SSA name alignment info accordingly.
753 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
755 PR tree-optimization/84114
756 * config/aarch64/aarch64.c (aarch64_reassociation_width)
757 Avoid reassociation of FLOAT_MODE addition.
759 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
761 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
762 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
763 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
764 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
765 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
766 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
767 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
769 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
770 __builtin_ia32_wbinvd): New builtins.
771 (SPECIAL_ARGS2): New.
772 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
773 (SPECIAL_ARGS2): New.
774 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
775 (ix86_valid_target_attribute_inner_p): Ditto.
776 (ix86_init_mmx_sse_builtins): Add special_args2.
777 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
778 TARGET_WBNOINVD_P): New.
779 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
780 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
781 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
782 * config/i386/immintrin.h (_wbinvd): New intrinsic.
783 * config/i386/pconfigintrin.h: New file.
784 * config/i386/wbnoinvdintrin.h: Ditto.
785 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
786 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
788 2018-03-05 Richard Biener <rguenther@suse.de>
790 PR tree-optimization/84670
791 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
793 (BB_VISITED_WITH_VISITED_SUCCS): New define.
794 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
795 (compute_antic_aux): Only assert the number of values in ANTIC_IN
796 doesn't grow if all successors (recursively) were visited at least
799 2018-03-05 Richard Biener <rguenther@suse.de>
801 PR tree-optimization/84650
802 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
803 if executed in the loop pipeline.
805 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
807 * doc/configfiles.texi (Configuration Files): Move info about
808 conditionalizing $target-protos.h to...
809 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
810 differs from $target-protos.h.
812 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
813 Chung-Ju Wu <jasonwucj@gmail.com>
815 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
816 * config/nds32/nds32-multiple.md (setmemsi): Define.
817 * config/nds32/nds32-memory-manipulation.c
818 (nds32_gen_dup_4_byte_to_word_value): New.
819 (emit_setmem_word_loop): New.
820 (emit_setmem_byte_loop): New.
821 (nds32_expand_setmem_loop): New.
822 (nds32_expand_setmem_loop_v3m): New.
823 (nds32_expand_setmem_unroll): New.
824 (nds32_expand_setmem): New.
826 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
827 Chung-Ju Wu <jasonwucj@gmail.com>
829 * config/nds32/nds32-memory-manipulation.c
830 (nds32_emit_load_store): New.
831 (nds32_emit_post_inc_load_store): New.
832 (nds32_emit_mem_move): New.
833 (nds32_emit_mem_move_block): New.
834 (nds32_expand_movmemsi_loop_unknown_size): New.
835 (nds32_expand_movmemsi_loop_known_size): New.
836 (nds32_expand_movmemsi_loop): New.
837 (nds32_expand_movmemsi_unroll): New.
838 (nds32_expand_movmemqi): Rename ...
839 (nds32_expand_movmemsi): ... to this.
840 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
841 (movmemsi): ... to this.
842 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
843 (nds32_expand_movmemsi): ... to this.
845 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
846 Monk Chiang <sh.chiang04@gmail.com>
847 Chung-Ju Wu <jasonwucj@gmail.com>
849 * config/nds32/nds32-protos.h
850 (nds32_expand_load_multiple): New arguments.
851 (nds32_expand_store_multiple): Ditto.
852 (nds32_valid_multiple_load_store): Rename ...
853 (nds32_valid_multiple_load_store_p): ... to this.
854 * config/nds32/nds32-memory-manipulation.c
855 (nds32_expand_load_multiple): Refine implementation.
856 (nds32_expand_store_multiple): Ditto.
857 * config/nds32/nds32-multiple.md
858 (load_multiple): Update nds32_expand_load_multiple interface.
859 (store_multiple): Update nds32_expand_store_multiple interface.
860 * config/nds32/nds32-predicates.c
861 (nds32_valid_multiple_load_store): Rename ...
862 (nds32_valid_multiple_load_store_p): ... to this and refine
864 * config/nds32/predicates.md
865 (nds32_load_multiple_and_update_address_operation): New predicate.
866 (nds32_store_multiple_and_update_address_operation): New predicate.
868 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
869 Chung-Ju Wu <jasonwucj@gmail.com>
871 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
872 (combo): New attribute.
873 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
875 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
877 * config/nds32/nds32.opt: Change -mcmodel= default value.
879 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
880 Monk Chiang <sh.chiang04@gmail.com>
881 Chung-Ju Wu <jasonwucj@gmail.com>
883 * config/nds32/constants.md (unspec_element): New enum.
884 * config/nds32/constraints.md (Umw): New constraint.
885 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
886 * config/nds32/nds32-intrinsic.md: Likewise.
887 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
888 (nds32_valid_smw_lwm_base_p): New.
889 (nds32_output_smw_single_word): New.
890 (nds32_output_lmw_single_word): New.
891 (nds32_expand_unaligned_load): New.
892 (nds32_expand_unaligned_store): New.
893 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
894 (nds32_output_smw_single_word): Declare.
895 (nds32_output_lmw_single_word): Declare.
896 (nds32_expand_unaligned_load): Declare.
897 (nds32_expand_unaligned_store): Declare.
898 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
899 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
900 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
901 NDS32_BUILTIN_UASTORE_DW.
902 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
905 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
906 Kito Cheng <kito.cheng@gmail.com>
907 Chung-Ju Wu <jasonwucj@gmail.com>
909 * config/nds32/nds32-intrinsic.c
910 (nds32_expand_builtin_null_ftype_reg): Delete.
911 (nds32_expand_builtin_reg_ftype_imm): Ditto.
912 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
913 (nds32_read_argument): New.
914 (nds32_legitimize_target): Ditto.
915 (nds32_legitimize_argument): Ditto.
916 (nds32_check_constant_argument): Ditto.
917 (nds32_expand_unop_builtin): Ditto.
918 (nds32_expand_unopimm_builtin): Ditto.
919 (nds32_expand_binop_builtin): Ditto.
920 (nds32_builtin_decl_impl): Ditto.
921 (builtin_description): Ditto.
922 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
923 (nds32_init_builtins_impl): Ditto.
924 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
925 (nds32_builtin_decl): New.
926 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
927 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
929 2018-03-02 Jeff Law <law@redhat.com>
931 * reorg.c (stop_search_p): Handle DEBUG_INSN.
932 (redundant_insn, fill_simple_delay_slots): Likewise.
933 (fill_slots_from_thread): Likewise.
934 * resource.c (mark_referenced_resources): Likewise.
935 (mark_set_resources, find_dead_or_set_registers): Likewise.
937 2018-03-02 Jakub Jelinek <jakub@redhat.com>
939 * substring-locations.h (format_warning_va): Formatting fix for
941 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
943 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
944 * substring-locations.c: Include intl.h.
945 (format_warning_va): Turned into small wrapper around
946 format_warning_n_va, renamed to ...
947 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
948 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
950 (format_warning_at_substring_n): New function.
951 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
952 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
953 format_warning_at_substring with just a shorter name instead of
954 const function pointer.
955 (fmtwarn_n): New function.
956 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
957 appropriate, get rid of all the fmtstr temporaries, move conditionals
958 with G_() wrapped string literals directly into fmtwarn arguments,
959 cast dir.len to (int), formatting fixes.
961 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
963 * doc/invoke.texi: Remove "Cilk Plus" references.
965 2018-03-02 Jakub Jelinek <jakub@redhat.com>
966 Richard Biener <rguenther@suse.de>
969 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
970 for error or warning attributes if CALL_FROM_THUNK_P is set.
973 2018-03-02 Jakub Jelinek <jakub@redhat.com>
976 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
977 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
980 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
981 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
983 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
984 instead of -1U in last predictors element's probability member.
986 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
989 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
990 arguments if they are comparable.
992 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
994 PR tree-optimization/84634
995 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
996 masks and masked_loop_p with a single loop_masks, making sure it's
997 null for bb vectorization.
999 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1001 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1002 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1003 loop->force_vectorize to check whether there is no alias.
1005 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1008 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1010 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1012 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1013 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1014 instead of a loop around prev_real_insn.
1015 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1019 * config/i386/i386.c (ix86_print_operand): Use conditional
1020 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1023 2018-03-02 Richard Biener <rguenther@suse.de>
1025 PR tree-optimization/84427
1026 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1027 (bitmap_set_subtract_values): Rewrite to handle multiple
1030 (prune_clobbered_mems): Likewise.
1031 (phi_translate): Take edge instead of pred/phiblock.
1032 (phi_translate_1): Likewise.
1033 (phi_translate_set): Likewise. Insert all translated
1034 exprs for a value into the set, keeping possibly multiple
1035 expressions per value.
1036 (compute_antic_aux): Adjust for phi_translate changes.
1037 When intersecting union the expressions and prune those
1038 not in the final value set, keeping possibly multiple
1039 expressions per value. Do not use value-insertion
1040 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1041 all expressions. Add verification that the value-sets
1042 only shrink during iteration.
1043 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1044 (do_pre_regular_insertion): Likewise.
1045 (do_pre_partial_partial_insertion): Likewise.
1047 2018-03-02 Richard Biener <rguenther@suse.de>
1050 * config/darwin.c (saved_debug_info_level): New static global.
1051 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1052 (darwin_asm_lto_end): Restore debug info generation settings.
1054 2018-03-01 Martin Liska <mliska@suse.cz>
1057 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1060 2018-03-01 Richard Biener <rguenther@suse.de>
1063 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1064 type annotation with LTO when debug was disabled at compile-time.
1066 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1068 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1070 (mips_final_postscan_insn): Likewise.
1072 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1074 PR rtl-optimization/84528
1075 * alias.c (init_alias_target): Add commentary.
1076 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1077 a unique base value if the frame pointer is not eliminated
1078 to the stack pointer.
1080 2018-03-01 Tom de Vries <tom@codesourcery.com>
1082 PR rtl-optimization/83327
1083 * lra-int.h (hard_regs_spilled_into): Declare.
1084 * lra.c (hard_regs_spilled_into): Define.
1085 (init_reg_info): Init hard_regs_spilled_into.
1086 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1087 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1088 (process_bb_lives): Handle hard_regs_spilled_into.
1089 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1090 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1092 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1094 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1095 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1096 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1097 * config/rs6000/aix72.h: New file.
1099 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1101 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1102 instead of warning_at with conditional singular and plural messages
1106 * stor-layout.c (update_alignment_for_field): For
1107 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1108 && !DECL_PACKED (field), do the alignment update, just use
1109 only desired_align instead of MAX (type_align, desired_align)
1111 (place_field): Don't do known_align < desired_align handling
1112 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1113 is non-NULL, instead do it after rli->prev_field handling and
1114 only if not within a bitfield word. For DECL_PACKED (field)
1115 use type_align of BITS_PER_UNIT.
1117 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1119 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1120 superfluous parentheses and trailing spaces.
1122 2018-02-28 Richard Biener <rguenther@suse.de>
1124 PR tree-optimization/84584
1125 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1126 SCoPs with fake exit edge.
1128 2018-02-28 Martin Liska <mliska@suse.cz>
1131 * timevar.c (timer::print): Fix format to properly print 100%
1134 2018-02-28 Richard Biener <rguenther@suse.de>
1137 * genmatch.c (capture_info::walk_match): Do not mark
1138 captured expressions without operands as expr_p given
1139 they act more like predicates and should be subject to
1140 "lost tail" side-effect preserving.
1142 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1144 PR rtl-optimization/81611
1145 * auto-inc-dec.c (attempt_change): Move dead note from
1146 mem_insn if it's the next use of regno
1147 (find_address): Take address use of reg holding
1148 non-incremented value. Add parm to limit search to the named
1150 (merge_in_block): Attempt to use a mem insn that is the next
1151 use of the original regno.
1153 2018-02-27 Martin Sebor <msebor@redhat.com>
1156 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1157 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1159 2018-02-27 Martin Sebor <msebor@redhat.com>
1161 PR translation/84207
1162 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1163 n argument to unsigned HOST_WIDE_INT.
1164 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1165 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1166 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1167 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1169 2018-02-27 Richard Biener <rguenther@suse.de>
1171 PR tree-optimization/84512
1172 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1173 Do not use the estimate returned from record_stmt_cost for
1174 the scalar iteration cost but sum properly using add_stmt_cost.
1176 2018-02-27 Richard Biener <rguenther@suse.de>
1178 PR tree-optimization/84466
1179 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1180 Adjust last change to less strictly validate use operands.
1182 2018-02-27 Martin Liska <mliska@suse.cz>
1184 PR gcov-profile/84548
1185 * gcov.c (process_file): Allow partial overlap and consider it
1186 also as group functions.
1187 (output_lines): Properly calculate range of lines for a group.
1189 2018-02-27 Martin Liska <mliska@suse.cz>
1191 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1192 'ggc' suffixes. Change first column width.
1193 (timer::print): Fix formatting of the column.
1195 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1197 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1198 preserve inline entry blocks for the sake of debug inline
1199 entry point markers alone.
1200 (remove_unused_locals): Suggest in comments a better place to
1201 force the preservation of inline entry blocks that are
1202 otherwise unused, but do not preserve them.
1204 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1206 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1208 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1211 * config/i386/constraints.md (Bs): Replace
1212 ix86_indirect_branch_register with
1213 TARGET_INDIRECT_BRANCH_REGISTER.
1215 * config/i386/i386.md (indirect_jump): Likewise.
1216 (tablejump): Likewise.
1217 (*sibcall_memory): Likewise.
1218 (*sibcall_value_memory): Likewise.
1219 Peepholes of indirect call and jump via memory: Likewise.
1220 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1221 (*sibcall_value_GOT_32): Likewise.
1222 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1223 (GOT_memory_operand): Likewise.
1224 (call_insn_operand): Likewise.
1225 (sibcall_insn_operand): Likewise.
1226 (GOT32_symbol_operand): Likewise.
1227 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1229 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1231 PR rtl-optimization/83496
1232 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1233 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1234 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1235 redundant insn, if any.
1236 (relax_delay_slots): Likewise.
1237 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1239 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1241 PR tree-optimization/83965
1242 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1243 that grouped statements are part of a reduction chain. Return
1244 true if the statement is not marked as a reduction itself but
1246 (vect_recog_dot_prod_pattern): Don't check whether the statement
1247 is part of a group here.
1248 (vect_recog_sad_pattern): Likewise.
1249 (vect_recog_widen_sum_pattern): Likewise.
1251 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1254 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1256 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1259 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1261 (ix86_output_indirect_function_return): New prototype.
1262 (ix86_split_simple_return_pop_internal): Likewise.
1263 * config/i386/i386.c (indirect_return_via_cx): New.
1264 (indirect_return_via_cx_bnd): Likewise.
1265 (indirect_thunk_name): Handle return va CX_REG.
1266 (output_indirect_thunk_function): Create alias for
1267 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1268 (ix86_output_indirect_jmp): Remove the bool argument.
1269 (ix86_output_indirect_function_return): New function.
1270 (ix86_split_simple_return_pop_internal): Likewise.
1271 * config/i386/i386.md (*indirect_jump): Don't pass false
1272 to ix86_output_indirect_jmp.
1273 (*tablejump_1): Likewise.
1274 (simple_return_pop_internal): Change it to define_insn_and_split.
1275 Call ix86_split_simple_return_pop_internal to split it for
1277 (simple_return_indirect_internal): Call
1278 ix86_output_indirect_function_return instead of
1279 ix86_output_indirect_jmp.
1281 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1284 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1285 memset and value initialization afterwards.
1287 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1289 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1291 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1294 * common/config/aarch64/aarch64-common.c
1295 (aarch_option_optimization_table[]): Switch
1296 off fomit-frame-pointer
1298 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1299 Chung-Ju Wu <jasonwucj@gmail.com>
1301 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1303 (store_multiple): Ditto.
1305 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1307 * config.gcc: Add --with-cpu support for nds32 target.
1308 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1309 * config/nds32/nds32.opt: Add -mcpu= option.
1311 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1313 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1314 isel=yes): Warn for these deprecated options.
1316 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1318 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1319 ISA_2_5_MASKS_EMBEDDED.
1321 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1323 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1324 p->max as pointers rather than using iterative_hash_expr.
1326 2018-02-23 Carl Love <cel@us.ibm.com>
1328 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1329 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1331 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1332 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1333 P8V_BUILTIN_VEC_VUNSIGNED2.
1335 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1338 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1339 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1341 * lra-constraints.c (curr_insn_transform): Set up
1342 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1344 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1345 * lra-eliminations.c (spill_pseudos): Ditto.
1346 (process_insn_for_elimination): Ditto.
1347 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1348 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1349 LRA_NON_CLOBBERED_ALT.
1351 2018-02-22 Martin Sebor <msebor@redhat.com>
1353 PR tree-optimization/84480
1354 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1355 to maybe_diag_stxncpy_trunc. Call it.
1356 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1357 from gimple_fold_builtin_strcpy. Print inlining stack.
1358 (handle_builtin_stxncpy): Print inlining stack.
1359 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1361 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1364 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1365 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1366 and -fcheck-pointer-bounds are used together.
1367 (indirect_thunk_prefix): New enum.
1368 (indirect_thunk_need_prefix): New function.
1369 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1370 "_nt" instead of "_bnd" for NOTRACK prefix.
1371 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1372 (output_indirect_thunk_function): Likewise.
1374 (ix86_code_end): Update output_indirect_thunk_function calls.
1375 (ix86_output_indirect_branch_via_reg): Replace
1376 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1377 (ix86_output_indirect_branch_via_push): Likewise.
1378 (ix86_output_function_return): Likewise.
1379 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1380 incompatible with -fcf-protection=branch and
1381 -fcheck-pointer-bounds.
1383 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1386 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1387 Change gcc_assert call to output_operand_lossage.
1389 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1391 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1393 2018-02-22 DJ Delorie <dj@redhat.com>
1394 Sebastian Perta <sebastian.perta@renesas.com>
1395 Oleg Endo <olegendo@gcc.gnu.org>
1397 * config/rx/rx.c (rx_rtx_costs): New function.
1398 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1400 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1402 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1404 2018-02-22 Martin Liska <mliska@suse.cz>
1407 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1408 Add "native" as a possible value.
1410 2018-02-22 Martin Liska <mliska@suse.cz>
1413 * config/i386/i386.c (ix86_option_override_internal):
1414 Add "native" as a possible value for -march and -mtune.
1416 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1419 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1420 to all type variants.
1422 PR tree-optimization/84503
1423 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1424 width as info->bitpos + info->bitsize - start.
1425 (merged_store_group::merge_overlapping): Simplify width computation.
1426 (check_no_overlap): New function.
1427 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1428 start + width and last_order of the group, fail if check_no_overlap
1430 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1431 to group if check_no_overlap fails.
1433 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1435 * config/rs6000/altivec.md: Delete contraint arguments to
1436 define_expand, define_split, and define_peephole2, and in
1437 define_insn_and_split if always unused.
1438 * config/rs6000/darwin.md: Ditto.
1439 * config/rs6000/dfp.md: Ditto.
1440 * config/rs6000/rs6000.md: Ditto.
1441 * config/rs6000/sync.md: Ditto.
1442 * config/rs6000/vector.md: Ditto.
1443 * config/rs6000/vsx.md: Ditto.
1445 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1447 * config/rs6000/altivec.md: Write output control strings as braced
1448 blocks instead of double-quoted strings.
1449 * config/rs6000/darwin.md: Ditto.
1450 * config/rs6000/rs6000.md: Ditto.
1451 * config/rs6000/vector.md: Ditto.
1452 * config/rs6000/vsx.md: Ditto.
1454 2018-02-21 Jason Merrill <jason@redhat.com>
1456 PR c++/84314 - ICE with templates and fastcall attribute.
1457 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1459 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1461 * ipa-cp.c (determine_versionability): Fix comment typos.
1463 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1466 * ipa-cp.c (determine_versionability): Do not version functions caling
1469 2018-02-21 Martin Liska <mliska@suse.cz>
1472 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1473 Add "native" as a possible value.
1474 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1475 the macro when native cpu detection is available.
1477 2018-02-21 Martin Liska <mliska@suse.cz>
1480 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1481 Add "native" as a possible value.
1482 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1483 when native cpu detection is available.
1485 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1486 Martin Sebor <msebor@redhat.com>
1488 PR tree-optimization/84478
1489 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1491 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1492 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1493 support which is conservatively correct, for 2 only stay conservative
1494 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1495 argument to the 2 argument get_range_strlen, adjust 6 arg
1496 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1498 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1499 (gimple_fold_builtin_strlen): Pass true as last argument to
1502 2018-02-20 Martin Sebor <msebor@redhat.com>
1505 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1506 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1507 (builtin_memref::builtin_memref): Factor out parts into
1508 set_base_and_offset and call it.
1510 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1513 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1514 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1515 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1516 search at the associated MODE_INT.
1518 2018-02-20 Jeff Law <law@redhat.com>
1521 PR tree-optimization/81592
1523 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1524 for range data rather than using global data.
1525 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1526 range data rather than using global data.
1527 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1528 pass it to children as needed.
1529 (struct directive::fmtresult): Similarly.
1530 (struct directive::set_width): Similarly.
1531 (struct directive::set_precision): Similarly.
1532 (format_integer, format_directive, parse_directive): Similarly.
1533 (format_none): Accept unnamed vr_values parameter.
1534 (format_percent, format_floating, format_character): Similarly.
1535 (format_string, format_plain): Similarly.
1536 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1537 the EVRP range analyzer for range data rather than using global data.
1538 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1539 gimple-ssa-evrp-analyze.h
1540 (class sprintf_dom_walker): Add after_dom_children member function.
1541 Add evrp_range_analyzer member.
1542 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1543 range analyzer as needed.
1544 (sprintf_dom_walker::after_dom_children): New member function.
1545 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1547 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1548 (evrp_range_analyzer::pop_to_marker): Likewise.
1550 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1552 PR tree-optimization/84419
1553 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1554 with the required type if its current type is compatible but
1557 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1560 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1561 after vectorization.
1563 2018-02-20 Martin Liska <mliska@suse.cz>
1566 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1567 possible values if we don't have a hint.
1569 2018-02-20 Martin Liska <mliska@suse.cz>
1573 * final.c (shorten_branches): Build align_tab array with one
1575 * opts.c (finish_options): Add alignment option limit check.
1576 (MAX_CODE_ALIGN): Likewise.
1577 (MAX_CODE_ALIGN_VALUE): Likewise.
1578 * doc/invoke.texi: Document maximum allowed option value for
1579 all -falign-* options.
1581 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1584 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1585 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1586 * var-tracking.c (emit_note_insn_var_location): Remove all references
1587 to NOTE_INSN_CALL_ARG_LOCATION.
1588 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1589 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1590 Use copy_rtx_if_shared.
1591 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1592 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1593 (dwarf2out_var_location): Remove handling of
1594 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1596 * final.c (final_scan_insn): Remove all references to
1597 NOTE_INSN_CALL_ARG_LOCATION.
1598 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1599 before dumping final insns.
1600 * except.c (emit_note_eh_region_end): Remove all references to
1601 NOTE_INSN_CALL_ARG_LOCATION.
1602 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1603 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1604 * config/arc/arc.c (hwloop_optimize): Likewise.
1605 * config/arm/arm.c (create_fix_barrier): Likewise.
1606 * config/s390/s390.c (s390_chunkify_start): Likewise.
1607 * config/sh/sh.c (find_barrier): Likewise.
1608 * config/i386/i386.c (rest_of_insert_endbranch,
1609 ix86_seh_fixup_eh_fallthru): Likewise.
1610 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1611 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1612 * config/frv/frv.c (frv_function_prologue): Likewise.
1613 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1615 (note_outside_basic_block_p): Remove all references to
1616 NOTE_INSN_CALL_ARG_LOCATION.
1617 * gengtype.c (adjust_field_rtx_def): Likewise.
1618 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1620 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1621 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1624 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1627 PR tree-optimization/84452
1628 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1629 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1632 2018-02-19 Martin Liska <mliska@suse.cz>
1635 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1637 2018-02-19 Martin Liska <mliska@suse.cz>
1638 Richard Sandiford <richard.sandiford@linaro.org>
1640 PR tree-optimization/82491
1641 * gimple-fold.c (get_base_constructor): Make earlier bail out
1644 2018-02-19 Carl Love <cel@us.ibm.com>
1646 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1647 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1649 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1650 P8V_BUILTIN_VEC_NEG.
1652 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1654 * config/rl78/rl78.md (movdf): New define expand.
1656 2018-02-19 Martin Liska <mliska@suse.cz>
1659 * doc/invoke.texi: Fix typo.
1660 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1662 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1664 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1665 handle rs6000_single_float and rs6000_double_float specially for
1668 2018-02-16 Jeff Law <law@redhat.com>
1670 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1673 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1676 * ipa-inline.c (inline_small_functions): Fix a typo.
1678 2018-02-16 Nathan Sidwell <nathan@acm.org>
1680 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1682 2018-02-16 Carl Love <cel@us.ibm.com>
1684 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1685 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1686 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1687 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1688 expansion to P8V_BUILTIN_VEC_FLOAT2.
1690 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1692 PR rtl-optimization/70023
1693 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1694 src_regno into account.
1696 2018-02-16 Carl Love <cel@us.ibm.com>
1698 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1699 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1700 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1701 * config/rs6000/rs6000.c: Remove case statements for
1702 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1703 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1704 and P9V_BUILTIN_VEC_VINSERT4B.
1705 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1706 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1707 * config/rs6000/vsx.md:
1708 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1711 2018-02-16 Carl Love <cel@us.ibm.com>
1713 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1715 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1717 * config/rs6000/rs6000-c.c: Add the definitions for
1718 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1719 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1720 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1721 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1722 definition for insert4b and define insn *insert3b_internal.
1723 * doc/extend.texi: Add documentation for vec_extract4b.
1725 2018-02-16 Nathan Sidwell <nathan@acm.org>
1727 * doc/extend.texi (Backwards Compatibility): Mention friend
1728 injection. Note for-scope is deprecated.
1729 * doc/invoke.texi (-ffriend-injection): Deprecate.
1731 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1733 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1734 that moved to I2, also allow destinations that are a paradoxical
1735 subreg (instead of a normal reg).
1737 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1740 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1743 2018-02-16 Richard Biener <rguenther@suse.de>
1745 PR tree-optimization/84037
1746 PR tree-optimization/84016
1748 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1749 Adjust vec_construct for the fact we need additional higher latency
1750 128bit inserts for AVX256 and AVX512 vector builds.
1751 (ix86_add_stmt_cost): Scale vector construction cost for
1754 2018-02-16 Richard Biener <rguenther@suse.de>
1756 PR tree-optimization/84417
1757 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1758 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1759 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1761 2018-02-16 Martin Liska <mliska@suse.cz>
1764 * internal-fn.def (ASAN_CHECK): Set proper flags.
1765 (ASAN_MARK): Likewise.
1767 2018-02-16 Julia Koval <julia.koval@intel.com>
1769 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1770 from PTA_CANNONLAKE.
1772 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1775 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1776 Use ++iter rather than iter++ for std::list iterators.
1777 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1778 defer deleting them until all nodes in the forest are processed. Do
1779 free even leaf nodes. Change to_process into auto_vec.
1782 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1783 * vec.h (vec_default_construct): Use memset instead of placement new
1784 if BROKEN_VALUE_INITIALIZATION is defined.
1785 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1786 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1789 PR rtl-optimization/83723
1790 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1791 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1792 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1794 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1796 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1798 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1800 PR rtl-optimization/81443
1801 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1802 from inner REGs to paradoxical SUBREGs.
1804 2018-02-16 Richard Biener <rguenther@suse.de>
1806 PR tree-optimization/84399
1807 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1808 For operands we can analyze at their definition make sure we can
1809 analyze them at each use as well.
1811 2018-02-16 Richard Biener <rguenther@suse.de>
1813 PR tree-optimization/84190
1814 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1815 volatile accesses if the decl isn't volatile.
1817 2018-02-15 Jason Merrill <jason@redhat.com>
1819 PR c++/84314 - ICE with templates and fastcall attribute.
1820 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1821 TYPE_CANONICAL on an existing type.
1823 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1825 PR tree-optimization/84383
1826 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1827 dstoff nor call operand_equal_p if dstbase is NULL.
1829 PR tree-optimization/84334
1830 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1831 also a CONSTANT_CLASS_P, punt.
1833 2018-02-14 Jim Wilson <jimw@sifive.com>
1835 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1836 first SMALL_OPERAND check. New local min_second_step. Move assert
1837 to where locals are set. Add TARGET_RVC support.
1838 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1840 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1842 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1844 2018-02-14 Martin Sebor <msebor@redhat.com>
1846 PR tree-optimization/83698
1847 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1848 arrays constrain the offset range to their bounds.
1849 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1850 (builtin_access::overlap): Avoid setting the size of overlap if it's
1852 (maybe_diag_overlap): Also consider arrays when deciding what values
1853 of offsets to include in diagnostics.
1855 2018-02-14 Martin Sebor <msebor@redhat.com>
1858 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1859 that correspond to the kind of a declaration.
1861 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1864 * config/pa/pa.md: Load address of PIC label using the linkage table
1865 if the label is nonlocal.
1867 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1869 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1870 warning message if user requests -maltivec=be.
1871 * doc/invoke.texi: Document deprecation of -maltivec=be.
1873 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1876 * config/rs6000/rs6000-c.c: Update definitions for
1877 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1878 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1880 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1883 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1884 add _get_ssp intrinsics. Remove argument from
1885 __builtin_ia32_rdssp[d|q].
1886 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1887 * config/i386/i386-builtin.def: Remove argument from
1888 __builtin_ia32_rdssp[d|q].
1889 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1890 ix86_expand_special_args_builtin for _rdssp[d|q].
1891 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1892 Clear register before usage.
1893 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1894 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1896 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1898 PR tree-optimization/84357
1899 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1900 operand 1 of an ARRAY_REF too.
1902 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1905 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1906 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1908 (set_of_reg): New struct.
1909 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1910 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1911 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1913 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1914 Split into bitclr, bitset, bitinvert patterns if appropriate.
1915 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1916 use rx_fuse_in_memory_bitop.
1917 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1918 to named insn, correct maximum insn length.
1920 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1923 * machmode.def: Define a complex mode for PARTIAL_INT.
1924 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1926 * doc/rtl.texi: Document CSPImode.
1927 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1929 (msp430_hard_regno_nregs_with_padding): Likewise.
1931 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1934 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1936 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1938 PR rtl-optimization/84169
1939 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1940 we generated a parallel as new i3 and we split that to new i2 and i3
1941 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1942 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1943 those to i2, not i1. Partially rewrite this scan code.
1945 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1948 * stor-layout.c (place_field): For variable length fields, adjust
1949 offset_align afterwards not just based on the field's alignment,
1950 but also on the size.
1953 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1954 of exps and logs in the use_exp2 case.
1956 2018-02-13 Jeff Law <law@redhat.com>
1958 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1961 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1964 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1968 * common.opt (gas-loc-support, gas-locview-support): New.
1969 (ginline-points, ginternal-reset-location-views): New.
1970 * doc/invoke.texi: Document them. Use @itemx where intended.
1971 (gvariable-location-views): Adjust.
1972 * target.def (reset_location_view): New.
1973 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1974 (TARGET_RESET_LOCATION_VIEW): New.
1975 * doc/tm.texi: Rebuilt.
1976 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1977 (dwarf2out_default_as_locview_support): New.
1978 (output_asm_line_debug_info): Use option variables.
1979 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1980 (output_loc_list): Likewise.
1981 (add_high_low_attributes): Check option variables.
1982 Don't output entry view attribute in strict mode.
1983 (gen_inlined_subroutine_die): Check option variables.
1984 (dwarf2out_inline_entry): Likewise.
1985 (init_sections_and_labels): Likewise.
1986 (dwarf2out_early_finish): Likewise.
1987 (maybe_reset_location_view): New, from...
1988 (dwarf2out_var_location): ... here. Call it.
1989 * debug.h (dwarf2out_default_as_loc_support): Declare.
1990 (dwarf2out_default_as_locview_support): Declare.
1991 * hooks.c (hook_int_rtx_insn_0): New.
1992 * hooks.h (hook_int_rtx_insn_0): Declare.
1993 * toplev.c (process_options): Take -gas-loc-support and
1994 -gas-locview-support from dwarf2out. Enable
1995 -gvariable-location-views by default only with locview
1996 assembler support. Enable -ginternal-reset-location-views by
1997 default only if the target defines the corresponding hook.
1998 Enable -ginline-points by default if location views are
1999 enabled; force it disabled if statement frontiers are
2001 * tree-inline.c (expand_call_inline): Check option variables.
2002 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2004 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2006 PR tree-optimization/84321
2007 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2008 handling. Also check whether the anti-range contains any values
2009 that satisfy the mask; switch to a VR_RANGE if not.
2011 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2014 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2016 2018-02-13 Martin Jambor <mjambor@suse.cz>
2019 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2020 of call statements, also set location of a load to a temporary.
2022 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2024 * config/rl78/rl78.c (add_vector_labels): New function.
2025 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2026 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2027 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2028 which checks that no arguments are passed.
2029 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2030 * doc/extend.texi: Documentation for the new attribute.
2032 2018-02-13 Andreas Schwab <schwab@suse.de>
2034 * config/riscv/linux.h (CPP_SPEC): Define.
2036 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2039 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2040 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2041 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2042 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2043 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2044 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2045 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2046 temporarily for AES and PCLMUL builtins.
2048 PR tree-optimization/84339
2049 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2050 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2054 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2055 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2056 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2058 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2060 * omp-simd-clone.h: New file.
2061 * omp-simd-clone.c: Include omp-simd-clone.h.
2062 (expand_simd_clones): No longer static.
2063 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2064 cgraph.h and omp-simd-clone.h.
2065 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2066 (vect_recog_widen_shift_pattern): Formatting fix.
2067 (vect_pattern_recog_1): Don't check optab for calls.
2070 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2071 operands[2] into a REG before using gen_lowpart on it.
2073 2018-02-12 Jeff Law <law@redhat.com>
2076 * config/sh/sh.c (find_barrier): Consider a sibling call
2079 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2080 successfully back substituting a reg.
2082 2018-02-12 Richard Biener <rguenther@suse.de>
2084 PR tree-optimization/84037
2085 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2086 parameter, move visited init to caller.
2087 (vect_slp_analyze_operations): Separate cost from validity
2088 check, initialize visited once for all instances.
2089 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2091 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2093 (vect_model_promotion_demotion_cost): Likewise.
2094 (vectorizable_bswap): Guard cost modeling with !slp_node
2095 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2097 (vectorizable_call): Likewise.
2098 (vectorizable_conversion): Likewise.
2099 (vectorizable_assignment): Likewise.
2100 (vectorizable_shift): Likewise.
2101 (vectorizable_operation): Likewise.
2102 (vectorizable_store): Likewise.
2103 (vectorizable_load): Likewise.
2104 (vectorizable_condition): Likewise.
2105 (vectorizable_comparison): Likewise.
2107 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2110 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2111 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2112 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2114 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2117 * params.def (inline-min-speedup): Increase from 8 to 15.
2118 (max-inline-insns-auto): Decrease from 40 to 30.
2119 * ipa-split.c (consider_split): Add some buffer for function to
2120 be considered inlining candidate.
2121 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2124 2018-02-12 Richard Biener <rguenther@suse.de>
2126 PR tree-optimization/84037
2127 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2128 matched stmts if we cannot swap the non-matched ones.
2130 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2132 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2133 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2134 _mm_maskz_scalef_round_ss): New intrinsics.
2135 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2136 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2137 __builtin_ia32_scalefss_round): Remove.
2138 (__builtin_ia32_scalefsd_mask_round,
2139 __builtin_ia32_scalefss_mask_round): New intrinsics.
2140 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2141 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2142 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2143 "<round_constraint>")): Changed to ...
2144 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2145 "<round_scalar_constraint>")): ... this.
2146 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2147 %0, %1, %2<round_op3>}"): Changed to ...
2148 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2149 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2150 %2<round_scalar_mask_op3>}"): ... this.
2151 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2153 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2155 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2156 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2157 (_mm_maskz_sqrt_round_ss): New intrinsics.
2158 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2159 (__builtin_ia32_sqrtsd_mask_round)
2160 (__builtin_ia32_sqrtss_mask_round): New builtins.
2161 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2162 (__builtin_ia32_sqrtss_round): Remove.
2163 (__builtin_ia32_sqrtsd_mask_round)
2164 (__builtin_ia32_sqrtss_mask_round): New builtins.
2165 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2166 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2167 ((match_operand:VF_128 1 "vector_operand"
2168 "xBm,<round_constraint>")): Changed to ...
2169 ((match_operand:VF_128 1 "vector_operand"
2170 "xBm,<round_scalar_constraint>")): ... this.
2171 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2172 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2173 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2174 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2175 %<iptr>1<round_scalar_mask_op3>}): ... this.
2176 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2177 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2179 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2182 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2183 Cast vec_cmpeq result to correct type.
2184 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2185 Cast vec_cmpgt result to correct type.
2187 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2189 * final.c (final_scan_insn_1): Renamed from...
2190 (final_scan_insn): ... this. New wrapper, to recover
2191 seen from the outermost call in recursive ones.
2192 * config/sparc/sparc.c (output_return): Drop seen from call.
2193 (output_sibcall): Likewise.
2194 * config/visium/visium.c (output_branch): Likewise.
2196 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2198 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2201 2018-02-10 Alan Modra <amodra@gmail.com>
2204 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2205 Specify LR as an input.
2207 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2210 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2211 remove_member_access_dummy_vars): New functions.
2212 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2213 lower_omp_1, execute_lower_omp): Use them.
2215 PR rtl-optimization/84308
2216 * shrink-wrap.c (spread_components): Release todo vector.
2218 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2220 PR rtl-optimization/57193
2221 * ira-color.c (struct allocno_color_data): Add member
2222 conflict_allocno_hard_prefs.
2223 (update_conflict_allocno_hard_prefs): New.
2224 (bucket_allocno_compare_func): Add a preference based on
2225 conflict_allocno_hard_prefs.
2226 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2227 (color_allocnos): Remove a dead code. Initiate
2228 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2230 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2233 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2234 constraint from =wa to wa. Avoid a subreg on the output operand,
2235 instead use a pseudo and subreg it in a move.
2236 (p9_xxbrd_<mode>): Changed to ...
2237 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2238 (p9_xxbrd_v2df): New expander.
2239 (p9_xxbrw_<mode>): Changed to ...
2240 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2241 (p9_xxbrw_v4sf): New expander.
2243 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2245 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2247 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2250 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2251 multiply in 32-bit mode.
2252 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2253 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2256 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2258 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2259 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2260 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2261 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2263 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2266 * dwarf2out.c (is_trivial_indirect_ref): New function.
2267 (dwarf2out_late_global_decl): Do not generate a location
2268 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2269 and that are not defined in the current unit.
2271 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2273 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2274 instead of a libcall for UNORDERED.
2276 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2279 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2280 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2282 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2285 * config/s390/s390.c (s390_set_current_function): Invoke
2286 s390_indirect_branch_settings also if fndecl didn't change.
2288 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2290 * config/rs6000/rs6000.md (blockage): Set length to zero.
2292 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2294 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2296 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2299 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2300 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2304 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2305 PARALLEL incoming that failed vt_get_decl_and_offset check.
2308 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2309 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2311 (get_variable_section): For decls in named .bss* sections pass true as
2312 second argument to bss_initializer_p.
2314 2018-02-09 Marek Polacek <polacek@redhat.com>
2315 Jakub Jelinek <jakub@redhat.com>
2318 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2319 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2320 Sync some changes from cxx_fold_indirect_ref.
2322 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2324 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2326 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2327 (BLOCK_INLINE_ENTRY_LABEL): New.
2328 (dwarf2out_var_location): Disregard inline entry markers.
2329 (inline_entry_data): New struct.
2330 (inline_entry_data_hasher): New hashtable type.
2331 (inline_entry_data_hasher::hash): New.
2332 (inline_entry_data_hasher::equal): New.
2333 (inline_entry_data_table): New variable.
2334 (add_high_low_attributes): Add DW_AT_entry_pc and
2335 DW_AT_GNU_entry_view attributes if a pending entry is found
2336 in inline_entry_data_table. Add old entry_pc attribute only
2337 if debug nonbinding markers are disabled.
2338 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2339 markers are enabled.
2340 (block_within_block_p, dwarf2out_inline_entry): New.
2341 (dwarf2out_finish): Check that no entries remained in
2342 inline_entry_data_table.
2343 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2344 (final_scan_insn, notice_source_line): Likewise.
2345 (rest_of_clean_state): Skip inline entry markers.
2346 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2348 * gimple.c (gimple_build_debug_inline_entry): New.
2349 * gimple.h (enum gimple_debug_subcode): Add
2350 GIMPLE_DEBUG_INLINE_ENTRY.
2351 (gimple_build_debug_inline_entry): Declare.
2352 (gimple_debug_inline_entry_p): New.
2353 (gimple_debug_nonbind_marker_p): Adjust.
2354 * insn-notes.def (INLINE_ENTRY): New.
2355 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2356 inline entry marker notes.
2357 (print_insn): Likewise.
2358 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2359 (INSN_DEBUG_MARKER_KIND): Likewise.
2360 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2361 * tree-inline.c (expand_call_inline): Build and insert
2362 debug_inline_entry stmt.
2363 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2364 inline entry blocks early, if nonbind markers are enabled.
2365 (dump_scope_block): Dump fragment info.
2366 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2367 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2368 (gimple_build_debug_inline_entry): New.
2369 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2370 Enable/disable inline entry points too.
2371 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2372 (DEBUG_INSN): Describe inline entry markers.
2374 * common.opt (gvariable-location-views): New.
2375 (gvariable-location-views=incompat5): New.
2376 * config.in: Rebuilt.
2377 * configure: Rebuilt.
2378 * configure.ac: Test assembler for view support.
2379 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2380 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2381 * dwarf2out.c (var_loc_view): New typedef.
2382 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2383 (dwarf2out_locviews_in_attribute): New.
2384 (dwarf2out_locviews_in_loclist): New.
2385 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2386 (enum dw_line_info_opcode): Add LI_adv_address.
2387 (struct dw_line_info_table): Add view.
2388 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2389 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2390 (zero_view_p): New variable.
2391 (ZERO_VIEW_P): New macro.
2392 (output_asm_line_debug_info): New.
2393 (struct var_loc_node): Add view.
2394 (add_AT_view_list, AT_loc_list): New.
2395 (add_var_loc_to_decl): Add view param. Test it against last.
2396 (new_loc_list): Add view params. Record them.
2397 (AT_loc_list_ptr): Handle loc and view lists.
2398 (view_list_to_loc_list_val_node): New.
2399 (print_dw_val): Handle dw_val_class_view_list.
2400 (size_of_die): Likewise.
2401 (value_format): Likewise.
2402 (loc_list_has_views): New.
2403 (gen_llsym): Set vl_symbol too.
2404 (maybe_gen_llsym, skip_loc_list_entry): New.
2405 (dwarf2out_maybe_output_loclist_view_pair): New.
2406 (output_loc_list): Output view list or entries too.
2407 (output_view_list_offset): New.
2408 (output_die): Handle dw_val_class_view_list.
2409 (output_dwarf_version): New.
2410 (output_compilation_unit_header): Use it.
2411 (output_skeleton_debug_sections): Likewise.
2412 (output_rnglists, output_line_info): Likewise.
2413 (output_pubnames, output_aranges): Update version comments.
2414 (output_one_line_info_table): Output view numbers in asm comments.
2415 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2416 Call maybe_gen_llsym.
2417 (loc_list_from_tree_1): Adjust.
2418 (add_AT_location_description): Create view list attribute if
2419 needed, check it's absent otherwise.
2420 (convert_cfa_to_fb_loc_list): Adjust.
2421 (maybe_emit_file): Call output_asm_line_debug_info for test.
2422 (dwarf2out_var_location): Reset views as needed. Precompute
2423 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2424 attribute. Set view.
2425 (new_line_info_table): Reset next view.
2426 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2427 (dwarf2out_source_line): Likewise. Output view resets and labels to
2428 the assembler, or select appropriate line info opcodes.
2429 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2430 (optimize_string_length): Catch it. Adjust.
2431 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2432 dw_val_class_view_list, and remove it if no longer needed.
2433 (hash_loc_list): Hash view numbers.
2434 (loc_list_hasher::equal): Compare them.
2435 (optimize_location_lists): Check whether a view list symbol is
2436 needed, and whether the locview attribute is present, and
2437 whether they match. Remove the locview attribute if no longer
2439 (index_location_lists): Call skip_loc_list_entry for test.
2440 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2441 Use output_dwarf_version.
2442 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2443 (struct dw_val_node): Add val_view_list.
2444 * final.c (SEEN_NEXT_VIEW): New.
2445 (set_next_view_needed): New.
2446 (clear_next_view_needed): New.
2447 (maybe_output_next_view): New.
2448 (final_start_function): Rename to...
2449 (final_start_function_1): ... this. Take pointer to FIRST,
2450 add SEEN parameter. Emit param bindings in the initial view.
2451 (final_start_function): Reintroduce SEEN-less interface.
2452 (final): Rename to...
2453 (final_1): ... this. Take SEEN parameter. Output final pending
2454 next view at the end.
2455 (final): Reintroduce seen-less interface.
2456 (final_scan_insn): Output pending next view before switching
2457 sections or ending a block. Mark the next view as needed when
2458 outputting variable locations. Notify debug backend of section
2459 changes, and of location view changes.
2460 (rest_of_handle_final): Adjust.
2461 * toplev.c (process_options): Autodetect value for debug variable
2462 location views option. Warn on incompat5 without -gdwarf-5.
2463 * doc/invoke.texi (gvariable-location-views): New.
2464 (gvariable-location-views=incompat5): New.
2465 (gno-variable-location-views): New.
2467 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2469 PR tree-optimization/84136
2470 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2471 that the result of find_edge is non-NULL.
2473 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2476 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2477 storing integer register in SImode. Fix cost of 256 and 512
2478 byte aligned SSE register store.
2480 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2482 * config/i386/i386.c (ix86_multiplication_cost): Fix
2483 multiplication cost for TARGET_AVX512DQ.
2485 2018-02-08 Marek Polacek <polacek@redhat.com>
2487 PR tree-optimization/84238
2488 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2491 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2493 PR tree-optimization/84265
2494 * tree-vect-stmts.c (vectorizable_store): Don't treat
2495 VMAT_CONTIGUOUS accesses as grouped.
2496 (vectorizable_load): Likewise.
2498 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2500 PR tree-optimization/81635
2501 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2502 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2503 (test_round_for_mask): New functions.
2504 (wide_int_cc_tests): Call test_round_for_mask.
2505 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2506 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2507 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2508 range returned by get_range_info.
2510 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2513 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2514 * symtab.c: Include builtins.h
2515 (symtab_node::output_to_lto_symbol_table_p): Move here
2516 from lto-streamer-out.c:output_symbol_p.
2517 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2518 (output_symbol_p): Move all logic to symtab.c
2519 (produce_symtab): Update.
2521 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2523 * config/s390/s390-opts.h (enum indirect_branch): Define.
2524 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2525 (s390_indirect_branch_via_thunk)
2526 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2527 (enum s390_indirect_branch_type): Define.
2528 * config/s390/s390.c (struct s390_frame_layout, struct
2529 machine_function): Remove.
2530 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2531 (indirect_branch_table_label_no, indirect_branch_table_name):
2533 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2534 (enum s390_indirect_branch_option): Define.
2535 (s390_return_addr_from_memory): New function.
2536 (s390_handle_string_attribute): New function.
2537 (s390_attribute_table): Add new attribute handler.
2538 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2539 (s390_indirect_branch_via_thunk): New function.
2540 (s390_indirect_branch_via_inline_thunk): New function.
2541 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2542 sibling call optimization for non z10 compiles.
2543 (s390_emit_call): Force indirect branch target to be a single
2544 register. Add r1 clobber for non-z10 compiles.
2545 (s390_emit_epilogue): Emit return jump via return_use expander.
2546 (s390_reorg): Handle JUMP_INSNs as execute targets.
2547 (s390_option_override_internal): Perform validity checks for the
2548 new command line options.
2549 (s390_indirect_branch_attrvalue): New function.
2550 (s390_indirect_branch_settings): New function.
2551 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2552 (s390_output_indirect_thunk_function): New function.
2553 (s390_code_end): Implement target hook.
2554 (s390_case_values_threshold): Implement target hook.
2555 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2557 * config/s390/s390.h (struct s390_frame_layout)
2558 (struct machine_function): Move here from s390.c.
2559 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2560 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2561 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2562 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2563 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2564 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2565 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2566 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2567 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2568 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2569 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2570 (mnemonic attribute): Add values which aren't recognized
2572 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2573 pattern for branch conversion. Fix mnemonic attribute.
2574 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2575 indirect branch via thunk if requested.
2576 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2577 ("*indirect_jump"): Disable for branch conversion using out of
2579 ("indirect_jump_via_thunk<mode>_z10")
2580 ("indirect_jump_via_thunk<mode>")
2581 ("indirect_jump_via_inlinethunk<mode>_z10")
2582 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2583 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2584 ("casesi_jump_via_inlinethunk<mode>_z10")
2585 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2586 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2587 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2588 ("*indirect2_jump"): Disable for branch conversion.
2589 ("casesi_jump"): Turn into expander and expand patterns for branch
2591 ("return_use"): New expander.
2592 ("*return"): Emit return via thunk and rename it to ...
2593 ("*return<mode>"): ... this one.
2594 * config/s390/s390.opt: Add new options and and enum for the
2597 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2599 * lra-constraints.c (match_reload): Unconditionally use
2600 gen_lowpart_SUBREG, rather than selecting between that
2601 and equivalent gen_rtx_SUBREG code.
2603 2018-02-08 Richard Biener <rguenther@suse.de>
2605 PR tree-optimization/84233
2606 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2607 changed flag instead of boguously re-using phi_inserted.
2609 2018-02-08 Martin Jambor <mjambor@suse.cz>
2611 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2612 static local variables.
2614 2018-02-08 Richard Biener <rguenther@suse.de>
2616 PR tree-optimization/84278
2617 * tree-vect-stmts.c (vectorizable_store): When looking for
2618 smaller vector types to perform grouped strided loads/stores
2619 make sure the mode is supported by the target.
2620 (vectorizable_load): Likewise.
2622 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2624 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2625 Increase LDP/STP opportunities by adding adjacent callee-saves.
2627 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2629 PR rtl-optimization/84068
2630 PR rtl-optimization/83459
2631 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2633 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2635 PR tree-optimization/84224
2636 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2637 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2640 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2643 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2644 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2645 count, remove one USE.
2647 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2649 * doc/install.texi (Configuration): Document the
2650 --with-long-double-format={ibm,ieee} PowerPC configuration
2654 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2655 Convert from define_expand to be define_insn_and_split. Rework
2656 float/double/_Float128 conversions to QI/HI/SImode to work with
2657 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2658 conversions to QI/HImode types did a store and then a load to
2659 truncate the value. For conversions to VSX registers, don't split
2660 the insn, instead emit the code directly. Use the code iterator
2661 any_fix to combine signed and unsigned conversions.
2662 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2663 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2664 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2665 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2666 (fix_<mode>di2_hw): Likewise.
2667 (fixuns_<mode>di2_hw): Likewise.
2668 (fix_<mode>si2_hw): Likewise.
2669 (fixuns_<mode>si2_hw): Likewise.
2670 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2671 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2672 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2673 fix<uns>_trunc<SFDF:mode>si2_p8.
2674 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2676 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2677 (fix<uns>_<mode>_mem): Likewise.
2678 (fctiw<u>z_<mode>_mem): Likewise.
2679 (fix<uns>_<mode>_mem): Likewise.
2680 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2681 the register allocator from doing a direct move to the GPRs to do
2682 a store, and instead use the ISA 3.0 store byte/half-word from
2683 vector register instruction. For IEEE 128-bit floating point,
2684 also optimize stores of 32-bit ints.
2685 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2687 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2689 * genextract.c (push_pathstr_operand): New function to support
2691 (walk_rtx): Call push_pathstr_operand.
2692 (print_path): Support [a-zA-Z].
2694 2018-02-07 Richard Biener <rguenther@suse.de>
2696 PR tree-optimization/84037
2697 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2698 (cse_and_gimplify_to_preheader): Declare.
2699 (vect_get_place_in_interleaving_chain): Likewise.
2700 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2702 (_loop_vec_info::~_loop_vec_info): Delete it.
2703 (cse_and_gimplify_to_preheader): New function.
2704 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2705 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2706 (vectorizable_load): Likewise. For grouped stores always base
2707 the IV on the first element.
2708 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2709 condition before gimplifying.
2711 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2713 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2714 *DIV_EXPR and *MOD_EXPR.
2716 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2719 * config/i386/i386.c (ix86_option_override_internal): Mask out
2720 the CF_SET bit when checking -fcf-protection.
2722 2018-02-07 Tom de Vries <tom@codesourcery.com>
2725 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2728 2018-02-07 Richard Biener <rguenther@suse.de>
2730 PR tree-optimization/84204
2731 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2734 PR tree-optimization/84205
2735 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2736 special-case isl_ast_op_zdiv_r.
2738 PR tree-optimization/84223
2739 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2740 Only add conditions from within the region.
2741 (gather_bbs::after_dom_children): Adjust.
2743 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2746 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2747 * config/avr/avr.md: Only post-reload split REG-REG moves if
2748 either register is GENERAL_REG_P.
2750 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2752 PR tree-optimization/84235
2753 * tree-ssa-scopedtables.c
2754 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2755 if the subtraction is performed in floating point type where NaNs are
2756 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2757 build 1. Formatting fix.
2759 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2762 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2763 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2764 and skip it regardless of bb boundaries. Use CALL_P macro,
2765 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2766 unnecessarily, formatting fix.
2768 2018-02-06 Michael Collison <michael.collison@arm.com>
2770 * config/arm/thumb2.md:
2771 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2772 (*thumb_mov_notscc): Ditto.
2774 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2777 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2778 unsigned_fix, not "s".
2780 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2782 * configure.ac (gcc_fn_eh_frame_ro): New function.
2783 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2784 correct .eh_frame permissions.
2785 * configure: Regenerate.
2787 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2789 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2792 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2794 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2795 Display warning message for -mno-speculate-indirect-jumps.
2797 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2799 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2801 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2803 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2805 PR tree-optimization/84225
2806 * tree-eh.c (find_trapping_overflow): Only call
2807 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2809 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2812 * config/i386/i386.c: Reimplement the check of possible options
2813 -mibt/-mshstk conbination. Change error messages.
2814 * doc/invoke.texi: Fix a typo: remove extra '='.
2816 2018-02-06 Marek Polacek <polacek@redhat.com>
2818 PR tree-optimization/84228
2819 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2821 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2824 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2825 emitted arch directives.
2826 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2827 __ARM_FEATURE_COPROC before changing architectures.
2829 2018-02-06 Richard Biener <rguenther@suse.de>
2831 * config/i386/i386.c (print_reg): Fix typo.
2832 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2834 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2836 * configure: Regenerate.
2838 2018-02-05 Martin Sebor <msebor@redhat.com>
2840 PR tree-optimization/83369
2841 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2844 2018-02-05 Martin Liska <mliska@suse.cz>
2846 * doc/invoke.texi: Cherry-pick upstream r323995.
2848 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2850 * ira.c (ira_init_register_move_cost): Adjust comment.
2852 2018-02-05 Martin Liska <mliska@suse.cz>
2854 PR gcov-profile/84137
2855 * doc/gcov.texi: Fix typo in documentation.
2857 2018-02-05 Martin Liska <mliska@suse.cz>
2859 PR gcov-profile/83879
2860 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2861 using dlopen functionality.
2863 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2865 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2866 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2867 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2868 _mm_maskz_range_round_ss): New intrinsics.
2869 (__builtin_ia32_rangesd128_round)
2870 (__builtin_ia32_rangess128_round): Remove.
2871 (__builtin_ia32_rangesd128_mask_round,
2872 __builtin_ia32_rangess128_mask_round): New builtins.
2873 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2874 __builtin_ia32_rangess128_round): Remove.
2875 (__builtin_ia32_rangesd128_mask_round,
2876 __builtin_ia32_rangess128_mask_round): New builtins.
2877 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2878 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2879 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2880 "<round_saeonly_constraint>")): Changed to ...
2881 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2882 "<round_saeonly_scalar_constraint>")): ... this.
2883 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2884 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2885 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2886 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2887 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2889 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2891 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2893 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2894 Remove all values except native, 8540 and 8548.
2896 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2898 * config/i386/i386.c (ix86_output_function_return): Pass
2899 INVALID_REGNUM, instead of -1, as invalid register number to
2900 indirect_thunk_name and output_indirect_thunk.
2902 2018-02-02 Julia Koval <julia.koval@intel.com>
2904 * config.gcc: Add -march=icelake.
2905 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2906 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2907 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2908 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2909 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2910 (processor_target_table): Add icelake.
2911 (ix86_option_override_internal): Handle new PTAs.
2912 (get_builtin_code_for_version): Handle icelake.
2913 (M_INTEL_COREI7_ICELAKE): New.
2914 (fold_builtin_cpu): Handle icelake.
2915 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2916 * doc/invoke.texi: Add -march=icelake.
2918 2018-02-02 Julia Koval <julia.koval@intel.com>
2920 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2921 to wide_int_bitmask.
2922 * wide-int-bitmask.h: New.
2924 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2927 * config/i386/i386.md: Replace Pmode with word_mode in
2928 builtin_setjmp_setup and builtin_longjmp to support x32.
2930 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2934 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2936 (rs6000_supported_cpu_names): New static variable.
2937 (linux_cpu_translation_table): Likewise.
2938 (elf_platform) <cpu>: Define new static variable and use it.
2939 Translate kernel AT_PLATFORM name to canonical name if needed.
2940 Error if platform name is unknown.
2942 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2945 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2947 2018-02-01 Jeff Law <law@redhat.com>
2950 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2951 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2952 the scratch if RELEASE_VIA_POP is false.
2953 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2954 If we have to save a temporary register, decrement SIZE appropriately.
2955 Pass new arguments to release_scratch_register_on_entry.
2956 (ix86_adjust_stack_and_probe): Likewise.
2957 (ix86_emit_probe_stack_range): Pass new arguments to
2958 release_scratch_register_on_entry.
2960 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2962 PR rtl-optimization/84157
2963 * combine.c (change_zero_ext): Use REG_P predicate in
2964 front of HARD_REGISTER_P predicate.
2966 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2968 * config/avr/avr.c (avr_option_override): Move disabling of
2969 -fdelete-null-pointer-checks to...
2970 * common/config/avr/avr-common.c (avr_option_optimization_table):
2973 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2975 PR tree-optimization/81635
2976 * tree-data-ref.c (split_constant_offset_1): For types that
2977 wrap on overflow, try to use range info to prove that wrapping
2980 2018-02-01 Renlin Li <renlin.li@arm.com>
2983 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2985 (aarch64_register_move_cost): Likewise.
2986 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2988 (REG_CLASS_NAMES): Likewise.
2989 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2990 TAILCALL_ADDR_REGS. Remove IP registers.
2991 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2993 2018-02-01 Richard Biener <rguenther@suse.de>
2995 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2996 for specifying RPO order and allow NULL for that.
2997 * domwalk.c (dom_walker::dom_walker): Likewise.
2998 (dom_walker::walk): Handle NULL RPO order.
2999 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3001 (rewrite_update_dom_walker): Likewise.
3002 (mark_def_dom_walker): Likewise.
3004 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3006 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3007 (aarch64_maybe_expand_sve_subreg_move): Declare.
3008 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3009 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3011 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3012 that are semantically a reverse operation.
3013 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3014 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3015 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3017 (aarch64_can_change_mode_class): For big-endian, forbid changes
3018 between two SVE modes if they have different element sizes.
3020 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3022 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3023 the TImode handling for big-endian targets.
3025 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3027 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3028 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3030 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3031 Remove BSWAP handing for big-endian targets and use the form of
3032 LD1RQ appropariate for the mode.
3034 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3036 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3037 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3040 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3043 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3044 check for operands that need to go through aarch64_sve_reload_be.
3046 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3048 PR tree-optimization/81661
3049 PR tree-optimization/84117
3050 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3051 * tree-eh.c: Include gimplify.h.
3052 (find_trapping_overflow, replace_trapping_overflow,
3053 rewrite_to_non_trapping_overflow): New functions.
3054 * tree-vect-loop.c: Include tree-eh.h.
3055 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3056 * tree-data-ref.c: Include tree-eh.h.
3057 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3059 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3061 PR rtl-optimization/84123
3062 * combine.c (change_zero_ext): Check if hard register satisfies
3063 can_change_dest_mode before calling gen_lowpart_SUBREG.
3065 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3068 * ira.c (ira_init_register_move_cost): Remove assert.
3070 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3072 PR rtl-optimization/84071
3073 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3074 * doc/tm.texi: Regenerate.
3076 2018-01-31 Richard Biener <rguenther@suse.de>
3078 PR tree-optimization/84132
3079 * tree-data-ref.c (analyze_miv_subscript): Properly
3080 check whether evolution_function_is_affine_multivariate_p
3081 before calling gcd_of_steps_may_divide_p.
3083 2018-01-31 Julia Koval <julia.koval@intel.com>
3086 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3087 * config/i386/i386.md (rdpid_rex64) New.
3088 (rdpid): Make 32bit only.
3090 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3093 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3094 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3096 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3099 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3101 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3103 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3105 PR rtl-optimization/84071
3106 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3107 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3109 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3111 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3112 (arc_attribute_table): Add 'aux' attribute.
3113 (arc_in_small_data_p): Consider aux like variables.
3114 (arc_is_aux_reg_p): New function.
3115 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3116 (arc_get_aux_arg): New function.
3117 (prepare_move_operands): Handle aux-register access.
3118 (arc_handle_aux_attribute): New function.
3119 * doc/extend.texi (ARC Variable attributes): Add subsection.
3121 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3123 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3124 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3125 (arc_attribute_table): Add 'uncached' attribute.
3126 (arc_print_operand): Print '.di' flag for uncached memory
3128 (arc_in_small_data_p): Do not consider for small data the uncached
3130 (arc_is_uncached_mem_p): New function.
3131 * config/arc/predicates.md (compact_store_memory_operand): Check
3132 for uncached memory accesses.
3133 (nonvol_nonimm_operand): Likewise.
3134 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3136 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3139 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3140 falign-loops=): Add Optimization flag.
3142 2018-01-30 Jeff Law <law@redhat.com>
3145 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3146 INT_REGISTERS_SAVED. Check it prior to calling
3147 get_scratch_register_on_entry.
3148 (ix86_adjust_stack_and_probe): Similarly.
3149 (ix86_emit_probe_stack_range): Similarly.
3150 (ix86_expand_prologue): Corresponding changes.
3152 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3155 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3156 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3158 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3161 * lra-constraints.c (curr_insn_transform): Process AND in the
3164 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3166 PR rtl-optimization/83986
3167 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3168 dependence against last_pending_memory_flush in addition to
3171 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3173 PR tree-optimization/81611
3174 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3177 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3180 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3183 2018-01-30 Richard Biener <rguenther@suse.de>
3184 Jakub Jelinek <jakub@redhat.com>
3186 PR tree-optimization/84111
3187 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3188 inner loops added during recursion, as they don't have up-to-date
3191 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3194 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3195 (can_inline_edge_by_limits_p): ... here.
3196 (can_early_inline_edge_p, check_callers,
3197 update_caller_keys, update_callee_keys, recursive_inlining,
3198 add_new_edges_to_heap, speculation_useful_p,
3199 inline_small_functions,
3200 inline_small_functions, flatten_function,
3201 inline_to_all_callers_1): Update.
3203 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3205 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3208 2018-01-30 Richard Biener <rguenther@suse.de>
3210 PR tree-optimization/83008
3211 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3212 invariant and constant vector uses in stmts when they need
3215 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3218 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3219 * configure: Regenerate.
3221 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3223 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3225 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3226 Use gen_rtx_REG rather than gen_lowpart.
3228 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3230 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3231 rather than 0 when creating partial subregs.
3233 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3235 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3238 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3241 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3242 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3243 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3244 flags. This restores the settings used before the 2017-07-24.
3245 Turning off pre increment/decrement/modify allows IVOPTS to
3246 optimize DF/SF loops where the index is an int.
3248 2018-01-29 Richard Biener <rguenther@suse.de>
3249 Kelvin Nilsen <kelvin@gcc.gnu.org>
3252 * tree-vect-stmts.c (vectorizable_call): Don't call
3253 targetm.vectorize_builtin_md_vectorized_function if callee is
3256 2018-01-22 Carl Love <cel@us.ibm.com>
3258 * doc/extend.tex: Fix typo in second arg in
3259 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3261 2018-01-29 Richard Biener <rguenther@suse.de>
3263 PR tree-optimization/84086
3264 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3265 (flush_ssaname_freelist): When SSA names were released reset
3266 the SCEV hash table.
3268 2018-01-29 Richard Biener <rguenther@suse.de>
3270 PR tree-optimization/84057
3271 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3272 removed paths when removing edges.
3274 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3276 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3277 -mfunction-return=@var{choice}.
3279 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3282 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3284 (layout::print_source_line): Likewise.
3285 (test_get_line_width_without_trailing_whitespace): Add test cases.
3287 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3290 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3293 2018-01-26 Jim Wilson <jimw@sifive.com>
3295 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3297 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3300 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3302 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3303 and CMP + SUB-immediate -> SUBS.
3305 2018-01-26 Martin Sebor <msebor@redhat.com>
3307 PR tree-optimization/83896
3308 * tree-ssa-strlen.c (get_string_len): Rename...
3309 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3310 Avoid assuming length is constant.
3311 (handle_char_store): Use HOST_WIDE_INT for string length.
3313 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3316 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3317 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3319 2018-01-26 Richard Biener <rguenther@suse.de>
3321 PR rtl-optimization/84003
3322 * dse.c (record_store): Only record redundant stores when
3323 the earlier store aliases at least all accesses the later one does.
3325 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3327 PR rtl-optimization/83985
3328 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3329 REG_CFA_RESTORE insns.
3330 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3331 REG_CFA_RESTORE insns here.
3334 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3335 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3337 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3339 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3340 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3341 (arc_init): Likewise.
3342 (arc_override_options): Likewise.
3343 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3345 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3347 * config/arc/arc.h (TARGET_DBNZ): Define.
3348 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3349 properly set the tune attribute.
3350 (dbnz): Use TARGET_DBNZ guard.
3351 * config/arc/arc.opt (mtune): Add core3 option.
3353 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3355 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3356 recognize new pic like addresses.
3357 (arc_delegitimize_address): Clean up.
3359 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3361 * config/arc/arc-arches.def: Option mrf16 valid for all
3363 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3364 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3365 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3366 * config/arc/arc-tables.opt: Regenerate.
3367 * config/arc/arc.c (arc_conditional_register_usage): Handle
3368 reduced register file case.
3369 (arc_file_start): Set must have build attributes.
3370 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3372 * config/arc/arc.opt (mrf16): Add new option.
3373 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3374 * config/arc/genmultilib.awk: Handle new mrf16 option.
3375 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3376 * config/arc/t-multilib: Regenerate.
3377 * doc/invoke.texi (ARC Options): Document mrf16 option.
3379 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3381 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3382 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3383 (arc_attribute_table): Add 'secure_call' attribute.
3384 (arc_print_operand): Print secure call operand.
3385 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3387 (arc_is_secure_call_p): New function. * config/arc/arc.md
3388 (call_i): Add support for sjli instruction.
3389 (call_value_i): Likewise.
3390 * config/arc/constraints.md (Csc): New constraint.
3392 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3393 John Eric Martin <John.Martin@emmicro-us.com>
3395 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3396 * config/arc/arc.c (_arc_jli_section): New struct.
3397 (arc_jli_section): New type.
3398 (rc_jli_sections): New static variable.
3399 (arc_handle_jli_attribute): New function.
3400 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3401 (arc_file_end): New function.
3402 (TARGET_ASM_FILE_END): Define.
3403 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3404 (arc_add_jli_section): New function.
3405 (jli_call_scan): Likewise.
3406 (arc_reorg): Call jli_call_scan.
3407 (arc_output_addsi): Remove 'S' from printing asm operand.
3408 (arc_is_jli_call_p): New function.
3409 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3411 (movhi_insn): Likewise.
3412 (movsi_insn): Likewise.
3413 (movsi_set_cc_insn): Likewise.
3414 (loadqi_update): Likewise.
3415 (load_zeroextendqisi_update): Likewise.
3416 (load_signextendqisi_update): Likewise.
3417 (loadhi_update): Likewise.
3418 (load_zeroextendhisi_update): Likewise.
3419 (load_signextendhisi_update): Likewise.
3420 (loadsi_update): Likewise.
3421 (loadsf_update): Likewise.
3422 (movsicc_insn): Likewise.
3423 (bset_insn): Likewise.
3424 (bxor_insn): Likewise.
3425 (bclr_insn): Likewise.
3426 (bmsk_insn): Likewise.
3427 (bicsi3_insn): Likewise.
3428 (cmpsi_cc_c_insn): Likewise.
3429 (movsi_ne): Likewise.
3430 (movsi_cond_exec): Likewise.
3431 (clrsbsi2): Likewise.
3442 (call_i): Remove 'S' asm letter, add jli instruction.
3443 (call_value_i): Likewise.
3444 * config/arc/arc.op (mjli-always): New option.
3445 * config/arc/constraints.md (Cji): New constraint.
3446 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3448 (subsf3_fpx): Likewise.
3449 (mulsf3_fpx): Likewise.
3450 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3452 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3454 * doc/invoke.texi (ARC): Document mjli-always option.
3456 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3458 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3459 avoid addition with 0 and use incw and decw where possible.
3461 2018-01-26 Richard Biener <rguenther@suse.de>
3463 PR tree-optimization/81082
3464 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3465 association if it requires casting to unsigned.
3466 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3467 from fold_plusminus_mult_expr to catch important cases late when
3468 range info is available.
3470 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3472 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3473 * configure.ac (hidden_linkonce): New test.
3474 * configure: Regenerate.
3475 * config.in: Regenerate.
3477 2018-01-26 Julia Koval <julia.koval@intel.com>
3479 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3480 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3481 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3482 _mm_mask_bitshuffle_epi64_mask): Fix type.
3483 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3484 USI_FTYPE_V4DI_V4DI_USI): Remove.
3485 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3486 __builtin_ia32_vpshufbitqmb256_mask,
3487 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3488 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3489 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3491 2018-01-26 Alan Modra <amodra@gmail.com>
3494 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3495 UNSPEC_VBPERMQ. Sort other unspecs.
3497 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3499 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3501 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3504 * predict.c (drop_profile): Do not push/pop cfun; update also
3506 (handle_missing_profiles): Fix logic looking for zero profiles.
3508 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3511 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3512 on functions with #pragma omp declare simd or functions with simd
3514 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3515 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3516 Remove trailing \n from warning_at calls.
3518 2018-01-25 Tom de Vries <tom@codesourcery.com>
3521 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3522 for neutered workers.
3524 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3527 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3528 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3530 2018-01-24 Jeff Law <law@redhat.com>
3533 * i386.c (get_probe_interval): Move to earlier point.
3534 (ix86_compute_frame_layout): If -fstack-clash-protection and
3535 the frame is larger than the probe interval, then use pushes
3536 to save registers rather than reg->mem moves.
3537 (ix86_expand_prologue): Remove conditional for int_registers_saved
3540 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3543 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3544 min/max for never referenced object.
3546 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3549 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3551 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3552 attributes from DECL_ATTRIBUTES (decl) without affecting
3553 DECL_ATTRIBUTES (current_function_decl).
3554 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3555 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3557 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3559 PR tree-optimization/83979
3560 * fold-const.c (fold_comparison): Use constant_boolean_node
3561 instead of boolean_{true,false}_node.
3563 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3565 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3568 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3570 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3571 Simplify the clause that sets the length attribute.
3572 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3573 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3574 clause that sets the length attribute.
3575 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3577 2018-01-24 Tom de Vries <tom@codesourcery.com>
3580 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3581 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3582 Add strict parameter.
3583 (prevent_branch_around_nothing): Insert dummy insn between branch to
3584 label and label with no ptx insn inbetween.
3585 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3587 2018-01-24 Tom de Vries <tom@codesourcery.com>
3590 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3591 for neutered threads in warp.
3592 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3594 2018-01-24 Richard Biener <rguenther@suse.de>
3596 PR tree-optimization/83176
3597 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3600 2018-01-24 Richard Biener <rguenther@suse.de>
3602 PR tree-optimization/82819
3603 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3604 code generating pluses that are no-ops in the target precision.
3606 2018-01-24 Richard Biener <rguenther@suse.de>
3609 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3611 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3613 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3614 to merge probabilities.
3615 * predict.c (probably_never_executed): Also mark as cold functions
3616 with global 0 profile and guessed local profile.
3617 * profile-count.c (profile_probability::combine_with_count): New
3619 * profile-count.h (profile_probability::operator*,
3620 profile_probability::operator*=, profile_probability::operator/,
3621 profile_probability::operator/=): Reduce precision to adjusted
3622 and set value to guessed on contradictory divisions.
3623 (profile_probability::combine_with_freq): Remove.
3624 (profile_probability::combine_wiht_count): Declare.
3625 (profile_count::force_nonzero):: Set to adjusted.
3626 (profile_count::probability_in):: Set quality to adjusted.
3627 * tree-ssa-tail-merge.c (replace_block_by): Use
3630 2018-01-23 Andrew Waterman <andrew@sifive.com>
3631 Jim Wilson <jimw@sifive.com>
3633 * config/riscv/riscv.c (riscv_stack_boundary): New.
3634 (riscv_option_override): Set riscv_stack_boundary. Handle
3635 riscv_preferred_stack_boundary_arg.
3636 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3637 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3638 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3639 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3640 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3641 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3643 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3646 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3647 of struct ix86_frame.
3648 (ix86_expand_epilogue): Likewise. Add a local variable for
3649 the reg_save_offset field in struct ix86_frame.
3651 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3653 PR tree-optimization/82604
3654 * tree-loop-distribution.c (enum partition_kind): New enum item
3655 PKIND_PARTIAL_MEMSET.
3656 (partition_builtin_p): Support above new enum item.
3657 (generate_code_for_partition): Ditto.
3658 (compute_access_range): Differentiate cases that equality can be
3659 proven at all loops, the innermost loops or no loops.
3660 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3661 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3662 (finalize_partitions, distribute_loop): Don't fuse partition of
3663 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3664 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3667 2018-01-23 Martin Liska <mliska@suse.cz>
3669 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3670 order to ignore the predictor.
3671 (PRED_POLYMORPHIC_CALL): Likewise.
3672 (PRED_RECURSIVE_CALL): Likewise.
3674 2018-01-23 Martin Liska <mliska@suse.cz>
3676 * tree-profile.c (tree_profiling): Print function header to
3677 aware reader which function we are working on.
3678 * value-prof.c (gimple_find_values_to_profile): Do not print
3679 not interesting value histograms.
3681 2018-01-23 Martin Liska <mliska@suse.cz>
3683 * profile-count.h (enum profile_quality): Add
3684 profile_uninitialized as the first value. Do not number values
3685 as they are zero based.
3686 (profile_count::verify): Update sanity check.
3687 (profile_probability::verify): Likewise.
3689 2018-01-23 Nathan Sidwell <nathan@acm.org>
3691 * doc/invoke.texi (ffor-scope): Deprecate.
3693 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3695 PR tree-optimization/83510
3696 * domwalk.c (set_all_edges_as_executable): New function.
3697 (dom_walker::dom_walker): Convert bool param
3698 "skip_unreachable_blocks" to enum reachability. Move setup of
3699 edge flags to set_all_edges_as_executable and only do it when
3700 reachability is REACHABLE_BLOCKS.
3701 * domwalk.h (enum dom_walker::reachability): New enum.
3702 (dom_walker::dom_walker): Convert bool param
3703 "skip_unreachable_blocks" to enum reachability.
3704 (set_all_edges_as_executable): New decl.
3705 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3706 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3708 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3709 but converting true to REACHABLE_BLOCKS.
3710 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3712 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3713 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3714 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3716 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3717 if check_all_array_refs will be called.
3719 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3721 * tree.c (selftest::test_location_wrappers): Add more test
3724 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3726 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3727 (selftest::test_bit_in_range): Likewise.
3729 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3732 * doc/sourcebuild.texi (vect_float): Say that the selector
3733 only describes the situation when -funsafe-math-optimizations is on.
3734 (vect_float_strict): Document.
3736 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3738 PR tree-optimization/83965
3739 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3740 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3741 instead of checking only for a reduction.
3742 (vect_recog_widen_sum_pattern): Likewise.
3744 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3746 * predict.c (probably_never_executed): Only use precise profile info.
3747 (compute_function_frequency): Skip after inlining hack since we now
3748 have quality checking.
3750 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3752 * profile-count.h (profile_probability::very_unlikely,
3753 profile_probability::unlikely, profile_probability::even): Set
3754 precision to guessed.
3756 2018-01-23 Richard Biener <rguenther@suse.de>
3758 PR tree-optimization/83963
3759 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3760 Properly terminate dominator walk when crossing the exit edge not
3761 when visiting its source block.
3763 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3766 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3767 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3769 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3771 PR tree-optimization/83957
3772 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3773 semicolon after for body surrounded by braces.
3775 PR tree-optimization/83081
3776 * profile-count.h (profile_probability::split): New method.
3777 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3778 Use profile_probability::split.
3779 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3780 when splitting a single conditional jump into 2.
3782 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3784 PR tree-optimization/69452
3785 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3788 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3790 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3791 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3792 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3794 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3796 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3797 * config/rl78/rl78.md (movdi): New define_expand.
3798 * config/rl78/rl78.c (rl78_split_movdi): New function.
3800 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3803 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3805 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3806 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3807 128-bit to produce an UNSPEC move to get the double word with the
3808 signbit and then a shift directly to do signbit.
3809 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3810 implementation with a new version that just does either a direct
3811 move or a regular move. Move memory interface to separate insns.
3812 Move insns so they are next to the expander.
3813 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3814 with signbit move. Split big and little endian case.
3815 (signbit<mode>2_dm_mem_le): Likewise.
3816 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3817 (signbit<mode>2_dm2): Likewise.
3819 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3821 * config/rl78/rl78.md (anddi3): New define_expand.
3823 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3825 * config/rl78/rl78.md (umindi3): New define_expand.
3827 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3829 * config/rl78/rl78.md (smindi3): New define_expand.
3831 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3833 * config/rl78/rl78.md (smaxdi3): New define_expand.
3835 2018-01-22 Carl Love <cel@us.ibm.com>
3837 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3838 LVX_V1TI): Add macro expansion.
3839 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3840 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3841 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3842 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3843 Change check to determine if the instruction is a byte reversing
3844 entry. Fix typo in comment.
3845 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3846 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3847 Add def_builtin calls for new builtins.
3848 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3849 Add define_insn expansion.
3851 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3853 * config/rl78/rl78.md (umaxdi3): New define_expand.
3855 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3857 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3858 for non-QImode registers.
3860 2018-01-22 Richard Biener <rguenther@suse.de>
3862 PR tree-optimization/83963
3863 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3864 including the loop exit block.
3865 (scop_detection::merge_sese): Likewise.
3866 (scop_detection::add_scop): Do it here instead.
3868 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3870 * doc/sourcebuild.texi (arm_softfloat): Document.
3872 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3875 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3876 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3877 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3879 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3880 David Edelsohn <dje.gcc@gmail.com>
3883 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3884 Change "crset eq" to "crset 2".
3885 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3886 (*call_indirect_aix<mode>_nospec): Likewise.
3887 (*call_value_indirect_aix<mode>_nospec): Likewise.
3888 (*call_indirect_elfv2<mode>_nospec): Likewise.
3889 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3890 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3891 change assembly output from . to $.
3892 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3893 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3894 (*tablejump<mode>_internal1_nospec): Likewise.
3896 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3899 * config/sh/sh_optimize_sett_clrt.cc:
3900 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3902 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3904 PR tree-optimization/83940
3905 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3906 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3907 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3908 use it to pass back the definition type.
3909 (vect_check_store_rhs): Likewise rhs_dt_out.
3910 (vect_build_gather_load_calls): Add a mask_dt argument and use
3911 it instead of a call to vect_is_simple_use.
3912 (vectorizable_store): Update calls to vect_check_load_store_mask
3913 and vect_check_store_rhs. Use the dt returned by the latter instead
3914 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3915 instead of calls to vect_is_simple_use. Pass the scalar rather
3916 than the vector operand to vect_is_simple_use when handling
3917 second and subsequent copies of an rhs value.
3918 (vectorizable_load): Update calls to vect_check_load_store_mask
3919 and vect_build_gather_load_calls. Use the cached mask_dt and
3920 gs_info.offset_dt instead of calls to vect_is_simple_use.
3922 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3925 * tree-emutls.c: Include gimplify.h.
3926 (lower_emutls_2): New function.
3927 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3928 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3929 it before further processing.
3932 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3933 UINTVAL (trueop1) instead of INTVAL (op1).
3935 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3939 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3940 INCOMING_FRAME_SP_OFFSET if not defined.
3941 (scan_trace): Add ENTRY argument. If true and
3942 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3943 emit a note to adjust the CFA offset.
3944 (create_cfi_notes): Adjust scan_trace callers.
3945 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3946 INCOMING_FRAME_SP_OFFSET in the CIE.
3947 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3948 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3950 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3951 * doc/tm.texi: Regenerated.
3953 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3955 PR rtl-optimization/83147
3956 * lra-constraints.c (remove_inheritance_pseudos): Use
3957 lra_substitute_pseudo_within_insn.
3959 2018-01-19 Tom de Vries <tom@codesourcery.com>
3960 Cesar Philippidis <cesar@codesourcery.com>
3963 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3965 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3968 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3969 spaces for function labels.
3971 2018-01-19 Martin Liska <mliska@suse.cz>
3973 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3974 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3975 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3976 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3977 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3978 (PRED_CONST_RETURN): Change from 69 to 65.
3979 (PRED_NULL_RETURN): Change from 91 to 71.
3980 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3981 (PRED_LOOP_GUARD): Change from 66 to 73.
3983 2018-01-19 Martin Liska <mliska@suse.cz>
3985 * predict.c (predict_insn_def): Add new assert.
3986 (struct branch_predictor): Change type to signed integer.
3987 (test_prediction_value_range): Amend test to cover
3989 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3990 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3991 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3992 (PRED_LOOP_IV_COMPARE): Likewise.
3993 * predict.h (PROB_UNINITIALIZED): Define new constant.
3995 2018-01-19 Martin Liska <mliska@suse.cz>
3997 * predict.c (dump_prediction): Add new format for
3998 analyze_brprob.py script which is enabled with -details
4000 * profile-count.h (precise_p): New function.
4002 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4004 PR tree-optimization/83922
4005 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4006 there are no statements that need masking.
4007 (vect_active_double_reduction_p): New function.
4008 (vect_analyze_loop_operations): Use it when handling phis that
4009 are not in the loop header.
4011 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4013 PR tree-optimization/83914
4014 * tree-vect-loop.c (vectorizable_induction): Don't convert
4015 init_expr or apply the peeling adjustment for inductions
4016 that are nested within the vectorized loop.
4018 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4020 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4023 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4027 * function.h (gimplify_parameters): Add gimple_seq * argument.
4028 * function.c: Include gimple.h and options.h.
4029 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4030 for the added local temporaries if needed.
4031 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4032 if there are any parameter cleanups, wrap whole body into a
4033 try/finally with the cleanups.
4035 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4038 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4039 Use GET_MODE_CLASS for scalar floating point.
4041 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4045 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4046 Fix call of call_cgraph_insertion_hooks.
4048 2018-01-18 Martin Sebor <msebor@redhat.com>
4050 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4052 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4055 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4058 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4061 * common.opt: (-ffile-prefix-map): New option.
4062 * opts.c (common_handle_option): Defer it.
4063 * opts-global.c (handle_common_deferred_options): Handle it.
4064 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4065 * file-prefix-map.h: New file.
4066 (remap_debug_filename, add_debug_prefix_map): ...here.
4067 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4068 * final.c (debug_prefix_map, add_debug_prefix_map
4069 remap_debug_filename): Move to...
4070 * file-prefix-map.c: New file.
4071 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4072 generalize, get rid of alloca(), use strrchr() instead of strchr().
4073 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4074 Implement in terms of add_prefix_map().
4075 (remap_macro_filename, remap_debug_filename): Implement in term of
4077 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4078 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4079 * dbxout.c: Include file-prefix-map.h.
4080 * varasm.c: Likewise.
4081 * vmsdbgout.c: Likewise.
4082 * xcoffout.c: Likewise.
4083 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4084 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4085 * doc/invoke.texi (-ffile-prefix-map): Document.
4086 (-fdebug-prefix-map): Update description.
4088 2018-01-18 Martin Liska <mliska@suse.cz>
4090 * config/i386/i386.c (indirect_thunk_name): Document that also
4092 (output_indirect_thunk): Document why both instructions
4093 (pause and lfence) are generated.
4095 2018-01-18 Richard Biener <rguenther@suse.de>
4097 PR tree-optimization/83887
4098 * graphite-scop-detection.c
4099 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4100 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4101 (scop_detection::merge_sese): Re-implement with a flood-fill
4102 algorithm that properly finds a SESE region if it exists.
4104 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4107 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4108 pointer_diff optimizations use view_convert instead of convert.
4110 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4112 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4113 Generate different code for -mno-speculate-indirect-jumps.
4114 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4115 (*call_indirect_aix<mode>): Disable for
4116 -mno-speculate-indirect-jumps.
4117 (*call_indirect_aix<mode>_nospec): New define_insn.
4118 (*call_value_indirect_aix<mode>): Disable for
4119 -mno-speculate-indirect-jumps.
4120 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4121 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4122 -mno-speculate-indirect-jumps.
4123 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4125 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4127 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4128 long double type, set the flags for noting the default long double
4129 type, even if we don't pass or return a long double type.
4131 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4134 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4137 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4139 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4140 support for merge[hl].
4141 (fold_mergehl_helper): New helper function.
4142 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4143 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4144 (altivec_vmrglw_direct): Add xxmrglw insn.
4146 2018-01-17 Andrew Waterman <andrew@sifive.com>
4148 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4149 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4151 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4154 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4155 call the lto_location_cache before reading the
4156 DECL_SOURCE_LOCATION of the types.
4158 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4159 Richard Sandiford <richard.sandiford@linaro.org>
4161 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4162 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4163 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4164 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4165 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4167 * config/aarch64/constraints.md (aarch64_movti_operand):
4169 * config/aarch64/predicates.md (Uti): Add new constraint.
4171 2018-01-17 Carl Love <cel@us.ibm.com>
4173 * config/rs6000/vsx.md (define_expand xl_len_r,
4174 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4175 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4177 (define_expand, define_insn): Move the shift left from the
4178 define_insn to the define_expand for lxvl and stxvl instructions.
4179 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4180 and XL_LEN_R definitions to PURE.
4182 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4184 * config/i386/i386.c (indirect_thunk_name): Declare regno
4185 as unsigned int. Compare regno with INVALID_REGNUM.
4186 (output_indirect_thunk): Ditto.
4187 (output_indirect_thunk_function): Ditto.
4188 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4189 in the call to output_indirect_thunk_function.
4191 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4194 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4195 rather than the size of inner_type to determine the stack slot size
4196 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4198 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4201 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4204 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4206 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4207 endian Linux systems to optionally enable multilibs for selecting
4208 the long double type if the user configured an explicit type.
4209 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4210 have no long double multilibs if not defined.
4211 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4212 warn if the user used -mabi={ieee,ibm}longdouble and we built
4213 multilibs for long double.
4214 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4215 appropriate multilib option.
4216 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4218 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4219 for building long double multilibs.
4220 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4222 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4224 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4227 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4229 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4232 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4235 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4238 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4240 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4241 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4243 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4245 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4246 different rtl trees depending on TARGET_64BIT.
4247 (rs6000_gen_lvx): Likewise.
4249 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4251 * config/visium/visium.md (nop): Tweak comment.
4252 (hazard_nop): Likewise.
4254 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4256 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4257 -mspeculate-indirect-jumps.
4258 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4259 for -mno-speculate-indirect-jumps.
4260 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4261 (*call_value_indirect_elfv2<mode>): Disable for
4262 -mno-speculate-indirect-jumps.
4263 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4264 (indirect_jump): Emit different RTL for
4265 -mno-speculate-indirect-jumps.
4266 (*indirect_jump<mode>): Disable for
4267 -mno-speculate-indirect-jumps.
4268 (*indirect_jump<mode>_nospec): New define_insn.
4269 (tablejump): Emit different RTL for
4270 -mno-speculate-indirect-jumps.
4271 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4272 (tablejumpsi_nospec): New define_expand.
4273 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4274 (tablejumpdi_nospec): New define_expand.
4275 (*tablejump<mode>_internal1): Disable for
4276 -mno-speculate-indirect-jumps.
4277 (*tablejump<mode>_internal1_nospec): New define_insn.
4278 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4281 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4283 * caller-save.c (insert_save): Drop unnecessary parameter. All
4286 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4287 Richard Biener <rguenth@suse.de>
4290 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4291 return early, inline manually is_gimple_sizepos. Make sure if we
4292 call gimplify_expr we don't end up with a gimple constant.
4293 * tree.c (variably_modified_type_p): Don't return true for
4294 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4295 * gimplify.h (is_gimple_sizepos): Remove.
4297 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4299 PR tree-optimization/83857
4300 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4301 vectorizable_live_operation for pure SLP statements.
4302 (vectorizable_live_operation): Handle PHIs.
4304 2018-01-16 Richard Biener <rguenther@suse.de>
4306 PR tree-optimization/83867
4307 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4308 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4310 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4313 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4314 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4315 If off is not INTEGER_CST, issue a may not be aligned warning
4316 rather than isn't aligned. Use isn%'t rather than isn't.
4317 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4319 <case MULT_EXPR>: Improve the case when bottom and one of the
4320 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4321 operand, in that case check if the other operand is multiple of
4322 bottom divided by the INTEGER_CST operand.
4324 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4327 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4328 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4329 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4330 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4331 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4332 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4333 (pa_function_arg_size): New function.
4335 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4337 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4338 in a separate statement.
4340 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4342 PR tree-optimization/83847
4343 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4344 group gathers and scatters.
4346 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4348 PR rtl-optimization/86620
4349 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4351 PR rtl-optimization/83213
4352 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4353 to last if both are JUMP_INSNs.
4355 PR tree-optimization/83843
4356 * gimple-ssa-store-merging.c
4357 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4358 store_immediate_info for bswap/nop orig_stores.
4360 2018-01-15 Andrew Waterman <andrew@sifive.com>
4362 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4364 <UDIV>: Increase cost if !TARGET_DIV.
4366 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4368 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4369 (define_attr "cr_logical_3op"): New.
4370 (cceq_ior_compare): Adjust.
4371 (cceq_ior_compare_complement): Adjust.
4372 (*cceq_rev_compare): Adjust.
4373 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4374 (is_cracked_insn): Adjust.
4375 (insn_must_be_first_in_group): Adjust.
4376 * config/rs6000/40x.md: Adjust.
4377 * config/rs6000/440.md: Adjust.
4378 * config/rs6000/476.md: Adjust.
4379 * config/rs6000/601.md: Adjust.
4380 * config/rs6000/603.md: Adjust.
4381 * config/rs6000/6xx.md: Adjust.
4382 * config/rs6000/7450.md: Adjust.
4383 * config/rs6000/7xx.md: Adjust.
4384 * config/rs6000/8540.md: Adjust.
4385 * config/rs6000/cell.md: Adjust.
4386 * config/rs6000/e300c2c3.md: Adjust.
4387 * config/rs6000/e500mc.md: Adjust.
4388 * config/rs6000/e500mc64.md: Adjust.
4389 * config/rs6000/e5500.md: Adjust.
4390 * config/rs6000/e6500.md: Adjust.
4391 * config/rs6000/mpc.md: Adjust.
4392 * config/rs6000/power4.md: Adjust.
4393 * config/rs6000/power5.md: Adjust.
4394 * config/rs6000/power6.md: Adjust.
4395 * config/rs6000/power7.md: Adjust.
4396 * config/rs6000/power8.md: Adjust.
4397 * config/rs6000/power9.md: Adjust.
4398 * config/rs6000/rs64.md: Adjust.
4399 * config/rs6000/titan.md: Adjust.
4401 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4403 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4404 ix86_indirect_branch_register logic.
4406 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4408 * config/i386/constraints.md (Bs): Update
4409 ix86_indirect_branch_register check. Don't check
4410 ix86_indirect_branch_register with GOT_memory_operand.
4412 * config/i386/predicates.md (GOT_memory_operand): Don't check
4413 ix86_indirect_branch_register here.
4414 (GOT32_symbol_operand): Likewise.
4416 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4418 * config/i386/predicates.md (constant_call_address_operand):
4419 Rewrite ix86_indirect_branch_register logic.
4420 (sibcall_insn_operand): Likewise.
4422 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4424 * config/i386/constraints.md (Bs): Replace
4425 ix86_indirect_branch_thunk_register with
4426 ix86_indirect_branch_register.
4428 * config/i386/i386.md (indirect_jump): Likewise.
4429 (tablejump): Likewise.
4430 (*sibcall_memory): Likewise.
4431 (*sibcall_value_memory): Likewise.
4432 Peepholes of indirect call and jump via memory: Likewise.
4433 * config/i386/i386.opt: Likewise.
4434 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4435 (GOT_memory_operand): Likewise.
4436 (call_insn_operand): Likewise.
4437 (sibcall_insn_operand): Likewise.
4438 (GOT32_symbol_operand): Likewise.
4440 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4443 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4444 type rather than type addr's type points to.
4445 (expand_omp_atomic_mutex): Likewise.
4446 (expand_omp_atomic): Likewise.
4448 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4451 * config/i386/i386.c (output_indirect_thunk_function): Use
4452 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4453 for __x86_return_thunk.
4455 2018-01-15 Richard Biener <rguenther@suse.de>
4458 * expmed.c (extract_bit_field_1): Fix typo.
4460 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4463 * config/arm/iterators.md (VF): New mode iterator.
4464 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4465 Remove integer-related logic from pattern.
4466 (neon_vabd<mode>_3): Likewise.
4468 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4471 * common.opt (fstrict-overflow): No longer an alias.
4472 (fwrapv-pointer): New option.
4473 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4474 also for pointer types based on flag_wrapv_pointer.
4475 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4476 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4477 opts->x_flag_wrapv got set.
4478 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4479 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4480 POINTER_TYPE_OVERFLOW_UNDEFINED.
4481 * match.pd: Likewise in address comparison pattern.
4482 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4484 2018-01-15 Richard Biener <rguenther@suse.de>
4487 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4488 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4489 Reset type names to their identifier if their TYPE_DECL doesn't
4490 have linkage (and thus is used for ODR and devirt).
4491 (save_debug_info_for_decl): Remove.
4492 (save_debug_info_for_type): Likewise.
4493 (add_tree_to_fld_list): Adjust.
4494 * tree-pretty-print.c (dump_generic_node): Make dumping of
4495 type names more robust.
4497 2018-01-15 Richard Biener <rguenther@suse.de>
4499 * BASE-VER: Bump to 8.0.1.
4501 2018-01-14 Martin Sebor <msebor@redhat.com>
4504 * builtins.c (check_access): Avoid warning when the no-warning bit
4507 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4509 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4510 * ira-color (allocno_hard_regs_compare): Likewise.
4512 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4515 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4516 Use .pushsection/.popsection.
4518 2018-01-14 Martin Sebor <msebor@redhat.com>
4521 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4523 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4525 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4526 entry from extra_headers.
4527 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4528 extra_headers, make the list bitwise identical to the i?86-*-* one.
4530 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4532 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4533 -mcmodel=large with -mindirect-branch=thunk,
4534 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4535 -mfunction-return=thunk-extern.
4536 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4537 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4538 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4540 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4542 * config/i386/i386.c (print_reg): Print the name of the full
4543 integer register without '%'.
4544 (ix86_print_operand): Handle 'V'.
4545 * doc/extend.texi: Document 'V' modifier.
4547 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4549 * config/i386/constraints.md (Bs): Disallow memory operand for
4550 -mindirect-branch-register.
4552 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4553 (GOT_memory_operand): Likewise.
4554 (call_insn_operand): Likewise.
4555 (sibcall_insn_operand): Likewise.
4556 (GOT32_symbol_operand): Likewise.
4557 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4558 for -mindirect-branch-register.
4559 (tablejump): Likewise.
4560 (*sibcall_memory): Likewise.
4561 (*sibcall_value_memory): Likewise.
4562 Disallow peepholes of indirect call and jump via memory for
4563 -mindirect-branch-register.
4564 (*call_pop): Replace m with Bw.
4565 (*call_value_pop): Likewise.
4566 (*sibcall_pop_memory): Replace m with Bs.
4567 * config/i386/i386.opt (mindirect-branch-register): New option.
4568 * doc/invoke.texi: Document -mindirect-branch-register option.
4570 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4572 * config/i386/i386-protos.h (ix86_output_function_return): New.
4573 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4574 set function_return_type.
4575 (indirect_thunk_name): Add ret_p to indicate thunk for function
4577 (output_indirect_thunk_function): Pass false to
4578 indirect_thunk_name.
4579 (ix86_output_indirect_branch_via_reg): Likewise.
4580 (ix86_output_indirect_branch_via_push): Likewise.
4581 (output_indirect_thunk_function): Create alias for function
4582 return thunk if regno < 0.
4583 (ix86_output_function_return): New function.
4584 (ix86_handle_fndecl_attribute): Handle function_return.
4585 (ix86_attribute_table): Add function_return.
4586 * config/i386/i386.h (machine_function): Add
4587 function_return_type.
4588 * config/i386/i386.md (simple_return_internal): Use
4589 ix86_output_function_return.
4590 (simple_return_internal_long): Likewise.
4591 * config/i386/i386.opt (mfunction-return=): New option.
4592 (indirect_branch): Mention -mfunction-return=.
4593 * doc/extend.texi: Document function_return function attribute.
4594 * doc/invoke.texi: Document -mfunction-return= option.
4596 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4598 * config/i386/i386-opts.h (indirect_branch): New.
4599 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4600 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4601 with local indirect jump when converting indirect call and jump.
4602 (ix86_set_indirect_branch_type): New.
4603 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4604 (indirectlabelno): New.
4605 (indirect_thunk_needed): Likewise.
4606 (indirect_thunk_bnd_needed): Likewise.
4607 (indirect_thunks_used): Likewise.
4608 (indirect_thunks_bnd_used): Likewise.
4609 (INDIRECT_LABEL): Likewise.
4610 (indirect_thunk_name): Likewise.
4611 (output_indirect_thunk): Likewise.
4612 (output_indirect_thunk_function): Likewise.
4613 (ix86_output_indirect_branch_via_reg): Likewise.
4614 (ix86_output_indirect_branch_via_push): Likewise.
4615 (ix86_output_indirect_branch): Likewise.
4616 (ix86_output_indirect_jmp): Likewise.
4617 (ix86_code_end): Call output_indirect_thunk_function if needed.
4618 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4620 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4621 (ix86_attribute_table): Add indirect_branch.
4622 * config/i386/i386.h (machine_function): Add indirect_branch_type
4623 and has_local_indirect_jump.
4624 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4626 (tablejump): Likewise.
4627 (*indirect_jump): Use ix86_output_indirect_jmp.
4628 (*tablejump_1): Likewise.
4629 (simple_return_indirect_internal): Likewise.
4630 * config/i386/i386.opt (mindirect-branch=): New option.
4631 (indirect_branch): New.
4634 (thunk-inline): Likewise.
4635 (thunk-extern): Likewise.
4636 * doc/extend.texi: Document indirect_branch function attribute.
4637 * doc/invoke.texi: Document -mindirect-branch= option.
4639 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4642 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4644 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4646 * ipa-inline.c (want_inline_small_function_p): Return false if
4647 inlining has already failed with CIF_FINAL_ERROR.
4648 (update_caller_keys): Call want_inline_small_function_p before
4650 (update_callee_keys): Likewise.
4652 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4654 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4656 (rs6000_quadword_masked_address_p): Likewise.
4657 (quad_aligned_load_p): Likewise.
4658 (quad_aligned_store_p): Likewise.
4659 (const_load_sequence_p): Add comment to describe the outer-most loop.
4660 (mimic_memory_attributes_and_flags): New function.
4661 (rs6000_gen_stvx): Likewise.
4662 (replace_swapped_aligned_store): Likewise.
4663 (rs6000_gen_lvx): Likewise.
4664 (replace_swapped_aligned_load): Likewise.
4665 (replace_swapped_load_constant): Capitalize argument name in
4666 comment describing this function.
4667 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4668 and stores that access quad-word aligned addresses and replace
4669 with stvx or lvx instructions when appropriate.
4670 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4671 New function prototype.
4672 (rs6000_quadword_masked_address_p): Likewise.
4673 (rs6000_gen_lvx): Likewise.
4674 (rs6000_gen_stvx): Likewise.
4675 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4676 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4677 when memory address is aligned.
4678 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4679 this split to select lvx instruction when memory address is aligned.
4680 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4681 instruction when memory address is aligned.
4682 (*vsx_le_perm_load_v16qi): Likewise.
4683 (four unnamed splitters): Modify to select the stvx instruction
4684 when memory is aligned.
4686 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4688 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4689 which appears in the queue multiple times.
4691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4692 Alan Hayward <alan.hayward@arm.com>
4693 David Sherwood <david.sherwood@arm.com>
4695 * tree-vectorizer.h (vec_lower_bound): New structure.
4696 (_loop_vec_info): Add check_nonzero and lower_bounds.
4697 (LOOP_VINFO_CHECK_NONZERO): New macro.
4698 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4699 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4700 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4701 fields. Make seg_len the distance travelled, not including the
4703 (dr_direction_indicator): Declare.
4704 (dr_zero_step_indicator): Likewise.
4705 (dr_known_forward_stride_p): Likewise.
4706 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4708 (runtime_alias_check_p): Allow runtime alias checks with
4710 (operator ==): Compare access_size and align.
4711 (prune_runtime_alias_test_list): Rework for new distinction between
4712 the access_size and seg_len.
4713 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4715 (get_segment_min_max): New function.
4716 (create_intersect_range_checks): Use it.
4717 (dr_step_indicator): New function.
4718 (dr_direction_indicator): Likewise.
4719 (dr_zero_step_indicator): Likewise.
4720 (dr_known_forward_stride_p): Likewise.
4721 * tree-loop-distribution.c (data_ref_segment_size): Return
4722 DR_STEP * (niters - 1).
4723 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4725 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4726 (vect_preserves_scalar_order_p): New function, split out from...
4727 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4728 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4729 (vect_vfa_access_size): New function.
4730 (vect_vfa_align): Likewise.
4731 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4732 (dump_lower_bound): New function.
4733 (vect_check_lower_bound): Likewise.
4734 (vect_small_gap_p): Likewise.
4735 (vectorizable_with_step_bound_p): Likewise.
4736 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4737 depencies if the vectorization factor is 1. Convert the checks
4738 for nonzero steps into checks on the bounds of DR_STEP. Try using
4739 a bunds check for variable steps if the minimum required step is
4740 relatively small. Update calls to the dr_with_seg_len
4741 constructor and to vect_compile_time_alias.
4742 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4744 (vect_loop_versioning): Call it.
4745 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4747 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4750 Alan Hayward <alan.hayward@arm.com>
4751 David Sherwood <david.sherwood@arm.com>
4753 * doc/sourcebuild.texi (vect_scatter_store): Document.
4754 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4756 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4758 * genopinit.c (main): Add supports_vec_scatter_store and
4759 supports_vec_scatter_store_cached to target_optabs.
4760 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4761 IFN_MASK_SCATTER_STORE.
4762 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4764 * internal-fn.h (internal_store_fn_p): Declare.
4765 (internal_fn_stored_value_index): Likewise.
4766 * internal-fn.c (scatter_store_direct): New macro.
4767 (expand_scatter_store_optab_fn): New function.
4768 (direct_scatter_store_optab_supported_p): New macro.
4769 (internal_store_fn_p): New function.
4770 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4771 IFN_MASK_SCATTER_STORE.
4772 (internal_fn_mask_index): Likewise.
4773 (internal_fn_stored_value_index): New function.
4774 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4776 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4777 * optabs-query.c (supports_vec_scatter_store_p): New function.
4778 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4779 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4780 true for scatter stores.
4781 (vect_gather_scatter_fn_p): Handle scatter stores too.
4782 (vect_check_gather_scatter): Consider using scatter stores if
4783 supports_vec_scatter_store_p.
4784 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4786 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4787 internal_fn_stored_value_index.
4788 (check_load_store_masking): Handle scatter stores too.
4789 (vect_get_store_rhs): Make public.
4790 (vectorizable_call): Use internal_store_fn_p.
4791 (vectorizable_store): Handle scatter store internal functions.
4792 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4793 when deciding whether the end of the group has been reached.
4794 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4795 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4796 (mask_scatter_store<mode>): New insns.
4798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4799 Alan Hayward <alan.hayward@arm.com>
4800 David Sherwood <david.sherwood@arm.com>
4802 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4803 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4804 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4806 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4807 Use vect_truncate_gather_scatter_offset if we can't treat the
4808 operation as a normal gather load or scatter store.
4809 (get_group_load_store_type): Take the gather_scatter_info
4810 as argument. Try using a gather load or scatter store for
4811 single-element groups.
4812 (get_load_store_type): Update calls to get_group_load_store_type
4813 and vect_use_strided_gather_scatters_p.
4815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4816 Alan Hayward <alan.hayward@arm.com>
4817 David Sherwood <david.sherwood@arm.com>
4819 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4820 optional tree argument.
4821 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4823 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4824 but continue to use the current value as a fallback.
4825 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4826 to compare the updates.
4827 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4828 (get_load_store_type): Use it when handling a strided access.
4829 (vect_get_strided_load_store_ops): New function.
4830 (vect_get_data_ptr_increment): Likewise.
4831 (vectorizable_load): Handle strided gather loads. Always pass
4832 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4835 Alan Hayward <alan.hayward@arm.com>
4836 David Sherwood <david.sherwood@arm.com>
4838 * doc/md.texi (gather_load@var{m}): Document.
4839 (mask_gather_load@var{m}): Likewise.
4840 * genopinit.c (main): Add supports_vec_gather_load and
4841 supports_vec_gather_load_cached to target_optabs.
4842 * optabs-tree.c (init_tree_optimization_optabs): Use
4843 ggc_cleared_alloc to allocate target_optabs.
4844 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4845 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4847 * internal-fn.h (internal_load_fn_p): Declare.
4848 (internal_gather_scatter_fn_p): Likewise.
4849 (internal_fn_mask_index): Likewise.
4850 (internal_gather_scatter_fn_supported_p): Likewise.
4851 * internal-fn.c (gather_load_direct): New macro.
4852 (expand_gather_load_optab_fn): New function.
4853 (direct_gather_load_optab_supported_p): New macro.
4854 (direct_internal_fn_optab): New function.
4855 (internal_load_fn_p): Likewise.
4856 (internal_gather_scatter_fn_p): Likewise.
4857 (internal_fn_mask_index): Likewise.
4858 (internal_gather_scatter_fn_supported_p): Likewise.
4859 * optabs-query.c (supports_at_least_one_mode_p): New function.
4860 (supports_vec_gather_load_p): Likewise.
4861 * optabs-query.h (supports_vec_gather_load_p): Declare.
4862 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4863 and memory_type field.
4864 (NUM_PATTERNS): Bump to 15.
4865 * tree-vect-data-refs.c: Include internal-fn.h.
4866 (vect_gather_scatter_fn_p): New function.
4867 (vect_describe_gather_scatter_call): Likewise.
4868 (vect_check_gather_scatter): Try using internal functions for
4869 gather loads. Recognize existing calls to a gather load function.
4870 (vect_analyze_data_refs): Consider using gather loads if
4871 supports_vec_gather_load_p.
4872 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4873 (vect_get_gather_scatter_offset_type): Likewise.
4874 (vect_convert_mask_for_vectype): Likewise.
4875 (vect_add_conversion_to_patterm): Likewise.
4876 (vect_try_gather_scatter_pattern): Likewise.
4877 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4878 (vect_vect_recog_func_ptrs): Add it.
4879 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4880 internal_fn_mask_index and internal_gather_scatter_fn_p.
4881 (check_load_store_masking): Take the gather_scatter_info as an
4882 argument and handle gather loads.
4883 (vect_get_gather_scatter_ops): New function.
4884 (vectorizable_call): Check internal_load_fn_p.
4885 (vectorizable_load): Likewise. Handle gather load internal
4887 (vectorizable_store): Update call to check_load_store_masking.
4888 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4889 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4890 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4891 (aarch64_gather_scale_operand_d): New predicates.
4892 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4893 (mask_gather_load<mode>): New insns.
4895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4896 Alan Hayward <alan.hayward@arm.com>
4897 David Sherwood <david.sherwood@arm.com>
4899 * optabs.def (fold_left_plus_optab): New optab.
4900 * doc/md.texi (fold_left_plus_@var{m}): Document.
4901 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4902 * internal-fn.c (fold_left_direct): Define.
4903 (expand_fold_left_optab_fn): Likewise.
4904 (direct_fold_left_optab_supported_p): Likewise.
4905 * fold-const-call.c (fold_const_fold_left): New function.
4906 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4907 * tree-parloops.c (valid_reduction_p): New function.
4908 (gather_scalar_reductions): Use it.
4909 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4910 (vect_finish_replace_stmt): Declare.
4911 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4912 (needs_fold_left_reduction_p): New function, split out from...
4913 (vect_is_simple_reduction): ...here. Accept reductions that
4914 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4915 (vect_force_simple_reduction): Also store the reduction type in
4916 the assignment's STMT_VINFO_REDUC_TYPE.
4917 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4918 (merge_with_identity): New function.
4919 (vect_expand_fold_left): Likewise.
4920 (vectorize_fold_left_reduction): Likewise.
4921 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4922 scalar phi in place for it. Check for target support and reject
4923 cases that would reassociate the operation. Defer the transform
4924 phase to vectorize_fold_left_reduction.
4925 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4926 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4927 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4931 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4932 call to ifc_temp_var.
4934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4935 Alan Hayward <alan.hayward@arm.com>
4936 David Sherwood <david.sherwood@arm.com>
4938 * target.def (legitimize_address_displacement): Take the original
4939 offset as a poly_int.
4940 * targhooks.h (default_legitimize_address_displacement): Update
4942 * targhooks.c (default_legitimize_address_displacement): Likewise.
4943 * doc/tm.texi: Regenerate.
4944 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4945 as an argument, moving assert of ad->disp == ad->disp_term to...
4946 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4947 Try calling targetm.legitimize_address_displacement before expanding
4948 the address rather than afterwards, and adjust for the new interface.
4949 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4950 Match the new hook interface. Handle SVE addresses.
4951 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4956 * Makefile.in (OBJS): Add early-remat.o.
4957 * target.def (select_early_remat_modes): New hook.
4958 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4959 * doc/tm.texi: Regenerate.
4960 * targhooks.h (default_select_early_remat_modes): Declare.
4961 * targhooks.c (default_select_early_remat_modes): New function.
4962 * timevar.def (TV_EARLY_REMAT): New timevar.
4963 * passes.def (pass_early_remat): New pass.
4964 * tree-pass.h (make_pass_early_remat): Declare.
4965 * early-remat.c: New file.
4966 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4968 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4971 Alan Hayward <alan.hayward@arm.com>
4972 David Sherwood <david.sherwood@arm.com>
4974 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4975 vfm1 with a bound_epilog parameter.
4976 (vect_do_peeling): Update calls accordingly, and move the prologue
4977 call earlier in the function. Treat the base bound_epilog as 0 for
4978 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4979 this base when peeling for gaps.
4980 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4981 with fully-masked loops.
4982 (vect_estimate_min_profitable_iters): Handle the single peeled
4983 iteration in that case.
4985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4986 Alan Hayward <alan.hayward@arm.com>
4987 David Sherwood <david.sherwood@arm.com>
4989 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4990 single-element interleaving even if the size is not a power of 2.
4991 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4992 accesses for single-element interleaving if the group size is
4995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4996 Alan Hayward <alan.hayward@arm.com>
4997 David Sherwood <david.sherwood@arm.com>
4999 * doc/md.texi (fold_extract_last_@var{m}): Document.
5000 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5001 * optabs.def (fold_extract_last_optab): New optab.
5002 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5003 * internal-fn.c (fold_extract_direct): New macro.
5004 (expand_fold_extract_optab_fn): Likewise.
5005 (direct_fold_extract_optab_supported_p): Likewise.
5006 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5007 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5008 EXTRACT_LAST_REDUCTION.
5009 (get_initial_def_for_reduction): Do not create an initial vector
5010 for EXTRACT_LAST_REDUCTION reductions.
5011 (vectorizable_reduction): Leave the scalar phi in place for
5012 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5013 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5014 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5015 transform phase to vectorizable_condition.
5016 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5018 (vect_finish_stmt_generation): ...here.
5019 (vect_finish_replace_stmt): New function.
5020 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5021 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5023 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5026 Alan Hayward <alan.hayward@arm.com>
5027 David Sherwood <david.sherwood@arm.com>
5029 * doc/md.texi (extract_last_@var{m}): Document.
5030 * optabs.def (extract_last_optab): New optab.
5031 * internal-fn.def (EXTRACT_LAST): New internal function.
5032 * internal-fn.c (cond_unary_direct): New macro.
5033 (expand_cond_unary_optab_fn): Likewise.
5034 (direct_cond_unary_optab_supported_p): Likewise.
5035 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5036 loops using EXTRACT_LAST.
5037 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5038 (extract_last_<mode>): ...this optab.
5039 (vec_extract<mode><Vel>): Update accordingly.
5041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5042 Alan Hayward <alan.hayward@arm.com>
5043 David Sherwood <david.sherwood@arm.com>
5045 * target.def (empty_mask_is_expensive): New hook.
5046 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5047 * doc/tm.texi: Regenerate.
5048 * targhooks.h (default_empty_mask_is_expensive): Declare.
5049 * targhooks.c (default_empty_mask_is_expensive): New function.
5050 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5051 if the target says that empty masks are expensive.
5052 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5054 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5057 Alan Hayward <alan.hayward@arm.com>
5058 David Sherwood <david.sherwood@arm.com>
5060 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5061 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5062 (vect_use_loop_mask_for_alignment_p): New function.
5063 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5064 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5065 niters_skip argument. Make sure that the first niters_skip elements
5066 of the first iteration are inactive.
5067 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5068 Update call to vect_set_loop_masks_directly.
5069 (get_misalign_in_elems): New function, split out from...
5070 (vect_gen_prolog_loop_niters): ...here.
5071 (vect_update_init_of_dr): Take a code argument that specifies whether
5072 the adjustment should be added or subtracted.
5073 (vect_update_init_of_drs): Likewise.
5074 (vect_prepare_for_masked_peels): New function.
5075 (vect_do_peeling): Skip prologue peeling if we're using a mask
5076 instead. Update call to vect_update_inits_of_drs.
5077 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5079 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5080 alignment. Do not include the number of peeled iterations in
5081 the minimum threshold in that case.
5082 (vectorizable_induction): Adjust the start value down by
5083 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5084 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5085 Take the number of skipped iterations into account when calculating
5087 * tree-vect-stmts.c (vect_gen_while_not): New function.
5089 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5090 Alan Hayward <alan.hayward@arm.com>
5091 David Sherwood <david.sherwood@arm.com>
5093 * doc/sourcebuild.texi (vect_fully_masked): Document.
5094 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5096 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5098 (vect_analyze_loop_2): ...here. Don't check the vectorization
5099 factor against the number of loop iterations if the loop is
5102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5103 Alan Hayward <alan.hayward@arm.com>
5104 David Sherwood <david.sherwood@arm.com>
5106 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5107 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5108 (dump_groups): Update accordingly.
5109 (iv_use::mem_type): New member variable.
5110 (address_p): New function.
5111 (record_use): Add a mem_type argument and initialize the new
5113 (record_group_use): Add a mem_type argument. Use address_p.
5114 Remove obsolete null checks of base_object. Update call to record_use.
5115 (find_interesting_uses_op): Update call to record_group_use.
5116 (find_interesting_uses_cond): Likewise.
5117 (find_interesting_uses_address): Likewise.
5118 (get_mem_type_for_internal_fn): New function.
5119 (find_address_like_use): Likewise.
5120 (find_interesting_uses_stmt): Try find_address_like_use before
5121 calling find_interesting_uses_op.
5122 (addr_offset_valid_p): Use the iv mem_type field as the type
5123 of the addressed memory.
5124 (add_autoinc_candidates): Likewise.
5125 (get_address_cost): Likewise.
5126 (split_small_address_groups_p): Use address_p.
5127 (split_address_groups): Likewise.
5128 (add_iv_candidate_for_use): Likewise.
5129 (autoinc_possible_for_pair): Likewise.
5130 (rewrite_groups): Likewise.
5131 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5132 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5133 (get_alias_ptr_type_for_ptr_address): New function.
5134 (rewrite_use_address): Rewrite address uses in calls that were
5135 identified by find_address_like_use.
5137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5138 Alan Hayward <alan.hayward@arm.com>
5139 David Sherwood <david.sherwood@arm.com>
5141 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5143 * gimple-expr.h (is_gimple_addressable: Likewise.
5144 * gimple-expr.c (is_gimple_address): Likewise.
5145 * internal-fn.c (expand_call_mem_ref): New function.
5146 (expand_mask_load_optab_fn): Use it.
5147 (expand_mask_store_optab_fn): Likewise.
5149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5150 Alan Hayward <alan.hayward@arm.com>
5151 David Sherwood <david.sherwood@arm.com>
5153 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5154 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5155 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5156 (cond_umax@var{mode}): Document.
5157 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5158 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5159 (cond_umin_optab, cond_umax_optab): New optabs.
5160 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5161 (COND_IOR, COND_XOR): New internal functions.
5162 * internal-fn.h (get_conditional_internal_fn): Declare.
5163 * internal-fn.c (cond_binary_direct): New macro.
5164 (expand_cond_binary_optab_fn): Likewise.
5165 (direct_cond_binary_optab_supported_p): Likewise.
5166 (get_conditional_internal_fn): New function.
5167 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5168 Cope with reduction statements that are vectorized as calls rather
5170 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5171 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5172 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5173 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5174 (UNSPEC_COND_EOR): New unspecs.
5175 (optab): Add mappings for them.
5176 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5177 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
5184 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5185 * internal-fn.def (WHILE_ULT): New internal function.
5186 * internal-fn.h (direct_internal_fn_supported_p): New override
5187 that takes two types as argument.
5188 * internal-fn.c (while_direct): New macro.
5189 (expand_while_optab_fn): New function.
5190 (convert_optab_supported_p): Likewise.
5191 (direct_while_optab_supported_p): New macro.
5192 * wide-int.h (wi::udiv_ceil): New function.
5193 * tree-vectorizer.h (rgroup_masks): New structure.
5194 (vec_loop_masks): New typedef.
5195 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5197 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5198 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5199 (vect_max_vf): New function.
5200 (slpeel_make_loop_iterate_ntimes): Delete.
5201 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5202 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5203 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5204 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5205 internal-fn.h, stor-layout.h and optabs-query.h.
5206 (vect_set_loop_mask): New function.
5207 (add_preheader_seq): Likewise.
5208 (add_header_seq): Likewise.
5209 (interleave_supported_p): Likewise.
5210 (vect_maybe_permute_loop_masks): Likewise.
5211 (vect_set_loop_masks_directly): Likewise.
5212 (vect_set_loop_condition_masked): Likewise.
5213 (vect_set_loop_condition_unmasked): New function, split out from
5214 slpeel_make_loop_iterate_ntimes.
5215 (slpeel_make_loop_iterate_ntimes): Rename to..
5216 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5217 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5218 (vect_do_peeling): Update call accordingly.
5219 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5221 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5222 mask_compare_type, can_fully_mask_p and fully_masked_p.
5223 (release_vec_loop_masks): New function.
5224 (_loop_vec_info): Use it to free the loop masks.
5225 (can_produce_all_loop_masks_p): New function.
5226 (vect_get_max_nscalars_per_iter): Likewise.
5227 (vect_verify_full_masking): Likewise.
5228 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5229 retries, and free the mask rgroups before retrying. Check loop-wide
5230 reasons for disallowing fully-masked loops. Make the final decision
5231 about whether use a fully-masked loop or not.
5232 (vect_estimate_min_profitable_iters): Do not assume that peeling
5233 for the number of iterations will be needed for fully-masked loops.
5234 (vectorizable_reduction): Disable fully-masked loops.
5235 (vectorizable_live_operation): Likewise.
5236 (vect_halve_mask_nunits): New function.
5237 (vect_double_mask_nunits): Likewise.
5238 (vect_record_loop_mask): Likewise.
5239 (vect_get_loop_mask): Likewise.
5240 (vect_transform_loop): Handle the case in which the final loop
5241 iteration might handle a partial vector. Call vect_set_loop_condition
5242 instead of slpeel_make_loop_iterate_ntimes.
5243 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5244 (check_load_store_masking): New function.
5245 (prepare_load_store_mask): Likewise.
5246 (vectorizable_store): Handle fully-masked loops.
5247 (vectorizable_load): Likewise.
5248 (supportable_widening_operation): Use vect_halve_mask_nunits for
5250 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5251 (vect_gen_while): New function.
5252 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5253 (aarch64_uqdec<mode>): New insn.
5255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5256 Alan Hayward <alan.hayward@arm.com>
5257 David Sherwood <david.sherwood@arm.com>
5259 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5260 (reduc_xor_scal_optab): New optabs.
5261 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5262 (reduc_xor_scal_@var{m}): Document.
5263 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5264 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5266 * fold-const-call.c (fold_const_call): Handle them.
5267 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5268 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5269 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5270 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5271 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5272 (UNSPEC_XORV): New unspecs.
5273 (optab): Add entries for them.
5274 (BITWISEV): New int iterator.
5275 (bit_reduc_op): New int attributes.
5277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5278 Alan Hayward <alan.hayward@arm.com>
5279 David Sherwood <david.sherwood@arm.com>
5281 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5282 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5283 * optabs.def (vec_shl_insert_optab): New optab.
5284 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5285 (duplicate_and_interleave): Likewise.
5286 * tree-vect-loop.c: Include internal-fn.h.
5287 (neutral_op_for_slp_reduction): New function, split out from
5288 get_initial_defs_for_reduction.
5289 (get_initial_def_for_reduction): Handle option 2 for variable-length
5290 vectors by loading the neutral value into a vector and then shifting
5291 the initial value into element 0.
5292 (get_initial_defs_for_reduction): Replace the code argument with
5293 the neutral value calculated by neutral_op_for_slp_reduction.
5294 Use gimple_build_vector for constant-length vectors.
5295 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5296 but the first group_size elements have a neutral value.
5297 Use duplicate_and_interleave otherwise.
5298 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5299 Update call to get_initial_defs_for_reduction. Handle SLP
5300 reductions for variable-length vectors by creating one vector
5301 result for each scalar result, with the elements associated
5302 with other scalar results stubbed out with the neutral value.
5303 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5304 Require IFN_VEC_SHL_INSERT for double reductions on
5305 variable-length vectors, or SLP reductions that have
5306 a neutral value. Require can_duplicate_and_interleave_p
5307 support for variable-length unchained SLP reductions if there
5308 is no neutral value, such as for MIN/MAX reductions. Also require
5309 the number of vector elements to be a multiple of the number of
5310 SLP statements when doing variable-length unchained SLP reductions.
5311 Update call to vect_create_epilog_for_reduction.
5312 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5313 and remove initial values.
5314 (duplicate_and_interleave): Make public.
5315 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5316 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5319 Alan Hayward <alan.hayward@arm.com>
5320 David Sherwood <david.sherwood@arm.com>
5322 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5323 (can_duplicate_and_interleave_p): New function.
5324 (vect_get_and_check_slp_defs): Take the vector of statements
5325 rather than just the current one. Remove excess parentheses.
5326 Restriction rejectinon of vect_constant_def and vect_external_def
5327 for variable-length vectors to boolean types, or types for which
5328 can_duplicate_and_interleave_p is false.
5329 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5330 (duplicate_and_interleave): New function.
5331 (vect_get_constant_vectors): Use gimple_build_vector for
5332 constant-length vectors and suitable variable-length constant
5333 vectors. Use duplicate_and_interleave for other variable-length
5334 vectors. Don't defer the update when inserting new statements.
5336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5337 Alan Hayward <alan.hayward@arm.com>
5338 David Sherwood <david.sherwood@arm.com>
5340 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5341 min_profitable_iters doesn't go negative.
5343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5344 Alan Hayward <alan.hayward@arm.com>
5345 David Sherwood <david.sherwood@arm.com>
5347 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5348 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5349 * optabs.def (vec_mask_load_lanes_optab): New optab.
5350 (vec_mask_store_lanes_optab): Likewise.
5351 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5352 (MASK_STORE_LANES): Likewise.
5353 * internal-fn.c (mask_load_lanes_direct): New macro.
5354 (mask_store_lanes_direct): Likewise.
5355 (expand_mask_load_optab_fn): Handle masked operations.
5356 (expand_mask_load_lanes_optab_fn): New macro.
5357 (expand_mask_store_optab_fn): Handle masked operations.
5358 (expand_mask_store_lanes_optab_fn): New macro.
5359 (direct_mask_load_lanes_optab_supported_p): Likewise.
5360 (direct_mask_store_lanes_optab_supported_p): Likewise.
5361 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5363 (vect_load_lanes_supported): Likewise.
5364 * tree-vect-data-refs.c (strip_conversion): New function.
5365 (can_group_stmts_p): Likewise.
5366 (vect_analyze_data_ref_accesses): Use it instead of checking
5367 for a pair of assignments.
5368 (vect_store_lanes_supported): Take a masked_p parameter.
5369 (vect_load_lanes_supported): Likewise.
5370 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5371 vect_store_lanes_supported and vect_load_lanes_supported.
5372 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5373 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5374 parameter. Don't allow gaps for masked accesses.
5375 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5376 and vect_load_lanes_supported.
5377 (get_load_store_type): Take a masked_p parameter and update
5378 call to get_group_load_store_type.
5379 (vectorizable_store): Update call to get_load_store_type.
5380 Handle IFN_MASK_STORE_LANES.
5381 (vectorizable_load): Update call to get_load_store_type.
5382 Handle IFN_MASK_LOAD_LANES.
5384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5385 Alan Hayward <alan.hayward@arm.com>
5386 David Sherwood <david.sherwood@arm.com>
5388 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5390 * config/aarch64/aarch64-protos.h
5391 (aarch64_sve_struct_memory_operand_p): Declare.
5392 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5393 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5394 (VPRED, vpred): Handle SVE structure modes.
5395 * config/aarch64/constraints.md (Utx): New constraint.
5396 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5397 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5398 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5399 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5400 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5401 structure modes. Split into pieces after RA.
5402 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5403 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5405 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5406 SVE structure modes.
5407 (aarch64_classify_address): Likewise.
5408 (sizetochar): Move earlier in file.
5409 (aarch64_print_operand): Handle SVE register lists.
5410 (aarch64_array_mode): New function.
5411 (aarch64_sve_struct_memory_operand_p): Likewise.
5412 (TARGET_ARRAY_MODE): Redefine.
5414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5415 Alan Hayward <alan.hayward@arm.com>
5416 David Sherwood <david.sherwood@arm.com>
5418 * target.def (array_mode): New target hook.
5419 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5420 * doc/tm.texi: Regenerate.
5421 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5422 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5423 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5425 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5429 Alan Hayward <alan.hayward@arm.com>
5430 David Sherwood <david.sherwood@arm.com>
5432 * fold-const.c (fold_binary_loc): Check the argument types
5433 rather than the result type when testing for a vector operation.
5435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5437 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5438 * doc/tm.texi: Regenerate.
5440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5441 Alan Hayward <alan.hayward@arm.com>
5442 David Sherwood <david.sherwood@arm.com>
5444 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5445 (sve): Document new AArch64 extension.
5446 * doc/md.texi (w): Extend the description of the AArch64
5447 constraint to include SVE vectors.
5448 (Upl, Upa): Document new AArch64 predicate constraints.
5449 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5451 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5452 (msve-vector-bits=): New option.
5453 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5454 SVE when these are disabled.
5455 (sve): New extension.
5456 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5457 modes. Adjust their number of units based on aarch64_sve_vg.
5458 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5459 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5460 aarch64_addr_query_type.
5461 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5462 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5463 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5464 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5465 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5466 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5467 (aarch64_simd_imm_zero_p): Delete.
5468 (aarch64_check_zero_based_sve_index_immediate): Declare.
5469 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5470 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5471 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5472 (aarch64_sve_float_mul_immediate_p): Likewise.
5473 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5475 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5476 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5477 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5478 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5479 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5480 (aarch64_regmode_natural_size): Likewise.
5481 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5482 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5484 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5485 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5486 for VG and the SVE predicate registers.
5487 (V_ALIASES): Add a "z"-prefixed alias.
5488 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5489 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5490 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5491 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5492 (REG_CLASS_NAMES): Add entries for them.
5493 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5494 and the predicate registers.
5495 (aarch64_sve_vg): Declare.
5496 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5497 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5498 (REGMODE_NATURAL_SIZE): Define.
5499 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5501 * config/aarch64/aarch64.c: Include cfgrtl.h.
5502 (simd_immediate_info): Add a constructor for series vectors,
5503 and an associated step field.
5504 (aarch64_sve_vg): New variable.
5505 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5506 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5507 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5508 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5509 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5510 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5511 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5512 (aarch64_get_mask_mode): New functions.
5513 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5514 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5515 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5516 predicate modes and predicate registers. Explicitly restrict
5517 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5518 to store a vector mode if it is recognized by
5519 aarch64_classify_vector_mode.
5520 (aarch64_regmode_natural_size): New function.
5521 (aarch64_hard_regno_caller_save_mode): Return the original mode
5523 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5524 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5525 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5526 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5528 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5529 does not overlap dest if the function is frame-related. Handle
5531 (aarch64_split_add_offset): New function.
5532 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5533 them aarch64_add_offset.
5534 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5535 and update call to aarch64_sub_sp.
5536 (aarch64_add_cfa_expression): New function.
5537 (aarch64_expand_prologue): Pass extra temporary registers to the
5538 functions above. Handle the case in which we need to emit new
5539 DW_CFA_expressions for registers that were originally saved
5540 relative to the stack pointer, but now have to be expressed
5541 relative to the frame pointer.
5542 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5544 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5545 IP0 and IP1 values for SVE frames.
5546 (aarch64_expand_vec_series): New function.
5547 (aarch64_expand_sve_widened_duplicate): Likewise.
5548 (aarch64_expand_sve_const_vector): Likewise.
5549 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5550 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5551 into the register, rather than emitting a SET directly.
5552 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5553 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5554 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5555 (offset_9bit_signed_scaled_p): New functions.
5556 (aarch64_replicate_bitmask_imm): New function.
5557 (aarch64_bitmask_imm): Use it.
5558 (aarch64_cannot_force_const_mem): Reject expressions involving
5559 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5560 (aarch64_classify_index): Handle SVE indices, by requiring
5561 a plain register index with a scale that matches the element size.
5562 (aarch64_classify_address): Handle SVE addresses. Assert that
5563 the mode of the address is VOIDmode or an integer mode.
5564 Update call to aarch64_classify_symbol.
5565 (aarch64_classify_symbolic_expression): Update call to
5566 aarch64_classify_symbol.
5567 (aarch64_const_vec_all_in_range_p): New function.
5568 (aarch64_print_vector_float_operand): Likewise.
5569 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5570 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5571 and the FP immediates 1.0 and 0.5.
5572 (aarch64_print_address_internal): Handle SVE addresses.
5573 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5574 (aarch64_regno_regclass): Handle predicate registers.
5575 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5577 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5578 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5579 (aarch64_convert_sve_vector_bits): New function.
5580 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5581 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5583 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5584 Handle SVE vector and predicate modes. Accept VL-based constants
5585 that need only one temporary register, and VL offsets that require
5586 no temporary registers.
5587 (aarch64_conditional_register_usage): Mark the predicate registers
5588 as fixed if SVE isn't available.
5589 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5590 Return true for SVE vector and predicate modes.
5591 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5592 rather than an unsigned int. Handle SVE modes.
5593 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5595 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5597 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5598 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5599 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5600 (aarch64_sve_float_mul_immediate_p): New functions.
5601 (aarch64_sve_valid_immediate): New function.
5602 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5603 Explicitly reject structure modes. Check for INDEX constants.
5604 Handle PTRUE and PFALSE constants.
5605 (aarch64_check_zero_based_sve_index_immediate): New function.
5606 (aarch64_simd_imm_zero_p): Delete.
5607 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5608 vector modes. Accept constants in the range of CNT[BHWD].
5609 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5610 ask for an Advanced SIMD mode.
5611 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5612 (aarch64_simd_vector_alignment): Handle SVE predicates.
5613 (aarch64_vectorize_preferred_vector_alignment): New function.
5614 (aarch64_simd_vector_alignment_reachable): Use it instead of
5616 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5617 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5619 (MAX_VECT_LEN): Delete.
5620 (expand_vec_perm_d): Add a vec_flags field.
5621 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5622 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5623 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5625 (aarch64_evpc_rev): Rename to...
5626 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5627 (aarch64_evpc_rev_global): New function.
5628 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5629 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5631 (aarch64_evpc_sve_tbl): New function.
5632 (aarch64_expand_vec_perm_const_1): Update after rename of
5633 aarch64_evpc_rev. Handle SVE permutes too, trying
5634 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5635 than aarch64_evpc_tbl.
5636 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5637 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5638 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5639 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5640 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5641 (aarch64_expand_sve_vcond): New functions.
5642 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5643 of aarch64_vector_mode_p.
5644 (aarch64_dwarf_poly_indeterminate_value): New function.
5645 (aarch64_compute_pressure_classes): Likewise.
5646 (aarch64_can_change_mode_class): Likewise.
5647 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5648 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5649 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5650 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5651 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5652 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5653 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5654 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5656 (Dn, Dl, Dr): Accept const as well as const_vector.
5657 (Dz): Likewise. Compare against CONST0_RTX.
5658 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5659 of "vector" where appropriate.
5660 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5661 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5662 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5663 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5664 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5665 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5666 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5667 (v_int_equiv): Extend to SVE modes.
5668 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5670 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5671 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5672 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5673 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5674 (SVE_COND_FP_CMP): New int iterators.
5675 (perm_hilo): Handle the new unpack unspecs.
5676 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5678 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5679 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5680 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5681 (aarch64_equality_operator, aarch64_constant_vector_operand)
5682 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5683 (aarch64_sve_nonimmediate_operand): Likewise.
5684 (aarch64_sve_general_operand): Likewise.
5685 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5686 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5687 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5688 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5689 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5690 (aarch64_sve_float_arith_immediate): Likewise.
5691 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5692 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5693 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5694 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5695 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5696 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5697 (aarch64_sve_float_arith_operand): Likewise.
5698 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5699 (aarch64_sve_float_mul_operand): Likewise.
5700 (aarch64_sve_vec_perm_operand): Likewise.
5701 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5702 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5703 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5704 as well as const_vector.
5705 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5706 in file. Use CONST0_RTX and CONSTM1_RTX.
5707 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5708 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5709 Use aarch64_simd_imm_zero.
5710 * config/aarch64/aarch64-sve.md: New file.
5711 * config/aarch64/aarch64.md: Include it.
5712 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5713 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5714 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5715 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5716 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5717 (sve): New attribute.
5718 (enabled): Disable instructions with the sve attribute unless
5720 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5721 aarch64_expand_mov_immediate.
5722 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5723 CNT[BHSD] immediates.
5724 (movti): Split CONST_POLY_INT moves into two halves.
5725 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5726 Split additions that need a temporary here if the destination
5727 is the stack pointer.
5728 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5729 (*add<mode>3_poly_1): New instruction.
5730 (set_clobber_cc): New expander.
5732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5734 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5735 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5736 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5737 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5738 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5739 Change innermode from fixed_mode_size to machine_mode.
5740 (simplify_subreg): Update call accordingly. Handle a constant-sized
5741 subreg of a variable-length CONST_VECTOR.
5743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5744 Alan Hayward <alan.hayward@arm.com>
5745 David Sherwood <david.sherwood@arm.com>
5747 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5748 (add_offset_to_base): New function, split out from...
5749 (create_mem_ref): ...here. When handling a scale other than 1,
5750 check first whether the address is valid without the offset.
5751 Add it into the base if so, leaving the index and scale as-is.
5753 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5756 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5757 fold_for_warn before checking if arg2 is INTEGER_CST.
5759 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5761 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5762 (store_multiple_operation): Delete.
5763 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5764 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5765 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5766 guarded by TARGET_STRING.
5767 (rs6000_output_load_multiple): Delete.
5768 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5769 OPTION_MASK_STRING / TARGET_STRING handling.
5770 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5771 (const rs6000_opt_masks) <"string">: Change mask to 0.
5772 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5773 (MASK_STRING): Delete.
5774 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5776 (load_multiple): Delete.
5783 (store_multiple): Delete.
5790 (movmemsi_8reg): Delete.
5791 (corresponding unnamed define_insn): Delete.
5792 (movmemsi_6reg): Delete.
5793 (corresponding unnamed define_insn): Delete.
5794 (movmemsi_4reg): Delete.
5795 (corresponding unnamed define_insn): Delete.
5796 (movmemsi_2reg): Delete.
5797 (corresponding unnamed define_insn): Delete.
5798 (movmemsi_1reg): Delete.
5799 (corresponding unnamed define_insn): Delete.
5800 * config/rs6000/rs6000.opt (mno-string): New.
5801 (mstring): Replace by deprecation warning stub.
5802 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5804 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5806 * regrename.c (regrename_do_replace): If replacing the same
5807 reg multiple times, try to reuse last created gen_raw_REG.
5810 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5811 main to workaround a bug in GDB.
5813 2018-01-12 Tom de Vries <tom@codesourcery.com>
5816 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5818 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5820 PR rtl-optimization/80481
5821 * ira-color.c (get_cap_member): New function.
5822 (allocnos_conflict_by_live_ranges_p): Use it.
5823 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5824 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5826 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5829 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5830 (*saddl_se_1): Ditto.
5832 (*ssubl_se_1): Ditto.
5834 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5836 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5837 rather than wi::to_widest for DR_INITs.
5838 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5839 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5840 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5842 (vect_analyze_group_access_1): Note that here.
5844 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5846 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5847 polynomial type sizes.
5849 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5851 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5852 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5853 (gimple_add_tmp_var): Likewise.
5855 2018-01-12 Martin Liska <mliska@suse.cz>
5857 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5858 (gimple_alloc_sizes): Likewise.
5859 (dump_gimple_statistics): Use PRIu64 in printf format.
5860 * gimple.h: Change uint64_t to int.
5862 2018-01-12 Martin Liska <mliska@suse.cz>
5864 * tree-core.h: Use uint64_t instead of int.
5865 * tree.c (tree_node_counts): Likewise.
5866 (tree_node_sizes): Likewise.
5867 (dump_tree_statistics): Use PRIu64 in printf format.
5869 2018-01-12 Martin Liska <mliska@suse.cz>
5871 * Makefile.in: As qsort_chk is implemented in vec.c, add
5872 vec.o to linkage of gencfn-macros.
5873 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5874 passing the info to record_node_allocation_statistics.
5875 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5877 * ggc-common.c (struct ggc_usage): Add operator== and use
5878 it in operator< and compare function.
5879 * mem-stats.h (struct mem_usage): Likewise.
5880 * vec.c (struct vec_usage): Remove operator< and compare
5881 function. Can be simply inherited.
5883 2018-01-12 Martin Jambor <mjambor@suse.cz>
5886 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5887 * tree-ssa-math-opts.c: Include domwalk.h.
5888 (convert_mult_to_fma_1): New function.
5889 (fma_transformation_info): New type.
5890 (fma_deferring_state): Likewise.
5891 (cancel_fma_deferring): New function.
5892 (result_of_phi): Likewise.
5893 (last_fma_candidate_feeds_initial_phi): Likewise.
5894 (convert_mult_to_fma): Added deferring logic, split actual
5895 transformation to convert_mult_to_fma_1.
5896 (math_opts_dom_walker): New type.
5897 (math_opts_dom_walker::after_dom_children): New method, body moved
5898 here from pass_optimize_widening_mul::execute, added deferring logic
5900 (pass_optimize_widening_mul::execute): Moved most of code to
5901 math_opts_dom_walker::after_dom_children.
5902 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5903 * config/i386/i386.c (ix86_option_override_internal): Added
5904 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5906 2018-01-12 Richard Biener <rguenther@suse.de>
5909 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5910 inline instance vars.
5912 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5915 * config/rx/rx.c (rx_is_restricted_memory_address):
5918 2018-01-12 Richard Biener <rguenther@suse.de>
5920 PR tree-optimization/80846
5921 * target.def (split_reduction): New target hook.
5922 * targhooks.c (default_split_reduction): New function.
5923 * targhooks.h (default_split_reduction): Declare.
5924 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5925 target requests first reduce vectors by combining low and high
5927 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5928 (get_vectype_for_scalar_type_and_size): Export.
5929 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5930 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5931 * doc/tm.texi: Regenerate.
5932 * config/i386/i386.c (ix86_split_reduction): Implement
5933 TARGET_VECTORIZE_SPLIT_REDUCTION.
5935 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5938 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5939 in PIC mode except for TARGET_VXWORKS_RTP.
5940 * config/sparc/sparc.c: Include cfgrtl.h.
5941 (TARGET_INIT_PIC_REG): Define.
5942 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5943 (sparc_pic_register_p): New predicate.
5944 (sparc_legitimate_address_p): Use it.
5945 (sparc_legitimize_pic_address): Likewise.
5946 (sparc_delegitimize_address): Likewise.
5947 (sparc_mode_dependent_address_p): Likewise.
5948 (gen_load_pcrel_sym): Remove 4th parameter.
5949 (load_got_register): Adjust call to above. Remove obsolete stuff.
5950 (sparc_expand_prologue): Do not call load_got_register here.
5951 (sparc_flat_expand_prologue): Likewise.
5952 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5953 (sparc_use_pseudo_pic_reg): New function.
5954 (sparc_init_pic_reg): Likewise.
5955 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5956 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5958 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5960 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5961 Add item for branch_cost.
5963 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5965 PR rtl-optimization/83565
5966 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5967 not extend the result to a larger mode for rotate operations.
5968 (num_sign_bit_copies1): Likewise.
5970 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5973 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5975 Use values-Xc.o for -pedantic.
5976 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5978 2018-01-12 Martin Liska <mliska@suse.cz>
5981 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5983 (possible_polymorphic_call_targets): Use it.
5984 (ipa_devirt): Likewise.
5986 2018-01-12 Martin Liska <mliska@suse.cz>
5988 * profile-count.h (enum profile_quality): Use 0 as invalid
5989 enum value of profile_quality.
5991 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5993 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5994 -mext-string options.
5996 2018-01-12 Richard Biener <rguenther@suse.de>
5998 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5999 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6000 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6002 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6004 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6006 * configure.ac (--with-long-double-format): Add support for the
6007 configuration option to change the default long double format on
6009 * config.gcc (powerpc*-linux*-*): Likewise.
6010 * configure: Regenerate.
6011 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6012 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6013 used without modification.
6015 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6017 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6018 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6019 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6020 MISC_BUILTIN_SPEC_BARRIER.
6021 (rs6000_init_builtins): Likewise.
6022 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6024 (speculation_barrier): New define_insn.
6025 * doc/extend.texi: Document __builtin_speculation_barrier.
6027 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6030 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6031 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6032 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6034 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6035 integral modes instead of "ss" and "sd".
6036 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6037 vectors with 32-bit and 64-bit elements.
6038 (vecdupssescalarmodesuffix): New mode attribute.
6039 (vec_dup<mode>): Use it.
6041 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6044 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6045 frame if argument is passed on stack.
6047 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6050 * ree.c (combine_reaching_defs): Optimize also
6051 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6052 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6054 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6057 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6059 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6062 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6063 after they are computed.
6065 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6067 PR tree-optimization/83695
6068 * gimple-loop-linterchange.cc
6069 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6070 reset cached scev information after interchange.
6071 (pass_linterchange::execute): Remove call to scev_reset_htab.
6073 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6075 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6076 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6077 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6078 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6079 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6080 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6081 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6082 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6083 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6084 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6085 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6086 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6087 (V_lane_reg): Likewise.
6088 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6090 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6091 (vfmal_lane_low<mode>_intrinsic,
6092 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6093 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6094 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6095 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6096 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6097 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6099 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6101 * config/arm/arm-cpus.in (fp16fml): New feature.
6102 (ALL_SIMD): Add fp16fml.
6103 (armv8.2-a): Add fp16fml as an option.
6104 (armv8.3-a): Likewise.
6105 (armv8.4-a): Add fp16fml as part of fp16.
6106 * config/arm/arm.h (TARGET_FP16FML): Define.
6107 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6109 * config/arm/arm-modes.def (V2HF): Define.
6110 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6111 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6112 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6113 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6114 vfmsl_low, vfmsl_high): New set of builtins.
6115 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6116 (vfml_op): New code attribute.
6117 (VFMLHALVES): New int iterator.
6118 (VFML, VFMLSEL): New mode attributes.
6119 (V_reg): Define mapping for V2HF.
6120 (V_hi, V_lo): New mode attributes.
6121 (VF_constraint): Likewise.
6122 (vfml_half, vfml_half_selector): New int attributes.
6123 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6125 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6126 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6128 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6129 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6130 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6131 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6133 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6134 Document new effective target and option set.
6136 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6138 * config/arm/arm-cpus.in (armv8_4): New feature.
6139 (ARMv8_4a): New fgroup.
6140 (armv8.4-a): New arch.
6141 * config/arm/arm-tables.opt: Regenerate.
6142 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6143 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6144 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6145 Add matching rules for -march=armv8.4-a and extensions.
6146 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6148 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6151 * config/rx/rx.md (BW): New mode attribute.
6152 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6154 2018-01-11 Richard Biener <rguenther@suse.de>
6156 PR tree-optimization/83435
6157 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6158 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6159 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6161 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6162 Alan Hayward <alan.hayward@arm.com>
6163 David Sherwood <david.sherwood@arm.com>
6165 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6167 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6168 (aarch64_print_address_internal): Use it to check for a zero offset.
6170 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6171 Alan Hayward <alan.hayward@arm.com>
6172 David Sherwood <david.sherwood@arm.com>
6174 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6175 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6176 Return a poly_int64 rather than a HOST_WIDE_INT.
6177 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6178 rather than a HOST_WIDE_INT.
6179 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6180 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6181 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6182 final_offset from HOST_WIDE_INT to poly_int64.
6183 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6184 to_constant when getting the number of units in an Advanced SIMD
6186 (aarch64_builtin_vectorized_function): Check for a constant number
6188 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6190 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6191 attribute instead of GET_MODE_NUNITS.
6192 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6193 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6194 GET_MODE_SIZE for fixed-size registers.
6195 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6196 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6197 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6198 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6199 (aarch64_print_operand, aarch64_print_address_internal)
6200 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6201 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6202 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6203 Handle polynomial GET_MODE_SIZE.
6204 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6205 wider than SImode without modification.
6206 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6207 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6208 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6209 passing and returning SVE modes.
6210 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6211 rather than GEN_INT.
6212 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6213 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6214 (aarch64_allocate_and_probe_stack_space): Likewise.
6215 (aarch64_layout_frame): Cope with polynomial offsets.
6216 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6217 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6219 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6220 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6221 poly_int64 rather than a HOST_WIDE_INT.
6222 (aarch64_get_separate_components, aarch64_process_components)
6223 (aarch64_expand_prologue, aarch64_expand_epilogue)
6224 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6225 (aarch64_anchor_offset): New function, split out from...
6226 (aarch64_legitimize_address): ...here.
6227 (aarch64_builtin_vectorization_cost): Handle polynomial
6228 TYPE_VECTOR_SUBPARTS.
6229 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6231 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6232 number of elements from the PARALLEL rather than the mode.
6233 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6234 rather than GET_MODE_BITSIZE.
6235 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6236 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6237 (aarch64_expand_vec_perm_const_1): Handle polynomial
6238 d->perm.length () and d->perm elements.
6239 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6240 Apply to_constant to d->perm elements.
6241 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6242 polynomial CONST_VECTOR_NUNITS.
6243 (aarch64_move_pointer): Take amount as a poly_int64 rather
6245 (aarch64_progress_pointer): Avoid temporary variable.
6246 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6247 the mode attribute instead of GET_MODE.
6249 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6250 Alan Hayward <alan.hayward@arm.com>
6251 David Sherwood <david.sherwood@arm.com>
6253 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6254 x exists before using it.
6255 (aarch64_add_constant_internal): Rename to...
6256 (aarch64_add_offset_1): ...this. Replace regnum with separate
6257 src and dest rtxes. Handle the case in which they're different,
6258 including when the offset is zero. Replace scratchreg with an rtx.
6259 Use 2 additions if there is no spare register into which we can
6260 move a 16-bit constant.
6261 (aarch64_add_constant): Delete.
6262 (aarch64_add_offset): Replace reg with separate src and dest
6263 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6264 Use aarch64_add_offset_1.
6265 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6266 an rtx rather than an int. Take the delta as a poly_int64
6267 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6268 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6269 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6270 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6271 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6273 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6274 aarch64_add_constant.
6276 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6278 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6279 Use scalar_float_mode.
6281 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6283 * config/aarch64/aarch64-simd.md
6284 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6285 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6286 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6287 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6288 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6289 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6290 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6291 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6292 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6293 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6295 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6298 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6299 targ_options->x_arm_arch_string is non NULL.
6301 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6303 * config/aarch64/aarch64.h
6304 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6306 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6309 * expmed.c (emit_store_flag_force): Swap if const op0
6310 and change VOIDmode to mode of op0.
6312 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6314 PR rtl-optimization/83761
6315 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6316 than bytes to mode_for_size.
6318 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6321 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6322 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6325 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6328 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6329 when in layout mode.
6330 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6331 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6334 2018-01-10 Michael Collison <michael.collison@arm.com>
6336 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6337 * config/aarch64/aarch64-option-extension.def: Add
6338 AARCH64_OPT_EXTENSION of 'fp16fml'.
6339 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6340 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6341 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6342 * config/aarch64/constraints.md (Ui7): New constraint.
6343 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6344 (VFMLA_SEL_W): Ditto.
6347 (VFMLA16_LOW): New int iterator.
6348 (VFMLA16_HIGH): Ditto.
6349 (UNSPEC_FMLAL): New unspec.
6350 (UNSPEC_FMLSL): Ditto.
6351 (UNSPEC_FMLAL2): Ditto.
6352 (UNSPEC_FMLSL2): Ditto.
6353 (f16mac): New code attribute.
6354 * config/aarch64/aarch64-simd-builtins.def
6355 (aarch64_fmlal_lowv2sf): Ditto.
6356 (aarch64_fmlsl_lowv2sf): Ditto.
6357 (aarch64_fmlalq_lowv4sf): Ditto.
6358 (aarch64_fmlslq_lowv4sf): Ditto.
6359 (aarch64_fmlal_highv2sf): Ditto.
6360 (aarch64_fmlsl_highv2sf): Ditto.
6361 (aarch64_fmlalq_highv4sf): Ditto.
6362 (aarch64_fmlslq_highv4sf): Ditto.
6363 (aarch64_fmlal_lane_lowv2sf): Ditto.
6364 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6365 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6366 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6367 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6368 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6369 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6370 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6371 (aarch64_fmlal_lane_highv2sf): Ditto.
6372 (aarch64_fmlsl_lane_highv2sf): Ditto.
6373 (aarch64_fmlal_laneq_highv2sf): Ditto.
6374 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6375 (aarch64_fmlalq_lane_highv4sf): Ditto.
6376 (aarch64_fmlsl_lane_highv4sf): Ditto.
6377 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6378 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6379 * config/aarch64/aarch64-simd.md:
6380 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6381 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6382 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6383 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6384 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6385 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6386 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6387 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6388 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6389 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6390 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6391 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6392 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6393 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6394 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6395 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6396 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6397 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6398 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6399 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6400 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6401 (vfmlsl_low_u32): Ditto.
6402 (vfmlalq_low_u32): Ditto.
6403 (vfmlslq_low_u32): Ditto.
6404 (vfmlal_high_u32): Ditto.
6405 (vfmlsl_high_u32): Ditto.
6406 (vfmlalq_high_u32): Ditto.
6407 (vfmlslq_high_u32): Ditto.
6408 (vfmlal_lane_low_u32): Ditto.
6409 (vfmlsl_lane_low_u32): Ditto.
6410 (vfmlal_laneq_low_u32): Ditto.
6411 (vfmlsl_laneq_low_u32): Ditto.
6412 (vfmlalq_lane_low_u32): Ditto.
6413 (vfmlslq_lane_low_u32): Ditto.
6414 (vfmlalq_laneq_low_u32): Ditto.
6415 (vfmlslq_laneq_low_u32): Ditto.
6416 (vfmlal_lane_high_u32): Ditto.
6417 (vfmlsl_lane_high_u32): Ditto.
6418 (vfmlal_laneq_high_u32): Ditto.
6419 (vfmlsl_laneq_high_u32): Ditto.
6420 (vfmlalq_lane_high_u32): Ditto.
6421 (vfmlslq_lane_high_u32): Ditto.
6422 (vfmlalq_laneq_high_u32): Ditto.
6423 (vfmlslq_laneq_high_u32): Ditto.
6424 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6425 (AARCH64_FL_FOR_ARCH8_4): New.
6426 (AARCH64_ISA_F16FML): New ISA flag.
6427 (TARGET_F16FML): New feature flag for fp16fml.
6428 (doc/invoke.texi): Document new fp16fml option.
6430 2018-01-10 Michael Collison <michael.collison@arm.com>
6432 * config/aarch64/aarch64-builtins.c:
6433 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6434 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6435 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6436 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6437 (AARCH64_ISA_SHA3): New ISA flag.
6438 (TARGET_SHA3): New feature flag for sha3.
6439 * config/aarch64/iterators.md (sha512_op): New int attribute.
6440 (CRYPTO_SHA512): New int iterator.
6441 (UNSPEC_SHA512H): New unspec.
6442 (UNSPEC_SHA512H2): Ditto.
6443 (UNSPEC_SHA512SU0): Ditto.
6444 (UNSPEC_SHA512SU1): Ditto.
6445 * config/aarch64/aarch64-simd-builtins.def
6446 (aarch64_crypto_sha512hqv2di): New builtin.
6447 (aarch64_crypto_sha512h2qv2di): Ditto.
6448 (aarch64_crypto_sha512su0qv2di): Ditto.
6449 (aarch64_crypto_sha512su1qv2di): Ditto.
6450 (aarch64_eor3qv8hi): Ditto.
6451 (aarch64_rax1qv2di): Ditto.
6452 (aarch64_xarqv2di): Ditto.
6453 (aarch64_bcaxqv8hi): Ditto.
6454 * config/aarch64/aarch64-simd.md:
6455 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6456 (aarch64_crypto_sha512su0qv2di): Ditto.
6457 (aarch64_crypto_sha512su1qv2di): Ditto.
6458 (aarch64_eor3qv8hi): Ditto.
6459 (aarch64_rax1qv2di): Ditto.
6460 (aarch64_xarqv2di): Ditto.
6461 (aarch64_bcaxqv8hi): Ditto.
6462 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6463 (vsha512h2q_u64): Ditto.
6464 (vsha512su0q_u64): Ditto.
6465 (vsha512su1q_u64): Ditto.
6466 (veor3q_u16): Ditto.
6467 (vrax1q_u64): Ditto.
6469 (vbcaxq_u16): Ditto.
6470 * config/arm/types.md (crypto_sha512): New type attribute.
6471 (crypto_sha3): Ditto.
6472 (doc/invoke.texi): Document new sha3 option.
6474 2018-01-10 Michael Collison <michael.collison@arm.com>
6476 * config/aarch64/aarch64-builtins.c:
6477 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6478 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6479 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6480 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6481 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6482 (AARCH64_ISA_SM4): New ISA flag.
6483 (TARGET_SM4): New feature flag for sm4.
6484 * config/aarch64/aarch64-simd-builtins.def
6485 (aarch64_sm3ss1qv4si): Ditto.
6486 (aarch64_sm3tt1aq4si): Ditto.
6487 (aarch64_sm3tt1bq4si): Ditto.
6488 (aarch64_sm3tt2aq4si): Ditto.
6489 (aarch64_sm3tt2bq4si): Ditto.
6490 (aarch64_sm3partw1qv4si): Ditto.
6491 (aarch64_sm3partw2qv4si): Ditto.
6492 (aarch64_sm4eqv4si): Ditto.
6493 (aarch64_sm4ekeyqv4si): Ditto.
6494 * config/aarch64/aarch64-simd.md:
6495 (aarch64_sm3ss1qv4si): Ditto.
6496 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6497 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6498 (aarch64_sm4eqv4si): Ditto.
6499 (aarch64_sm4ekeyqv4si): Ditto.
6500 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6501 (sm3part_op): Ditto.
6502 (CRYPTO_SM3TT): Ditto.
6503 (CRYPTO_SM3PART): Ditto.
6504 (UNSPEC_SM3SS1): New unspec.
6505 (UNSPEC_SM3TT1A): Ditto.
6506 (UNSPEC_SM3TT1B): Ditto.
6507 (UNSPEC_SM3TT2A): Ditto.
6508 (UNSPEC_SM3TT2B): Ditto.
6509 (UNSPEC_SM3PARTW1): Ditto.
6510 (UNSPEC_SM3PARTW2): Ditto.
6511 (UNSPEC_SM4E): Ditto.
6512 (UNSPEC_SM4EKEY): Ditto.
6513 * config/aarch64/constraints.md (Ui2): New constraint.
6514 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6515 * config/arm/types.md (crypto_sm3): New type attribute.
6516 (crypto_sm4): Ditto.
6517 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6518 (vsm3tt1aq_u32): Ditto.
6519 (vsm3tt1bq_u32): Ditto.
6520 (vsm3tt2aq_u32): Ditto.
6521 (vsm3tt2bq_u32): Ditto.
6522 (vsm3partw1q_u32): Ditto.
6523 (vsm3partw2q_u32): Ditto.
6524 (vsm4eq_u32): Ditto.
6525 (vsm4ekeyq_u32): Ditto.
6526 (doc/invoke.texi): Document new sm4 option.
6528 2018-01-10 Michael Collison <michael.collison@arm.com>
6530 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6531 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6532 (AARCH64_FL_FOR_ARCH8_4): New.
6533 (AARCH64_FL_V8_4): New flag.
6534 (doc/invoke.texi): Document new armv8.4-a option.
6536 2018-01-10 Michael Collison <michael.collison@arm.com>
6538 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6539 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6540 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6541 * config/aarch64/aarch64-option-extension.def: Add
6542 AARCH64_OPT_EXTENSION of 'sha2'.
6543 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6544 (crypto): Disable sha2 and aes if crypto disabled.
6545 (crypto): Enable aes and sha2 if enabled.
6546 (simd): Disable sha2 and aes if simd disabled.
6547 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6549 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6550 (TARGET_SHA2): New feature flag for sha2.
6551 (TARGET_AES): New feature flag for aes.
6552 * config/aarch64/aarch64-simd.md:
6553 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6554 conditional on TARGET_AES.
6555 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6556 (aarch64_crypto_sha1hsi): Make pattern conditional
6558 (aarch64_crypto_sha1hv4si): Ditto.
6559 (aarch64_be_crypto_sha1hv4si): Ditto.
6560 (aarch64_crypto_sha1su1v4si): Ditto.
6561 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6562 (aarch64_crypto_sha1su0v4si): Ditto.
6563 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6564 (aarch64_crypto_sha256su0v4si): Ditto.
6565 (aarch64_crypto_sha256su1v4si): Ditto.
6566 (doc/invoke.texi): Document new aes and sha2 options.
6568 2018-01-10 Martin Sebor <msebor@redhat.com>
6570 PR tree-optimization/83781
6571 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6574 2018-01-11 Martin Sebor <msebor@gmail.com>
6575 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6577 PR tree-optimization/83501
6578 PR tree-optimization/81703
6580 * tree-ssa-strlen.c (get_string_cst): Rename...
6581 (get_string_len): ...to this. Handle global constants.
6582 (handle_char_store): Adjust.
6584 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6585 Jim Wilson <jimw@sifive.com>
6587 * config/riscv/riscv-protos.h (riscv_output_return): New.
6588 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6589 (riscv_attribute_table, riscv_output_return),
6590 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6591 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6592 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6594 (riscv_expand_prologue): Add early return for naked function.
6595 (riscv_expand_epilogue): Likewise.
6596 (riscv_function_ok_for_sibcall): Return false for naked function.
6597 (riscv_set_current_function): New.
6598 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6599 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6600 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6601 * doc/extend.texi (RISC-V Function Attributes): New.
6603 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6605 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6606 check for 128-bit long double before checking TCmode.
6607 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6608 128-bit long doubles before checking TFmode or TCmode.
6609 (FLOAT128_IBM_P): Likewise.
6611 2018-01-10 Martin Sebor <msebor@redhat.com>
6613 PR tree-optimization/83671
6614 * builtins.c (c_strlen): Unconditionally return zero for the empty
6616 Use -Warray-bounds for warnings.
6617 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6618 for non-constant array indices with COMPONENT_REF, arrays of
6619 arrays, and pointers to arrays.
6620 (gimple_fold_builtin_strlen): Determine and set length range for
6621 non-constant character arrays.
6623 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6626 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6629 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6631 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6633 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6636 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6637 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6638 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6639 indexed_or_indirect_operand predicate.
6640 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6641 (*vsx_le_perm_load_v8hi): Likewise.
6642 (*vsx_le_perm_load_v16qi): Likewise.
6643 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6644 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6645 (*vsx_le_perm_store_v8hi): Likewise.
6646 (*vsx_le_perm_store_v16qi): Likewise.
6647 (eight unnamed splitters): Likewise.
6649 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6651 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6652 * config/rs6000/emmintrin.h: Likewise.
6653 * config/rs6000/mmintrin.h: Likewise.
6654 * config/rs6000/xmmintrin.h: Likewise.
6656 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6659 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6661 * tree.c (tree_nop_conversion): Return true for location wrapper
6663 (maybe_wrap_with_location): New function.
6664 (selftest::check_strip_nops): New function.
6665 (selftest::test_location_wrappers): New function.
6666 (selftest::tree_c_tests): Call it.
6667 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6668 (maybe_wrap_with_location): New decl.
6669 (EXPR_LOCATION_WRAPPER_P): New macro.
6670 (location_wrapper_p): New inline function.
6671 (tree_strip_any_location_wrapper): New inline function.
6673 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6676 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6677 stack_realign_offset for the largest alignment of stack slot
6679 (ix86_find_max_used_stack_alignment): New function.
6680 (ix86_finalize_stack_frame_flags): Use it. Set
6681 max_used_stack_alignment if we don't realign stack.
6682 * config/i386/i386.h (machine_function): Add
6683 max_used_stack_alignment.
6685 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6687 * config/arm/arm.opt (-mbranch-cost): New option.
6688 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6691 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6694 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6695 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6697 2018-01-10 Richard Biener <rguenther@suse.de>
6700 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6701 early out so it also covers the case where we have a non-NULL
6704 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6706 PR tree-optimization/83753
6707 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6708 for non-strided grouped accesses if the number of elements is 1.
6710 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6713 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6714 * i386.h (TARGET_USE_GATHER): Define.
6715 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6717 2018-01-10 Martin Liska <mliska@suse.cz>
6720 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6721 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6723 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6724 CLEANUP_NO_PARTITIONING is not set.
6726 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6728 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6729 for vectors, as a partial revert of r254296.
6730 * rtl.h (const_vec_p): Delete.
6731 (const_vec_duplicate_p): Don't test for vector CONSTs.
6732 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6733 * expmed.c (make_tree): Likewise.
6736 * common.md (E, F): Use CONSTANT_P instead of checking for
6738 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6739 checking for CONST_VECTOR.
6741 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6744 * predict.c (force_edge_cold): Handle in more sane way edges
6747 2018-01-09 Carl Love <cel@us.ibm.com>
6749 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6751 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6752 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6753 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6754 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6755 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6756 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6757 * config/rs6000/rs6000-protos.h: Add extern defition for
6758 rs6000_generate_float2_double_code.
6759 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6761 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6762 (float2_v2df): Add define_expand.
6764 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6767 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6768 op_mode in the force_to_mode call.
6770 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6772 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6773 instead of checking each element individually.
6774 (aarch64_evpc_uzp): Likewise.
6775 (aarch64_evpc_zip): Likewise.
6776 (aarch64_evpc_ext): Likewise.
6777 (aarch64_evpc_rev): Likewise.
6778 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6779 instead of checking each element individually. Return true without
6781 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6782 whether all selected elements come from the same input, instead of
6783 checking each element individually. Remove calls to gen_rtx_REG,
6784 start_sequence and end_sequence and instead assert that no rtl is
6787 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6789 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6790 order of HIGH and CONST checks.
6792 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6794 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6795 if the destination isn't an SSA_NAME.
6797 2018-01-09 Richard Biener <rguenther@suse.de>
6799 PR tree-optimization/83668
6800 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6802 (canonicalize_loop_form): ... here, renamed from ...
6803 (canonicalize_loop_closed_ssa_form): ... this and amended to
6804 swap successor edges for loop exit blocks to make us use
6805 the RPO order we need for initial schedule generation.
6807 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6809 PR tree-optimization/64811
6810 * match.pd: When optimizing comparisons with Inf, avoid
6811 introducing or losing exceptions from comparisons with NaN.
6813 2018-01-09 Martin Liska <mliska@suse.cz>
6816 * asan.c (shadow_mem_size): Add gcc_assert.
6818 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6820 Don't save registers in main().
6823 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6824 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6825 * config/avr/avr.c (avr_set_current_function): Don't error if
6826 naked, OS_task or OS_main are specified at the same time.
6827 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6829 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6831 * common/config/avr/avr-common.c (avr_option_optimization_table):
6832 Switch on -mmain-is-OS_task for optimizing compilations.
6834 2018-01-09 Richard Biener <rguenther@suse.de>
6836 PR tree-optimization/83572
6837 * graphite.c: Include cfganal.h.
6838 (graphite_transform_loops): Connect infinite loops to exit
6839 and remove fake edges at the end.
6841 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6843 * ipa-inline.c (edge_badness): Revert accidental checkin.
6845 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6848 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6849 symbols; not inline clones.
6851 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6854 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6855 hard registers. Formatting fixes.
6857 PR preprocessor/83722
6858 * gcc.c (try_generate_repro): Pass
6859 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6860 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6863 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6864 Kito Cheng <kito.cheng@gmail.com>
6866 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6867 (riscv_leaf_function_p): Delete.
6868 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6870 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6872 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6874 (do_ifelse): New function.
6875 (do_isel): New function.
6876 (do_sub3): New function.
6877 (do_add3): New function.
6878 (do_load_mask_compare): New function.
6879 (do_overlap_load_compare): New function.
6880 (expand_compare_loop): New function.
6881 (expand_block_compare): Call expand_compare_loop() when appropriate.
6882 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6884 (-mblock-compare-inline-loop-limit): New option.
6886 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6889 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6890 Reverse order of second and third operands in first alternative.
6891 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6892 of first and second elements in UNSPEC_VPERMR vector.
6893 (altivec_expand_vec_perm_le): Likewise.
6895 2018-01-08 Jeff Law <law@redhat.com>
6897 PR rtl-optimizatin/81308
6898 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6899 (process_switch): If group_case_labels makes a change, then set
6901 (pass_convert_switch::execute): If a switch is converted, then
6902 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6904 PR rtl-optimization/81308
6905 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6908 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6910 PR target/83663 - Revert r255946
6911 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6912 generation for cases where splatting a value is not useful.
6913 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6914 across a vec_duplicate and a paradoxical subreg forming a vector
6915 mode to a vec_concat.
6917 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6919 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6920 -march=armv8.3-a variants.
6921 * config/arm/t-multilib: Likewise.
6922 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6924 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6926 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6928 (cceq_ior_compare_complement): Give it a name so I can use it, and
6929 change boolean_or_operator predicate to boolean_operator so it can
6930 be used to generate a crand.
6931 (eqne): New code iterator.
6932 (bd/bd_neg): New code_attrs.
6933 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6934 a single define_insn.
6935 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6936 decrement (bdnzt/bdnzf/bdzt/bdzf).
6937 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6938 with the new names of the branch decrement patterns, and added the
6939 names of the branch decrement conditional patterns.
6941 2018-01-08 Richard Biener <rguenther@suse.de>
6943 PR tree-optimization/83563
6944 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6947 2018-01-08 Richard Biener <rguenther@suse.de>
6950 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6952 2018-01-08 Richard Biener <rguenther@suse.de>
6954 PR tree-optimization/83685
6955 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6956 references to abnormals.
6958 2018-01-08 Richard Biener <rguenther@suse.de>
6961 * dwarf2out.c (output_indirect_strings): Handle empty
6962 skeleton_debug_str_hash.
6963 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6965 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6967 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6968 (emit_store_direct): Likewise.
6969 (arc_trampoline_adjust_address): Likewise.
6970 (arc_asm_trampoline_template): New function.
6971 (arc_initialize_trampoline): Use asm_trampoline_template.
6972 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6973 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6974 * config/arc/arc.md (flush_icache): Delete pattern.
6976 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6978 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6979 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6982 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6985 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6986 by not USED_FOR_TARGET.
6987 (make_pass_resolve_sw_modes): Likewise.
6989 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6991 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6994 2018-01-08 Richard Biener <rguenther@suse.de>
6997 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6999 2018-01-08 Richard Biener <rguenther@suse.de>
7002 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7004 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7007 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7008 basic blocks with a small number of successors.
7009 (convert_control_dep_chain_into_preds): Improve handling of
7011 (dump_predicates): Split apart into...
7012 (dump_pred_chain): ...here...
7013 (dump_pred_info): ...and here.
7014 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7015 (can_chain_union_be_invalidated_p): Improve check for invalidation
7017 (uninit_uses_cannot_happen): Avoid unnecessary if
7018 convert_control_dep_chain_into_preds yielded nothing.
7020 2018-01-06 Martin Sebor <msebor@redhat.com>
7022 PR tree-optimization/83640
7023 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7024 subtracting negative offset from size.
7025 (builtin_access::overlap): Adjust offset bounds of the access to fall
7026 within the size of the object if possible.
7028 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7030 PR rtl-optimization/83699
7031 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7032 extract_bit_field_as_subreg to cases in which the extracted
7033 value is also a vector.
7035 * lra-constraints.c (process_alt_operands): Test for the equivalence
7036 substitutions when detecting a possible reload cycle.
7038 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7041 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7042 by default if flag_selective_schedling{,2}. Formatting fixes.
7044 PR rtl-optimization/83682
7045 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7046 if it has non-VECTOR_MODE element mode.
7047 (vec_duplicate_p): Likewise.
7050 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7051 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7053 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7056 * config/i386/i386-builtin.def
7057 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7058 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7059 Require also OPTION_MASK_ISA_AVX512F in addition to
7060 OPTION_MASK_ISA_GFNI.
7061 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7062 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7063 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7064 to OPTION_MASK_ISA_GFNI.
7065 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7066 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7067 OPTION_MASK_ISA_AVX512BW.
7068 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7069 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7070 addition to OPTION_MASK_ISA_GFNI.
7071 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7072 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7073 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7074 to OPTION_MASK_ISA_GFNI.
7075 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7076 a requirement for all ISAs rather than any of them with a few
7078 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7080 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7081 bitmasks to be enabled with 3 exceptions, instead of requiring any
7082 enabled ISA with lots of exceptions.
7083 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7084 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7085 Change avx512bw in isa attribute to avx512f.
7086 * config/i386/sgxintrin.h: Add license boilerplate.
7087 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7088 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7089 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7090 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7092 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7093 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7094 temporarily sse2 rather than sse if not enabled already.
7097 * config/i386/sse.md (VI248_VLBW): Rename to ...
7098 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7099 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7100 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7101 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7102 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7103 mode iterator instead of VI248_VLBW.
7105 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7107 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7108 (record_modified): Skip clobbers; add debug output.
7109 (param_change_prob): Use sreal frequencies.
7111 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7113 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7114 punt for user-aligned variables.
7116 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7118 * tree-chrec.c (chrec_contains_symbols): Return true for
7121 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7124 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7125 of (x|y) == x for BICS pattern.
7127 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7129 PR tree-optimization/83605
7130 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7131 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7134 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7136 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7137 * config/epiphany/rtems.h: New file.
7139 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7140 Uros Bizjak <ubizjak@gmail.com>
7143 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7144 QIreg_operand instead of register_operand predicate.
7145 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7146 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7147 comments instead of -fmitigate[-_]rop.
7149 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7152 * cgraphunit.c (symbol_table::compile): Switch to text_section
7153 before calling assembly_start debug hook.
7154 * run-rtl-passes.c (run_rtl_passes): Likewise.
7157 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7159 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7160 range_int_cst_p rather than !symbolic_range_p before calling
7161 extract_range_from_multiplicative_op_1.
7163 2018-01-04 Jeff Law <law@redhat.com>
7165 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7166 redundant test in assertion.
7168 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7170 * doc/rtl.texi: Document machine_mode wrapper classes.
7172 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7174 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7177 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7179 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7180 the VEC_PERM_EXPR fold to fail.
7182 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7185 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7186 to switched_sections.
7188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7191 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7194 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7197 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7198 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7200 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7203 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7204 is BLKmode and bitpos not zero or mode change is needed.
7206 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7209 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7212 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7215 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7216 instead of MULT rtx. Update all corresponding splitters.
7218 (*ssub<modesuffix>): Ditto.
7220 (*cmp_sadd_di): Update split patterns.
7221 (*cmp_sadd_si): Ditto.
7222 (*cmp_sadd_sidi): Ditto.
7223 (*cmp_ssub_di): Ditto.
7224 (*cmp_ssub_si): Ditto.
7225 (*cmp_ssub_sidi): Ditto.
7226 * config/alpha/predicates.md (const23_operand): New predicate.
7227 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7228 Look for ASHIFT, not MULT inner operand.
7229 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7231 2018-01-04 Martin Liska <mliska@suse.cz>
7233 PR gcov-profile/83669
7234 * gcov.c (output_intermediate_file): Add version to intermediate
7236 * doc/gcov.texi: Document new field 'version' in intermediate
7237 file format. Fix location of '-k' option of gcov command.
7239 2018-01-04 Martin Liska <mliska@suse.cz>
7242 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7244 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7246 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7248 2018-01-03 Martin Sebor <msebor@redhat.com>
7250 PR tree-optimization/83655
7251 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7252 checking calls with invalid arguments.
7254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7256 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7257 (vectorizable_mask_load_store): Delete.
7258 (vectorizable_call): Return false for masked loads and stores.
7259 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7260 instead of gimple_assign_rhs1.
7261 (vectorizable_load): Handle IFN_MASK_LOAD.
7262 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7266 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7268 (vectorizable_mask_load_store): ...here.
7269 (vectorizable_load): ...and here.
7271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7273 * tree-vect-stmts.c (vect_build_all_ones_mask)
7274 (vect_build_zero_merge_argument): New functions, split out from...
7275 (vectorizable_load): ...here.
7277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7279 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7281 (vectorizable_mask_load_store): ...here.
7282 (vectorizable_store): ...and here.
7284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7286 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7288 (vectorizable_mask_load_store): ...here.
7290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7292 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7293 (vect_model_store_cost): Take a vec_load_store_type instead of a
7295 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7296 (vect_model_store_cost): Take a vec_load_store_type instead of a
7298 (vectorizable_mask_load_store): Update accordingly.
7299 (vectorizable_store): Likewise.
7300 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7304 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7305 IFN_MASK_LOAD calls here rather than...
7306 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7309 Alan Hayward <alan.hayward@arm.com>
7310 David Sherwood <david.sherwood@arm.com>
7312 * expmed.c (extract_bit_field_1): For vector extracts,
7313 fall back to extract_bit_field_as_subreg if vec_extract
7316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7317 Alan Hayward <alan.hayward@arm.com>
7318 David Sherwood <david.sherwood@arm.com>
7320 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7321 they are variable or constant sized.
7322 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7323 slots for constant-sized data.
7325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7326 Alan Hayward <alan.hayward@arm.com>
7327 David Sherwood <david.sherwood@arm.com>
7329 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7330 handling COND_EXPRs with boolean comparisons, try to find a better
7331 basis for the mask type than the boolean itself.
7333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7335 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7336 is calculated and how it can be overridden.
7337 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7338 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7340 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7344 Alan Hayward <alan.hayward@arm.com>
7345 David Sherwood <david.sherwood@arm.com>
7347 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7348 Remove the mode argument.
7349 (aarch64_simd_valid_immediate): Remove the mode and inverse
7351 * config/aarch64/iterators.md (bitsize): New iterator.
7352 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7353 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7354 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7355 aarch64_simd_valid_immediate.
7356 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7357 (aarch64_reg_or_bic_imm): Likewise.
7358 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7359 with an insn_type enum and msl with a modifier_type enum.
7360 Replace element_width with a scalar_mode. Change the shift
7361 to unsigned int. Add constructors for scalar_float_mode and
7362 scalar_int_mode elements.
7363 (aarch64_vect_float_const_representable_p): Delete.
7364 (aarch64_can_const_movi_rtx_p)
7365 (aarch64_simd_scalar_immediate_valid_for_move)
7366 (aarch64_simd_make_constant): Update call to
7367 aarch64_simd_valid_immediate.
7368 (aarch64_advsimd_valid_immediate_hs): New function.
7369 (aarch64_advsimd_valid_immediate): Likewise.
7370 (aarch64_simd_valid_immediate): Remove mode and inverse
7371 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7372 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7373 and aarch64_float_const_representable_p on the result.
7374 (aarch64_output_simd_mov_immediate): Remove mode argument.
7375 Update call to aarch64_simd_valid_immediate and use of
7376 simd_immediate_info.
7377 (aarch64_output_scalar_simd_mov_immediate): Update call
7380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7381 Alan Hayward <alan.hayward@arm.com>
7382 David Sherwood <david.sherwood@arm.com>
7384 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7385 (mode_nunits): Likewise CONST_MODE_NUNITS.
7386 * machmode.def (ADJUST_NUNITS): Document.
7387 * genmodes.c (mode_data::need_nunits_adj): New field.
7388 (blank_mode): Update accordingly.
7389 (adj_nunits): New variable.
7390 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7392 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7393 listed in adj_nunits.
7394 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7395 listed in adj_nunits. Don't emit case statements for such modes.
7396 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7397 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7398 nothing if adj_nunits is nonnull.
7399 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7400 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7401 (emit_mode_fbit): Update use of print_maybe_const_decl.
7402 (emit_move_size): Likewise. Treat the array as non-const
7404 (emit_mode_adjustments): Handle adj_nunits.
7406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7408 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7409 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7410 (VECTOR_MODES): Use it.
7411 (make_vector_modes): Take the prefix as an argument.
7413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7414 Alan Hayward <alan.hayward@arm.com>
7415 David Sherwood <david.sherwood@arm.com>
7417 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7418 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7419 for MODE_VECTOR_BOOL.
7420 * machmode.def (VECTOR_BOOL_MODE): Document.
7421 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7422 (make_vector_bool_mode): New function.
7423 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7425 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7426 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7428 * stor-layout.c (int_mode_for_mode): Likewise.
7429 * tree.c (build_vector_type_for_mode): Likewise.
7430 * varasm.c (output_constant_pool_2): Likewise.
7431 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7432 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7433 for MODE_VECTOR_BOOL.
7434 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7435 of mode class checks.
7436 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7437 instead of a list of mode class checks.
7438 (expand_vector_scalar_condition): Likewise.
7439 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7442 Alan Hayward <alan.hayward@arm.com>
7443 David Sherwood <david.sherwood@arm.com>
7445 * machmode.h (mode_size): Change from unsigned short to
7447 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7448 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7449 or if measurement_type is not polynomial.
7450 (fixed_size_mode::includes_p): Check for constant-sized modes.
7451 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7452 return a poly_uint16 rather than an unsigned short.
7453 (emit_mode_size): Change the type of mode_size from unsigned short
7454 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7455 (emit_mode_adjustments): Cope with polynomial vector sizes.
7456 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7458 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7460 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7461 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7462 * caller-save.c (setup_save_areas): Likewise.
7463 (replace_reg_with_saved_mem): Likewise.
7464 * calls.c (emit_library_call_value_1): Likewise.
7465 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7466 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7467 (gen_lowpart_for_combine): Likewise.
7468 * convert.c (convert_to_integer_1): Likewise.
7469 * cse.c (equiv_constant, cse_insn): Likewise.
7470 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7471 (cselib_subst_to_values): Likewise.
7472 * dce.c (word_dce_process_block): Likewise.
7473 * df-problems.c (df_word_lr_mark_ref): Likewise.
7474 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7475 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7476 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7477 (rtl_for_decl_location): Likewise.
7478 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7479 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7480 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7481 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7482 (expand_expr_real_1): Likewise.
7483 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7484 (pad_below): Likewise.
7485 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7486 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7487 * ira.c (get_subreg_tracking_sizes): Likewise.
7488 * ira-build.c (ira_create_allocno_objects): Likewise.
7489 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7490 (ira_sort_regnos_for_alter_reg): Likewise.
7491 * ira-costs.c (record_operand_costs): Likewise.
7492 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7493 (resolve_simple_move): Likewise.
7494 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7495 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7496 (lra_constraints): Likewise.
7497 (CONST_POOL_OK_P): Reject variable-sized modes.
7498 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7499 (add_pseudo_to_slot, lra_spill): Likewise.
7500 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7501 * optabs-query.c (get_best_extraction_insn): Likewise.
7502 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7503 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7504 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7505 * recog.c (offsettable_address_addr_space_p): Likewise.
7506 * regcprop.c (maybe_mode_change): Likewise.
7507 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7508 * regrename.c (build_def_use): Likewise.
7509 * regstat.c (dump_reg_info): Likewise.
7510 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7511 (find_reloads, find_reloads_subreg_address): Likewise.
7512 * reload1.c (eliminate_regs_1): Likewise.
7513 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7514 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7515 (simplify_binary_operation_1, simplify_subreg): Likewise.
7516 * targhooks.c (default_function_arg_padding): Likewise.
7517 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7518 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7519 (verify_gimple_assign_ternary): Likewise.
7520 * tree-inline.c (estimate_move_cost): Likewise.
7521 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7522 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7523 (get_address_cost_ainc): Likewise.
7524 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7525 (vect_supportable_dr_alignment): Likewise.
7526 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7527 (vectorizable_reduction): Likewise.
7528 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7529 (vectorizable_operation, vectorizable_load): Likewise.
7530 * tree.c (build_same_sized_truth_vector_type): Likewise.
7531 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7532 * var-tracking.c (emit_note_insn_var_location): Likewise.
7533 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7534 (ADDR_VEC_ALIGN): Likewise.
7536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7537 Alan Hayward <alan.hayward@arm.com>
7538 David Sherwood <david.sherwood@arm.com>
7540 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7542 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7543 or if measurement_type is polynomial.
7544 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7545 * combine.c (make_extraction): Likewise.
7546 * dse.c (find_shift_sequence): Likewise.
7547 * dwarf2out.c (mem_loc_descriptor): Likewise.
7548 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7549 (extract_bit_field, extract_low_bits): Likewise.
7550 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7551 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7552 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7553 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7554 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7555 * reload.c (find_reloads): Likewise.
7556 * reload1.c (alter_reg): Likewise.
7557 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7558 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7559 * tree-if-conv.c (predicate_mem_writes): Likewise.
7560 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7561 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7562 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7563 * valtrack.c (dead_debug_insert_temp): Likewise.
7564 * varasm.c (mergeable_constant_section): Likewise.
7565 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * expr.c (expand_assignment): Cope with polynomial mode sizes
7572 when assigning to a CONCAT.
7574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7575 Alan Hayward <alan.hayward@arm.com>
7576 David Sherwood <david.sherwood@arm.com>
7578 * machmode.h (mode_precision): Change from unsigned short to
7580 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7582 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7583 or if measurement_type is not polynomial.
7584 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7585 in which the mode is already known to be a scalar_int_mode.
7586 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7587 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7589 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7590 for GET_MODE_PRECISION.
7591 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7592 for GET_MODE_PRECISION.
7593 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7595 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7596 (expand_field_assignment, make_extraction): Likewise.
7597 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7598 (get_last_value): Likewise.
7599 * convert.c (convert_to_integer_1): Likewise.
7600 * cse.c (cse_insn): Likewise.
7601 * expr.c (expand_expr_real_1): Likewise.
7602 * lra-constraints.c (simplify_operand_subreg): Likewise.
7603 * optabs-query.c (can_atomic_load_p): Likewise.
7604 * optabs.c (expand_atomic_load): Likewise.
7605 (expand_atomic_store): Likewise.
7606 * ree.c (combine_reaching_defs): Likewise.
7607 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7608 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7609 * tree.h (type_has_mode_precision_p): Likewise.
7610 * ubsan.c (instrument_si_overflow): Likewise.
7612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7613 Alan Hayward <alan.hayward@arm.com>
7614 David Sherwood <david.sherwood@arm.com>
7616 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7617 polynomial numbers of units.
7618 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7619 (valid_vector_subparts_p): New function.
7620 (build_vector_type): Remove temporary shim and take the number
7621 of units as a poly_uint64 rather than an int.
7622 (build_opaque_vector_type): Take the number of units as a
7623 poly_uint64 rather than an int.
7624 * tree.c (build_vector_from_ctor): Handle polynomial
7625 TYPE_VECTOR_SUBPARTS.
7626 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7627 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7628 (build_vector_from_val): If the number of units is variable,
7629 use build_vec_duplicate_cst for constant operands and
7630 VEC_DUPLICATE_EXPR otherwise.
7631 (make_vector_type): Remove temporary is_constant ().
7632 (build_vector_type, build_opaque_vector_type): Take the number of
7633 units as a poly_uint64 rather than an int.
7634 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7636 * cfgexpand.c (expand_debug_expr): Likewise.
7637 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7638 (store_constructor, expand_expr_real_1): Likewise.
7639 (const_scalar_mask_from_tree): Likewise.
7640 * fold-const-call.c (fold_const_reduction): Likewise.
7641 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7642 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7643 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7644 (fold_relational_const): Likewise.
7645 (native_interpret_vector): Likewise. Change the size from an
7646 int to an unsigned int.
7647 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7648 TYPE_VECTOR_SUBPARTS.
7649 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7650 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7651 duplicating a non-constant operand into a variable-length vector.
7652 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7653 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7654 * ipa-icf.c (sem_variable::equals): Likewise.
7655 * match.pd: Likewise.
7656 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7657 * print-tree.c (print_node): Likewise.
7658 * stor-layout.c (layout_type): Likewise.
7659 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7660 * tree-cfg.c (verify_gimple_comparison): Likewise.
7661 (verify_gimple_assign_binary): Likewise.
7662 (verify_gimple_assign_ternary): Likewise.
7663 (verify_gimple_assign_single): Likewise.
7664 * tree-pretty-print.c (dump_generic_node): Likewise.
7665 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7666 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7667 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7668 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7669 (vect_shift_permute_load_chain): Likewise.
7670 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7671 (expand_vector_condition, optimize_vector_constructor): Likewise.
7672 (lower_vec_perm, get_compute_type): Likewise.
7673 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7674 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7675 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7676 (vect_recog_mask_conversion_pattern): Likewise.
7677 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7678 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7679 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7680 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7681 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7682 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7683 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7684 (supportable_widening_operation): Likewise.
7685 (supportable_narrowing_operation): Likewise.
7686 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7688 * varasm.c (output_constant): Likewise.
7690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7691 Alan Hayward <alan.hayward@arm.com>
7692 David Sherwood <david.sherwood@arm.com>
7694 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7695 so that both the length == 3 and length != 3 cases set up their
7696 own permute vectors. Add comments explaining why we know the
7697 number of elements is constant.
7698 (vect_permute_load_chain): Likewise.
7700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7701 Alan Hayward <alan.hayward@arm.com>
7702 David Sherwood <david.sherwood@arm.com>
7704 * machmode.h (mode_nunits): Change from unsigned char to
7706 (ONLY_FIXED_SIZE_MODES): New macro.
7707 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7708 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7709 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7711 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7712 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7713 or if measurement_type is not polynomial.
7714 * genmodes.c (ZERO_COEFFS): New macro.
7715 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7717 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7718 Use ZERO_COEFFS when emitting initializers.
7719 * data-streamer.h (bp_pack_poly_value): New function.
7720 (bp_unpack_poly_value): Likewise.
7721 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7722 for GET_MODE_NUNITS.
7723 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7724 for GET_MODE_NUNITS.
7725 * tree.c (make_vector_type): Remove temporary shim and make
7726 the real function take the number of units as a poly_uint64
7728 (build_vector_type_for_mode): Handle polynomial nunits.
7729 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7730 * emit-rtl.c (const_vec_series_p_1): Likewise.
7731 (gen_rtx_CONST_VECTOR): Likewise.
7732 * fold-const.c (test_vec_duplicate_folding): Likewise.
7733 * genrecog.c (validate_pattern): Likewise.
7734 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7735 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7736 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7737 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7738 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7739 * rtlanal.c (subreg_get_info): Likewise.
7740 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7741 (vect_grouped_load_supported): Likewise.
7742 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7743 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7744 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7745 (simplify_const_unary_operation, simplify_binary_operation_1)
7746 (simplify_const_binary_operation, simplify_ternary_operation)
7747 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7748 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7749 instead of CONST_VECTOR_NUNITS.
7750 * varasm.c (output_constant_pool_2): Likewise.
7751 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7752 explicit-encoded elements in the XVEC for variable-length vectors.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7756 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7759 Alan Hayward <alan.hayward@arm.com>
7760 David Sherwood <david.sherwood@arm.com>
7762 * coretypes.h (fixed_size_mode): Declare.
7763 (fixed_size_mode_pod): New typedef.
7764 * builtins.h (target_builtins::x_apply_args_mode)
7765 (target_builtins::x_apply_result_mode): Change type to
7766 fixed_size_mode_pod.
7767 * builtins.c (apply_args_size, apply_result_size, result_vector)
7768 (expand_builtin_apply_args_1, expand_builtin_apply)
7769 (expand_builtin_return): Update accordingly.
7771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7773 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7774 * cselib.c (cselib_hash_rtx): Likewise.
7775 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7776 CONST_VECTOR encoding.
7778 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7779 Jeff Law <law@redhat.com>
7782 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7783 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7784 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7785 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7788 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7789 explicitly probe *sp in a noreturn function if there were any callee
7790 register saves or frame pointer is needed.
7792 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7795 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7796 BLKmode for ternary, binary or unary expressions.
7799 * var-tracking.c (delete_vta_debug_insn): New inline function.
7800 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7801 insns from get_insns () to NULL instead of each bb separately.
7802 Use delete_vta_debug_insn. No longer static.
7803 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7804 delete_vta_debug_insns callers.
7805 * rtl.h (delete_vta_debug_insns): Declare.
7806 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7807 instead of variable_tracking_main.
7809 2018-01-03 Martin Sebor <msebor@redhat.com>
7811 PR tree-optimization/83603
7812 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7813 arguments past the endof the argument list in functions declared
7814 without a prototype.
7815 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7816 Avoid checking when arguments are null.
7818 2018-01-03 Martin Sebor <msebor@redhat.com>
7821 * doc/extend.texi (attribute const): Fix a typo.
7822 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7823 issuing -Wsuggest-attribute for void functions.
7825 2018-01-03 Martin Sebor <msebor@redhat.com>
7827 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7828 offset_int::from instead of wide_int::to_shwi.
7829 (maybe_diag_overlap): Remove assertion.
7830 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7831 * gimple-ssa-sprintf.c (format_directive): Same.
7832 (parse_directive): Same.
7833 (sprintf_dom_walker::compute_format_length): Same.
7834 (try_substitute_return_value): Same.
7836 2018-01-03 Jeff Law <law@redhat.com>
7839 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7840 non-constant residual for zero at runtime and avoid probing in
7841 that case. Reorganize code for trailing problem to mirror handling
7844 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7846 PR tree-optimization/83501
7847 * tree-ssa-strlen.c (get_string_cst): New.
7848 (handle_char_store): Call get_string_cst.
7850 2018-01-03 Martin Liska <mliska@suse.cz>
7852 PR tree-optimization/83593
7853 * tree-ssa-strlen.c: Include tree-cfg.h.
7854 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7855 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7856 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7858 (strlen_dom_walker::before_dom_children): Call
7859 gimple_purge_dead_eh_edges. Dump tranformation with details
7861 (strlen_dom_walker::before_dom_children): Update call by adding
7862 new argument cleanup_eh.
7863 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7865 2018-01-03 Martin Liska <mliska@suse.cz>
7868 * cif-code.def (VARIADIC_THUNK): New enum value.
7869 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7872 2018-01-03 Jan Beulich <jbeulich@suse.com>
7874 * sse.md (mov<mode>_internal): Tighten condition for when to use
7875 vmovdqu<ssescalarsize> for TI and OI modes.
7877 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7879 Update copyright years.
7881 2018-01-03 Martin Liska <mliska@suse.cz>
7884 * ipa-visibility.c (function_and_variable_visibility): Skip
7885 functions with noipa attribure.
7887 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7889 * gcc.c (process_command): Update copyright notice dates.
7890 * gcov-dump.c (print_version): Ditto.
7891 * gcov.c (print_version): Ditto.
7892 * gcov-tool.c (print_version): Ditto.
7893 * gengtype.c (create_file): Ditto.
7894 * doc/cpp.texi: Bump @copying's copyright year.
7895 * doc/cppinternals.texi: Ditto.
7896 * doc/gcc.texi: Ditto.
7897 * doc/gccint.texi: Ditto.
7898 * doc/gcov.texi: Ditto.
7899 * doc/install.texi: Ditto.
7900 * doc/invoke.texi: Ditto.
7902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7904 * vector-builder.h (vector_builder::m_full_nelts): Change from
7905 unsigned int to poly_uint64.
7906 (vector_builder::full_nelts): Update prototype accordingly.
7907 (vector_builder::new_vector): Likewise.
7908 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7909 (vector_builder::operator ==): Likewise.
7910 (vector_builder::finalize): Likewise.
7911 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7912 Take the number of elements as a poly_uint64 rather than an
7914 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7915 from unsigned int to poly_uint64.
7916 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7917 (vec_perm_indices::new_vector): Likewise.
7918 (vec_perm_indices::length): Likewise.
7919 (vec_perm_indices::nelts_per_input): Likewise.
7920 (vec_perm_indices::input_nelts): Likewise.
7921 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7922 number of elements per input as a poly_uint64 rather than an
7923 unsigned int. Use the original encoding for variable-length
7924 vectors, rather than clamping each individual element.
7925 For the second and subsequent elements in each pattern,
7926 clamp the step and base before clamping their sum.
7927 (vec_perm_indices::series_p): Handle polynomial element counts.
7928 (vec_perm_indices::all_in_range_p): Likewise.
7929 (vec_perm_indices_to_tree): Likewise.
7930 (vec_perm_indices_to_rtx): Likewise.
7931 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7932 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7933 (tree_vector_builder::new_binary_operation): Handle polynomial
7934 element counts. Return false if we need to know the number
7935 of elements at compile time.
7936 * fold-const.c (fold_vec_perm): Punt if the number of elements
7937 isn't known at compile time.
7939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7941 * vec-perm-indices.h (vec_perm_builder): Change element type
7942 from HOST_WIDE_INT to poly_int64.
7943 (vec_perm_indices::element_type): Update accordingly.
7944 (vec_perm_indices::clamp): Handle polynomial element_types.
7945 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7946 (vec_perm_indices::all_in_range_p): Likewise.
7947 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7949 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7950 polynomial vec_perm_indices element types.
7951 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7952 * fold-const.c (fold_vec_perm): Likewise.
7953 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7954 * tree-vect-generic.c (lower_vec_perm): Likewise.
7955 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7956 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7957 element type to HOST_WIDE_INT.
7959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7960 Alan Hayward <alan.hayward@arm.com>
7961 David Sherwood <david.sherwood@arm.com>
7963 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7964 rather than an int. Use plus_constant.
7965 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7966 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7969 Alan Hayward <alan.hayward@arm.com>
7970 David Sherwood <david.sherwood@arm.com>
7972 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7973 a HOST_WIDE_INT to a poly_int64.
7975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7976 Alan Hayward <alan.hayward@arm.com>
7977 David Sherwood <david.sherwood@arm.com>
7979 * calls.c (load_register_parameters): Cope with polynomial
7980 mode sizes. Require a constant size for BLKmode parameters
7981 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7982 forces a parameter to be padded at the lsb end in order to
7983 fill a complete number of words, require the parameter size
7984 to be ordered wrt UNITS_PER_WORD.
7986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 Alan Hayward <alan.hayward@arm.com>
7988 David Sherwood <david.sherwood@arm.com>
7990 * reload1.c (spill_stack_slot_width): Change element type
7991 from unsigned int to poly_uint64_pod.
7992 (alter_reg): Treat mode sizes as polynomial.
7994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7995 Alan Hayward <alan.hayward@arm.com>
7996 David Sherwood <david.sherwood@arm.com>
7998 * reload.c (complex_word_subreg_p): New function.
7999 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
8006 operands whose sizes aren't ordered.
8007 (match_reload): Refer to this check here.
8009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8010 Alan Hayward <alan.hayward@arm.com>
8011 David Sherwood <david.sherwood@arm.com>
8013 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8014 that the mode size is in the set {1, 2, 4, 8, 16}.
8016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8017 Alan Hayward <alan.hayward@arm.com>
8018 David Sherwood <david.sherwood@arm.com>
8020 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8021 Use plus_constant instead of gen_rtx_PLUS.
8023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8024 Alan Hayward <alan.hayward@arm.com>
8025 David Sherwood <david.sherwood@arm.com>
8027 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8028 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8029 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8030 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8031 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8032 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8033 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8034 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8035 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8036 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8038 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8039 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8040 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8041 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8042 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8043 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8044 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8045 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8046 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8047 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8049 * expr.c (emit_move_resolve_push): Treat the input and result
8050 of PUSH_ROUNDING as a poly_int64.
8051 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8052 (emit_push_insn): Likewise.
8053 * lra-eliminations.c (mark_not_eliminable): Likewise.
8054 * recog.c (push_operand): Likewise.
8055 * reload1.c (elimination_effects): Likewise.
8056 * rtlanal.c (nonzero_bits1): Likewise.
8057 * calls.c (store_one_arg): Likewise. Require the padding to be
8058 known at compile time.
8060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8061 Alan Hayward <alan.hayward@arm.com>
8062 David Sherwood <david.sherwood@arm.com>
8064 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8065 Use plus_constant instead of gen_rtx_PLUS.
8067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8068 Alan Hayward <alan.hayward@arm.com>
8069 David Sherwood <david.sherwood@arm.com>
8071 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8075 Alan Hayward <alan.hayward@arm.com>
8076 David Sherwood <david.sherwood@arm.com>
8078 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8079 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8080 via stack temporaries. Treat the mode size as polynomial too.
8082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8083 Alan Hayward <alan.hayward@arm.com>
8084 David Sherwood <david.sherwood@arm.com>
8086 * expr.c (expand_expr_real_2): When handling conversions involving
8087 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8088 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8089 as a poly_uint64 too.
8091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8092 Alan Hayward <alan.hayward@arm.com>
8093 David Sherwood <david.sherwood@arm.com>
8095 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8098 Alan Hayward <alan.hayward@arm.com>
8099 David Sherwood <david.sherwood@arm.com>
8101 * combine.c (can_change_dest_mode): Handle polynomial
8102 REGMODE_NATURAL_SIZE.
8103 * expmed.c (store_bit_field_1): Likewise.
8104 * expr.c (store_constructor): Likewise.
8105 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8106 and polynomial REGMODE_NATURAL_SIZE.
8107 (gen_lowpart_common): Likewise.
8108 * reginfo.c (record_subregs_of_mode): Likewise.
8109 * rtlanal.c (read_modify_subreg_p): Likewise.
8111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8112 Alan Hayward <alan.hayward@arm.com>
8113 David Sherwood <david.sherwood@arm.com>
8115 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8116 numbers of elements.
8118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8119 Alan Hayward <alan.hayward@arm.com>
8120 David Sherwood <david.sherwood@arm.com>
8122 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8129 in a POINTER_PLUS_EXPR.
8131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8132 Alan Hayward <alan.hayward@arm.com>
8133 David Sherwood <david.sherwood@arm.com>
8135 * omp-simd-clone.c (simd_clone_subparts): New function.
8136 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8137 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8144 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8145 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8146 (expand_vector_condition, vector_element): Likewise.
8147 (subparts_gt): New function.
8148 (get_compute_type): Use subparts_gt.
8149 (count_type_subparts): Delete.
8150 (expand_vector_operations_1): Use subparts_gt instead of
8151 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8158 (vect_compile_time_alias): ...this new function. Do the calculation
8159 on poly_ints rather than trees.
8160 (vect_prune_runtime_alias_test_list): Update call accordingly.
8162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8163 Alan Hayward <alan.hayward@arm.com>
8164 David Sherwood <david.sherwood@arm.com>
8166 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8168 (vect_schedule_slp_instance): Likewise.
8170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8171 Alan Hayward <alan.hayward@arm.com>
8172 David Sherwood <david.sherwood@arm.com>
8174 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8175 constant and extern definitions for variable-length vectors.
8176 (vect_get_constant_vectors): Note that the number of units
8177 is known to be constant.
8179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8180 Alan Hayward <alan.hayward@arm.com>
8181 David Sherwood <david.sherwood@arm.com>
8183 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8184 of units as polynomial. Choose between WIDE and NARROW based
8187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8188 Alan Hayward <alan.hayward@arm.com>
8189 David Sherwood <david.sherwood@arm.com>
8191 * tree-vect-stmts.c (simd_clone_subparts): New function.
8192 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8195 Alan Hayward <alan.hayward@arm.com>
8196 David Sherwood <david.sherwood@arm.com>
8198 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8199 vectors as polynomial. Use build_index_vector for
8202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8203 Alan Hayward <alan.hayward@arm.com>
8204 David Sherwood <david.sherwood@arm.com>
8206 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8207 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8208 for variable-length vectors.
8209 (vectorizable_mask_load_store): Treat the number of units as
8210 polynomial, asserting that it is constant if the condition has
8211 already been enforced.
8212 (vectorizable_store, vectorizable_load): Likewise.
8214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8215 Alan Hayward <alan.hayward@arm.com>
8216 David Sherwood <david.sherwood@arm.com>
8218 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8219 of units as polynomial. Punt if we can't tell at compile time
8220 which vector contains the final result.
8222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8223 Alan Hayward <alan.hayward@arm.com>
8224 David Sherwood <david.sherwood@arm.com>
8226 * tree-vect-loop.c (vectorizable_induction): Treat the number
8227 of units as polynomial. Punt on SLP inductions. Use an integer
8228 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8229 cast of such a series for variable-length floating-point
8232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8233 Alan Hayward <alan.hayward@arm.com>
8234 David Sherwood <david.sherwood@arm.com>
8236 * tree.h (build_index_vector): Declare.
8237 * tree.c (build_index_vector): New function.
8238 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8239 of units as polynomial, forcibly converting it to a constant if
8240 vectorizable_reduction has already enforced the condition.
8241 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8242 to create a {1,2,3,...} vector.
8243 (vectorizable_reduction): Treat the number of units as polynomial.
8244 Choose vectype_in based on the largest scalar element size rather
8245 than the smallest number of units. Enforce the restrictions
8248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8249 Alan Hayward <alan.hayward@arm.com>
8250 David Sherwood <david.sherwood@arm.com>
8252 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8253 number of units as polynomial.
8255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8256 Alan Hayward <alan.hayward@arm.com>
8257 David Sherwood <david.sherwood@arm.com>
8259 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8260 * target.def (autovectorize_vector_sizes): Return the vector sizes
8261 by pointer, using vector_sizes rather than a bitmask.
8262 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8263 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8264 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8266 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8267 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8268 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8269 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8270 * omp-general.c (omp_max_vf): Likewise.
8271 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8272 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8273 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8274 * tree-vect-slp.c (vect_slp_bb): Likewise.
8275 * doc/tm.texi: Regenerate.
8276 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8278 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8279 the vector size as a poly_uint64 rather than an unsigned int.
8280 (current_vector_size): Change from an unsigned int to a poly_uint64.
8281 (get_vectype_for_scalar_type): Update accordingly.
8282 * tree.h (build_truth_vector_type): Take the size and number of
8283 units as a poly_uint64 rather than an unsigned int.
8284 (build_vector_type): Add a temporary overload that takes
8285 the number of units as a poly_uint64 rather than an unsigned int.
8286 * tree.c (make_vector_type): Likewise.
8287 (build_truth_vector_type): Take the number of units as a poly_uint64
8288 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
8295 as poly_uint64s rather than unsigned ints.
8296 * targhooks.h (default_get_mask_mode): Update accordingly.
8297 * targhooks.c (default_get_mask_mode): Likewise.
8298 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8299 * doc/tm.texi: Regenerate.
8301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8302 Alan Hayward <alan.hayward@arm.com>
8303 David Sherwood <david.sherwood@arm.com>
8305 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8306 * omp-general.c (omp_max_vf): Likewise.
8307 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8308 (expand_omp_simd): Handle polynomial safelen.
8309 * omp-low.c (omplow_simd_context): Add a default constructor.
8310 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8311 (lower_rec_simd_input_clauses): Update accordingly.
8312 (lower_rec_input_clauses): Likewise.
8314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8315 Alan Hayward <alan.hayward@arm.com>
8316 David Sherwood <david.sherwood@arm.com>
8318 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8319 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8320 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8321 (vect_analyze_slp_cost): Likewise.
8322 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8323 (vect_model_load_cost): Likewise.
8325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8326 Alan Hayward <alan.hayward@arm.com>
8327 David Sherwood <david.sherwood@arm.com>
8329 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8330 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8331 from an unsigned int * to a poly_uint64_pod *.
8332 (calculate_unrolling_factor): New function.
8333 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8336 Alan Hayward <alan.hayward@arm.com>
8337 David Sherwood <david.sherwood@arm.com>
8339 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8340 from an unsigned int to a poly_uint64.
8341 (_loop_vec_info::slp_unrolling_factor): Likewise.
8342 (_loop_vec_info::vectorization_factor): Change from an int
8344 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8345 (vect_get_num_vectors): New function.
8346 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8347 (vect_get_num_copies): Use vect_get_num_vectors.
8348 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8349 to an unsigned int *.
8350 (vect_analyze_data_refs): Change min_vf from an int * to a
8352 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8353 than an unsigned HOST_WIDE_INT.
8354 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8355 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8356 to an unsigned int *.
8357 (vect_analyze_data_ref_dependences): Likewise.
8358 (vect_compute_data_ref_alignment): Handle polynomial vf.
8359 (vect_enhance_data_refs_alignment): Likewise.
8360 (vect_prune_runtime_alias_test_list): Likewise.
8361 (vect_shift_permute_load_chain): Likewise.
8362 (vect_supportable_dr_alignment): Likewise.
8363 (dependence_distance_ge_vf): Take the vectorization factor as a
8364 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8365 (vect_analyze_data_refs): Change min_vf from an int * to a
8367 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8368 vfm1 as a poly_uint64 rather than an int. Make the same change
8369 for the returned bound_scalar.
8370 (vect_gen_vector_loop_niters): Handle polynomial vf.
8371 (vect_do_peeling): Likewise. Update call to
8372 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8373 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8375 * tree-vect-loop.c (vect_determine_vectorization_factor)
8376 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8377 (vect_get_known_peeling_cost): Likewise.
8378 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8379 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8380 (vect_transform_loop): Likewise. Use the lowest possible VF when
8381 updating the upper bounds of the loop.
8382 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8384 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8385 polynomial unroll factors.
8386 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8387 (vect_make_slp_decision): Likewise.
8388 (vect_supported_load_permutation_p): Likewise, and polynomial
8390 (vect_analyze_slp_cost): Handle polynomial vf.
8391 (vect_slp_analyze_node_operations): Likewise.
8392 (vect_slp_analyze_bb_1): Likewise.
8393 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8394 than an unsigned HOST_WIDE_INT.
8395 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8396 (vectorizable_load): Handle polynomial vf.
8397 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8399 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8402 Alan Hayward <alan.hayward@arm.com>
8403 David Sherwood <david.sherwood@arm.com>
8405 * match.pd: Handle bit operations involving three constants
8406 and try to fold one pair.
8408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8410 * tree-vect-loop-manip.c: Include gimple-fold.h.
8411 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8412 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8413 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8414 Add a path that uses a step of VF instead of 1, but disable it
8416 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8417 and niters_no_overflow parameters. Update calls to
8418 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8419 Create a new SSA name if the latter choses to use a ste other
8420 than zero, and return it via niters_vector_mult_vf_var.
8421 * tree-vect-loop.c (vect_transform_loop): Update calls to
8422 vect_do_peeling, vect_gen_vector_loop_niters and
8423 slpeel_make_loop_iterate_ntimes.
8424 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8425 (vect_gen_vector_loop_niters): Update declarations after above changes.
8427 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8429 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8430 128-bit round to integer instructions.
8431 (ceil<mode>2): Likewise.
8432 (btrunc<mode>2): Likewise.
8433 (round<mode>2): Likewise.
8435 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8437 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8438 unaligned VSX load/store on P8/P9.
8439 (expand_block_clear): Allow the use of unaligned VSX
8440 load/store on P8/P9.
8442 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8444 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8446 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8447 swap associated with both a load and a store.
8449 2018-01-02 Andrew Waterman <andrew@sifive.com>
8451 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8452 * config/riscv/riscv.md (clear_cache): Use it.
8454 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8456 * web.c: Remove out-of-date comment.
8458 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8460 * expr.c (fixup_args_size_notes): Check that any existing
8461 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8462 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8463 (emit_single_push_insn): ...here.
8465 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8467 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8468 (const_vector_encoded_nelts): New function.
8469 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8470 (const_vector_int_elt, const_vector_elt): Declare.
8471 * emit-rtl.c (const_vector_int_elt_1): New function.
8472 (const_vector_elt): Likewise.
8473 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8474 of CONST_VECTOR_ELT.
8476 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8478 * expr.c: Include rtx-vector-builder.h.
8479 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8480 directly on the tree encoding.
8481 (const_vector_from_tree): Likewise.
8482 * optabs.c: Include rtx-vector-builder.h.
8483 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8484 sequence of "u" values.
8485 * vec-perm-indices.c: Include rtx-vector-builder.h.
8486 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8487 directly on the vec_perm_indices encoding.
8489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8491 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8492 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8493 * rtx-vector-builder.h: New file.
8494 * rtx-vector-builder.c: Likewise.
8495 * rtl.h (rtx_def::u2): Add a const_vector field.
8496 (CONST_VECTOR_NPATTERNS): New macro.
8497 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8498 (CONST_VECTOR_DUPLICATE_P): Likewise.
8499 (CONST_VECTOR_STEPPED_P): Likewise.
8500 (CONST_VECTOR_ENCODED_ELT): Likewise.
8501 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8502 (unwrap_const_vec_duplicate): Likewise.
8503 (const_vec_series_p): Check for a non-duplicated vector encoding.
8504 Say that the function only returns true for integer vectors.
8505 * emit-rtl.c: Include rtx-vector-builder.h.
8506 (gen_const_vec_duplicate_1): Delete.
8507 (gen_const_vector): Call gen_const_vec_duplicate instead of
8508 gen_const_vec_duplicate_1.
8509 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8510 (gen_const_vec_duplicate): Use rtx_vector_builder.
8511 (gen_const_vec_series): Likewise.
8512 (gen_rtx_CONST_VECTOR): Likewise.
8513 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8514 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8515 Build a new vector rather than modifying a CONST_VECTOR in-place.
8516 (handle_special_swappables): Update call accordingly.
8517 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8518 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8519 Build a new vector rather than modifying a CONST_VECTOR in-place.
8520 (handle_special_swappables): Update call accordingly.
8522 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8524 * simplify-rtx.c (simplify_const_binary_operation): Use
8525 CONST_VECTOR_ELT instead of XVECEXP.
8527 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8529 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8530 the selector elements to be different from the data elements
8531 if the selector is a VECTOR_CST.
8532 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8533 ssizetype for the selector.
8535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8537 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8538 before testing each element individually.
8539 * tree-vect-generic.c (lower_vec_perm): Likewise.
8541 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8543 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8544 * selftest-run-tests.c (selftest::run_tests): Call it.
8545 * vector-builder.h (vector_builder::operator ==): New function.
8546 (vector_builder::operator !=): Likewise.
8547 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8548 (vec_perm_indices::all_from_input_p): New function.
8549 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8550 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8551 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8552 instead of reading the VECTOR_CST directly. Detect whether both
8553 vector inputs are the same before constructing the vec_perm_indices,
8554 and update the number of inputs argument accordingly. Use the
8555 utility functions added above. Only construct sel2 if we need to.
8557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8559 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8560 the broadcast of the low byte.
8561 (expand_mult_highpart): Use an explicit encoding for the permutes.
8562 * optabs-query.c (can_mult_highpart_p): Likewise.
8563 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8564 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8565 (vectorizable_bswap): Likewise.
8566 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8567 explicit encoding for the power-of-2 permutes.
8568 (vect_permute_store_chain): Likewise.
8569 (vect_grouped_load_supported): Likewise.
8570 (vect_permute_load_chain): Likewise.
8572 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8574 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8575 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8576 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8577 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8578 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8579 (vect_gen_perm_mask_any): Likewise.
8581 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8583 * int-vector-builder.h: New file.
8584 * vec-perm-indices.h: Include int-vector-builder.h.
8585 (vec_perm_indices): Redefine as an int_vector_builder.
8586 (auto_vec_perm_indices): Delete.
8587 (vec_perm_builder): Redefine as a stand-alone class.
8588 (vec_perm_indices::vec_perm_indices): New function.
8589 (vec_perm_indices::clamp): Likewise.
8590 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8591 (vec_perm_indices::new_vector): New function.
8592 (vec_perm_indices::new_expanded_vector): Update for new
8593 vec_perm_indices class.
8594 (vec_perm_indices::rotate_inputs): New function.
8595 (vec_perm_indices::all_in_range_p): Operate directly on the
8596 encoded form, without computing elided elements.
8597 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8598 encoding. Update for new vec_perm_indices class.
8599 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8600 the given vec_perm_builder.
8601 (expand_vec_perm_var): Update vec_perm_builder constructor.
8602 (expand_mult_highpart): Use vec_perm_builder instead of
8603 auto_vec_perm_indices.
8604 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8605 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8606 or double series encoding as appropriate.
8607 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8608 vec_perm_indices instead of auto_vec_perm_indices.
8609 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8610 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8611 (vect_permute_store_chain): Likewise.
8612 (vect_grouped_load_supported): Likewise.
8613 (vect_permute_load_chain): Likewise.
8614 (vect_shift_permute_load_chain): Likewise.
8615 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8616 (vect_transform_slp_perm_load): Likewise.
8617 (vect_schedule_slp_instance): Likewise.
8618 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8619 (vectorizable_mask_load_store): Likewise.
8620 (vectorizable_bswap): Likewise.
8621 (vectorizable_store): Likewise.
8622 (vectorizable_load): Likewise.
8623 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8624 vec_perm_indices instead of auto_vec_perm_indices. Use
8625 tree_to_vec_perm_builder to read the vector from a tree.
8626 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8627 vec_perm_builder instead of a vec_perm_indices.
8628 (have_whole_vector_shift): Use vec_perm_builder and
8629 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8630 truncation to calc_vec_perm_mask_for_shift.
8631 (vect_create_epilog_for_reduction): Likewise.
8632 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8633 from auto_vec_perm_indices to vec_perm_indices.
8634 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8635 instead of changing individual elements.
8636 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8637 the vector in d.perm.
8638 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8639 from auto_vec_perm_indices to vec_perm_indices.
8640 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8641 instead of changing individual elements.
8642 (arm_vectorize_vec_perm_const): Use new_vector to install
8643 the vector in d.perm.
8644 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8645 Update vec_perm_builder constructor.
8646 (rs6000_expand_interleave): Likewise.
8647 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8648 (rs6000_expand_interleave): Likewise.
8650 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8652 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8653 to qimode could truncate the indices.
8654 * optabs.c (expand_vec_perm_var): Likewise.
8656 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8658 * Makefile.in (OBJS): Add vec-perm-indices.o.
8659 * vec-perm-indices.h: New file.
8660 * vec-perm-indices.c: Likewise.
8661 * target.h (vec_perm_indices): Replace with a forward class
8663 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8664 * optabs.h: Include vec-perm-indices.h.
8665 (expand_vec_perm): Delete.
8666 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8667 (expand_vec_perm_const): Declare.
8668 * target.def (vec_perm_const_ok): Replace with...
8669 (vec_perm_const): ...this new hook.
8670 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8671 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8672 * doc/tm.texi: Regenerate.
8673 * optabs.def (vec_perm_const): Delete.
8674 * doc/md.texi (vec_perm_const): Likewise.
8675 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8676 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8677 expand_vec_perm for constant permutation vectors. Assert that
8678 the mode of variable permutation vectors is the integer equivalent
8679 of the mode that is being permuted.
8680 * optabs-query.h (selector_fits_mode_p): Declare.
8681 * optabs-query.c: Include vec-perm-indices.h.
8682 (selector_fits_mode_p): New function.
8683 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8684 is defined, instead of checking whether the vec_perm_const_optab
8685 exists. Use targetm.vectorize.vec_perm_const instead of
8686 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8687 fit in the vector mode before using a variable permute.
8688 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8689 vec_perm_indices instead of an rtx.
8690 (expand_vec_perm): Replace with...
8691 (expand_vec_perm_const): ...this new function. Take the selector
8692 as a vec_perm_indices rather than an rtx. Also take the mode of
8693 the selector. Update call to shift_amt_for_vec_perm_mask.
8694 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8695 Use vec_perm_indices::new_expanded_vector to expand the original
8696 selector into bytes. Check whether the indices fit in the vector
8697 mode before using a variable permute.
8698 (expand_vec_perm_var): Make global.
8699 (expand_mult_highpart): Use expand_vec_perm_const.
8700 * fold-const.c: Includes vec-perm-indices.h.
8701 * tree-ssa-forwprop.c: Likewise.
8702 * tree-vect-data-refs.c: Likewise.
8703 * tree-vect-generic.c: Likewise.
8704 * tree-vect-loop.c: Likewise.
8705 * tree-vect-slp.c: Likewise.
8706 * tree-vect-stmts.c: Likewise.
8707 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8709 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8710 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8711 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8712 (aarch64_vectorize_vec_perm_const): ...this new function.
8713 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8714 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8715 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8716 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8717 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8718 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8719 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8721 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8722 check for NEON modes.
8723 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8724 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8725 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8726 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8728 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8729 the old VEC_PERM_CONST conditions.
8730 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8731 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8732 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8733 (ia64_vectorize_vec_perm_const_ok): Merge into...
8734 (ia64_vectorize_vec_perm_const): ...this new function.
8735 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8736 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8737 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8738 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8739 * config/mips/mips.c (mips_expand_vec_perm_const)
8740 (mips_vectorize_vec_perm_const_ok): Merge into...
8741 (mips_vectorize_vec_perm_const): ...this new function.
8742 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8743 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8744 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8745 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8746 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8747 (rs6000_expand_vec_perm_const): Delete.
8748 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8750 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8751 (altivec_expand_vec_perm_const_le): Take each operand individually.
8752 Operate on constant selectors rather than rtxes.
8753 (altivec_expand_vec_perm_const): Likewise. Update call to
8754 altivec_expand_vec_perm_const_le.
8755 (rs6000_expand_vec_perm_const): Delete.
8756 (rs6000_vectorize_vec_perm_const_ok): Delete.
8757 (rs6000_vectorize_vec_perm_const): New function.
8758 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8759 an element count and rtx array.
8760 (rs6000_expand_extract_even): Update call accordingly.
8761 (rs6000_expand_interleave): Likewise.
8762 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8763 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8764 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8765 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8766 (rs6000_expand_vec_perm_const): Delete.
8767 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8768 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8769 (altivec_expand_vec_perm_const_le): Take each operand individually.
8770 Operate on constant selectors rather than rtxes.
8771 (altivec_expand_vec_perm_const): Likewise. Update call to
8772 altivec_expand_vec_perm_const_le.
8773 (rs6000_expand_vec_perm_const): Delete.
8774 (rs6000_vectorize_vec_perm_const_ok): Delete.
8775 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8776 reference to the SPE evmerge intructions.
8777 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8778 an element count and rtx array.
8779 (rs6000_expand_extract_even): Update call accordingly.
8780 (rs6000_expand_interleave): Likewise.
8781 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8782 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8784 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8788 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8789 vector mode and that that mode matches the mode of the data
8791 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8792 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8793 directly using expand_vec_perm_1 when forcing selectors into
8795 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8797 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8799 * optabs-query.h (can_vec_perm_p): Delete.
8800 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8801 * optabs-query.c (can_vec_perm_p): Split into...
8802 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8803 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8804 particular selector is valid.
8805 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8806 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8807 (vect_grouped_load_supported): Likewise.
8808 (vect_shift_permute_load_chain): Likewise.
8809 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8810 (vect_transform_slp_perm_load): Likewise.
8811 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8812 (vectorizable_bswap): Likewise.
8813 (vect_gen_perm_mask_checked): Likewise.
8814 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8815 implementations of variable permutation vectors into account
8816 when deciding which selector to use.
8817 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8818 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8819 with a false third argument.
8820 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8821 to test whether the constant selector is valid and can_vec_perm_var_p
8822 to test whether a variable selector is valid.
8824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8826 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8827 * optabs-query.c (can_vec_perm_p): Likewise.
8828 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8829 instead of vec_perm_indices.
8830 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8831 (vect_gen_perm_mask_checked): Likewise,
8832 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8833 (vect_gen_perm_mask_checked): Likewise,
8835 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8837 * optabs-query.h (qimode_for_vec_perm): Declare.
8838 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8839 (qimode_for_vec_perm): ...this new function.
8840 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8842 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8844 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8845 does not have a conditional at the top.
8847 2018-01-02 Richard Biener <rguenther@suse.de>
8849 * ipa-inline.c (big_speedup_p): Fix expression.
8851 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8854 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8857 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8861 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8862 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8863 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8864 cond_taken_branch_cost 3->4.
8866 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8868 PR tree-optimization/83581
8869 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8870 TODO_cleanup_cfg if any changes have been made.
8873 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8874 convert_modes if target mode has the right side, but different mode
8878 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8879 last argument when extracting from CONCAT. If either from_real or
8880 from_imag is NULL, use expansion through memory. If result is not
8881 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8882 the parts directly to inner mode, if even that fails, use expansion
8886 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8887 check for bswap in mode rather than HImode and use that in expand_unop
8890 Copyright (C) 2018 Free Software Foundation, Inc.
8892 Copying and distribution of this file, with or without modification,
8893 are permitted in any medium without royalty provided the copyright
8894 notice and this notice are preserved.