1 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4 * diagnostic-show-locus.c (class layout_point): Convert m_line
5 from int to linenum_type.
6 (line_span::comparator): Use linenum "compare" function when
7 comparing line numbers.
8 (test_line_span): New function.
9 (layout_range::contains_point): Convert param "row" from int to
11 (layout_range::intersects_line_p): Likewise.
12 (layout::will_show_line_p): Likewise.
13 (layout::print_source_line): Likewise.
14 (layout::should_print_annotation_line_p): Likewise.
15 (layout::print_annotation_line): Likewise.
16 (layout::print_leading_fixits): Likewise.
17 (layout::annotation_line_showed_range_p): Likewise.
18 (struct line_corrections): Likewise for field m_row.
19 (line_corrections::line_corrections): Likewise for param "row".
20 (layout::print_trailing_fixits): Likewise.
21 (layout::get_state_at_point): Likewise.
22 (layout::get_x_bound_for_row): Likewise.
23 (layout::print_line): Likewise.
24 (diagnostic_show_locus): Likewise for locals "last_line" and
26 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
27 * input.c (selftest::test_linenum_comparisons): New function.
28 (selftest::input_c_tests): Call it.
29 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
30 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
31 * selftest.h (ASSERT_GT): New macro.
32 (ASSERT_GT_AT): New macro.
33 (ASSERT_LT): New macro.
34 (ASSERT_LT_AT): New macro.
36 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
38 PR rtl-optimization/84780
39 * combine.c (distribute_links): Don't make a link based on pc_rtx.
41 2018-03-14 Martin Liska <mliska@suse.cz>
43 * tree.c (record_node_allocation_statistics): Use
45 (get_stats_node_kind): New function extracted from
46 record_node_allocation_statistics.
47 (free_node): Use get_stats_node_kind.
49 2018-03-14 Richard Biener <rguenther@suse.de>
51 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
52 that the value-set of ANTIC_IN doesn't grow.
55 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
57 (BB_VISITED_WITH_VISITED_SUCCS): New define.
58 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
60 2018-03-14 Julia Koval <julia.koval@intel.com>
62 * config.gcc (icelake-client, icelake-server): New.
64 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
65 (initial_ix86_arch_features): Ditto.
66 (PTA_SKYLAKE): Add SGX.
67 (PTA_ICELAKE): Remove.
68 (PTA_ICELAKE_CLIENT): New.
69 (PTA_ICELAKE_SERVER): New.
70 (ix86_option_override_internal): Split up icelake on icelake client and
72 (get_builtin_code_for_version): Ditto.
73 (fold_builtin_cpu): Ditto.
74 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
75 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
76 * config/i386/i386.h (processor_type): Ditto.
77 * doc/invoke.texi: Ditto.
79 2018-03-14 Jakub Jelinek <jakub@redhat.com>
82 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
83 INTEGER_CST offset, add it together with bitpos / 8 and
84 sign extend based on POINTER_SIZE.
88 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
91 * config/i386/constraints.md (Yc): New register constraint.
92 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
93 Use Yc constraint for alternative 2 of operand 0. Remove
94 preferred_for_speed attribute.
96 2018-03-14 Richard Biener <rguenther@suse.de>
98 PR tree-optimization/84830
99 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
100 with the old one to avoid oscillations.
102 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
105 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
107 (assign_by_spills): Return a flag of reload assignment failure.
108 Do not process the reload assignment failures. Do not spill other
109 reload pseudos if they has the same reg class. Update n if
111 (lra_assign): Add a return arg. Set up from the result of
112 assign_by_spills call.
113 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
114 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
115 usage_insns if it is not NULL.
116 (spill_hard_reg_in_range): New function.
117 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
118 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
120 (lra_assign): Change prototype.
121 * lra.c (lra): Add code to deal with fails by splitting hard reg
124 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
126 * config/riscv/riscv.opt (mrelax): New option.
127 * config/riscv/riscv.c (riscv_file_start): Emit ".option
128 "norelax" when riscv_mrelax is disabled.
129 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
131 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
134 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
135 reassociation for int modes.
137 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
139 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
140 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
142 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
143 * config/aarch64/aarch64-sve.md
144 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
145 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
146 (*extend<mode><Vwide>2): Rename to...
147 (aarch64_sve_extend<mode><Vwide>2): ...this.
148 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
149 renaming the old pattern to...
150 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
152 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
153 define_expand, renaming the old pattern to...
154 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
155 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
156 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
157 account when deciding which SVE instruction the optab should use.
158 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
160 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
162 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
163 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
164 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
165 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
166 (tlsdesc_small_<mode>): Turn a define_expand and use
167 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
168 (tlsdesc_small_advsimd_<mode>): ...this.
169 (tlsdesc_small_sve_<mode>): New pattern.
171 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
173 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
174 (UNSPEC_UMUL_HIGHPART): New constants.
175 (MUL_HIGHPART): New int iteraor.
176 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
177 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
179 (*<su>mul<mode>3_highpart): New define_insn.
181 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
184 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
187 2018-03-13 Martin Liska <mliska@suse.cz>
190 * (sem_item_optimizer::sem_item_optimizer): Initialize new
192 (sem_item_optimizer::~sem_item_optimizer): Release it.
193 (sem_item_optimizer::merge_classes): Register variable aliases.
194 (sem_item_optimizer::fixup_pt_set): New function.
195 (sem_item_optimizer::fixup_points_to_sets): Likewise.
196 * ipa-icf.h: Declare new variables and functions.
198 2018-03-13 Jakub Jelinek <jakub@redhat.com>
201 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
202 integer_pow2p@2 and test integer_pow2p in condition.
203 (A < 0 ? C : 0): Similarly for @1.
206 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
207 characters starting at p contain '\0' character, don't look beyond
211 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
212 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
215 * reg-stack.c (change_stack): Change update_end var from int to
216 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
217 also call set_block_for_insn on the newly added insns and rescan.
220 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
224 * tree.c (stabilize_reference_1): Return save_expr (e) for
225 STATEMENT_LIST even if it doesn't have side-effects.
227 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
229 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
231 2018-03-12 Renlin Li <renlin.li@arm.com>
233 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
234 aarch64_output_scalar_simd_mov_immediate.
236 2018-03-12 Martin Sebor <msebor@redhat.com>
238 PR tree-optimization/83456
239 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
240 for perfectly overlapping calls to memcpy.
241 (gimple_fold_builtin_memory_chk): Same.
242 (gimple_fold_builtin_strcpy): Handle no-warning.
243 (gimple_fold_builtin_stxcpy_chk): Same.
244 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
246 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
248 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
249 parameter. Use it for SFmode.
250 (rs6000_function_arg_advance_1): Adjust.
251 (rs6000_function_arg): Adjust.
252 (rs6000_gimplify_va_arg): Pass false for that new parameter.
254 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
256 PR rtl-optimization/84169
257 PR rtl-optimization/84780
258 * combine.c (can_combine_p): Check for a 2-insn combination whether
259 the destination register is used between the two insns, too.
261 2018-03-12 Richard Biener <rguenther@suse.de>
263 PR tree-optimization/84803
264 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
265 for refs DR analysis didn't process.
267 2018-03-12 Richard Biener <rguenther@suse.de>
269 PR tree-optimization/84777
270 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
271 force-vectorize loops ignore whether we are optimizing for size.
273 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
275 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
276 (TARGET_MD_ASM_ADJUST): Define.
278 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
279 Kito Cheng <kito.cheng@gmail.com>
280 Chung-Ju Wu <jasonwucj@gmail.com>
282 * config/nds32/nds32.c (nds32_compute_stack_frame,
283 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
284 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
285 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
286 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
287 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
288 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
289 * config/nds32/nds32.md (prologue, epilogue): Use macro
290 NDS32_V3PUSH_AVAILABLE_P to do checking.
292 2018-03-11 Jakub Jelinek <jakub@redhat.com>
295 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
296 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
297 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
298 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
299 addition of most attributes on !orig_type_die or the attribute not
300 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
302 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
303 Chung-Ju Wu <jasonwucj@gmail.com>
305 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
307 * config/nds32/nds32.opt (mvh): New option.
309 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
310 Chung-Ju Wu <jasonwucj@gmail.com>
312 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
314 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
315 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
318 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
319 Chung-Ju Wu <jasonwucj@gmail.com>
321 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
323 * config/nds32/nds32-multiple.md (strlensi): New pattern.
324 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
326 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
327 Kito Cheng <kito.cheng@gmail.com>
328 Chung-Ju Wu <jasonwucj@gmail.com>
330 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
331 UNSPEC_FFMISM and UNSPEC_FLMISM.
332 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
333 for ffb, ffmism and flmism.
334 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
335 (unspec_ffmism): Ditto.
336 (unspec_flmism): Ditto.
337 (nds32_expand_builtin_impl): Check if string extension is available.
338 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
339 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
341 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
344 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
347 * lra-assigns.c (assign_by_spills): Return a flag of reload
348 assignment failure. Do not process the reload assignment
349 failures. Do not spill other reload pseudos if they has the same
351 (lra_assign): Add a return arg. Set up from the result of
352 assign_by_spills call.
353 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
354 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
355 usage_insns if it is not NULL.
356 (spill_hard_reg_in_range): New function.
357 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
358 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
360 (lra_assign): Change prototype.
361 * lra.c (lra): Add code to deal with fails by splitting hard reg
364 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
367 * config/i386/i386.opt: Replace Enforcment with Enforcement.
369 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
372 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
373 (dw_val_node): Add val_symbolic_view.
374 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
375 (symview_upper_bound): New.
376 (new_line_info_table): Initialize symviews_since_reset.
377 (dwarf2out_source_line): Count symviews_since_reset and set
379 (dw_val_equal_p): Handle symview.
380 (add_AT_symview): New.
381 (print_dw_val): Handle symview.
382 (attr_checksum, attr_checksum_ordered): Likewise.
383 (same_dw_val_p, size_of_die): Likewise.
384 (value_format, output_die): Likewise.
385 (add_high_low_attributes): Use add_AT_symview for entry_view.
386 (dwarf2out_finish): Reset symview_upper_bound, clear
389 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
392 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
393 Add strict argument and use it.
394 (rs6000_split_multireg_move): Update for new strict argument.
395 (mem_operand_gpr): Disallow all non-offsettable addresses.
396 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
398 2018-03-09 Jakub Jelinek <jakub@redhat.com>
401 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
402 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
403 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
406 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
407 decl, use remap_type if we want to use the type.
409 2018-03-09 Martin Sebor <msebor@redhat.com>
411 PR tree-optimization/84526
412 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
414 (builtin_access::generic_overlap): Be prepared to handle non-array
417 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
419 PR rtl-optimization/84682
420 * lra-constraints.c (process_address_1): Check is_address flag
421 for address constraints.
422 (process_alt_operands): Likewise.
423 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
424 preprocess_constraints.
425 * recog.h (preprocess_constraints): Add oploc parameter.
427 * recog.c (preprocess_constraints): Test address_operand for
428 CT_ADDRESS constraints.
430 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
433 * lra-assigns.c (assign_by_spills): Return a flag of reload
434 assignment failure. Do not process the reload assignment
435 failures. Do not spill other reload pseudos if they has the same
437 (lra_assign): Add a return arg. Set up from the result of
438 assign_by_spills call.
439 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
440 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
441 usage_insns if it is not NULL.
442 (spill_hard_reg_in_range): New function.
443 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
444 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
446 (lra_assign): Change prototype.
447 * lra.c (lra): Add code to deal with fails by splitting hard reg
450 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
453 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
454 Accept complain bool parameter. Only emit errors if it is true.
455 (arm_parse_cpu_option_name): Likewise.
456 (arm_target_thumb_only): Adjust callers of the above.
457 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
458 prototype to take a default true bool parameter.
459 (arm_parse_arch_option_name): Likewise.
461 2018-03-09 David Malcolm <dmalcolm@redhat.com>
462 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
466 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
467 * configure: Regenerate.
468 * configure.ac ("linker --version-script option"): New.
469 ("linker soname option"): New.
471 2018-03-09 Richard Biener <rguenther@suse.de>
473 PR tree-optimization/84775
474 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
475 immediate uses of predicate stmts and mark them modified.
478 PR tree-optimization/84178
479 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
481 (version_loop_for_if_conversion): Delay update_ssa call.
482 (tree_if_conversion): Delay update_ssa until after predicate
485 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
488 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
489 when the function accesses prior frames.
491 2018-03-08 Jakub Jelinek <jakub@redhat.com>
494 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
495 gen_llsym, otherwise call maybe_gen_llsym.
498 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
499 has ',' character inside of it.
501 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
504 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
505 as clobbering CC_REGNUM.
507 2018-03-08 Richard Biener <rguenther@suse.de>
510 * tree-scalar-evolution.c: Include tree-into-ssa.h.
511 (follow_copies_to_constant): Do not follow SSA names registered
514 2018-03-08 Richard Biener <rguenther@suse.de>
516 PR tree-optimization/84178
517 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
519 (version_loop_for_if_conversion): Delay update_ssa call.
520 (tree_if_conversion): Delay update_ssa until after predicate
523 2018-03-08 David Malcolm <dmalcolm@redhat.com>
525 PR tree-optimization/84178
526 * tree-if-conv.c (release_bb_predicate): Remove the
527 the assertion that the stmts have NULL use_ops.
528 Discard the statements, asserting that they haven't
529 yet been added to a BB.
531 2018-03-08 Richard Biener <rguenther@suse.de>
533 PR tree-optimization/84746
534 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
535 (phi_translate): Pass in destination ANTIC_OUT set.
536 (phi_translate_1): Likewise. For a simplified result lookup
537 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
538 (phi_translate_set): Adjust.
539 (do_pre_regular_insertion): Likewise.
540 (do_pre_partial_partial_insertion): Likewise.
542 2018-03-08 Martin Liska <mliska@suse.cz>
544 PR gcov-profile/84735
545 * doc/gcov.texi: Document usage of profile files.
546 * gcov-io.h: Document changes in the format.
548 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
552 * dwarf2out.c (struct dw_line_info_table): Update comments for
554 (FORCE_RESET_NEXT_VIEW): New.
555 (FORCE_RESETTING_VIEW_P): New.
556 (RESETTING_VIEW_P): Check for -1 too.
557 (ZERO_VIEW_P): Likewise.
558 (new_line_info_table): Force-reset next view.
559 (dwarf2out_begin_function): Likewise.
560 (dwarf2out_source_line): Simplify zero_view_p initialization.
561 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
562 view directly. Omit view when omitting .loc at line 0.
564 2018-03-08 Jakub Jelinek <jakub@redhat.com>
566 PR tree-optimization/84740
567 * tree-switch-conversion.c (process_switch): Call build_constructors
568 only if info.phi_count is non-zero.
570 PR tree-optimization/84739
571 * tree-tailcall.c (find_tail_calls): Check call arguments against
572 DECL_ARGUMENTS (current_function_decl) rather than
573 DECL_ARGUMENTS (func) when checking for tail recursion.
575 2018-03-07 Jakub Jelinek <jakub@redhat.com>
577 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
578 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
579 Volker Reichelt's entry and add entries for people that perform
580 GCC fuzzy testing and report numerous bugs.
582 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
585 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
586 readonly data in sdata, if that is disabled.
587 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
588 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
589 -mreadonly-in-sdata option.
591 2018-03-07 Martin Sebor <msebor@redhat.com>
593 PR tree-optimization/84468
594 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
595 basic block when looking for nul assignment.
597 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
600 * except.h (output_function_exception_table): Adjust prototype.
601 * except.c (output_function_exception_table): Remove FNNAME parameter
602 and add SECTION parameter. Ouput one part of the table at a time.
603 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
604 the first part of the exception table and emit unwind directives.
605 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
606 (i386_pe_seh_cold_init): Likewise.
607 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
608 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
609 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
610 (ix86_output_call_insn): Emit a nop in one more case for SEH.
611 * config/i386/winnt.c: Include except.h.
612 (struct seh_frame_state): Add reg_offset, after_prologue and
613 in_cold_section fields.
614 (i386_pe_seh_end_prologue): Set seh->after_prologue.
615 (i386_pe_seh_cold_init): New function.
616 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
617 to seh->in_cold_section.
618 (seh_emit_push): Record the offset of the push.
619 (seh_emit_save): Record the offet of the save.
620 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
621 Test seh->after_prologue to disregard the epilogue.
622 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
623 (i386_pe_end_cold_function): New function.
625 2018-03-07 Jakub Jelinek <jakub@redhat.com>
628 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
629 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
632 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
634 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
635 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
638 * multiple_target.c: Include tree-inline.h and intl.h.
639 (expand_target_clones): Diagnose and fail if node->definition and
640 !tree_versionable_function_p (node->decl).
642 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
644 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
646 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
647 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
648 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
650 2018-03-06 Jakub Jelinek <jakub@redhat.com>
653 * combine.c (try_combine): Use reg_or_subregno instead of handling
654 just paradoxical SUBREGs and REGs.
656 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
658 * config/arc/arc.c (arc_finalize_pic): Remove function.
659 (arc_must_save_register): We use single base PIC register, remove
660 checks to save/restore the PIC register.
661 (arc_expand_prologue): Likewise.
662 * config/arc/arc-protos.h (arc_set_default_type_attributes):
664 (arc_verify_short): Likewise.
665 (arc_attr_type): Likewise.
666 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
667 (walk_stores): Likewise.
668 (arc_address_cost): Make it static.
669 (arc_verify_short): Likewise.
670 (branch_dest): Likewise.
671 (arc_attr_type): Likewise.
672 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
673 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
674 (arc_final_prescan_insn): Remove inserting the nops due to
675 hardware hazards. It is done in reorg step.
676 (insn_length_variant_t): Remove.
677 (insn_length_parameters_t): Likewise.
678 (arc_insn_length_parameters): Likewise.
679 (arc_get_insn_variants): Likewise.
680 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
682 2018-03-06 Jakub Jelinek <jakub@redhat.com>
685 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
688 PR tree-optimization/84687
689 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
691 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
693 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
695 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
696 Rename to ppc_speculation_barrier.
697 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
698 __builtin_ppc_speculation_barrier.
700 2018-03-05 Jakub Jelinek <jakub@redhat.com>
703 * combine.c (combine_simplify_rtx): Don't try to simplify if
704 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
707 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
709 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
710 to 32 bytes when compiling for POWER9.
712 2018-03-05 Jakub Jelinek <jakub@redhat.com>
715 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
716 regparm >= 3 with no arg reg available also for calls with
717 flag_force_indirect_call. Pass decl to ix86_function_regparm.
720 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
722 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
724 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
727 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
729 2018-03-05 Richard Biener <rguenther@suse.de>
731 PR tree-optimization/84486
732 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
733 When inserting a __builtin_assume_aligned call set the LHS
734 SSA name alignment info accordingly.
736 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
738 PR tree-optimization/84114
739 * config/aarch64/aarch64.c (aarch64_reassociation_width)
740 Avoid reassociation of FLOAT_MODE addition.
742 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
744 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
745 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
746 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
747 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
748 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
749 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
750 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
752 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
753 __builtin_ia32_wbinvd): New builtins.
754 (SPECIAL_ARGS2): New.
755 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
756 (SPECIAL_ARGS2): New.
757 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
758 (ix86_valid_target_attribute_inner_p): Ditto.
759 (ix86_init_mmx_sse_builtins): Add special_args2.
760 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
761 TARGET_WBNOINVD_P): New.
762 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
763 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
764 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
765 * config/i386/immintrin.h (_wbinvd): New intrinsic.
766 * config/i386/pconfigintrin.h: New file.
767 * config/i386/wbnoinvdintrin.h: Ditto.
768 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
769 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
771 2018-03-05 Richard Biener <rguenther@suse.de>
773 PR tree-optimization/84670
774 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
776 (BB_VISITED_WITH_VISITED_SUCCS): New define.
777 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
778 (compute_antic_aux): Only assert the number of values in ANTIC_IN
779 doesn't grow if all successors (recursively) were visited at least
782 2018-03-05 Richard Biener <rguenther@suse.de>
784 PR tree-optimization/84650
785 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
786 if executed in the loop pipeline.
788 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
790 * doc/configfiles.texi (Configuration Files): Move info about
791 conditionalizing $target-protos.h to...
792 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
793 differs from $target-protos.h.
795 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
796 Chung-Ju Wu <jasonwucj@gmail.com>
798 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
799 * config/nds32/nds32-multiple.md (setmemsi): Define.
800 * config/nds32/nds32-memory-manipulation.c
801 (nds32_gen_dup_4_byte_to_word_value): New.
802 (emit_setmem_word_loop): New.
803 (emit_setmem_byte_loop): New.
804 (nds32_expand_setmem_loop): New.
805 (nds32_expand_setmem_loop_v3m): New.
806 (nds32_expand_setmem_unroll): New.
807 (nds32_expand_setmem): New.
809 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
810 Chung-Ju Wu <jasonwucj@gmail.com>
812 * config/nds32/nds32-memory-manipulation.c
813 (nds32_emit_load_store): New.
814 (nds32_emit_post_inc_load_store): New.
815 (nds32_emit_mem_move): New.
816 (nds32_emit_mem_move_block): New.
817 (nds32_expand_movmemsi_loop_unknown_size): New.
818 (nds32_expand_movmemsi_loop_known_size): New.
819 (nds32_expand_movmemsi_loop): New.
820 (nds32_expand_movmemsi_unroll): New.
821 (nds32_expand_movmemqi): Rename ...
822 (nds32_expand_movmemsi): ... to this.
823 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
824 (movmemsi): ... to this.
825 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
826 (nds32_expand_movmemsi): ... to this.
828 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
829 Monk Chiang <sh.chiang04@gmail.com>
830 Chung-Ju Wu <jasonwucj@gmail.com>
832 * config/nds32/nds32-protos.h
833 (nds32_expand_load_multiple): New arguments.
834 (nds32_expand_store_multiple): Ditto.
835 (nds32_valid_multiple_load_store): Rename ...
836 (nds32_valid_multiple_load_store_p): ... to this.
837 * config/nds32/nds32-memory-manipulation.c
838 (nds32_expand_load_multiple): Refine implementation.
839 (nds32_expand_store_multiple): Ditto.
840 * config/nds32/nds32-multiple.md
841 (load_multiple): Update nds32_expand_load_multiple interface.
842 (store_multiple): Update nds32_expand_store_multiple interface.
843 * config/nds32/nds32-predicates.c
844 (nds32_valid_multiple_load_store): Rename ...
845 (nds32_valid_multiple_load_store_p): ... to this and refine
847 * config/nds32/predicates.md
848 (nds32_load_multiple_and_update_address_operation): New predicate.
849 (nds32_store_multiple_and_update_address_operation): New predicate.
851 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
852 Chung-Ju Wu <jasonwucj@gmail.com>
854 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
855 (combo): New attribute.
856 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
858 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
860 * config/nds32/nds32.opt: Change -mcmodel= default value.
862 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
863 Monk Chiang <sh.chiang04@gmail.com>
864 Chung-Ju Wu <jasonwucj@gmail.com>
866 * config/nds32/constants.md (unspec_element): New enum.
867 * config/nds32/constraints.md (Umw): New constraint.
868 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
869 * config/nds32/nds32-intrinsic.md: Likewise.
870 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
871 (nds32_valid_smw_lwm_base_p): New.
872 (nds32_output_smw_single_word): New.
873 (nds32_output_lmw_single_word): New.
874 (nds32_expand_unaligned_load): New.
875 (nds32_expand_unaligned_store): New.
876 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
877 (nds32_output_smw_single_word): Declare.
878 (nds32_output_lmw_single_word): Declare.
879 (nds32_expand_unaligned_load): Declare.
880 (nds32_expand_unaligned_store): Declare.
881 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
882 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
883 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
884 NDS32_BUILTIN_UASTORE_DW.
885 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
888 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
889 Kito Cheng <kito.cheng@gmail.com>
890 Chung-Ju Wu <jasonwucj@gmail.com>
892 * config/nds32/nds32-intrinsic.c
893 (nds32_expand_builtin_null_ftype_reg): Delete.
894 (nds32_expand_builtin_reg_ftype_imm): Ditto.
895 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
896 (nds32_read_argument): New.
897 (nds32_legitimize_target): Ditto.
898 (nds32_legitimize_argument): Ditto.
899 (nds32_check_constant_argument): Ditto.
900 (nds32_expand_unop_builtin): Ditto.
901 (nds32_expand_unopimm_builtin): Ditto.
902 (nds32_expand_binop_builtin): Ditto.
903 (nds32_builtin_decl_impl): Ditto.
904 (builtin_description): Ditto.
905 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
906 (nds32_init_builtins_impl): Ditto.
907 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
908 (nds32_builtin_decl): New.
909 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
910 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
912 2018-03-02 Jeff Law <law@redhat.com>
914 * reorg.c (stop_search_p): Handle DEBUG_INSN.
915 (redundant_insn, fill_simple_delay_slots): Likewise.
916 (fill_slots_from_thread): Likewise.
917 * resource.c (mark_referenced_resources): Likewise.
918 (mark_set_resources, find_dead_or_set_registers): Likewise.
920 2018-03-02 Jakub Jelinek <jakub@redhat.com>
922 * substring-locations.h (format_warning_va): Formatting fix for
924 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
926 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
927 * substring-locations.c: Include intl.h.
928 (format_warning_va): Turned into small wrapper around
929 format_warning_n_va, renamed to ...
930 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
931 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
933 (format_warning_at_substring_n): New function.
934 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
935 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
936 format_warning_at_substring with just a shorter name instead of
937 const function pointer.
938 (fmtwarn_n): New function.
939 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
940 appropriate, get rid of all the fmtstr temporaries, move conditionals
941 with G_() wrapped string literals directly into fmtwarn arguments,
942 cast dir.len to (int), formatting fixes.
944 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
946 * doc/invoke.texi: Remove "Cilk Plus" references.
948 2018-03-02 Jakub Jelinek <jakub@redhat.com>
949 Richard Biener <rguenther@suse.de>
952 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
953 for error or warning attributes if CALL_FROM_THUNK_P is set.
956 2018-03-02 Jakub Jelinek <jakub@redhat.com>
959 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
960 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
963 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
964 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
966 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
967 instead of -1U in last predictors element's probability member.
969 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
972 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
973 arguments if they are comparable.
975 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
977 PR tree-optimization/84634
978 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
979 masks and masked_loop_p with a single loop_masks, making sure it's
980 null for bb vectorization.
982 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
984 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
985 (vect_analyze_data_ref_access): Use loop->safe_len rather than
986 loop->force_vectorize to check whether there is no alias.
988 2018-03-02 Jakub Jelinek <jakub@redhat.com>
991 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
993 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
995 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
996 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
997 instead of a loop around prev_real_insn.
998 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1002 * config/i386/i386.c (ix86_print_operand): Use conditional
1003 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1006 2018-03-02 Richard Biener <rguenther@suse.de>
1008 PR tree-optimization/84427
1009 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1010 (bitmap_set_subtract_values): Rewrite to handle multiple
1013 (prune_clobbered_mems): Likewise.
1014 (phi_translate): Take edge instead of pred/phiblock.
1015 (phi_translate_1): Likewise.
1016 (phi_translate_set): Likewise. Insert all translated
1017 exprs for a value into the set, keeping possibly multiple
1018 expressions per value.
1019 (compute_antic_aux): Adjust for phi_translate changes.
1020 When intersecting union the expressions and prune those
1021 not in the final value set, keeping possibly multiple
1022 expressions per value. Do not use value-insertion
1023 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1024 all expressions. Add verification that the value-sets
1025 only shrink during iteration.
1026 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1027 (do_pre_regular_insertion): Likewise.
1028 (do_pre_partial_partial_insertion): Likewise.
1030 2018-03-02 Richard Biener <rguenther@suse.de>
1033 * config/darwin.c (saved_debug_info_level): New static global.
1034 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1035 (darwin_asm_lto_end): Restore debug info generation settings.
1037 2018-03-01 Martin Liska <mliska@suse.cz>
1040 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1043 2018-03-01 Richard Biener <rguenther@suse.de>
1046 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1047 type annotation with LTO when debug was disabled at compile-time.
1049 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1051 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1053 (mips_final_postscan_insn): Likewise.
1055 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1057 PR rtl-optimization/84528
1058 * alias.c (init_alias_target): Add commentary.
1059 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1060 a unique base value if the frame pointer is not eliminated
1061 to the stack pointer.
1063 2018-03-01 Tom de Vries <tom@codesourcery.com>
1065 PR rtl-optimization/83327
1066 * lra-int.h (hard_regs_spilled_into): Declare.
1067 * lra.c (hard_regs_spilled_into): Define.
1068 (init_reg_info): Init hard_regs_spilled_into.
1069 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1070 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1071 (process_bb_lives): Handle hard_regs_spilled_into.
1072 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1073 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1075 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1077 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1078 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1079 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1080 * config/rs6000/aix72.h: New file.
1082 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1084 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1085 instead of warning_at with conditional singular and plural messages
1089 * stor-layout.c (update_alignment_for_field): For
1090 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1091 && !DECL_PACKED (field), do the alignment update, just use
1092 only desired_align instead of MAX (type_align, desired_align)
1094 (place_field): Don't do known_align < desired_align handling
1095 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1096 is non-NULL, instead do it after rli->prev_field handling and
1097 only if not within a bitfield word. For DECL_PACKED (field)
1098 use type_align of BITS_PER_UNIT.
1100 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1102 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1103 superfluous parentheses and trailing spaces.
1105 2018-02-28 Richard Biener <rguenther@suse.de>
1107 PR tree-optimization/84584
1108 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1109 SCoPs with fake exit edge.
1111 2018-02-28 Martin Liska <mliska@suse.cz>
1114 * timevar.c (timer::print): Fix format to properly print 100%
1117 2018-02-28 Richard Biener <rguenther@suse.de>
1120 * genmatch.c (capture_info::walk_match): Do not mark
1121 captured expressions without operands as expr_p given
1122 they act more like predicates and should be subject to
1123 "lost tail" side-effect preserving.
1125 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1127 PR rtl-optimization/81611
1128 * auto-inc-dec.c (attempt_change): Move dead note from
1129 mem_insn if it's the next use of regno
1130 (find_address): Take address use of reg holding
1131 non-incremented value. Add parm to limit search to the named
1133 (merge_in_block): Attempt to use a mem insn that is the next
1134 use of the original regno.
1136 2018-02-27 Martin Sebor <msebor@redhat.com>
1139 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1140 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1142 2018-02-27 Martin Sebor <msebor@redhat.com>
1144 PR translation/84207
1145 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1146 n argument to unsigned HOST_WIDE_INT.
1147 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1148 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1149 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1150 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1152 2018-02-27 Richard Biener <rguenther@suse.de>
1154 PR tree-optimization/84512
1155 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1156 Do not use the estimate returned from record_stmt_cost for
1157 the scalar iteration cost but sum properly using add_stmt_cost.
1159 2018-02-27 Richard Biener <rguenther@suse.de>
1161 PR tree-optimization/84466
1162 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1163 Adjust last change to less strictly validate use operands.
1165 2018-02-27 Martin Liska <mliska@suse.cz>
1167 PR gcov-profile/84548
1168 * gcov.c (process_file): Allow partial overlap and consider it
1169 also as group functions.
1170 (output_lines): Properly calculate range of lines for a group.
1172 2018-02-27 Martin Liska <mliska@suse.cz>
1174 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1175 'ggc' suffixes. Change first column width.
1176 (timer::print): Fix formatting of the column.
1178 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1180 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1181 preserve inline entry blocks for the sake of debug inline
1182 entry point markers alone.
1183 (remove_unused_locals): Suggest in comments a better place to
1184 force the preservation of inline entry blocks that are
1185 otherwise unused, but do not preserve them.
1187 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1189 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1191 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1194 * config/i386/constraints.md (Bs): Replace
1195 ix86_indirect_branch_register with
1196 TARGET_INDIRECT_BRANCH_REGISTER.
1198 * config/i386/i386.md (indirect_jump): Likewise.
1199 (tablejump): Likewise.
1200 (*sibcall_memory): Likewise.
1201 (*sibcall_value_memory): Likewise.
1202 Peepholes of indirect call and jump via memory: Likewise.
1203 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1204 (*sibcall_value_GOT_32): Likewise.
1205 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1206 (GOT_memory_operand): Likewise.
1207 (call_insn_operand): Likewise.
1208 (sibcall_insn_operand): Likewise.
1209 (GOT32_symbol_operand): Likewise.
1210 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1212 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1214 PR rtl-optimization/83496
1215 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1216 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1217 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1218 redundant insn, if any.
1219 (relax_delay_slots): Likewise.
1220 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1222 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1224 PR tree-optimization/83965
1225 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1226 that grouped statements are part of a reduction chain. Return
1227 true if the statement is not marked as a reduction itself but
1229 (vect_recog_dot_prod_pattern): Don't check whether the statement
1230 is part of a group here.
1231 (vect_recog_sad_pattern): Likewise.
1232 (vect_recog_widen_sum_pattern): Likewise.
1234 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1237 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1239 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1242 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1244 (ix86_output_indirect_function_return): New prototype.
1245 (ix86_split_simple_return_pop_internal): Likewise.
1246 * config/i386/i386.c (indirect_return_via_cx): New.
1247 (indirect_return_via_cx_bnd): Likewise.
1248 (indirect_thunk_name): Handle return va CX_REG.
1249 (output_indirect_thunk_function): Create alias for
1250 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1251 (ix86_output_indirect_jmp): Remove the bool argument.
1252 (ix86_output_indirect_function_return): New function.
1253 (ix86_split_simple_return_pop_internal): Likewise.
1254 * config/i386/i386.md (*indirect_jump): Don't pass false
1255 to ix86_output_indirect_jmp.
1256 (*tablejump_1): Likewise.
1257 (simple_return_pop_internal): Change it to define_insn_and_split.
1258 Call ix86_split_simple_return_pop_internal to split it for
1260 (simple_return_indirect_internal): Call
1261 ix86_output_indirect_function_return instead of
1262 ix86_output_indirect_jmp.
1264 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1267 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1268 memset and value initialization afterwards.
1270 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1272 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1274 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1277 * common/config/aarch64/aarch64-common.c
1278 (aarch_option_optimization_table[]): Switch
1279 off fomit-frame-pointer
1281 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1282 Chung-Ju Wu <jasonwucj@gmail.com>
1284 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1286 (store_multiple): Ditto.
1288 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1290 * config.gcc: Add --with-cpu support for nds32 target.
1291 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1292 * config/nds32/nds32.opt: Add -mcpu= option.
1294 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1296 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1297 isel=yes): Warn for these deprecated options.
1299 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1301 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1302 ISA_2_5_MASKS_EMBEDDED.
1304 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1306 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1307 p->max as pointers rather than using iterative_hash_expr.
1309 2018-02-23 Carl Love <cel@us.ibm.com>
1311 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1312 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1314 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1315 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1316 P8V_BUILTIN_VEC_VUNSIGNED2.
1318 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1321 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1322 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1324 * lra-constraints.c (curr_insn_transform): Set up
1325 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1327 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1328 * lra-eliminations.c (spill_pseudos): Ditto.
1329 (process_insn_for_elimination): Ditto.
1330 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1331 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1332 LRA_NON_CLOBBERED_ALT.
1334 2018-02-22 Martin Sebor <msebor@redhat.com>
1336 PR tree-optimization/84480
1337 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1338 to maybe_diag_stxncpy_trunc. Call it.
1339 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1340 from gimple_fold_builtin_strcpy. Print inlining stack.
1341 (handle_builtin_stxncpy): Print inlining stack.
1342 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1344 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1347 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1348 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1349 and -fcheck-pointer-bounds are used together.
1350 (indirect_thunk_prefix): New enum.
1351 (indirect_thunk_need_prefix): New function.
1352 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1353 "_nt" instead of "_bnd" for NOTRACK prefix.
1354 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1355 (output_indirect_thunk_function): Likewise.
1357 (ix86_code_end): Update output_indirect_thunk_function calls.
1358 (ix86_output_indirect_branch_via_reg): Replace
1359 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1360 (ix86_output_indirect_branch_via_push): Likewise.
1361 (ix86_output_function_return): Likewise.
1362 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1363 incompatible with -fcf-protection=branch and
1364 -fcheck-pointer-bounds.
1366 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1369 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1370 Change gcc_assert call to output_operand_lossage.
1372 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1374 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1376 2018-02-22 DJ Delorie <dj@redhat.com>
1377 Sebastian Perta <sebastian.perta@renesas.com>
1378 Oleg Endo <olegendo@gcc.gnu.org>
1380 * config/rx/rx.c (rx_rtx_costs): New function.
1381 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1383 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1385 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1387 2018-02-22 Martin Liska <mliska@suse.cz>
1390 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1391 Add "native" as a possible value.
1393 2018-02-22 Martin Liska <mliska@suse.cz>
1396 * config/i386/i386.c (ix86_option_override_internal):
1397 Add "native" as a possible value for -march and -mtune.
1399 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1402 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1403 to all type variants.
1405 PR tree-optimization/84503
1406 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1407 width as info->bitpos + info->bitsize - start.
1408 (merged_store_group::merge_overlapping): Simplify width computation.
1409 (check_no_overlap): New function.
1410 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1411 start + width and last_order of the group, fail if check_no_overlap
1413 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1414 to group if check_no_overlap fails.
1416 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1418 * config/rs6000/altivec.md: Delete contraint arguments to
1419 define_expand, define_split, and define_peephole2, and in
1420 define_insn_and_split if always unused.
1421 * config/rs6000/darwin.md: Ditto.
1422 * config/rs6000/dfp.md: Ditto.
1423 * config/rs6000/rs6000.md: Ditto.
1424 * config/rs6000/sync.md: Ditto.
1425 * config/rs6000/vector.md: Ditto.
1426 * config/rs6000/vsx.md: Ditto.
1428 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1430 * config/rs6000/altivec.md: Write output control strings as braced
1431 blocks instead of double-quoted strings.
1432 * config/rs6000/darwin.md: Ditto.
1433 * config/rs6000/rs6000.md: Ditto.
1434 * config/rs6000/vector.md: Ditto.
1435 * config/rs6000/vsx.md: Ditto.
1437 2018-02-21 Jason Merrill <jason@redhat.com>
1439 PR c++/84314 - ICE with templates and fastcall attribute.
1440 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1442 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1444 * ipa-cp.c (determine_versionability): Fix comment typos.
1446 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1449 * ipa-cp.c (determine_versionability): Do not version functions caling
1452 2018-02-21 Martin Liska <mliska@suse.cz>
1455 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1456 Add "native" as a possible value.
1457 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1458 the macro when native cpu detection is available.
1460 2018-02-21 Martin Liska <mliska@suse.cz>
1463 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1464 Add "native" as a possible value.
1465 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1466 when native cpu detection is available.
1468 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1469 Martin Sebor <msebor@redhat.com>
1471 PR tree-optimization/84478
1472 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1474 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1475 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1476 support which is conservatively correct, for 2 only stay conservative
1477 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1478 argument to the 2 argument get_range_strlen, adjust 6 arg
1479 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1481 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1482 (gimple_fold_builtin_strlen): Pass true as last argument to
1485 2018-02-20 Martin Sebor <msebor@redhat.com>
1488 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1489 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1490 (builtin_memref::builtin_memref): Factor out parts into
1491 set_base_and_offset and call it.
1493 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1496 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1497 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1498 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1499 search at the associated MODE_INT.
1501 2018-02-20 Jeff Law <law@redhat.com>
1504 PR tree-optimization/81592
1506 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1507 for range data rather than using global data.
1508 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1509 range data rather than using global data.
1510 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1511 pass it to children as needed.
1512 (struct directive::fmtresult): Similarly.
1513 (struct directive::set_width): Similarly.
1514 (struct directive::set_precision): Similarly.
1515 (format_integer, format_directive, parse_directive): Similarly.
1516 (format_none): Accept unnamed vr_values parameter.
1517 (format_percent, format_floating, format_character): Similarly.
1518 (format_string, format_plain): Similarly.
1519 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1520 the EVRP range analyzer for range data rather than using global data.
1521 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1522 gimple-ssa-evrp-analyze.h
1523 (class sprintf_dom_walker): Add after_dom_children member function.
1524 Add evrp_range_analyzer member.
1525 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1526 range analyzer as needed.
1527 (sprintf_dom_walker::after_dom_children): New member function.
1528 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1530 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1531 (evrp_range_analyzer::pop_to_marker): Likewise.
1533 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1535 PR tree-optimization/84419
1536 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1537 with the required type if its current type is compatible but
1540 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1543 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1544 after vectorization.
1546 2018-02-20 Martin Liska <mliska@suse.cz>
1549 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1550 possible values if we don't have a hint.
1552 2018-02-20 Martin Liska <mliska@suse.cz>
1556 * final.c (shorten_branches): Build align_tab array with one
1558 * opts.c (finish_options): Add alignment option limit check.
1559 (MAX_CODE_ALIGN): Likewise.
1560 (MAX_CODE_ALIGN_VALUE): Likewise.
1561 * doc/invoke.texi: Document maximum allowed option value for
1562 all -falign-* options.
1564 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1567 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1568 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1569 * var-tracking.c (emit_note_insn_var_location): Remove all references
1570 to NOTE_INSN_CALL_ARG_LOCATION.
1571 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1572 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1573 Use copy_rtx_if_shared.
1574 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1575 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1576 (dwarf2out_var_location): Remove handling of
1577 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1579 * final.c (final_scan_insn): Remove all references to
1580 NOTE_INSN_CALL_ARG_LOCATION.
1581 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1582 before dumping final insns.
1583 * except.c (emit_note_eh_region_end): Remove all references to
1584 NOTE_INSN_CALL_ARG_LOCATION.
1585 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1586 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1587 * config/arc/arc.c (hwloop_optimize): Likewise.
1588 * config/arm/arm.c (create_fix_barrier): Likewise.
1589 * config/s390/s390.c (s390_chunkify_start): Likewise.
1590 * config/sh/sh.c (find_barrier): Likewise.
1591 * config/i386/i386.c (rest_of_insert_endbranch,
1592 ix86_seh_fixup_eh_fallthru): Likewise.
1593 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1594 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1595 * config/frv/frv.c (frv_function_prologue): Likewise.
1596 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1598 (note_outside_basic_block_p): Remove all references to
1599 NOTE_INSN_CALL_ARG_LOCATION.
1600 * gengtype.c (adjust_field_rtx_def): Likewise.
1601 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1603 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1604 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1607 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1610 PR tree-optimization/84452
1611 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1612 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1615 2018-02-19 Martin Liska <mliska@suse.cz>
1618 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1620 2018-02-19 Martin Liska <mliska@suse.cz>
1621 Richard Sandiford <richard.sandiford@linaro.org>
1623 PR tree-optimization/82491
1624 * gimple-fold.c (get_base_constructor): Make earlier bail out
1627 2018-02-19 Carl Love <cel@us.ibm.com>
1629 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1630 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1632 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1633 P8V_BUILTIN_VEC_NEG.
1635 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1637 * config/rl78/rl78.md (movdf): New define expand.
1639 2018-02-19 Martin Liska <mliska@suse.cz>
1642 * doc/invoke.texi: Fix typo.
1643 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1645 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1647 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1648 handle rs6000_single_float and rs6000_double_float specially for
1651 2018-02-16 Jeff Law <law@redhat.com>
1653 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1656 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1659 * ipa-inline.c (inline_small_functions): Fix a typo.
1661 2018-02-16 Nathan Sidwell <nathan@acm.org>
1663 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1665 2018-02-16 Carl Love <cel@us.ibm.com>
1667 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1668 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1669 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1670 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1671 expansion to P8V_BUILTIN_VEC_FLOAT2.
1673 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1675 PR rtl-optimization/70023
1676 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1677 src_regno into account.
1679 2018-02-16 Carl Love <cel@us.ibm.com>
1681 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1682 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1683 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1684 * config/rs6000/rs6000.c: Remove case statements for
1685 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1686 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1687 and P9V_BUILTIN_VEC_VINSERT4B.
1688 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1689 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1690 * config/rs6000/vsx.md:
1691 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1694 2018-02-16 Carl Love <cel@us.ibm.com>
1696 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1698 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1700 * config/rs6000/rs6000-c.c: Add the definitions for
1701 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1702 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1703 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1704 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1705 definition for insert4b and define insn *insert3b_internal.
1706 * doc/extend.texi: Add documentation for vec_extract4b.
1708 2018-02-16 Nathan Sidwell <nathan@acm.org>
1710 * doc/extend.texi (Backwards Compatibility): Mention friend
1711 injection. Note for-scope is deprecated.
1712 * doc/invoke.texi (-ffriend-injection): Deprecate.
1714 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1716 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1717 that moved to I2, also allow destinations that are a paradoxical
1718 subreg (instead of a normal reg).
1720 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1723 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1726 2018-02-16 Richard Biener <rguenther@suse.de>
1728 PR tree-optimization/84037
1729 PR tree-optimization/84016
1731 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1732 Adjust vec_construct for the fact we need additional higher latency
1733 128bit inserts for AVX256 and AVX512 vector builds.
1734 (ix86_add_stmt_cost): Scale vector construction cost for
1737 2018-02-16 Richard Biener <rguenther@suse.de>
1739 PR tree-optimization/84417
1740 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1741 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1742 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1744 2018-02-16 Martin Liska <mliska@suse.cz>
1747 * internal-fn.def (ASAN_CHECK): Set proper flags.
1748 (ASAN_MARK): Likewise.
1750 2018-02-16 Julia Koval <julia.koval@intel.com>
1752 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1753 from PTA_CANNONLAKE.
1755 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1758 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1759 Use ++iter rather than iter++ for std::list iterators.
1760 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1761 defer deleting them until all nodes in the forest are processed. Do
1762 free even leaf nodes. Change to_process into auto_vec.
1765 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1766 * vec.h (vec_default_construct): Use memset instead of placement new
1767 if BROKEN_VALUE_INITIALIZATION is defined.
1768 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1769 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1772 PR rtl-optimization/83723
1773 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1774 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1775 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1777 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1779 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1781 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1783 PR rtl-optimization/81443
1784 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1785 from inner REGs to paradoxical SUBREGs.
1787 2018-02-16 Richard Biener <rguenther@suse.de>
1789 PR tree-optimization/84399
1790 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1791 For operands we can analyze at their definition make sure we can
1792 analyze them at each use as well.
1794 2018-02-16 Richard Biener <rguenther@suse.de>
1796 PR tree-optimization/84190
1797 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1798 volatile accesses if the decl isn't volatile.
1800 2018-02-15 Jason Merrill <jason@redhat.com>
1802 PR c++/84314 - ICE with templates and fastcall attribute.
1803 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1804 TYPE_CANONICAL on an existing type.
1806 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1808 PR tree-optimization/84383
1809 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1810 dstoff nor call operand_equal_p if dstbase is NULL.
1812 PR tree-optimization/84334
1813 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1814 also a CONSTANT_CLASS_P, punt.
1816 2018-02-14 Jim Wilson <jimw@sifive.com>
1818 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1819 first SMALL_OPERAND check. New local min_second_step. Move assert
1820 to where locals are set. Add TARGET_RVC support.
1821 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1823 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1825 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1827 2018-02-14 Martin Sebor <msebor@redhat.com>
1829 PR tree-optimization/83698
1830 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1831 arrays constrain the offset range to their bounds.
1832 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1833 (builtin_access::overlap): Avoid setting the size of overlap if it's
1835 (maybe_diag_overlap): Also consider arrays when deciding what values
1836 of offsets to include in diagnostics.
1838 2018-02-14 Martin Sebor <msebor@redhat.com>
1841 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1842 that correspond to the kind of a declaration.
1844 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1847 * config/pa/pa.md: Load address of PIC label using the linkage table
1848 if the label is nonlocal.
1850 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1852 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1853 warning message if user requests -maltivec=be.
1854 * doc/invoke.texi: Document deprecation of -maltivec=be.
1856 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1859 * config/rs6000/rs6000-c.c: Update definitions for
1860 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1861 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1863 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1866 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1867 add _get_ssp intrinsics. Remove argument from
1868 __builtin_ia32_rdssp[d|q].
1869 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1870 * config/i386/i386-builtin.def: Remove argument from
1871 __builtin_ia32_rdssp[d|q].
1872 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1873 ix86_expand_special_args_builtin for _rdssp[d|q].
1874 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1875 Clear register before usage.
1876 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1877 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1879 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1881 PR tree-optimization/84357
1882 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1883 operand 1 of an ARRAY_REF too.
1885 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1888 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1889 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1891 (set_of_reg): New struct.
1892 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1893 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1894 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1896 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1897 Split into bitclr, bitset, bitinvert patterns if appropriate.
1898 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1899 use rx_fuse_in_memory_bitop.
1900 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1901 to named insn, correct maximum insn length.
1903 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1906 * machmode.def: Define a complex mode for PARTIAL_INT.
1907 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1909 * doc/rtl.texi: Document CSPImode.
1910 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1912 (msp430_hard_regno_nregs_with_padding): Likewise.
1914 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1917 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1919 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1921 PR rtl-optimization/84169
1922 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1923 we generated a parallel as new i3 and we split that to new i2 and i3
1924 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1925 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1926 those to i2, not i1. Partially rewrite this scan code.
1928 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1931 * stor-layout.c (place_field): For variable length fields, adjust
1932 offset_align afterwards not just based on the field's alignment,
1933 but also on the size.
1936 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1937 of exps and logs in the use_exp2 case.
1939 2018-02-13 Jeff Law <law@redhat.com>
1941 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1944 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1947 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1951 * common.opt (gas-loc-support, gas-locview-support): New.
1952 (ginline-points, ginternal-reset-location-views): New.
1953 * doc/invoke.texi: Document them. Use @itemx where intended.
1954 (gvariable-location-views): Adjust.
1955 * target.def (reset_location_view): New.
1956 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1957 (TARGET_RESET_LOCATION_VIEW): New.
1958 * doc/tm.texi: Rebuilt.
1959 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1960 (dwarf2out_default_as_locview_support): New.
1961 (output_asm_line_debug_info): Use option variables.
1962 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1963 (output_loc_list): Likewise.
1964 (add_high_low_attributes): Check option variables.
1965 Don't output entry view attribute in strict mode.
1966 (gen_inlined_subroutine_die): Check option variables.
1967 (dwarf2out_inline_entry): Likewise.
1968 (init_sections_and_labels): Likewise.
1969 (dwarf2out_early_finish): Likewise.
1970 (maybe_reset_location_view): New, from...
1971 (dwarf2out_var_location): ... here. Call it.
1972 * debug.h (dwarf2out_default_as_loc_support): Declare.
1973 (dwarf2out_default_as_locview_support): Declare.
1974 * hooks.c (hook_int_rtx_insn_0): New.
1975 * hooks.h (hook_int_rtx_insn_0): Declare.
1976 * toplev.c (process_options): Take -gas-loc-support and
1977 -gas-locview-support from dwarf2out. Enable
1978 -gvariable-location-views by default only with locview
1979 assembler support. Enable -ginternal-reset-location-views by
1980 default only if the target defines the corresponding hook.
1981 Enable -ginline-points by default if location views are
1982 enabled; force it disabled if statement frontiers are
1984 * tree-inline.c (expand_call_inline): Check option variables.
1985 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1987 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1989 PR tree-optimization/84321
1990 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1991 handling. Also check whether the anti-range contains any values
1992 that satisfy the mask; switch to a VR_RANGE if not.
1994 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1997 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1999 2018-02-13 Martin Jambor <mjambor@suse.cz>
2002 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2003 of call statements, also set location of a load to a temporary.
2005 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2007 * config/rl78/rl78.c (add_vector_labels): New function.
2008 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2009 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2010 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2011 which checks that no arguments are passed.
2012 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2013 * doc/extend.texi: Documentation for the new attribute.
2015 2018-02-13 Andreas Schwab <schwab@suse.de>
2017 * config/riscv/linux.h (CPP_SPEC): Define.
2019 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2022 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2023 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2024 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2025 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2026 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2027 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2028 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2029 temporarily for AES and PCLMUL builtins.
2031 PR tree-optimization/84339
2032 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2033 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2037 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2038 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2039 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2041 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2043 * omp-simd-clone.h: New file.
2044 * omp-simd-clone.c: Include omp-simd-clone.h.
2045 (expand_simd_clones): No longer static.
2046 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2047 cgraph.h and omp-simd-clone.h.
2048 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2049 (vect_recog_widen_shift_pattern): Formatting fix.
2050 (vect_pattern_recog_1): Don't check optab for calls.
2053 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2054 operands[2] into a REG before using gen_lowpart on it.
2056 2018-02-12 Jeff Law <law@redhat.com>
2059 * config/sh/sh.c (find_barrier): Consider a sibling call
2062 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2063 successfully back substituting a reg.
2065 2018-02-12 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/84037
2068 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2069 parameter, move visited init to caller.
2070 (vect_slp_analyze_operations): Separate cost from validity
2071 check, initialize visited once for all instances.
2072 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2074 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2076 (vect_model_promotion_demotion_cost): Likewise.
2077 (vectorizable_bswap): Guard cost modeling with !slp_node
2078 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2080 (vectorizable_call): Likewise.
2081 (vectorizable_conversion): Likewise.
2082 (vectorizable_assignment): Likewise.
2083 (vectorizable_shift): Likewise.
2084 (vectorizable_operation): Likewise.
2085 (vectorizable_store): Likewise.
2086 (vectorizable_load): Likewise.
2087 (vectorizable_condition): Likewise.
2088 (vectorizable_comparison): Likewise.
2090 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2093 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2094 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2095 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2097 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2100 * params.def (inline-min-speedup): Increase from 8 to 15.
2101 (max-inline-insns-auto): Decrease from 40 to 30.
2102 * ipa-split.c (consider_split): Add some buffer for function to
2103 be considered inlining candidate.
2104 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2107 2018-02-12 Richard Biener <rguenther@suse.de>
2109 PR tree-optimization/84037
2110 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2111 matched stmts if we cannot swap the non-matched ones.
2113 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2115 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2116 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2117 _mm_maskz_scalef_round_ss): New intrinsics.
2118 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2119 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2120 __builtin_ia32_scalefss_round): Remove.
2121 (__builtin_ia32_scalefsd_mask_round,
2122 __builtin_ia32_scalefss_mask_round): New intrinsics.
2123 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2124 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2125 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2126 "<round_constraint>")): Changed to ...
2127 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2128 "<round_scalar_constraint>")): ... this.
2129 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2130 %0, %1, %2<round_op3>}"): Changed to ...
2131 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2132 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2133 %2<round_scalar_mask_op3>}"): ... this.
2134 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2136 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2138 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2139 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2140 (_mm_maskz_sqrt_round_ss): New intrinsics.
2141 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2142 (__builtin_ia32_sqrtsd_mask_round)
2143 (__builtin_ia32_sqrtss_mask_round): New builtins.
2144 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2145 (__builtin_ia32_sqrtss_round): Remove.
2146 (__builtin_ia32_sqrtsd_mask_round)
2147 (__builtin_ia32_sqrtss_mask_round): New builtins.
2148 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2149 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2150 ((match_operand:VF_128 1 "vector_operand"
2151 "xBm,<round_constraint>")): Changed to ...
2152 ((match_operand:VF_128 1 "vector_operand"
2153 "xBm,<round_scalar_constraint>")): ... this.
2154 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2155 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2156 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2157 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2158 %<iptr>1<round_scalar_mask_op3>}): ... this.
2159 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2160 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2162 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2165 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2166 Cast vec_cmpeq result to correct type.
2167 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2168 Cast vec_cmpgt result to correct type.
2170 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2172 * final.c (final_scan_insn_1): Renamed from...
2173 (final_scan_insn): ... this. New wrapper, to recover
2174 seen from the outermost call in recursive ones.
2175 * config/sparc/sparc.c (output_return): Drop seen from call.
2176 (output_sibcall): Likewise.
2177 * config/visium/visium.c (output_branch): Likewise.
2179 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2181 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2184 2018-02-10 Alan Modra <amodra@gmail.com>
2187 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2188 Specify LR as an input.
2190 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2193 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2194 remove_member_access_dummy_vars): New functions.
2195 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2196 lower_omp_1, execute_lower_omp): Use them.
2198 PR rtl-optimization/84308
2199 * shrink-wrap.c (spread_components): Release todo vector.
2201 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2203 PR rtl-optimization/57193
2204 * ira-color.c (struct allocno_color_data): Add member
2205 conflict_allocno_hard_prefs.
2206 (update_conflict_allocno_hard_prefs): New.
2207 (bucket_allocno_compare_func): Add a preference based on
2208 conflict_allocno_hard_prefs.
2209 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2210 (color_allocnos): Remove a dead code. Initiate
2211 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2213 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2216 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2217 constraint from =wa to wa. Avoid a subreg on the output operand,
2218 instead use a pseudo and subreg it in a move.
2219 (p9_xxbrd_<mode>): Changed to ...
2220 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2221 (p9_xxbrd_v2df): New expander.
2222 (p9_xxbrw_<mode>): Changed to ...
2223 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2224 (p9_xxbrw_v4sf): New expander.
2226 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2228 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2230 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2233 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2234 multiply in 32-bit mode.
2235 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2236 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2239 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2241 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2242 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2243 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2244 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2246 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2249 * dwarf2out.c (is_trivial_indirect_ref): New function.
2250 (dwarf2out_late_global_decl): Do not generate a location
2251 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2252 and that are not defined in the current unit.
2254 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2256 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2257 instead of a libcall for UNORDERED.
2259 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2262 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2263 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2265 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2268 * config/s390/s390.c (s390_set_current_function): Invoke
2269 s390_indirect_branch_settings also if fndecl didn't change.
2271 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2273 * config/rs6000/rs6000.md (blockage): Set length to zero.
2275 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2277 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2279 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2282 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2283 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2287 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2288 PARALLEL incoming that failed vt_get_decl_and_offset check.
2291 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2292 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2294 (get_variable_section): For decls in named .bss* sections pass true as
2295 second argument to bss_initializer_p.
2297 2018-02-09 Marek Polacek <polacek@redhat.com>
2298 Jakub Jelinek <jakub@redhat.com>
2301 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2302 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2303 Sync some changes from cxx_fold_indirect_ref.
2305 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2307 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2309 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2310 (BLOCK_INLINE_ENTRY_LABEL): New.
2311 (dwarf2out_var_location): Disregard inline entry markers.
2312 (inline_entry_data): New struct.
2313 (inline_entry_data_hasher): New hashtable type.
2314 (inline_entry_data_hasher::hash): New.
2315 (inline_entry_data_hasher::equal): New.
2316 (inline_entry_data_table): New variable.
2317 (add_high_low_attributes): Add DW_AT_entry_pc and
2318 DW_AT_GNU_entry_view attributes if a pending entry is found
2319 in inline_entry_data_table. Add old entry_pc attribute only
2320 if debug nonbinding markers are disabled.
2321 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2322 markers are enabled.
2323 (block_within_block_p, dwarf2out_inline_entry): New.
2324 (dwarf2out_finish): Check that no entries remained in
2325 inline_entry_data_table.
2326 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2327 (final_scan_insn, notice_source_line): Likewise.
2328 (rest_of_clean_state): Skip inline entry markers.
2329 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2331 * gimple.c (gimple_build_debug_inline_entry): New.
2332 * gimple.h (enum gimple_debug_subcode): Add
2333 GIMPLE_DEBUG_INLINE_ENTRY.
2334 (gimple_build_debug_inline_entry): Declare.
2335 (gimple_debug_inline_entry_p): New.
2336 (gimple_debug_nonbind_marker_p): Adjust.
2337 * insn-notes.def (INLINE_ENTRY): New.
2338 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2339 inline entry marker notes.
2340 (print_insn): Likewise.
2341 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2342 (INSN_DEBUG_MARKER_KIND): Likewise.
2343 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2344 * tree-inline.c (expand_call_inline): Build and insert
2345 debug_inline_entry stmt.
2346 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2347 inline entry blocks early, if nonbind markers are enabled.
2348 (dump_scope_block): Dump fragment info.
2349 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2350 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2351 (gimple_build_debug_inline_entry): New.
2352 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2353 Enable/disable inline entry points too.
2354 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2355 (DEBUG_INSN): Describe inline entry markers.
2357 * common.opt (gvariable-location-views): New.
2358 (gvariable-location-views=incompat5): New.
2359 * config.in: Rebuilt.
2360 * configure: Rebuilt.
2361 * configure.ac: Test assembler for view support.
2362 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2363 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2364 * dwarf2out.c (var_loc_view): New typedef.
2365 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2366 (dwarf2out_locviews_in_attribute): New.
2367 (dwarf2out_locviews_in_loclist): New.
2368 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2369 (enum dw_line_info_opcode): Add LI_adv_address.
2370 (struct dw_line_info_table): Add view.
2371 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2372 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2373 (zero_view_p): New variable.
2374 (ZERO_VIEW_P): New macro.
2375 (output_asm_line_debug_info): New.
2376 (struct var_loc_node): Add view.
2377 (add_AT_view_list, AT_loc_list): New.
2378 (add_var_loc_to_decl): Add view param. Test it against last.
2379 (new_loc_list): Add view params. Record them.
2380 (AT_loc_list_ptr): Handle loc and view lists.
2381 (view_list_to_loc_list_val_node): New.
2382 (print_dw_val): Handle dw_val_class_view_list.
2383 (size_of_die): Likewise.
2384 (value_format): Likewise.
2385 (loc_list_has_views): New.
2386 (gen_llsym): Set vl_symbol too.
2387 (maybe_gen_llsym, skip_loc_list_entry): New.
2388 (dwarf2out_maybe_output_loclist_view_pair): New.
2389 (output_loc_list): Output view list or entries too.
2390 (output_view_list_offset): New.
2391 (output_die): Handle dw_val_class_view_list.
2392 (output_dwarf_version): New.
2393 (output_compilation_unit_header): Use it.
2394 (output_skeleton_debug_sections): Likewise.
2395 (output_rnglists, output_line_info): Likewise.
2396 (output_pubnames, output_aranges): Update version comments.
2397 (output_one_line_info_table): Output view numbers in asm comments.
2398 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2399 Call maybe_gen_llsym.
2400 (loc_list_from_tree_1): Adjust.
2401 (add_AT_location_description): Create view list attribute if
2402 needed, check it's absent otherwise.
2403 (convert_cfa_to_fb_loc_list): Adjust.
2404 (maybe_emit_file): Call output_asm_line_debug_info for test.
2405 (dwarf2out_var_location): Reset views as needed. Precompute
2406 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2407 attribute. Set view.
2408 (new_line_info_table): Reset next view.
2409 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2410 (dwarf2out_source_line): Likewise. Output view resets and labels to
2411 the assembler, or select appropriate line info opcodes.
2412 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2413 (optimize_string_length): Catch it. Adjust.
2414 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2415 dw_val_class_view_list, and remove it if no longer needed.
2416 (hash_loc_list): Hash view numbers.
2417 (loc_list_hasher::equal): Compare them.
2418 (optimize_location_lists): Check whether a view list symbol is
2419 needed, and whether the locview attribute is present, and
2420 whether they match. Remove the locview attribute if no longer
2422 (index_location_lists): Call skip_loc_list_entry for test.
2423 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2424 Use output_dwarf_version.
2425 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2426 (struct dw_val_node): Add val_view_list.
2427 * final.c (SEEN_NEXT_VIEW): New.
2428 (set_next_view_needed): New.
2429 (clear_next_view_needed): New.
2430 (maybe_output_next_view): New.
2431 (final_start_function): Rename to...
2432 (final_start_function_1): ... this. Take pointer to FIRST,
2433 add SEEN parameter. Emit param bindings in the initial view.
2434 (final_start_function): Reintroduce SEEN-less interface.
2435 (final): Rename to...
2436 (final_1): ... this. Take SEEN parameter. Output final pending
2437 next view at the end.
2438 (final): Reintroduce seen-less interface.
2439 (final_scan_insn): Output pending next view before switching
2440 sections or ending a block. Mark the next view as needed when
2441 outputting variable locations. Notify debug backend of section
2442 changes, and of location view changes.
2443 (rest_of_handle_final): Adjust.
2444 * toplev.c (process_options): Autodetect value for debug variable
2445 location views option. Warn on incompat5 without -gdwarf-5.
2446 * doc/invoke.texi (gvariable-location-views): New.
2447 (gvariable-location-views=incompat5): New.
2448 (gno-variable-location-views): New.
2450 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2452 PR tree-optimization/84136
2453 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2454 that the result of find_edge is non-NULL.
2456 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2459 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2460 storing integer register in SImode. Fix cost of 256 and 512
2461 byte aligned SSE register store.
2463 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2465 * config/i386/i386.c (ix86_multiplication_cost): Fix
2466 multiplication cost for TARGET_AVX512DQ.
2468 2018-02-08 Marek Polacek <polacek@redhat.com>
2470 PR tree-optimization/84238
2471 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2474 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2476 PR tree-optimization/84265
2477 * tree-vect-stmts.c (vectorizable_store): Don't treat
2478 VMAT_CONTIGUOUS accesses as grouped.
2479 (vectorizable_load): Likewise.
2481 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2483 PR tree-optimization/81635
2484 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2485 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2486 (test_round_for_mask): New functions.
2487 (wide_int_cc_tests): Call test_round_for_mask.
2488 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2489 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2490 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2491 range returned by get_range_info.
2493 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2496 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2497 * symtab.c: Include builtins.h
2498 (symtab_node::output_to_lto_symbol_table_p): Move here
2499 from lto-streamer-out.c:output_symbol_p.
2500 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2501 (output_symbol_p): Move all logic to symtab.c
2502 (produce_symtab): Update.
2504 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2506 * config/s390/s390-opts.h (enum indirect_branch): Define.
2507 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2508 (s390_indirect_branch_via_thunk)
2509 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2510 (enum s390_indirect_branch_type): Define.
2511 * config/s390/s390.c (struct s390_frame_layout, struct
2512 machine_function): Remove.
2513 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2514 (indirect_branch_table_label_no, indirect_branch_table_name):
2516 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2517 (enum s390_indirect_branch_option): Define.
2518 (s390_return_addr_from_memory): New function.
2519 (s390_handle_string_attribute): New function.
2520 (s390_attribute_table): Add new attribute handler.
2521 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2522 (s390_indirect_branch_via_thunk): New function.
2523 (s390_indirect_branch_via_inline_thunk): New function.
2524 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2525 sibling call optimization for non z10 compiles.
2526 (s390_emit_call): Force indirect branch target to be a single
2527 register. Add r1 clobber for non-z10 compiles.
2528 (s390_emit_epilogue): Emit return jump via return_use expander.
2529 (s390_reorg): Handle JUMP_INSNs as execute targets.
2530 (s390_option_override_internal): Perform validity checks for the
2531 new command line options.
2532 (s390_indirect_branch_attrvalue): New function.
2533 (s390_indirect_branch_settings): New function.
2534 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2535 (s390_output_indirect_thunk_function): New function.
2536 (s390_code_end): Implement target hook.
2537 (s390_case_values_threshold): Implement target hook.
2538 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2540 * config/s390/s390.h (struct s390_frame_layout)
2541 (struct machine_function): Move here from s390.c.
2542 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2543 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2544 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2545 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2546 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2547 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2548 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2549 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2550 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2551 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2552 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2553 (mnemonic attribute): Add values which aren't recognized
2555 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2556 pattern for branch conversion. Fix mnemonic attribute.
2557 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2558 indirect branch via thunk if requested.
2559 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2560 ("*indirect_jump"): Disable for branch conversion using out of
2562 ("indirect_jump_via_thunk<mode>_z10")
2563 ("indirect_jump_via_thunk<mode>")
2564 ("indirect_jump_via_inlinethunk<mode>_z10")
2565 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2566 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2567 ("casesi_jump_via_inlinethunk<mode>_z10")
2568 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2569 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2570 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2571 ("*indirect2_jump"): Disable for branch conversion.
2572 ("casesi_jump"): Turn into expander and expand patterns for branch
2574 ("return_use"): New expander.
2575 ("*return"): Emit return via thunk and rename it to ...
2576 ("*return<mode>"): ... this one.
2577 * config/s390/s390.opt: Add new options and and enum for the
2580 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2582 * lra-constraints.c (match_reload): Unconditionally use
2583 gen_lowpart_SUBREG, rather than selecting between that
2584 and equivalent gen_rtx_SUBREG code.
2586 2018-02-08 Richard Biener <rguenther@suse.de>
2588 PR tree-optimization/84233
2589 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2590 changed flag instead of boguously re-using phi_inserted.
2592 2018-02-08 Martin Jambor <mjambor@suse.cz>
2594 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2595 static local variables.
2597 2018-02-08 Richard Biener <rguenther@suse.de>
2599 PR tree-optimization/84278
2600 * tree-vect-stmts.c (vectorizable_store): When looking for
2601 smaller vector types to perform grouped strided loads/stores
2602 make sure the mode is supported by the target.
2603 (vectorizable_load): Likewise.
2605 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2607 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2608 Increase LDP/STP opportunities by adding adjacent callee-saves.
2610 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2612 PR rtl-optimization/84068
2613 PR rtl-optimization/83459
2614 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2616 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2618 PR tree-optimization/84224
2619 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2620 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2623 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2626 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2627 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2628 count, remove one USE.
2630 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2632 * doc/install.texi (Configuration): Document the
2633 --with-long-double-format={ibm,ieee} PowerPC configuration
2637 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2638 Convert from define_expand to be define_insn_and_split. Rework
2639 float/double/_Float128 conversions to QI/HI/SImode to work with
2640 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2641 conversions to QI/HImode types did a store and then a load to
2642 truncate the value. For conversions to VSX registers, don't split
2643 the insn, instead emit the code directly. Use the code iterator
2644 any_fix to combine signed and unsigned conversions.
2645 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2646 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2647 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2648 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2649 (fix_<mode>di2_hw): Likewise.
2650 (fixuns_<mode>di2_hw): Likewise.
2651 (fix_<mode>si2_hw): Likewise.
2652 (fixuns_<mode>si2_hw): Likewise.
2653 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2654 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2655 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2656 fix<uns>_trunc<SFDF:mode>si2_p8.
2657 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2659 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2660 (fix<uns>_<mode>_mem): Likewise.
2661 (fctiw<u>z_<mode>_mem): Likewise.
2662 (fix<uns>_<mode>_mem): Likewise.
2663 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2664 the register allocator from doing a direct move to the GPRs to do
2665 a store, and instead use the ISA 3.0 store byte/half-word from
2666 vector register instruction. For IEEE 128-bit floating point,
2667 also optimize stores of 32-bit ints.
2668 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2670 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2672 * genextract.c (push_pathstr_operand): New function to support
2674 (walk_rtx): Call push_pathstr_operand.
2675 (print_path): Support [a-zA-Z].
2677 2018-02-07 Richard Biener <rguenther@suse.de>
2679 PR tree-optimization/84037
2680 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2681 (cse_and_gimplify_to_preheader): Declare.
2682 (vect_get_place_in_interleaving_chain): Likewise.
2683 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2685 (_loop_vec_info::~_loop_vec_info): Delete it.
2686 (cse_and_gimplify_to_preheader): New function.
2687 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2688 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2689 (vectorizable_load): Likewise. For grouped stores always base
2690 the IV on the first element.
2691 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2692 condition before gimplifying.
2694 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2696 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2697 *DIV_EXPR and *MOD_EXPR.
2699 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2702 * config/i386/i386.c (ix86_option_override_internal): Mask out
2703 the CF_SET bit when checking -fcf-protection.
2705 2018-02-07 Tom de Vries <tom@codesourcery.com>
2708 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2711 2018-02-07 Richard Biener <rguenther@suse.de>
2713 PR tree-optimization/84204
2714 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2717 PR tree-optimization/84205
2718 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2719 special-case isl_ast_op_zdiv_r.
2721 PR tree-optimization/84223
2722 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2723 Only add conditions from within the region.
2724 (gather_bbs::after_dom_children): Adjust.
2726 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2729 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2730 * config/avr/avr.md: Only post-reload split REG-REG moves if
2731 either register is GENERAL_REG_P.
2733 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2735 PR tree-optimization/84235
2736 * tree-ssa-scopedtables.c
2737 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2738 if the subtraction is performed in floating point type where NaNs are
2739 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2740 build 1. Formatting fix.
2742 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2745 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2746 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2747 and skip it regardless of bb boundaries. Use CALL_P macro,
2748 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2749 unnecessarily, formatting fix.
2751 2018-02-06 Michael Collison <michael.collison@arm.com>
2753 * config/arm/thumb2.md:
2754 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2755 (*thumb_mov_notscc): Ditto.
2757 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2760 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2761 unsigned_fix, not "s".
2763 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2765 * configure.ac (gcc_fn_eh_frame_ro): New function.
2766 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2767 correct .eh_frame permissions.
2768 * configure: Regenerate.
2770 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2772 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2775 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2777 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2778 Display warning message for -mno-speculate-indirect-jumps.
2780 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2782 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2784 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2786 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2788 PR tree-optimization/84225
2789 * tree-eh.c (find_trapping_overflow): Only call
2790 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2792 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2795 * config/i386/i386.c: Reimplement the check of possible options
2796 -mibt/-mshstk conbination. Change error messages.
2797 * doc/invoke.texi: Fix a typo: remove extra '='.
2799 2018-02-06 Marek Polacek <polacek@redhat.com>
2801 PR tree-optimization/84228
2802 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2804 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2807 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2808 emitted arch directives.
2809 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2810 __ARM_FEATURE_COPROC before changing architectures.
2812 2018-02-06 Richard Biener <rguenther@suse.de>
2814 * config/i386/i386.c (print_reg): Fix typo.
2815 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2817 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2819 * configure: Regenerate.
2821 2018-02-05 Martin Sebor <msebor@redhat.com>
2823 PR tree-optimization/83369
2824 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2827 2018-02-05 Martin Liska <mliska@suse.cz>
2829 * doc/invoke.texi: Cherry-pick upstream r323995.
2831 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2833 * ira.c (ira_init_register_move_cost): Adjust comment.
2835 2018-02-05 Martin Liska <mliska@suse.cz>
2837 PR gcov-profile/84137
2838 * doc/gcov.texi: Fix typo in documentation.
2840 2018-02-05 Martin Liska <mliska@suse.cz>
2842 PR gcov-profile/83879
2843 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2844 using dlopen functionality.
2846 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2848 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2849 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2850 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2851 _mm_maskz_range_round_ss): New intrinsics.
2852 (__builtin_ia32_rangesd128_round)
2853 (__builtin_ia32_rangess128_round): Remove.
2854 (__builtin_ia32_rangesd128_mask_round,
2855 __builtin_ia32_rangess128_mask_round): New builtins.
2856 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2857 __builtin_ia32_rangess128_round): Remove.
2858 (__builtin_ia32_rangesd128_mask_round,
2859 __builtin_ia32_rangess128_mask_round): New builtins.
2860 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2861 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2862 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2863 "<round_saeonly_constraint>")): Changed to ...
2864 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2865 "<round_saeonly_scalar_constraint>")): ... this.
2866 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2867 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2868 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2869 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2870 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2872 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2874 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2876 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2877 Remove all values except native, 8540 and 8548.
2879 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2881 * config/i386/i386.c (ix86_output_function_return): Pass
2882 INVALID_REGNUM, instead of -1, as invalid register number to
2883 indirect_thunk_name and output_indirect_thunk.
2885 2018-02-02 Julia Koval <julia.koval@intel.com>
2887 * config.gcc: Add -march=icelake.
2888 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2889 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2890 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2891 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2892 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2893 (processor_target_table): Add icelake.
2894 (ix86_option_override_internal): Handle new PTAs.
2895 (get_builtin_code_for_version): Handle icelake.
2896 (M_INTEL_COREI7_ICELAKE): New.
2897 (fold_builtin_cpu): Handle icelake.
2898 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2899 * doc/invoke.texi: Add -march=icelake.
2901 2018-02-02 Julia Koval <julia.koval@intel.com>
2903 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2904 to wide_int_bitmask.
2905 * wide-int-bitmask.h: New.
2907 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2910 * config/i386/i386.md: Replace Pmode with word_mode in
2911 builtin_setjmp_setup and builtin_longjmp to support x32.
2913 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2917 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2919 (rs6000_supported_cpu_names): New static variable.
2920 (linux_cpu_translation_table): Likewise.
2921 (elf_platform) <cpu>: Define new static variable and use it.
2922 Translate kernel AT_PLATFORM name to canonical name if needed.
2923 Error if platform name is unknown.
2925 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2928 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2930 2018-02-01 Jeff Law <law@redhat.com>
2933 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2934 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2935 the scratch if RELEASE_VIA_POP is false.
2936 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2937 If we have to save a temporary register, decrement SIZE appropriately.
2938 Pass new arguments to release_scratch_register_on_entry.
2939 (ix86_adjust_stack_and_probe): Likewise.
2940 (ix86_emit_probe_stack_range): Pass new arguments to
2941 release_scratch_register_on_entry.
2943 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2945 PR rtl-optimization/84157
2946 * combine.c (change_zero_ext): Use REG_P predicate in
2947 front of HARD_REGISTER_P predicate.
2949 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2951 * config/avr/avr.c (avr_option_override): Move disabling of
2952 -fdelete-null-pointer-checks to...
2953 * common/config/avr/avr-common.c (avr_option_optimization_table):
2956 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2958 PR tree-optimization/81635
2959 * tree-data-ref.c (split_constant_offset_1): For types that
2960 wrap on overflow, try to use range info to prove that wrapping
2963 2018-02-01 Renlin Li <renlin.li@arm.com>
2966 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2968 (aarch64_register_move_cost): Likewise.
2969 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2971 (REG_CLASS_NAMES): Likewise.
2972 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2973 TAILCALL_ADDR_REGS. Remove IP registers.
2974 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2976 2018-02-01 Richard Biener <rguenther@suse.de>
2978 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2979 for specifying RPO order and allow NULL for that.
2980 * domwalk.c (dom_walker::dom_walker): Likewise.
2981 (dom_walker::walk): Handle NULL RPO order.
2982 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2984 (rewrite_update_dom_walker): Likewise.
2985 (mark_def_dom_walker): Likewise.
2987 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2989 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2990 (aarch64_maybe_expand_sve_subreg_move): Declare.
2991 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2992 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2994 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2995 that are semantically a reverse operation.
2996 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2997 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2998 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3000 (aarch64_can_change_mode_class): For big-endian, forbid changes
3001 between two SVE modes if they have different element sizes.
3003 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3005 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3006 the TImode handling for big-endian targets.
3008 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3010 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3011 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3013 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3014 Remove BSWAP handing for big-endian targets and use the form of
3015 LD1RQ appropariate for the mode.
3017 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3019 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3020 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3023 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3026 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3027 check for operands that need to go through aarch64_sve_reload_be.
3029 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3031 PR tree-optimization/81661
3032 PR tree-optimization/84117
3033 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3034 * tree-eh.c: Include gimplify.h.
3035 (find_trapping_overflow, replace_trapping_overflow,
3036 rewrite_to_non_trapping_overflow): New functions.
3037 * tree-vect-loop.c: Include tree-eh.h.
3038 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3039 * tree-data-ref.c: Include tree-eh.h.
3040 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3042 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3044 PR rtl-optimization/84123
3045 * combine.c (change_zero_ext): Check if hard register satisfies
3046 can_change_dest_mode before calling gen_lowpart_SUBREG.
3048 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3051 * ira.c (ira_init_register_move_cost): Remove assert.
3053 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3055 PR rtl-optimization/84071
3056 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3057 * doc/tm.texi: Regenerate.
3059 2018-01-31 Richard Biener <rguenther@suse.de>
3061 PR tree-optimization/84132
3062 * tree-data-ref.c (analyze_miv_subscript): Properly
3063 check whether evolution_function_is_affine_multivariate_p
3064 before calling gcd_of_steps_may_divide_p.
3066 2018-01-31 Julia Koval <julia.koval@intel.com>
3069 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3070 * config/i386/i386.md (rdpid_rex64) New.
3071 (rdpid): Make 32bit only.
3073 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3076 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3077 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3079 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3082 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3084 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3086 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3088 PR rtl-optimization/84071
3089 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3090 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3092 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3094 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3095 (arc_attribute_table): Add 'aux' attribute.
3096 (arc_in_small_data_p): Consider aux like variables.
3097 (arc_is_aux_reg_p): New function.
3098 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3099 (arc_get_aux_arg): New function.
3100 (prepare_move_operands): Handle aux-register access.
3101 (arc_handle_aux_attribute): New function.
3102 * doc/extend.texi (ARC Variable attributes): Add subsection.
3104 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3106 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3107 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3108 (arc_attribute_table): Add 'uncached' attribute.
3109 (arc_print_operand): Print '.di' flag for uncached memory
3111 (arc_in_small_data_p): Do not consider for small data the uncached
3113 (arc_is_uncached_mem_p): New function.
3114 * config/arc/predicates.md (compact_store_memory_operand): Check
3115 for uncached memory accesses.
3116 (nonvol_nonimm_operand): Likewise.
3117 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3119 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3122 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3123 falign-loops=): Add Optimization flag.
3125 2018-01-30 Jeff Law <law@redhat.com>
3128 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3129 INT_REGISTERS_SAVED. Check it prior to calling
3130 get_scratch_register_on_entry.
3131 (ix86_adjust_stack_and_probe): Similarly.
3132 (ix86_emit_probe_stack_range): Similarly.
3133 (ix86_expand_prologue): Corresponding changes.
3135 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3138 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3139 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3141 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3144 * lra-constraints.c (curr_insn_transform): Process AND in the
3147 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3149 PR rtl-optimization/83986
3150 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3151 dependence against last_pending_memory_flush in addition to
3154 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3156 PR tree-optimization/81611
3157 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3160 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3163 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3166 2018-01-30 Richard Biener <rguenther@suse.de>
3167 Jakub Jelinek <jakub@redhat.com>
3169 PR tree-optimization/84111
3170 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3171 inner loops added during recursion, as they don't have up-to-date
3174 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3177 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3178 (can_inline_edge_by_limits_p): ... here.
3179 (can_early_inline_edge_p, check_callers,
3180 update_caller_keys, update_callee_keys, recursive_inlining,
3181 add_new_edges_to_heap, speculation_useful_p,
3182 inline_small_functions,
3183 inline_small_functions, flatten_function,
3184 inline_to_all_callers_1): Update.
3186 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3188 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3191 2018-01-30 Richard Biener <rguenther@suse.de>
3193 PR tree-optimization/83008
3194 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3195 invariant and constant vector uses in stmts when they need
3198 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3201 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3202 * configure: Regenerate.
3204 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3206 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3208 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3209 Use gen_rtx_REG rather than gen_lowpart.
3211 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3213 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3214 rather than 0 when creating partial subregs.
3216 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3218 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3221 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3224 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3225 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3226 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3227 flags. This restores the settings used before the 2017-07-24.
3228 Turning off pre increment/decrement/modify allows IVOPTS to
3229 optimize DF/SF loops where the index is an int.
3231 2018-01-29 Richard Biener <rguenther@suse.de>
3232 Kelvin Nilsen <kelvin@gcc.gnu.org>
3235 * tree-vect-stmts.c (vectorizable_call): Don't call
3236 targetm.vectorize_builtin_md_vectorized_function if callee is
3239 2018-01-22 Carl Love <cel@us.ibm.com>
3241 * doc/extend.tex: Fix typo in second arg in
3242 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3244 2018-01-29 Richard Biener <rguenther@suse.de>
3246 PR tree-optimization/84086
3247 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3248 (flush_ssaname_freelist): When SSA names were released reset
3249 the SCEV hash table.
3251 2018-01-29 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/84057
3254 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3255 removed paths when removing edges.
3257 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3259 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3260 -mfunction-return=@var{choice}.
3262 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3265 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3267 (layout::print_source_line): Likewise.
3268 (test_get_line_width_without_trailing_whitespace): Add test cases.
3270 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3273 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3276 2018-01-26 Jim Wilson <jimw@sifive.com>
3278 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3280 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3283 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3285 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3286 and CMP + SUB-immediate -> SUBS.
3288 2018-01-26 Martin Sebor <msebor@redhat.com>
3290 PR tree-optimization/83896
3291 * tree-ssa-strlen.c (get_string_len): Rename...
3292 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3293 Avoid assuming length is constant.
3294 (handle_char_store): Use HOST_WIDE_INT for string length.
3296 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3299 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3300 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3302 2018-01-26 Richard Biener <rguenther@suse.de>
3304 PR rtl-optimization/84003
3305 * dse.c (record_store): Only record redundant stores when
3306 the earlier store aliases at least all accesses the later one does.
3308 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3310 PR rtl-optimization/83985
3311 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3312 REG_CFA_RESTORE insns.
3313 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3314 REG_CFA_RESTORE insns here.
3317 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3318 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3320 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3322 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3323 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3324 (arc_init): Likewise.
3325 (arc_override_options): Likewise.
3326 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3328 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3330 * config/arc/arc.h (TARGET_DBNZ): Define.
3331 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3332 properly set the tune attribute.
3333 (dbnz): Use TARGET_DBNZ guard.
3334 * config/arc/arc.opt (mtune): Add core3 option.
3336 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3338 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3339 recognize new pic like addresses.
3340 (arc_delegitimize_address): Clean up.
3342 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3344 * config/arc/arc-arches.def: Option mrf16 valid for all
3346 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3347 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3348 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3349 * config/arc/arc-tables.opt: Regenerate.
3350 * config/arc/arc.c (arc_conditional_register_usage): Handle
3351 reduced register file case.
3352 (arc_file_start): Set must have build attributes.
3353 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3355 * config/arc/arc.opt (mrf16): Add new option.
3356 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3357 * config/arc/genmultilib.awk: Handle new mrf16 option.
3358 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3359 * config/arc/t-multilib: Regenerate.
3360 * doc/invoke.texi (ARC Options): Document mrf16 option.
3362 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3364 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3365 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3366 (arc_attribute_table): Add 'secure_call' attribute.
3367 (arc_print_operand): Print secure call operand.
3368 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3370 (arc_is_secure_call_p): New function. * config/arc/arc.md
3371 (call_i): Add support for sjli instruction.
3372 (call_value_i): Likewise.
3373 * config/arc/constraints.md (Csc): New constraint.
3375 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3376 John Eric Martin <John.Martin@emmicro-us.com>
3378 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3379 * config/arc/arc.c (_arc_jli_section): New struct.
3380 (arc_jli_section): New type.
3381 (rc_jli_sections): New static variable.
3382 (arc_handle_jli_attribute): New function.
3383 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3384 (arc_file_end): New function.
3385 (TARGET_ASM_FILE_END): Define.
3386 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3387 (arc_add_jli_section): New function.
3388 (jli_call_scan): Likewise.
3389 (arc_reorg): Call jli_call_scan.
3390 (arc_output_addsi): Remove 'S' from printing asm operand.
3391 (arc_is_jli_call_p): New function.
3392 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3394 (movhi_insn): Likewise.
3395 (movsi_insn): Likewise.
3396 (movsi_set_cc_insn): Likewise.
3397 (loadqi_update): Likewise.
3398 (load_zeroextendqisi_update): Likewise.
3399 (load_signextendqisi_update): Likewise.
3400 (loadhi_update): Likewise.
3401 (load_zeroextendhisi_update): Likewise.
3402 (load_signextendhisi_update): Likewise.
3403 (loadsi_update): Likewise.
3404 (loadsf_update): Likewise.
3405 (movsicc_insn): Likewise.
3406 (bset_insn): Likewise.
3407 (bxor_insn): Likewise.
3408 (bclr_insn): Likewise.
3409 (bmsk_insn): Likewise.
3410 (bicsi3_insn): Likewise.
3411 (cmpsi_cc_c_insn): Likewise.
3412 (movsi_ne): Likewise.
3413 (movsi_cond_exec): Likewise.
3414 (clrsbsi2): Likewise.
3425 (call_i): Remove 'S' asm letter, add jli instruction.
3426 (call_value_i): Likewise.
3427 * config/arc/arc.op (mjli-always): New option.
3428 * config/arc/constraints.md (Cji): New constraint.
3429 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3431 (subsf3_fpx): Likewise.
3432 (mulsf3_fpx): Likewise.
3433 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3435 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3437 * doc/invoke.texi (ARC): Document mjli-always option.
3439 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3441 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3442 avoid addition with 0 and use incw and decw where possible.
3444 2018-01-26 Richard Biener <rguenther@suse.de>
3446 PR tree-optimization/81082
3447 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3448 association if it requires casting to unsigned.
3449 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3450 from fold_plusminus_mult_expr to catch important cases late when
3451 range info is available.
3453 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3455 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3456 * configure.ac (hidden_linkonce): New test.
3457 * configure: Regenerate.
3458 * config.in: Regenerate.
3460 2018-01-26 Julia Koval <julia.koval@intel.com>
3462 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3463 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3464 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3465 _mm_mask_bitshuffle_epi64_mask): Fix type.
3466 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3467 USI_FTYPE_V4DI_V4DI_USI): Remove.
3468 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3469 __builtin_ia32_vpshufbitqmb256_mask,
3470 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3471 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3472 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3474 2018-01-26 Alan Modra <amodra@gmail.com>
3477 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3478 UNSPEC_VBPERMQ. Sort other unspecs.
3480 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3482 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3484 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3487 * predict.c (drop_profile): Do not push/pop cfun; update also
3489 (handle_missing_profiles): Fix logic looking for zero profiles.
3491 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3494 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3495 on functions with #pragma omp declare simd or functions with simd
3497 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3498 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3499 Remove trailing \n from warning_at calls.
3501 2018-01-25 Tom de Vries <tom@codesourcery.com>
3504 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3505 for neutered workers.
3507 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3510 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3511 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3513 2018-01-24 Jeff Law <law@redhat.com>
3516 * i386.c (get_probe_interval): Move to earlier point.
3517 (ix86_compute_frame_layout): If -fstack-clash-protection and
3518 the frame is larger than the probe interval, then use pushes
3519 to save registers rather than reg->mem moves.
3520 (ix86_expand_prologue): Remove conditional for int_registers_saved
3523 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3526 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3527 min/max for never referenced object.
3529 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3532 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3534 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3535 attributes from DECL_ATTRIBUTES (decl) without affecting
3536 DECL_ATTRIBUTES (current_function_decl).
3537 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3538 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3540 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3542 PR tree-optimization/83979
3543 * fold-const.c (fold_comparison): Use constant_boolean_node
3544 instead of boolean_{true,false}_node.
3546 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3548 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3551 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3553 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3554 Simplify the clause that sets the length attribute.
3555 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3556 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3557 clause that sets the length attribute.
3558 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3560 2018-01-24 Tom de Vries <tom@codesourcery.com>
3563 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3564 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3565 Add strict parameter.
3566 (prevent_branch_around_nothing): Insert dummy insn between branch to
3567 label and label with no ptx insn inbetween.
3568 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3570 2018-01-24 Tom de Vries <tom@codesourcery.com>
3573 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3574 for neutered threads in warp.
3575 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3577 2018-01-24 Richard Biener <rguenther@suse.de>
3579 PR tree-optimization/83176
3580 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3583 2018-01-24 Richard Biener <rguenther@suse.de>
3585 PR tree-optimization/82819
3586 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3587 code generating pluses that are no-ops in the target precision.
3589 2018-01-24 Richard Biener <rguenther@suse.de>
3592 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3594 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3596 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3597 to merge probabilities.
3598 * predict.c (probably_never_executed): Also mark as cold functions
3599 with global 0 profile and guessed local profile.
3600 * profile-count.c (profile_probability::combine_with_count): New
3602 * profile-count.h (profile_probability::operator*,
3603 profile_probability::operator*=, profile_probability::operator/,
3604 profile_probability::operator/=): Reduce precision to adjusted
3605 and set value to guessed on contradictory divisions.
3606 (profile_probability::combine_with_freq): Remove.
3607 (profile_probability::combine_wiht_count): Declare.
3608 (profile_count::force_nonzero):: Set to adjusted.
3609 (profile_count::probability_in):: Set quality to adjusted.
3610 * tree-ssa-tail-merge.c (replace_block_by): Use
3613 2018-01-23 Andrew Waterman <andrew@sifive.com>
3614 Jim Wilson <jimw@sifive.com>
3616 * config/riscv/riscv.c (riscv_stack_boundary): New.
3617 (riscv_option_override): Set riscv_stack_boundary. Handle
3618 riscv_preferred_stack_boundary_arg.
3619 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3620 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3621 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3622 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3623 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3624 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3626 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3629 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3630 of struct ix86_frame.
3631 (ix86_expand_epilogue): Likewise. Add a local variable for
3632 the reg_save_offset field in struct ix86_frame.
3634 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3636 PR tree-optimization/82604
3637 * tree-loop-distribution.c (enum partition_kind): New enum item
3638 PKIND_PARTIAL_MEMSET.
3639 (partition_builtin_p): Support above new enum item.
3640 (generate_code_for_partition): Ditto.
3641 (compute_access_range): Differentiate cases that equality can be
3642 proven at all loops, the innermost loops or no loops.
3643 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3644 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3645 (finalize_partitions, distribute_loop): Don't fuse partition of
3646 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3647 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3650 2018-01-23 Martin Liska <mliska@suse.cz>
3652 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3653 order to ignore the predictor.
3654 (PRED_POLYMORPHIC_CALL): Likewise.
3655 (PRED_RECURSIVE_CALL): Likewise.
3657 2018-01-23 Martin Liska <mliska@suse.cz>
3659 * tree-profile.c (tree_profiling): Print function header to
3660 aware reader which function we are working on.
3661 * value-prof.c (gimple_find_values_to_profile): Do not print
3662 not interesting value histograms.
3664 2018-01-23 Martin Liska <mliska@suse.cz>
3666 * profile-count.h (enum profile_quality): Add
3667 profile_uninitialized as the first value. Do not number values
3668 as they are zero based.
3669 (profile_count::verify): Update sanity check.
3670 (profile_probability::verify): Likewise.
3672 2018-01-23 Nathan Sidwell <nathan@acm.org>
3674 * doc/invoke.texi (ffor-scope): Deprecate.
3676 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3678 PR tree-optimization/83510
3679 * domwalk.c (set_all_edges_as_executable): New function.
3680 (dom_walker::dom_walker): Convert bool param
3681 "skip_unreachable_blocks" to enum reachability. Move setup of
3682 edge flags to set_all_edges_as_executable and only do it when
3683 reachability is REACHABLE_BLOCKS.
3684 * domwalk.h (enum dom_walker::reachability): New enum.
3685 (dom_walker::dom_walker): Convert bool param
3686 "skip_unreachable_blocks" to enum reachability.
3687 (set_all_edges_as_executable): New decl.
3688 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3689 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3691 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3692 but converting true to REACHABLE_BLOCKS.
3693 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3695 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3696 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3697 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3699 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3700 if check_all_array_refs will be called.
3702 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3704 * tree.c (selftest::test_location_wrappers): Add more test
3707 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3709 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3710 (selftest::test_bit_in_range): Likewise.
3712 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3715 * doc/sourcebuild.texi (vect_float): Say that the selector
3716 only describes the situation when -funsafe-math-optimizations is on.
3717 (vect_float_strict): Document.
3719 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3721 PR tree-optimization/83965
3722 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3723 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3724 instead of checking only for a reduction.
3725 (vect_recog_widen_sum_pattern): Likewise.
3727 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3729 * predict.c (probably_never_executed): Only use precise profile info.
3730 (compute_function_frequency): Skip after inlining hack since we now
3731 have quality checking.
3733 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3735 * profile-count.h (profile_probability::very_unlikely,
3736 profile_probability::unlikely, profile_probability::even): Set
3737 precision to guessed.
3739 2018-01-23 Richard Biener <rguenther@suse.de>
3741 PR tree-optimization/83963
3742 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3743 Properly terminate dominator walk when crossing the exit edge not
3744 when visiting its source block.
3746 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3749 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3750 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3752 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3754 PR tree-optimization/83957
3755 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3756 semicolon after for body surrounded by braces.
3758 PR tree-optimization/83081
3759 * profile-count.h (profile_probability::split): New method.
3760 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3761 Use profile_probability::split.
3762 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3763 when splitting a single conditional jump into 2.
3765 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3767 PR tree-optimization/69452
3768 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3771 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3773 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3774 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3775 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3777 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3779 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3780 * config/rl78/rl78.md (movdi): New define_expand.
3781 * config/rl78/rl78.c (rl78_split_movdi): New function.
3783 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3786 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3788 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3789 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3790 128-bit to produce an UNSPEC move to get the double word with the
3791 signbit and then a shift directly to do signbit.
3792 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3793 implementation with a new version that just does either a direct
3794 move or a regular move. Move memory interface to separate insns.
3795 Move insns so they are next to the expander.
3796 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3797 with signbit move. Split big and little endian case.
3798 (signbit<mode>2_dm_mem_le): Likewise.
3799 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3800 (signbit<mode>2_dm2): Likewise.
3802 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3804 * config/rl78/rl78.md (anddi3): New define_expand.
3806 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3808 * config/rl78/rl78.md (umindi3): New define_expand.
3810 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3812 * config/rl78/rl78.md (smindi3): New define_expand.
3814 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3816 * config/rl78/rl78.md (smaxdi3): New define_expand.
3818 2018-01-22 Carl Love <cel@us.ibm.com>
3820 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3821 LVX_V1TI): Add macro expansion.
3822 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3823 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3824 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3825 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3826 Change check to determine if the instruction is a byte reversing
3827 entry. Fix typo in comment.
3828 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3829 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3830 Add def_builtin calls for new builtins.
3831 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3832 Add define_insn expansion.
3834 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3836 * config/rl78/rl78.md (umaxdi3): New define_expand.
3838 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3840 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3841 for non-QImode registers.
3843 2018-01-22 Richard Biener <rguenther@suse.de>
3845 PR tree-optimization/83963
3846 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3847 including the loop exit block.
3848 (scop_detection::merge_sese): Likewise.
3849 (scop_detection::add_scop): Do it here instead.
3851 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3853 * doc/sourcebuild.texi (arm_softfloat): Document.
3855 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3858 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3859 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3860 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3862 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3863 David Edelsohn <dje.gcc@gmail.com>
3866 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3867 Change "crset eq" to "crset 2".
3868 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3869 (*call_indirect_aix<mode>_nospec): Likewise.
3870 (*call_value_indirect_aix<mode>_nospec): Likewise.
3871 (*call_indirect_elfv2<mode>_nospec): Likewise.
3872 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3873 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3874 change assembly output from . to $.
3875 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3876 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3877 (*tablejump<mode>_internal1_nospec): Likewise.
3879 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3882 * config/sh/sh_optimize_sett_clrt.cc:
3883 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3885 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3887 PR tree-optimization/83940
3888 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3889 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3890 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3891 use it to pass back the definition type.
3892 (vect_check_store_rhs): Likewise rhs_dt_out.
3893 (vect_build_gather_load_calls): Add a mask_dt argument and use
3894 it instead of a call to vect_is_simple_use.
3895 (vectorizable_store): Update calls to vect_check_load_store_mask
3896 and vect_check_store_rhs. Use the dt returned by the latter instead
3897 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3898 instead of calls to vect_is_simple_use. Pass the scalar rather
3899 than the vector operand to vect_is_simple_use when handling
3900 second and subsequent copies of an rhs value.
3901 (vectorizable_load): Update calls to vect_check_load_store_mask
3902 and vect_build_gather_load_calls. Use the cached mask_dt and
3903 gs_info.offset_dt instead of calls to vect_is_simple_use.
3905 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3908 * tree-emutls.c: Include gimplify.h.
3909 (lower_emutls_2): New function.
3910 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3911 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3912 it before further processing.
3915 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3916 UINTVAL (trueop1) instead of INTVAL (op1).
3918 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3922 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3923 INCOMING_FRAME_SP_OFFSET if not defined.
3924 (scan_trace): Add ENTRY argument. If true and
3925 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3926 emit a note to adjust the CFA offset.
3927 (create_cfi_notes): Adjust scan_trace callers.
3928 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3929 INCOMING_FRAME_SP_OFFSET in the CIE.
3930 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3931 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3933 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3934 * doc/tm.texi: Regenerated.
3936 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3938 PR rtl-optimization/83147
3939 * lra-constraints.c (remove_inheritance_pseudos): Use
3940 lra_substitute_pseudo_within_insn.
3942 2018-01-19 Tom de Vries <tom@codesourcery.com>
3943 Cesar Philippidis <cesar@codesourcery.com>
3946 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3948 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3951 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3952 spaces for function labels.
3954 2018-01-19 Martin Liska <mliska@suse.cz>
3956 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3957 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3958 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3959 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3960 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3961 (PRED_CONST_RETURN): Change from 69 to 65.
3962 (PRED_NULL_RETURN): Change from 91 to 71.
3963 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3964 (PRED_LOOP_GUARD): Change from 66 to 73.
3966 2018-01-19 Martin Liska <mliska@suse.cz>
3968 * predict.c (predict_insn_def): Add new assert.
3969 (struct branch_predictor): Change type to signed integer.
3970 (test_prediction_value_range): Amend test to cover
3972 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3973 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3974 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3975 (PRED_LOOP_IV_COMPARE): Likewise.
3976 * predict.h (PROB_UNINITIALIZED): Define new constant.
3978 2018-01-19 Martin Liska <mliska@suse.cz>
3980 * predict.c (dump_prediction): Add new format for
3981 analyze_brprob.py script which is enabled with -details
3983 * profile-count.h (precise_p): New function.
3985 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3987 PR tree-optimization/83922
3988 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3989 there are no statements that need masking.
3990 (vect_active_double_reduction_p): New function.
3991 (vect_analyze_loop_operations): Use it when handling phis that
3992 are not in the loop header.
3994 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3996 PR tree-optimization/83914
3997 * tree-vect-loop.c (vectorizable_induction): Don't convert
3998 init_expr or apply the peeling adjustment for inductions
3999 that are nested within the vectorized loop.
4001 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4003 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4006 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4010 * function.h (gimplify_parameters): Add gimple_seq * argument.
4011 * function.c: Include gimple.h and options.h.
4012 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4013 for the added local temporaries if needed.
4014 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4015 if there are any parameter cleanups, wrap whole body into a
4016 try/finally with the cleanups.
4018 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4021 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4022 Use GET_MODE_CLASS for scalar floating point.
4024 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4028 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4029 Fix call of call_cgraph_insertion_hooks.
4031 2018-01-18 Martin Sebor <msebor@redhat.com>
4033 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4035 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4038 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4041 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4044 * common.opt: (-ffile-prefix-map): New option.
4045 * opts.c (common_handle_option): Defer it.
4046 * opts-global.c (handle_common_deferred_options): Handle it.
4047 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4048 * file-prefix-map.h: New file.
4049 (remap_debug_filename, add_debug_prefix_map): ...here.
4050 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4051 * final.c (debug_prefix_map, add_debug_prefix_map
4052 remap_debug_filename): Move to...
4053 * file-prefix-map.c: New file.
4054 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4055 generalize, get rid of alloca(), use strrchr() instead of strchr().
4056 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4057 Implement in terms of add_prefix_map().
4058 (remap_macro_filename, remap_debug_filename): Implement in term of
4060 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4061 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4062 * dbxout.c: Include file-prefix-map.h.
4063 * varasm.c: Likewise.
4064 * vmsdbgout.c: Likewise.
4065 * xcoffout.c: Likewise.
4066 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4067 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4068 * doc/invoke.texi (-ffile-prefix-map): Document.
4069 (-fdebug-prefix-map): Update description.
4071 2018-01-18 Martin Liska <mliska@suse.cz>
4073 * config/i386/i386.c (indirect_thunk_name): Document that also
4075 (output_indirect_thunk): Document why both instructions
4076 (pause and lfence) are generated.
4078 2018-01-18 Richard Biener <rguenther@suse.de>
4080 PR tree-optimization/83887
4081 * graphite-scop-detection.c
4082 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4083 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4084 (scop_detection::merge_sese): Re-implement with a flood-fill
4085 algorithm that properly finds a SESE region if it exists.
4087 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4090 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4091 pointer_diff optimizations use view_convert instead of convert.
4093 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4095 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4096 Generate different code for -mno-speculate-indirect-jumps.
4097 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4098 (*call_indirect_aix<mode>): Disable for
4099 -mno-speculate-indirect-jumps.
4100 (*call_indirect_aix<mode>_nospec): New define_insn.
4101 (*call_value_indirect_aix<mode>): Disable for
4102 -mno-speculate-indirect-jumps.
4103 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4104 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4105 -mno-speculate-indirect-jumps.
4106 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4108 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4110 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4111 long double type, set the flags for noting the default long double
4112 type, even if we don't pass or return a long double type.
4114 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4117 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4120 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4122 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4123 support for merge[hl].
4124 (fold_mergehl_helper): New helper function.
4125 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4126 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4127 (altivec_vmrglw_direct): Add xxmrglw insn.
4129 2018-01-17 Andrew Waterman <andrew@sifive.com>
4131 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4132 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4134 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4137 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4138 call the lto_location_cache before reading the
4139 DECL_SOURCE_LOCATION of the types.
4141 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4142 Richard Sandiford <richard.sandiford@linaro.org>
4144 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4145 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4146 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4147 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4148 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4150 * config/aarch64/constraints.md (aarch64_movti_operand):
4152 * config/aarch64/predicates.md (Uti): Add new constraint.
4154 2018-01-17 Carl Love <cel@us.ibm.com>
4156 * config/rs6000/vsx.md (define_expand xl_len_r,
4157 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4158 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4160 (define_expand, define_insn): Move the shift left from the
4161 define_insn to the define_expand for lxvl and stxvl instructions.
4162 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4163 and XL_LEN_R definitions to PURE.
4165 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4167 * config/i386/i386.c (indirect_thunk_name): Declare regno
4168 as unsigned int. Compare regno with INVALID_REGNUM.
4169 (output_indirect_thunk): Ditto.
4170 (output_indirect_thunk_function): Ditto.
4171 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4172 in the call to output_indirect_thunk_function.
4174 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4177 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4178 rather than the size of inner_type to determine the stack slot size
4179 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4181 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4184 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4187 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4189 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4190 endian Linux systems to optionally enable multilibs for selecting
4191 the long double type if the user configured an explicit type.
4192 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4193 have no long double multilibs if not defined.
4194 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4195 warn if the user used -mabi={ieee,ibm}longdouble and we built
4196 multilibs for long double.
4197 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4198 appropriate multilib option.
4199 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4201 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4202 for building long double multilibs.
4203 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4205 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4207 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4210 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4212 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4215 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4218 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4221 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4223 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4224 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4226 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4228 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4229 different rtl trees depending on TARGET_64BIT.
4230 (rs6000_gen_lvx): Likewise.
4232 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4234 * config/visium/visium.md (nop): Tweak comment.
4235 (hazard_nop): Likewise.
4237 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4239 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4240 -mspeculate-indirect-jumps.
4241 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4242 for -mno-speculate-indirect-jumps.
4243 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4244 (*call_value_indirect_elfv2<mode>): Disable for
4245 -mno-speculate-indirect-jumps.
4246 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4247 (indirect_jump): Emit different RTL for
4248 -mno-speculate-indirect-jumps.
4249 (*indirect_jump<mode>): Disable for
4250 -mno-speculate-indirect-jumps.
4251 (*indirect_jump<mode>_nospec): New define_insn.
4252 (tablejump): Emit different RTL for
4253 -mno-speculate-indirect-jumps.
4254 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4255 (tablejumpsi_nospec): New define_expand.
4256 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4257 (tablejumpdi_nospec): New define_expand.
4258 (*tablejump<mode>_internal1): Disable for
4259 -mno-speculate-indirect-jumps.
4260 (*tablejump<mode>_internal1_nospec): New define_insn.
4261 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4264 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4266 * caller-save.c (insert_save): Drop unnecessary parameter. All
4269 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4270 Richard Biener <rguenth@suse.de>
4273 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4274 return early, inline manually is_gimple_sizepos. Make sure if we
4275 call gimplify_expr we don't end up with a gimple constant.
4276 * tree.c (variably_modified_type_p): Don't return true for
4277 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4278 * gimplify.h (is_gimple_sizepos): Remove.
4280 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4282 PR tree-optimization/83857
4283 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4284 vectorizable_live_operation for pure SLP statements.
4285 (vectorizable_live_operation): Handle PHIs.
4287 2018-01-16 Richard Biener <rguenther@suse.de>
4289 PR tree-optimization/83867
4290 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4291 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4293 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4296 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4297 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4298 If off is not INTEGER_CST, issue a may not be aligned warning
4299 rather than isn't aligned. Use isn%'t rather than isn't.
4300 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4302 <case MULT_EXPR>: Improve the case when bottom and one of the
4303 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4304 operand, in that case check if the other operand is multiple of
4305 bottom divided by the INTEGER_CST operand.
4307 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4310 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4311 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4312 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4313 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4314 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4315 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4316 (pa_function_arg_size): New function.
4318 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4320 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4321 in a separate statement.
4323 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4325 PR tree-optimization/83847
4326 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4327 group gathers and scatters.
4329 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4331 PR rtl-optimization/86620
4332 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4334 PR rtl-optimization/83213
4335 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4336 to last if both are JUMP_INSNs.
4338 PR tree-optimization/83843
4339 * gimple-ssa-store-merging.c
4340 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4341 store_immediate_info for bswap/nop orig_stores.
4343 2018-01-15 Andrew Waterman <andrew@sifive.com>
4345 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4347 <UDIV>: Increase cost if !TARGET_DIV.
4349 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4351 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4352 (define_attr "cr_logical_3op"): New.
4353 (cceq_ior_compare): Adjust.
4354 (cceq_ior_compare_complement): Adjust.
4355 (*cceq_rev_compare): Adjust.
4356 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4357 (is_cracked_insn): Adjust.
4358 (insn_must_be_first_in_group): Adjust.
4359 * config/rs6000/40x.md: Adjust.
4360 * config/rs6000/440.md: Adjust.
4361 * config/rs6000/476.md: Adjust.
4362 * config/rs6000/601.md: Adjust.
4363 * config/rs6000/603.md: Adjust.
4364 * config/rs6000/6xx.md: Adjust.
4365 * config/rs6000/7450.md: Adjust.
4366 * config/rs6000/7xx.md: Adjust.
4367 * config/rs6000/8540.md: Adjust.
4368 * config/rs6000/cell.md: Adjust.
4369 * config/rs6000/e300c2c3.md: Adjust.
4370 * config/rs6000/e500mc.md: Adjust.
4371 * config/rs6000/e500mc64.md: Adjust.
4372 * config/rs6000/e5500.md: Adjust.
4373 * config/rs6000/e6500.md: Adjust.
4374 * config/rs6000/mpc.md: Adjust.
4375 * config/rs6000/power4.md: Adjust.
4376 * config/rs6000/power5.md: Adjust.
4377 * config/rs6000/power6.md: Adjust.
4378 * config/rs6000/power7.md: Adjust.
4379 * config/rs6000/power8.md: Adjust.
4380 * config/rs6000/power9.md: Adjust.
4381 * config/rs6000/rs64.md: Adjust.
4382 * config/rs6000/titan.md: Adjust.
4384 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4386 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4387 ix86_indirect_branch_register logic.
4389 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4391 * config/i386/constraints.md (Bs): Update
4392 ix86_indirect_branch_register check. Don't check
4393 ix86_indirect_branch_register with GOT_memory_operand.
4395 * config/i386/predicates.md (GOT_memory_operand): Don't check
4396 ix86_indirect_branch_register here.
4397 (GOT32_symbol_operand): Likewise.
4399 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4401 * config/i386/predicates.md (constant_call_address_operand):
4402 Rewrite ix86_indirect_branch_register logic.
4403 (sibcall_insn_operand): Likewise.
4405 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4407 * config/i386/constraints.md (Bs): Replace
4408 ix86_indirect_branch_thunk_register with
4409 ix86_indirect_branch_register.
4411 * config/i386/i386.md (indirect_jump): Likewise.
4412 (tablejump): Likewise.
4413 (*sibcall_memory): Likewise.
4414 (*sibcall_value_memory): Likewise.
4415 Peepholes of indirect call and jump via memory: Likewise.
4416 * config/i386/i386.opt: Likewise.
4417 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4418 (GOT_memory_operand): Likewise.
4419 (call_insn_operand): Likewise.
4420 (sibcall_insn_operand): Likewise.
4421 (GOT32_symbol_operand): Likewise.
4423 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4426 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4427 type rather than type addr's type points to.
4428 (expand_omp_atomic_mutex): Likewise.
4429 (expand_omp_atomic): Likewise.
4431 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4434 * config/i386/i386.c (output_indirect_thunk_function): Use
4435 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4436 for __x86_return_thunk.
4438 2018-01-15 Richard Biener <rguenther@suse.de>
4441 * expmed.c (extract_bit_field_1): Fix typo.
4443 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4446 * config/arm/iterators.md (VF): New mode iterator.
4447 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4448 Remove integer-related logic from pattern.
4449 (neon_vabd<mode>_3): Likewise.
4451 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4454 * common.opt (fstrict-overflow): No longer an alias.
4455 (fwrapv-pointer): New option.
4456 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4457 also for pointer types based on flag_wrapv_pointer.
4458 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4459 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4460 opts->x_flag_wrapv got set.
4461 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4462 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4463 POINTER_TYPE_OVERFLOW_UNDEFINED.
4464 * match.pd: Likewise in address comparison pattern.
4465 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4467 2018-01-15 Richard Biener <rguenther@suse.de>
4470 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4471 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4472 Reset type names to their identifier if their TYPE_DECL doesn't
4473 have linkage (and thus is used for ODR and devirt).
4474 (save_debug_info_for_decl): Remove.
4475 (save_debug_info_for_type): Likewise.
4476 (add_tree_to_fld_list): Adjust.
4477 * tree-pretty-print.c (dump_generic_node): Make dumping of
4478 type names more robust.
4480 2018-01-15 Richard Biener <rguenther@suse.de>
4482 * BASE-VER: Bump to 8.0.1.
4484 2018-01-14 Martin Sebor <msebor@redhat.com>
4487 * builtins.c (check_access): Avoid warning when the no-warning bit
4490 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4492 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4493 * ira-color (allocno_hard_regs_compare): Likewise.
4495 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4498 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4499 Use .pushsection/.popsection.
4501 2018-01-14 Martin Sebor <msebor@redhat.com>
4504 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4506 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4508 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4509 entry from extra_headers.
4510 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4511 extra_headers, make the list bitwise identical to the i?86-*-* one.
4513 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4515 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4516 -mcmodel=large with -mindirect-branch=thunk,
4517 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4518 -mfunction-return=thunk-extern.
4519 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4520 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4521 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4523 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4525 * config/i386/i386.c (print_reg): Print the name of the full
4526 integer register without '%'.
4527 (ix86_print_operand): Handle 'V'.
4528 * doc/extend.texi: Document 'V' modifier.
4530 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4532 * config/i386/constraints.md (Bs): Disallow memory operand for
4533 -mindirect-branch-register.
4535 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4536 (GOT_memory_operand): Likewise.
4537 (call_insn_operand): Likewise.
4538 (sibcall_insn_operand): Likewise.
4539 (GOT32_symbol_operand): Likewise.
4540 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4541 for -mindirect-branch-register.
4542 (tablejump): Likewise.
4543 (*sibcall_memory): Likewise.
4544 (*sibcall_value_memory): Likewise.
4545 Disallow peepholes of indirect call and jump via memory for
4546 -mindirect-branch-register.
4547 (*call_pop): Replace m with Bw.
4548 (*call_value_pop): Likewise.
4549 (*sibcall_pop_memory): Replace m with Bs.
4550 * config/i386/i386.opt (mindirect-branch-register): New option.
4551 * doc/invoke.texi: Document -mindirect-branch-register option.
4553 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4555 * config/i386/i386-protos.h (ix86_output_function_return): New.
4556 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4557 set function_return_type.
4558 (indirect_thunk_name): Add ret_p to indicate thunk for function
4560 (output_indirect_thunk_function): Pass false to
4561 indirect_thunk_name.
4562 (ix86_output_indirect_branch_via_reg): Likewise.
4563 (ix86_output_indirect_branch_via_push): Likewise.
4564 (output_indirect_thunk_function): Create alias for function
4565 return thunk if regno < 0.
4566 (ix86_output_function_return): New function.
4567 (ix86_handle_fndecl_attribute): Handle function_return.
4568 (ix86_attribute_table): Add function_return.
4569 * config/i386/i386.h (machine_function): Add
4570 function_return_type.
4571 * config/i386/i386.md (simple_return_internal): Use
4572 ix86_output_function_return.
4573 (simple_return_internal_long): Likewise.
4574 * config/i386/i386.opt (mfunction-return=): New option.
4575 (indirect_branch): Mention -mfunction-return=.
4576 * doc/extend.texi: Document function_return function attribute.
4577 * doc/invoke.texi: Document -mfunction-return= option.
4579 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4581 * config/i386/i386-opts.h (indirect_branch): New.
4582 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4583 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4584 with local indirect jump when converting indirect call and jump.
4585 (ix86_set_indirect_branch_type): New.
4586 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4587 (indirectlabelno): New.
4588 (indirect_thunk_needed): Likewise.
4589 (indirect_thunk_bnd_needed): Likewise.
4590 (indirect_thunks_used): Likewise.
4591 (indirect_thunks_bnd_used): Likewise.
4592 (INDIRECT_LABEL): Likewise.
4593 (indirect_thunk_name): Likewise.
4594 (output_indirect_thunk): Likewise.
4595 (output_indirect_thunk_function): Likewise.
4596 (ix86_output_indirect_branch_via_reg): Likewise.
4597 (ix86_output_indirect_branch_via_push): Likewise.
4598 (ix86_output_indirect_branch): Likewise.
4599 (ix86_output_indirect_jmp): Likewise.
4600 (ix86_code_end): Call output_indirect_thunk_function if needed.
4601 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4603 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4604 (ix86_attribute_table): Add indirect_branch.
4605 * config/i386/i386.h (machine_function): Add indirect_branch_type
4606 and has_local_indirect_jump.
4607 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4609 (tablejump): Likewise.
4610 (*indirect_jump): Use ix86_output_indirect_jmp.
4611 (*tablejump_1): Likewise.
4612 (simple_return_indirect_internal): Likewise.
4613 * config/i386/i386.opt (mindirect-branch=): New option.
4614 (indirect_branch): New.
4617 (thunk-inline): Likewise.
4618 (thunk-extern): Likewise.
4619 * doc/extend.texi: Document indirect_branch function attribute.
4620 * doc/invoke.texi: Document -mindirect-branch= option.
4622 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4625 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4627 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4629 * ipa-inline.c (want_inline_small_function_p): Return false if
4630 inlining has already failed with CIF_FINAL_ERROR.
4631 (update_caller_keys): Call want_inline_small_function_p before
4633 (update_callee_keys): Likewise.
4635 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4637 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4639 (rs6000_quadword_masked_address_p): Likewise.
4640 (quad_aligned_load_p): Likewise.
4641 (quad_aligned_store_p): Likewise.
4642 (const_load_sequence_p): Add comment to describe the outer-most loop.
4643 (mimic_memory_attributes_and_flags): New function.
4644 (rs6000_gen_stvx): Likewise.
4645 (replace_swapped_aligned_store): Likewise.
4646 (rs6000_gen_lvx): Likewise.
4647 (replace_swapped_aligned_load): Likewise.
4648 (replace_swapped_load_constant): Capitalize argument name in
4649 comment describing this function.
4650 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4651 and stores that access quad-word aligned addresses and replace
4652 with stvx or lvx instructions when appropriate.
4653 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4654 New function prototype.
4655 (rs6000_quadword_masked_address_p): Likewise.
4656 (rs6000_gen_lvx): Likewise.
4657 (rs6000_gen_stvx): Likewise.
4658 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4659 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4660 when memory address is aligned.
4661 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4662 this split to select lvx instruction when memory address is aligned.
4663 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4664 instruction when memory address is aligned.
4665 (*vsx_le_perm_load_v16qi): Likewise.
4666 (four unnamed splitters): Modify to select the stvx instruction
4667 when memory is aligned.
4669 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4671 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4672 which appears in the queue multiple times.
4674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4675 Alan Hayward <alan.hayward@arm.com>
4676 David Sherwood <david.sherwood@arm.com>
4678 * tree-vectorizer.h (vec_lower_bound): New structure.
4679 (_loop_vec_info): Add check_nonzero and lower_bounds.
4680 (LOOP_VINFO_CHECK_NONZERO): New macro.
4681 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4682 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4683 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4684 fields. Make seg_len the distance travelled, not including the
4686 (dr_direction_indicator): Declare.
4687 (dr_zero_step_indicator): Likewise.
4688 (dr_known_forward_stride_p): Likewise.
4689 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4691 (runtime_alias_check_p): Allow runtime alias checks with
4693 (operator ==): Compare access_size and align.
4694 (prune_runtime_alias_test_list): Rework for new distinction between
4695 the access_size and seg_len.
4696 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4698 (get_segment_min_max): New function.
4699 (create_intersect_range_checks): Use it.
4700 (dr_step_indicator): New function.
4701 (dr_direction_indicator): Likewise.
4702 (dr_zero_step_indicator): Likewise.
4703 (dr_known_forward_stride_p): Likewise.
4704 * tree-loop-distribution.c (data_ref_segment_size): Return
4705 DR_STEP * (niters - 1).
4706 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4708 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4709 (vect_preserves_scalar_order_p): New function, split out from...
4710 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4711 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4712 (vect_vfa_access_size): New function.
4713 (vect_vfa_align): Likewise.
4714 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4715 (dump_lower_bound): New function.
4716 (vect_check_lower_bound): Likewise.
4717 (vect_small_gap_p): Likewise.
4718 (vectorizable_with_step_bound_p): Likewise.
4719 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4720 depencies if the vectorization factor is 1. Convert the checks
4721 for nonzero steps into checks on the bounds of DR_STEP. Try using
4722 a bunds check for variable steps if the minimum required step is
4723 relatively small. Update calls to the dr_with_seg_len
4724 constructor and to vect_compile_time_alias.
4725 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4727 (vect_loop_versioning): Call it.
4728 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4730 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4733 Alan Hayward <alan.hayward@arm.com>
4734 David Sherwood <david.sherwood@arm.com>
4736 * doc/sourcebuild.texi (vect_scatter_store): Document.
4737 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4739 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4741 * genopinit.c (main): Add supports_vec_scatter_store and
4742 supports_vec_scatter_store_cached to target_optabs.
4743 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4744 IFN_MASK_SCATTER_STORE.
4745 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4747 * internal-fn.h (internal_store_fn_p): Declare.
4748 (internal_fn_stored_value_index): Likewise.
4749 * internal-fn.c (scatter_store_direct): New macro.
4750 (expand_scatter_store_optab_fn): New function.
4751 (direct_scatter_store_optab_supported_p): New macro.
4752 (internal_store_fn_p): New function.
4753 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4754 IFN_MASK_SCATTER_STORE.
4755 (internal_fn_mask_index): Likewise.
4756 (internal_fn_stored_value_index): New function.
4757 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4759 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4760 * optabs-query.c (supports_vec_scatter_store_p): New function.
4761 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4762 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4763 true for scatter stores.
4764 (vect_gather_scatter_fn_p): Handle scatter stores too.
4765 (vect_check_gather_scatter): Consider using scatter stores if
4766 supports_vec_scatter_store_p.
4767 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4769 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4770 internal_fn_stored_value_index.
4771 (check_load_store_masking): Handle scatter stores too.
4772 (vect_get_store_rhs): Make public.
4773 (vectorizable_call): Use internal_store_fn_p.
4774 (vectorizable_store): Handle scatter store internal functions.
4775 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4776 when deciding whether the end of the group has been reached.
4777 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4778 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4779 (mask_scatter_store<mode>): New insns.
4781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4782 Alan Hayward <alan.hayward@arm.com>
4783 David Sherwood <david.sherwood@arm.com>
4785 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4786 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4787 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4789 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4790 Use vect_truncate_gather_scatter_offset if we can't treat the
4791 operation as a normal gather load or scatter store.
4792 (get_group_load_store_type): Take the gather_scatter_info
4793 as argument. Try using a gather load or scatter store for
4794 single-element groups.
4795 (get_load_store_type): Update calls to get_group_load_store_type
4796 and vect_use_strided_gather_scatters_p.
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_create_data_ref_ptr): Take an extra
4803 optional tree argument.
4804 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4806 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4807 but continue to use the current value as a fallback.
4808 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4809 to compare the updates.
4810 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4811 (get_load_store_type): Use it when handling a strided access.
4812 (vect_get_strided_load_store_ops): New function.
4813 (vect_get_data_ptr_increment): Likewise.
4814 (vectorizable_load): Handle strided gather loads. Always pass
4815 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4818 Alan Hayward <alan.hayward@arm.com>
4819 David Sherwood <david.sherwood@arm.com>
4821 * doc/md.texi (gather_load@var{m}): Document.
4822 (mask_gather_load@var{m}): Likewise.
4823 * genopinit.c (main): Add supports_vec_gather_load and
4824 supports_vec_gather_load_cached to target_optabs.
4825 * optabs-tree.c (init_tree_optimization_optabs): Use
4826 ggc_cleared_alloc to allocate target_optabs.
4827 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4828 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4830 * internal-fn.h (internal_load_fn_p): Declare.
4831 (internal_gather_scatter_fn_p): Likewise.
4832 (internal_fn_mask_index): Likewise.
4833 (internal_gather_scatter_fn_supported_p): Likewise.
4834 * internal-fn.c (gather_load_direct): New macro.
4835 (expand_gather_load_optab_fn): New function.
4836 (direct_gather_load_optab_supported_p): New macro.
4837 (direct_internal_fn_optab): New function.
4838 (internal_load_fn_p): Likewise.
4839 (internal_gather_scatter_fn_p): Likewise.
4840 (internal_fn_mask_index): Likewise.
4841 (internal_gather_scatter_fn_supported_p): Likewise.
4842 * optabs-query.c (supports_at_least_one_mode_p): New function.
4843 (supports_vec_gather_load_p): Likewise.
4844 * optabs-query.h (supports_vec_gather_load_p): Declare.
4845 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4846 and memory_type field.
4847 (NUM_PATTERNS): Bump to 15.
4848 * tree-vect-data-refs.c: Include internal-fn.h.
4849 (vect_gather_scatter_fn_p): New function.
4850 (vect_describe_gather_scatter_call): Likewise.
4851 (vect_check_gather_scatter): Try using internal functions for
4852 gather loads. Recognize existing calls to a gather load function.
4853 (vect_analyze_data_refs): Consider using gather loads if
4854 supports_vec_gather_load_p.
4855 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4856 (vect_get_gather_scatter_offset_type): Likewise.
4857 (vect_convert_mask_for_vectype): Likewise.
4858 (vect_add_conversion_to_patterm): Likewise.
4859 (vect_try_gather_scatter_pattern): Likewise.
4860 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4861 (vect_vect_recog_func_ptrs): Add it.
4862 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4863 internal_fn_mask_index and internal_gather_scatter_fn_p.
4864 (check_load_store_masking): Take the gather_scatter_info as an
4865 argument and handle gather loads.
4866 (vect_get_gather_scatter_ops): New function.
4867 (vectorizable_call): Check internal_load_fn_p.
4868 (vectorizable_load): Likewise. Handle gather load internal
4870 (vectorizable_store): Update call to check_load_store_masking.
4871 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4872 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4873 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4874 (aarch64_gather_scale_operand_d): New predicates.
4875 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4876 (mask_gather_load<mode>): New insns.
4878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4879 Alan Hayward <alan.hayward@arm.com>
4880 David Sherwood <david.sherwood@arm.com>
4882 * optabs.def (fold_left_plus_optab): New optab.
4883 * doc/md.texi (fold_left_plus_@var{m}): Document.
4884 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4885 * internal-fn.c (fold_left_direct): Define.
4886 (expand_fold_left_optab_fn): Likewise.
4887 (direct_fold_left_optab_supported_p): Likewise.
4888 * fold-const-call.c (fold_const_fold_left): New function.
4889 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4890 * tree-parloops.c (valid_reduction_p): New function.
4891 (gather_scalar_reductions): Use it.
4892 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4893 (vect_finish_replace_stmt): Declare.
4894 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4895 (needs_fold_left_reduction_p): New function, split out from...
4896 (vect_is_simple_reduction): ...here. Accept reductions that
4897 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4898 (vect_force_simple_reduction): Also store the reduction type in
4899 the assignment's STMT_VINFO_REDUC_TYPE.
4900 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4901 (merge_with_identity): New function.
4902 (vect_expand_fold_left): Likewise.
4903 (vectorize_fold_left_reduction): Likewise.
4904 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4905 scalar phi in place for it. Check for target support and reject
4906 cases that would reassociate the operation. Defer the transform
4907 phase to vectorize_fold_left_reduction.
4908 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4909 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4910 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4914 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4915 call to ifc_temp_var.
4917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4918 Alan Hayward <alan.hayward@arm.com>
4919 David Sherwood <david.sherwood@arm.com>
4921 * target.def (legitimize_address_displacement): Take the original
4922 offset as a poly_int.
4923 * targhooks.h (default_legitimize_address_displacement): Update
4925 * targhooks.c (default_legitimize_address_displacement): Likewise.
4926 * doc/tm.texi: Regenerate.
4927 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4928 as an argument, moving assert of ad->disp == ad->disp_term to...
4929 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4930 Try calling targetm.legitimize_address_displacement before expanding
4931 the address rather than afterwards, and adjust for the new interface.
4932 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4933 Match the new hook interface. Handle SVE addresses.
4934 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4939 * Makefile.in (OBJS): Add early-remat.o.
4940 * target.def (select_early_remat_modes): New hook.
4941 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4942 * doc/tm.texi: Regenerate.
4943 * targhooks.h (default_select_early_remat_modes): Declare.
4944 * targhooks.c (default_select_early_remat_modes): New function.
4945 * timevar.def (TV_EARLY_REMAT): New timevar.
4946 * passes.def (pass_early_remat): New pass.
4947 * tree-pass.h (make_pass_early_remat): Declare.
4948 * early-remat.c: New file.
4949 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4951 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4954 Alan Hayward <alan.hayward@arm.com>
4955 David Sherwood <david.sherwood@arm.com>
4957 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4958 vfm1 with a bound_epilog parameter.
4959 (vect_do_peeling): Update calls accordingly, and move the prologue
4960 call earlier in the function. Treat the base bound_epilog as 0 for
4961 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4962 this base when peeling for gaps.
4963 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4964 with fully-masked loops.
4965 (vect_estimate_min_profitable_iters): Handle the single peeled
4966 iteration in that case.
4968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4969 Alan Hayward <alan.hayward@arm.com>
4970 David Sherwood <david.sherwood@arm.com>
4972 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4973 single-element interleaving even if the size is not a power of 2.
4974 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4975 accesses for single-element interleaving if the group size is
4978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4979 Alan Hayward <alan.hayward@arm.com>
4980 David Sherwood <david.sherwood@arm.com>
4982 * doc/md.texi (fold_extract_last_@var{m}): Document.
4983 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4984 * optabs.def (fold_extract_last_optab): New optab.
4985 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4986 * internal-fn.c (fold_extract_direct): New macro.
4987 (expand_fold_extract_optab_fn): Likewise.
4988 (direct_fold_extract_optab_supported_p): Likewise.
4989 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4990 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4991 EXTRACT_LAST_REDUCTION.
4992 (get_initial_def_for_reduction): Do not create an initial vector
4993 for EXTRACT_LAST_REDUCTION reductions.
4994 (vectorizable_reduction): Leave the scalar phi in place for
4995 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4996 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4997 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4998 transform phase to vectorizable_condition.
4999 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5001 (vect_finish_stmt_generation): ...here.
5002 (vect_finish_replace_stmt): New function.
5003 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5004 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5006 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5009 Alan Hayward <alan.hayward@arm.com>
5010 David Sherwood <david.sherwood@arm.com>
5012 * doc/md.texi (extract_last_@var{m}): Document.
5013 * optabs.def (extract_last_optab): New optab.
5014 * internal-fn.def (EXTRACT_LAST): New internal function.
5015 * internal-fn.c (cond_unary_direct): New macro.
5016 (expand_cond_unary_optab_fn): Likewise.
5017 (direct_cond_unary_optab_supported_p): Likewise.
5018 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5019 loops using EXTRACT_LAST.
5020 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5021 (extract_last_<mode>): ...this optab.
5022 (vec_extract<mode><Vel>): Update accordingly.
5024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5025 Alan Hayward <alan.hayward@arm.com>
5026 David Sherwood <david.sherwood@arm.com>
5028 * target.def (empty_mask_is_expensive): New hook.
5029 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5030 * doc/tm.texi: Regenerate.
5031 * targhooks.h (default_empty_mask_is_expensive): Declare.
5032 * targhooks.c (default_empty_mask_is_expensive): New function.
5033 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5034 if the target says that empty masks are expensive.
5035 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5037 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5040 Alan Hayward <alan.hayward@arm.com>
5041 David Sherwood <david.sherwood@arm.com>
5043 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5044 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5045 (vect_use_loop_mask_for_alignment_p): New function.
5046 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5047 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5048 niters_skip argument. Make sure that the first niters_skip elements
5049 of the first iteration are inactive.
5050 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5051 Update call to vect_set_loop_masks_directly.
5052 (get_misalign_in_elems): New function, split out from...
5053 (vect_gen_prolog_loop_niters): ...here.
5054 (vect_update_init_of_dr): Take a code argument that specifies whether
5055 the adjustment should be added or subtracted.
5056 (vect_update_init_of_drs): Likewise.
5057 (vect_prepare_for_masked_peels): New function.
5058 (vect_do_peeling): Skip prologue peeling if we're using a mask
5059 instead. Update call to vect_update_inits_of_drs.
5060 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5062 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5063 alignment. Do not include the number of peeled iterations in
5064 the minimum threshold in that case.
5065 (vectorizable_induction): Adjust the start value down by
5066 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5067 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5068 Take the number of skipped iterations into account when calculating
5070 * tree-vect-stmts.c (vect_gen_while_not): New function.
5072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5073 Alan Hayward <alan.hayward@arm.com>
5074 David Sherwood <david.sherwood@arm.com>
5076 * doc/sourcebuild.texi (vect_fully_masked): Document.
5077 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5079 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5081 (vect_analyze_loop_2): ...here. Don't check the vectorization
5082 factor against the number of loop iterations if the loop is
5085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5086 Alan Hayward <alan.hayward@arm.com>
5087 David Sherwood <david.sherwood@arm.com>
5089 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5090 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5091 (dump_groups): Update accordingly.
5092 (iv_use::mem_type): New member variable.
5093 (address_p): New function.
5094 (record_use): Add a mem_type argument and initialize the new
5096 (record_group_use): Add a mem_type argument. Use address_p.
5097 Remove obsolete null checks of base_object. Update call to record_use.
5098 (find_interesting_uses_op): Update call to record_group_use.
5099 (find_interesting_uses_cond): Likewise.
5100 (find_interesting_uses_address): Likewise.
5101 (get_mem_type_for_internal_fn): New function.
5102 (find_address_like_use): Likewise.
5103 (find_interesting_uses_stmt): Try find_address_like_use before
5104 calling find_interesting_uses_op.
5105 (addr_offset_valid_p): Use the iv mem_type field as the type
5106 of the addressed memory.
5107 (add_autoinc_candidates): Likewise.
5108 (get_address_cost): Likewise.
5109 (split_small_address_groups_p): Use address_p.
5110 (split_address_groups): Likewise.
5111 (add_iv_candidate_for_use): Likewise.
5112 (autoinc_possible_for_pair): Likewise.
5113 (rewrite_groups): Likewise.
5114 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5115 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5116 (get_alias_ptr_type_for_ptr_address): New function.
5117 (rewrite_use_address): Rewrite address uses in calls that were
5118 identified by find_address_like_use.
5120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5121 Alan Hayward <alan.hayward@arm.com>
5122 David Sherwood <david.sherwood@arm.com>
5124 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5126 * gimple-expr.h (is_gimple_addressable: Likewise.
5127 * gimple-expr.c (is_gimple_address): Likewise.
5128 * internal-fn.c (expand_call_mem_ref): New function.
5129 (expand_mask_load_optab_fn): Use it.
5130 (expand_mask_store_optab_fn): Likewise.
5132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5133 Alan Hayward <alan.hayward@arm.com>
5134 David Sherwood <david.sherwood@arm.com>
5136 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5137 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5138 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5139 (cond_umax@var{mode}): Document.
5140 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5141 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5142 (cond_umin_optab, cond_umax_optab): New optabs.
5143 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5144 (COND_IOR, COND_XOR): New internal functions.
5145 * internal-fn.h (get_conditional_internal_fn): Declare.
5146 * internal-fn.c (cond_binary_direct): New macro.
5147 (expand_cond_binary_optab_fn): Likewise.
5148 (direct_cond_binary_optab_supported_p): Likewise.
5149 (get_conditional_internal_fn): New function.
5150 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5151 Cope with reduction statements that are vectorized as calls rather
5153 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5154 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5155 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5156 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5157 (UNSPEC_COND_EOR): New unspecs.
5158 (optab): Add mappings for them.
5159 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5160 (sve_int_op, sve_fp_op): New int attributes.
5162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5163 Alan Hayward <alan.hayward@arm.com>
5164 David Sherwood <david.sherwood@arm.com>
5166 * optabs.def (while_ult_optab): New optab.
5167 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5168 * internal-fn.def (WHILE_ULT): New internal function.
5169 * internal-fn.h (direct_internal_fn_supported_p): New override
5170 that takes two types as argument.
5171 * internal-fn.c (while_direct): New macro.
5172 (expand_while_optab_fn): New function.
5173 (convert_optab_supported_p): Likewise.
5174 (direct_while_optab_supported_p): New macro.
5175 * wide-int.h (wi::udiv_ceil): New function.
5176 * tree-vectorizer.h (rgroup_masks): New structure.
5177 (vec_loop_masks): New typedef.
5178 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5180 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5181 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5182 (vect_max_vf): New function.
5183 (slpeel_make_loop_iterate_ntimes): Delete.
5184 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5185 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5186 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5187 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5188 internal-fn.h, stor-layout.h and optabs-query.h.
5189 (vect_set_loop_mask): New function.
5190 (add_preheader_seq): Likewise.
5191 (add_header_seq): Likewise.
5192 (interleave_supported_p): Likewise.
5193 (vect_maybe_permute_loop_masks): Likewise.
5194 (vect_set_loop_masks_directly): Likewise.
5195 (vect_set_loop_condition_masked): Likewise.
5196 (vect_set_loop_condition_unmasked): New function, split out from
5197 slpeel_make_loop_iterate_ntimes.
5198 (slpeel_make_loop_iterate_ntimes): Rename to..
5199 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5200 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5201 (vect_do_peeling): Update call accordingly.
5202 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5204 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5205 mask_compare_type, can_fully_mask_p and fully_masked_p.
5206 (release_vec_loop_masks): New function.
5207 (_loop_vec_info): Use it to free the loop masks.
5208 (can_produce_all_loop_masks_p): New function.
5209 (vect_get_max_nscalars_per_iter): Likewise.
5210 (vect_verify_full_masking): Likewise.
5211 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5212 retries, and free the mask rgroups before retrying. Check loop-wide
5213 reasons for disallowing fully-masked loops. Make the final decision
5214 about whether use a fully-masked loop or not.
5215 (vect_estimate_min_profitable_iters): Do not assume that peeling
5216 for the number of iterations will be needed for fully-masked loops.
5217 (vectorizable_reduction): Disable fully-masked loops.
5218 (vectorizable_live_operation): Likewise.
5219 (vect_halve_mask_nunits): New function.
5220 (vect_double_mask_nunits): Likewise.
5221 (vect_record_loop_mask): Likewise.
5222 (vect_get_loop_mask): Likewise.
5223 (vect_transform_loop): Handle the case in which the final loop
5224 iteration might handle a partial vector. Call vect_set_loop_condition
5225 instead of slpeel_make_loop_iterate_ntimes.
5226 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5227 (check_load_store_masking): New function.
5228 (prepare_load_store_mask): Likewise.
5229 (vectorizable_store): Handle fully-masked loops.
5230 (vectorizable_load): Likewise.
5231 (supportable_widening_operation): Use vect_halve_mask_nunits for
5233 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5234 (vect_gen_while): New function.
5235 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5236 (aarch64_uqdec<mode>): New insn.
5238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5239 Alan Hayward <alan.hayward@arm.com>
5240 David Sherwood <david.sherwood@arm.com>
5242 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5243 (reduc_xor_scal_optab): New optabs.
5244 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5245 (reduc_xor_scal_@var{m}): Document.
5246 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5247 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5249 * fold-const-call.c (fold_const_call): Handle them.
5250 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5251 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5252 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5253 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5254 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5255 (UNSPEC_XORV): New unspecs.
5256 (optab): Add entries for them.
5257 (BITWISEV): New int iterator.
5258 (bit_reduc_op): New int attributes.
5260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5261 Alan Hayward <alan.hayward@arm.com>
5262 David Sherwood <david.sherwood@arm.com>
5264 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5265 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5266 * optabs.def (vec_shl_insert_optab): New optab.
5267 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5268 (duplicate_and_interleave): Likewise.
5269 * tree-vect-loop.c: Include internal-fn.h.
5270 (neutral_op_for_slp_reduction): New function, split out from
5271 get_initial_defs_for_reduction.
5272 (get_initial_def_for_reduction): Handle option 2 for variable-length
5273 vectors by loading the neutral value into a vector and then shifting
5274 the initial value into element 0.
5275 (get_initial_defs_for_reduction): Replace the code argument with
5276 the neutral value calculated by neutral_op_for_slp_reduction.
5277 Use gimple_build_vector for constant-length vectors.
5278 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5279 but the first group_size elements have a neutral value.
5280 Use duplicate_and_interleave otherwise.
5281 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5282 Update call to get_initial_defs_for_reduction. Handle SLP
5283 reductions for variable-length vectors by creating one vector
5284 result for each scalar result, with the elements associated
5285 with other scalar results stubbed out with the neutral value.
5286 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5287 Require IFN_VEC_SHL_INSERT for double reductions on
5288 variable-length vectors, or SLP reductions that have
5289 a neutral value. Require can_duplicate_and_interleave_p
5290 support for variable-length unchained SLP reductions if there
5291 is no neutral value, such as for MIN/MAX reductions. Also require
5292 the number of vector elements to be a multiple of the number of
5293 SLP statements when doing variable-length unchained SLP reductions.
5294 Update call to vect_create_epilog_for_reduction.
5295 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5296 and remove initial values.
5297 (duplicate_and_interleave): Make public.
5298 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5299 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5302 Alan Hayward <alan.hayward@arm.com>
5303 David Sherwood <david.sherwood@arm.com>
5305 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5306 (can_duplicate_and_interleave_p): New function.
5307 (vect_get_and_check_slp_defs): Take the vector of statements
5308 rather than just the current one. Remove excess parentheses.
5309 Restriction rejectinon of vect_constant_def and vect_external_def
5310 for variable-length vectors to boolean types, or types for which
5311 can_duplicate_and_interleave_p is false.
5312 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5313 (duplicate_and_interleave): New function.
5314 (vect_get_constant_vectors): Use gimple_build_vector for
5315 constant-length vectors and suitable variable-length constant
5316 vectors. Use duplicate_and_interleave for other variable-length
5317 vectors. Don't defer the update when inserting new statements.
5319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5320 Alan Hayward <alan.hayward@arm.com>
5321 David Sherwood <david.sherwood@arm.com>
5323 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5324 min_profitable_iters doesn't go negative.
5326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5327 Alan Hayward <alan.hayward@arm.com>
5328 David Sherwood <david.sherwood@arm.com>
5330 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5331 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5332 * optabs.def (vec_mask_load_lanes_optab): New optab.
5333 (vec_mask_store_lanes_optab): Likewise.
5334 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5335 (MASK_STORE_LANES): Likewise.
5336 * internal-fn.c (mask_load_lanes_direct): New macro.
5337 (mask_store_lanes_direct): Likewise.
5338 (expand_mask_load_optab_fn): Handle masked operations.
5339 (expand_mask_load_lanes_optab_fn): New macro.
5340 (expand_mask_store_optab_fn): Handle masked operations.
5341 (expand_mask_store_lanes_optab_fn): New macro.
5342 (direct_mask_load_lanes_optab_supported_p): Likewise.
5343 (direct_mask_store_lanes_optab_supported_p): Likewise.
5344 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5346 (vect_load_lanes_supported): Likewise.
5347 * tree-vect-data-refs.c (strip_conversion): New function.
5348 (can_group_stmts_p): Likewise.
5349 (vect_analyze_data_ref_accesses): Use it instead of checking
5350 for a pair of assignments.
5351 (vect_store_lanes_supported): Take a masked_p parameter.
5352 (vect_load_lanes_supported): Likewise.
5353 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5354 vect_store_lanes_supported and vect_load_lanes_supported.
5355 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5356 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5357 parameter. Don't allow gaps for masked accesses.
5358 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5359 and vect_load_lanes_supported.
5360 (get_load_store_type): Take a masked_p parameter and update
5361 call to get_group_load_store_type.
5362 (vectorizable_store): Update call to get_load_store_type.
5363 Handle IFN_MASK_STORE_LANES.
5364 (vectorizable_load): Update call to get_load_store_type.
5365 Handle IFN_MASK_LOAD_LANES.
5367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5368 Alan Hayward <alan.hayward@arm.com>
5369 David Sherwood <david.sherwood@arm.com>
5371 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5373 * config/aarch64/aarch64-protos.h
5374 (aarch64_sve_struct_memory_operand_p): Declare.
5375 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5376 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5377 (VPRED, vpred): Handle SVE structure modes.
5378 * config/aarch64/constraints.md (Utx): New constraint.
5379 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5380 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5381 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5382 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5383 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5384 structure modes. Split into pieces after RA.
5385 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5386 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5388 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5389 SVE structure modes.
5390 (aarch64_classify_address): Likewise.
5391 (sizetochar): Move earlier in file.
5392 (aarch64_print_operand): Handle SVE register lists.
5393 (aarch64_array_mode): New function.
5394 (aarch64_sve_struct_memory_operand_p): Likewise.
5395 (TARGET_ARRAY_MODE): Redefine.
5397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5398 Alan Hayward <alan.hayward@arm.com>
5399 David Sherwood <david.sherwood@arm.com>
5401 * target.def (array_mode): New target hook.
5402 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5403 * doc/tm.texi: Regenerate.
5404 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5405 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5406 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5408 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5412 Alan Hayward <alan.hayward@arm.com>
5413 David Sherwood <david.sherwood@arm.com>
5415 * fold-const.c (fold_binary_loc): Check the argument types
5416 rather than the result type when testing for a vector operation.
5418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5420 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5421 * doc/tm.texi: Regenerate.
5423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5424 Alan Hayward <alan.hayward@arm.com>
5425 David Sherwood <david.sherwood@arm.com>
5427 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5428 (sve): Document new AArch64 extension.
5429 * doc/md.texi (w): Extend the description of the AArch64
5430 constraint to include SVE vectors.
5431 (Upl, Upa): Document new AArch64 predicate constraints.
5432 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5434 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5435 (msve-vector-bits=): New option.
5436 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5437 SVE when these are disabled.
5438 (sve): New extension.
5439 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5440 modes. Adjust their number of units based on aarch64_sve_vg.
5441 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5442 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5443 aarch64_addr_query_type.
5444 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5445 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5446 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5447 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5448 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5449 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5450 (aarch64_simd_imm_zero_p): Delete.
5451 (aarch64_check_zero_based_sve_index_immediate): Declare.
5452 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5453 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5454 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5455 (aarch64_sve_float_mul_immediate_p): Likewise.
5456 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5458 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5459 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5460 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5461 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5462 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5463 (aarch64_regmode_natural_size): Likewise.
5464 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5465 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5467 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5468 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5469 for VG and the SVE predicate registers.
5470 (V_ALIASES): Add a "z"-prefixed alias.
5471 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5472 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5473 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5474 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5475 (REG_CLASS_NAMES): Add entries for them.
5476 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5477 and the predicate registers.
5478 (aarch64_sve_vg): Declare.
5479 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5480 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5481 (REGMODE_NATURAL_SIZE): Define.
5482 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5484 * config/aarch64/aarch64.c: Include cfgrtl.h.
5485 (simd_immediate_info): Add a constructor for series vectors,
5486 and an associated step field.
5487 (aarch64_sve_vg): New variable.
5488 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5489 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5490 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5491 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5492 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5493 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5494 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5495 (aarch64_get_mask_mode): New functions.
5496 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5497 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5498 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5499 predicate modes and predicate registers. Explicitly restrict
5500 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5501 to store a vector mode if it is recognized by
5502 aarch64_classify_vector_mode.
5503 (aarch64_regmode_natural_size): New function.
5504 (aarch64_hard_regno_caller_save_mode): Return the original mode
5506 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5507 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5508 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5509 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5511 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5512 does not overlap dest if the function is frame-related. Handle
5514 (aarch64_split_add_offset): New function.
5515 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5516 them aarch64_add_offset.
5517 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5518 and update call to aarch64_sub_sp.
5519 (aarch64_add_cfa_expression): New function.
5520 (aarch64_expand_prologue): Pass extra temporary registers to the
5521 functions above. Handle the case in which we need to emit new
5522 DW_CFA_expressions for registers that were originally saved
5523 relative to the stack pointer, but now have to be expressed
5524 relative to the frame pointer.
5525 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5527 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5528 IP0 and IP1 values for SVE frames.
5529 (aarch64_expand_vec_series): New function.
5530 (aarch64_expand_sve_widened_duplicate): Likewise.
5531 (aarch64_expand_sve_const_vector): Likewise.
5532 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5533 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5534 into the register, rather than emitting a SET directly.
5535 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5536 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5537 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5538 (offset_9bit_signed_scaled_p): New functions.
5539 (aarch64_replicate_bitmask_imm): New function.
5540 (aarch64_bitmask_imm): Use it.
5541 (aarch64_cannot_force_const_mem): Reject expressions involving
5542 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5543 (aarch64_classify_index): Handle SVE indices, by requiring
5544 a plain register index with a scale that matches the element size.
5545 (aarch64_classify_address): Handle SVE addresses. Assert that
5546 the mode of the address is VOIDmode or an integer mode.
5547 Update call to aarch64_classify_symbol.
5548 (aarch64_classify_symbolic_expression): Update call to
5549 aarch64_classify_symbol.
5550 (aarch64_const_vec_all_in_range_p): New function.
5551 (aarch64_print_vector_float_operand): Likewise.
5552 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5553 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5554 and the FP immediates 1.0 and 0.5.
5555 (aarch64_print_address_internal): Handle SVE addresses.
5556 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5557 (aarch64_regno_regclass): Handle predicate registers.
5558 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5560 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5561 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5562 (aarch64_convert_sve_vector_bits): New function.
5563 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5564 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5566 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5567 Handle SVE vector and predicate modes. Accept VL-based constants
5568 that need only one temporary register, and VL offsets that require
5569 no temporary registers.
5570 (aarch64_conditional_register_usage): Mark the predicate registers
5571 as fixed if SVE isn't available.
5572 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5573 Return true for SVE vector and predicate modes.
5574 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5575 rather than an unsigned int. Handle SVE modes.
5576 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5578 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5580 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5581 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5582 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5583 (aarch64_sve_float_mul_immediate_p): New functions.
5584 (aarch64_sve_valid_immediate): New function.
5585 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5586 Explicitly reject structure modes. Check for INDEX constants.
5587 Handle PTRUE and PFALSE constants.
5588 (aarch64_check_zero_based_sve_index_immediate): New function.
5589 (aarch64_simd_imm_zero_p): Delete.
5590 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5591 vector modes. Accept constants in the range of CNT[BHWD].
5592 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5593 ask for an Advanced SIMD mode.
5594 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5595 (aarch64_simd_vector_alignment): Handle SVE predicates.
5596 (aarch64_vectorize_preferred_vector_alignment): New function.
5597 (aarch64_simd_vector_alignment_reachable): Use it instead of
5599 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5600 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5602 (MAX_VECT_LEN): Delete.
5603 (expand_vec_perm_d): Add a vec_flags field.
5604 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5605 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5606 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5608 (aarch64_evpc_rev): Rename to...
5609 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5610 (aarch64_evpc_rev_global): New function.
5611 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5612 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5614 (aarch64_evpc_sve_tbl): New function.
5615 (aarch64_expand_vec_perm_const_1): Update after rename of
5616 aarch64_evpc_rev. Handle SVE permutes too, trying
5617 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5618 than aarch64_evpc_tbl.
5619 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5620 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5621 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5622 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5623 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5624 (aarch64_expand_sve_vcond): New functions.
5625 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5626 of aarch64_vector_mode_p.
5627 (aarch64_dwarf_poly_indeterminate_value): New function.
5628 (aarch64_compute_pressure_classes): Likewise.
5629 (aarch64_can_change_mode_class): Likewise.
5630 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5631 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5632 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5633 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5634 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5635 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5636 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5637 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5639 (Dn, Dl, Dr): Accept const as well as const_vector.
5640 (Dz): Likewise. Compare against CONST0_RTX.
5641 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5642 of "vector" where appropriate.
5643 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5644 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5645 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5646 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5647 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5648 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5649 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5650 (v_int_equiv): Extend to SVE modes.
5651 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5653 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5654 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5655 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5656 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5657 (SVE_COND_FP_CMP): New int iterators.
5658 (perm_hilo): Handle the new unpack unspecs.
5659 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5661 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5662 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5663 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5664 (aarch64_equality_operator, aarch64_constant_vector_operand)
5665 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5666 (aarch64_sve_nonimmediate_operand): Likewise.
5667 (aarch64_sve_general_operand): Likewise.
5668 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5669 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5670 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5671 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5672 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5673 (aarch64_sve_float_arith_immediate): Likewise.
5674 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5675 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5676 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5677 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5678 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5679 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5680 (aarch64_sve_float_arith_operand): Likewise.
5681 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5682 (aarch64_sve_float_mul_operand): Likewise.
5683 (aarch64_sve_vec_perm_operand): Likewise.
5684 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5685 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5686 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5687 as well as const_vector.
5688 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5689 in file. Use CONST0_RTX and CONSTM1_RTX.
5690 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5691 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5692 Use aarch64_simd_imm_zero.
5693 * config/aarch64/aarch64-sve.md: New file.
5694 * config/aarch64/aarch64.md: Include it.
5695 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5696 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5697 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5698 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5699 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5700 (sve): New attribute.
5701 (enabled): Disable instructions with the sve attribute unless
5703 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5704 aarch64_expand_mov_immediate.
5705 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5706 CNT[BHSD] immediates.
5707 (movti): Split CONST_POLY_INT moves into two halves.
5708 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5709 Split additions that need a temporary here if the destination
5710 is the stack pointer.
5711 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5712 (*add<mode>3_poly_1): New instruction.
5713 (set_clobber_cc): New expander.
5715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5717 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5718 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5719 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5720 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5721 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5722 Change innermode from fixed_mode_size to machine_mode.
5723 (simplify_subreg): Update call accordingly. Handle a constant-sized
5724 subreg of a variable-length CONST_VECTOR.
5726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5727 Alan Hayward <alan.hayward@arm.com>
5728 David Sherwood <david.sherwood@arm.com>
5730 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5731 (add_offset_to_base): New function, split out from...
5732 (create_mem_ref): ...here. When handling a scale other than 1,
5733 check first whether the address is valid without the offset.
5734 Add it into the base if so, leaving the index and scale as-is.
5736 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5739 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5740 fold_for_warn before checking if arg2 is INTEGER_CST.
5742 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5744 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5745 (store_multiple_operation): Delete.
5746 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5747 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5748 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5749 guarded by TARGET_STRING.
5750 (rs6000_output_load_multiple): Delete.
5751 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5752 OPTION_MASK_STRING / TARGET_STRING handling.
5753 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5754 (const rs6000_opt_masks) <"string">: Change mask to 0.
5755 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5756 (MASK_STRING): Delete.
5757 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5759 (load_multiple): Delete.
5766 (store_multiple): Delete.
5773 (movmemsi_8reg): Delete.
5774 (corresponding unnamed define_insn): Delete.
5775 (movmemsi_6reg): Delete.
5776 (corresponding unnamed define_insn): Delete.
5777 (movmemsi_4reg): Delete.
5778 (corresponding unnamed define_insn): Delete.
5779 (movmemsi_2reg): Delete.
5780 (corresponding unnamed define_insn): Delete.
5781 (movmemsi_1reg): Delete.
5782 (corresponding unnamed define_insn): Delete.
5783 * config/rs6000/rs6000.opt (mno-string): New.
5784 (mstring): Replace by deprecation warning stub.
5785 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5787 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5789 * regrename.c (regrename_do_replace): If replacing the same
5790 reg multiple times, try to reuse last created gen_raw_REG.
5793 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5794 main to workaround a bug in GDB.
5796 2018-01-12 Tom de Vries <tom@codesourcery.com>
5799 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5801 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5803 PR rtl-optimization/80481
5804 * ira-color.c (get_cap_member): New function.
5805 (allocnos_conflict_by_live_ranges_p): Use it.
5806 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5807 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5809 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5812 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5813 (*saddl_se_1): Ditto.
5815 (*ssubl_se_1): Ditto.
5817 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5819 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5820 rather than wi::to_widest for DR_INITs.
5821 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5822 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5823 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5825 (vect_analyze_group_access_1): Note that here.
5827 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5829 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5830 polynomial type sizes.
5832 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5834 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5835 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5836 (gimple_add_tmp_var): Likewise.
5838 2018-01-12 Martin Liska <mliska@suse.cz>
5840 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5841 (gimple_alloc_sizes): Likewise.
5842 (dump_gimple_statistics): Use PRIu64 in printf format.
5843 * gimple.h: Change uint64_t to int.
5845 2018-01-12 Martin Liska <mliska@suse.cz>
5847 * tree-core.h: Use uint64_t instead of int.
5848 * tree.c (tree_node_counts): Likewise.
5849 (tree_node_sizes): Likewise.
5850 (dump_tree_statistics): Use PRIu64 in printf format.
5852 2018-01-12 Martin Liska <mliska@suse.cz>
5854 * Makefile.in: As qsort_chk is implemented in vec.c, add
5855 vec.o to linkage of gencfn-macros.
5856 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5857 passing the info to record_node_allocation_statistics.
5858 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5860 * ggc-common.c (struct ggc_usage): Add operator== and use
5861 it in operator< and compare function.
5862 * mem-stats.h (struct mem_usage): Likewise.
5863 * vec.c (struct vec_usage): Remove operator< and compare
5864 function. Can be simply inherited.
5866 2018-01-12 Martin Jambor <mjambor@suse.cz>
5869 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5870 * tree-ssa-math-opts.c: Include domwalk.h.
5871 (convert_mult_to_fma_1): New function.
5872 (fma_transformation_info): New type.
5873 (fma_deferring_state): Likewise.
5874 (cancel_fma_deferring): New function.
5875 (result_of_phi): Likewise.
5876 (last_fma_candidate_feeds_initial_phi): Likewise.
5877 (convert_mult_to_fma): Added deferring logic, split actual
5878 transformation to convert_mult_to_fma_1.
5879 (math_opts_dom_walker): New type.
5880 (math_opts_dom_walker::after_dom_children): New method, body moved
5881 here from pass_optimize_widening_mul::execute, added deferring logic
5883 (pass_optimize_widening_mul::execute): Moved most of code to
5884 math_opts_dom_walker::after_dom_children.
5885 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5886 * config/i386/i386.c (ix86_option_override_internal): Added
5887 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5889 2018-01-12 Richard Biener <rguenther@suse.de>
5892 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5893 inline instance vars.
5895 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5898 * config/rx/rx.c (rx_is_restricted_memory_address):
5901 2018-01-12 Richard Biener <rguenther@suse.de>
5903 PR tree-optimization/80846
5904 * target.def (split_reduction): New target hook.
5905 * targhooks.c (default_split_reduction): New function.
5906 * targhooks.h (default_split_reduction): Declare.
5907 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5908 target requests first reduce vectors by combining low and high
5910 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5911 (get_vectype_for_scalar_type_and_size): Export.
5912 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5913 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5914 * doc/tm.texi: Regenerate.
5915 * config/i386/i386.c (ix86_split_reduction): Implement
5916 TARGET_VECTORIZE_SPLIT_REDUCTION.
5918 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5921 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5922 in PIC mode except for TARGET_VXWORKS_RTP.
5923 * config/sparc/sparc.c: Include cfgrtl.h.
5924 (TARGET_INIT_PIC_REG): Define.
5925 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5926 (sparc_pic_register_p): New predicate.
5927 (sparc_legitimate_address_p): Use it.
5928 (sparc_legitimize_pic_address): Likewise.
5929 (sparc_delegitimize_address): Likewise.
5930 (sparc_mode_dependent_address_p): Likewise.
5931 (gen_load_pcrel_sym): Remove 4th parameter.
5932 (load_got_register): Adjust call to above. Remove obsolete stuff.
5933 (sparc_expand_prologue): Do not call load_got_register here.
5934 (sparc_flat_expand_prologue): Likewise.
5935 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5936 (sparc_use_pseudo_pic_reg): New function.
5937 (sparc_init_pic_reg): Likewise.
5938 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5939 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5941 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5943 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5944 Add item for branch_cost.
5946 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5948 PR rtl-optimization/83565
5949 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5950 not extend the result to a larger mode for rotate operations.
5951 (num_sign_bit_copies1): Likewise.
5953 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5956 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5958 Use values-Xc.o for -pedantic.
5959 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5961 2018-01-12 Martin Liska <mliska@suse.cz>
5964 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5966 (possible_polymorphic_call_targets): Use it.
5967 (ipa_devirt): Likewise.
5969 2018-01-12 Martin Liska <mliska@suse.cz>
5971 * profile-count.h (enum profile_quality): Use 0 as invalid
5972 enum value of profile_quality.
5974 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5976 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5977 -mext-string options.
5979 2018-01-12 Richard Biener <rguenther@suse.de>
5981 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5982 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5983 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5985 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5987 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5989 * configure.ac (--with-long-double-format): Add support for the
5990 configuration option to change the default long double format on
5992 * config.gcc (powerpc*-linux*-*): Likewise.
5993 * configure: Regenerate.
5994 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5995 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5996 used without modification.
5998 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6000 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6001 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6002 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6003 MISC_BUILTIN_SPEC_BARRIER.
6004 (rs6000_init_builtins): Likewise.
6005 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6007 (speculation_barrier): New define_insn.
6008 * doc/extend.texi: Document __builtin_speculation_barrier.
6010 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6013 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6014 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6015 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6017 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6018 integral modes instead of "ss" and "sd".
6019 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6020 vectors with 32-bit and 64-bit elements.
6021 (vecdupssescalarmodesuffix): New mode attribute.
6022 (vec_dup<mode>): Use it.
6024 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6027 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6028 frame if argument is passed on stack.
6030 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6033 * ree.c (combine_reaching_defs): Optimize also
6034 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6035 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6037 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6040 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6042 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6045 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6046 after they are computed.
6048 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6050 PR tree-optimization/83695
6051 * gimple-loop-linterchange.cc
6052 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6053 reset cached scev information after interchange.
6054 (pass_linterchange::execute): Remove call to scev_reset_htab.
6056 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6058 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6059 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6060 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6061 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6062 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6063 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6064 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6065 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6066 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6067 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6068 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6069 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6070 (V_lane_reg): Likewise.
6071 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6073 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6074 (vfmal_lane_low<mode>_intrinsic,
6075 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6076 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6077 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6078 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6079 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6080 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6082 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6084 * config/arm/arm-cpus.in (fp16fml): New feature.
6085 (ALL_SIMD): Add fp16fml.
6086 (armv8.2-a): Add fp16fml as an option.
6087 (armv8.3-a): Likewise.
6088 (armv8.4-a): Add fp16fml as part of fp16.
6089 * config/arm/arm.h (TARGET_FP16FML): Define.
6090 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6092 * config/arm/arm-modes.def (V2HF): Define.
6093 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6094 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6095 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6096 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6097 vfmsl_low, vfmsl_high): New set of builtins.
6098 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6099 (vfml_op): New code attribute.
6100 (VFMLHALVES): New int iterator.
6101 (VFML, VFMLSEL): New mode attributes.
6102 (V_reg): Define mapping for V2HF.
6103 (V_hi, V_lo): New mode attributes.
6104 (VF_constraint): Likewise.
6105 (vfml_half, vfml_half_selector): New int attributes.
6106 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6108 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6109 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6111 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6112 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6113 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6114 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6116 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6117 Document new effective target and option set.
6119 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6121 * config/arm/arm-cpus.in (armv8_4): New feature.
6122 (ARMv8_4a): New fgroup.
6123 (armv8.4-a): New arch.
6124 * config/arm/arm-tables.opt: Regenerate.
6125 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6126 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6127 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6128 Add matching rules for -march=armv8.4-a and extensions.
6129 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6131 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6134 * config/rx/rx.md (BW): New mode attribute.
6135 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6137 2018-01-11 Richard Biener <rguenther@suse.de>
6139 PR tree-optimization/83435
6140 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6141 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6142 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6144 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6145 Alan Hayward <alan.hayward@arm.com>
6146 David Sherwood <david.sherwood@arm.com>
6148 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6150 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6151 (aarch64_print_address_internal): Use it to check for a zero offset.
6153 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6154 Alan Hayward <alan.hayward@arm.com>
6155 David Sherwood <david.sherwood@arm.com>
6157 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6158 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6159 Return a poly_int64 rather than a HOST_WIDE_INT.
6160 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6161 rather than a HOST_WIDE_INT.
6162 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6163 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6164 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6165 final_offset from HOST_WIDE_INT to poly_int64.
6166 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6167 to_constant when getting the number of units in an Advanced SIMD
6169 (aarch64_builtin_vectorized_function): Check for a constant number
6171 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6173 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6174 attribute instead of GET_MODE_NUNITS.
6175 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6176 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6177 GET_MODE_SIZE for fixed-size registers.
6178 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6179 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6180 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6181 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6182 (aarch64_print_operand, aarch64_print_address_internal)
6183 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6184 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6185 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6186 Handle polynomial GET_MODE_SIZE.
6187 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6188 wider than SImode without modification.
6189 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6190 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6191 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6192 passing and returning SVE modes.
6193 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6194 rather than GEN_INT.
6195 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6196 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6197 (aarch64_allocate_and_probe_stack_space): Likewise.
6198 (aarch64_layout_frame): Cope with polynomial offsets.
6199 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6200 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6202 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6203 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6204 poly_int64 rather than a HOST_WIDE_INT.
6205 (aarch64_get_separate_components, aarch64_process_components)
6206 (aarch64_expand_prologue, aarch64_expand_epilogue)
6207 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6208 (aarch64_anchor_offset): New function, split out from...
6209 (aarch64_legitimize_address): ...here.
6210 (aarch64_builtin_vectorization_cost): Handle polynomial
6211 TYPE_VECTOR_SUBPARTS.
6212 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6214 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6215 number of elements from the PARALLEL rather than the mode.
6216 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6217 rather than GET_MODE_BITSIZE.
6218 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6219 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6220 (aarch64_expand_vec_perm_const_1): Handle polynomial
6221 d->perm.length () and d->perm elements.
6222 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6223 Apply to_constant to d->perm elements.
6224 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6225 polynomial CONST_VECTOR_NUNITS.
6226 (aarch64_move_pointer): Take amount as a poly_int64 rather
6228 (aarch64_progress_pointer): Avoid temporary variable.
6229 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6230 the mode attribute instead of GET_MODE.
6232 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6233 Alan Hayward <alan.hayward@arm.com>
6234 David Sherwood <david.sherwood@arm.com>
6236 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6237 x exists before using it.
6238 (aarch64_add_constant_internal): Rename to...
6239 (aarch64_add_offset_1): ...this. Replace regnum with separate
6240 src and dest rtxes. Handle the case in which they're different,
6241 including when the offset is zero. Replace scratchreg with an rtx.
6242 Use 2 additions if there is no spare register into which we can
6243 move a 16-bit constant.
6244 (aarch64_add_constant): Delete.
6245 (aarch64_add_offset): Replace reg with separate src and dest
6246 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6247 Use aarch64_add_offset_1.
6248 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6249 an rtx rather than an int. Take the delta as a poly_int64
6250 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6251 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6252 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6253 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6254 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6256 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6257 aarch64_add_constant.
6259 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6261 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6262 Use scalar_float_mode.
6264 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6266 * config/aarch64/aarch64-simd.md
6267 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6268 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6269 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6270 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6271 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6272 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6273 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6274 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6275 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6276 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6278 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6281 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6282 targ_options->x_arm_arch_string is non NULL.
6284 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6286 * config/aarch64/aarch64.h
6287 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6289 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6292 * expmed.c (emit_store_flag_force): Swap if const op0
6293 and change VOIDmode to mode of op0.
6295 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6297 PR rtl-optimization/83761
6298 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6299 than bytes to mode_for_size.
6301 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6304 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6305 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6308 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6311 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6312 when in layout mode.
6313 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6314 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6317 2018-01-10 Michael Collison <michael.collison@arm.com>
6319 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6320 * config/aarch64/aarch64-option-extension.def: Add
6321 AARCH64_OPT_EXTENSION of 'fp16fml'.
6322 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6323 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6324 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6325 * config/aarch64/constraints.md (Ui7): New constraint.
6326 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6327 (VFMLA_SEL_W): Ditto.
6330 (VFMLA16_LOW): New int iterator.
6331 (VFMLA16_HIGH): Ditto.
6332 (UNSPEC_FMLAL): New unspec.
6333 (UNSPEC_FMLSL): Ditto.
6334 (UNSPEC_FMLAL2): Ditto.
6335 (UNSPEC_FMLSL2): Ditto.
6336 (f16mac): New code attribute.
6337 * config/aarch64/aarch64-simd-builtins.def
6338 (aarch64_fmlal_lowv2sf): Ditto.
6339 (aarch64_fmlsl_lowv2sf): Ditto.
6340 (aarch64_fmlalq_lowv4sf): Ditto.
6341 (aarch64_fmlslq_lowv4sf): Ditto.
6342 (aarch64_fmlal_highv2sf): Ditto.
6343 (aarch64_fmlsl_highv2sf): Ditto.
6344 (aarch64_fmlalq_highv4sf): Ditto.
6345 (aarch64_fmlslq_highv4sf): Ditto.
6346 (aarch64_fmlal_lane_lowv2sf): Ditto.
6347 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6348 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6349 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6350 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6351 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6352 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6353 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6354 (aarch64_fmlal_lane_highv2sf): Ditto.
6355 (aarch64_fmlsl_lane_highv2sf): Ditto.
6356 (aarch64_fmlal_laneq_highv2sf): Ditto.
6357 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6358 (aarch64_fmlalq_lane_highv4sf): Ditto.
6359 (aarch64_fmlsl_lane_highv4sf): Ditto.
6360 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6361 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6362 * config/aarch64/aarch64-simd.md:
6363 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6364 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6365 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6366 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6367 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6368 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6369 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6370 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6371 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6372 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6373 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6374 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6375 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6376 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6377 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6378 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6379 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6380 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6381 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6382 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6383 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6384 (vfmlsl_low_u32): Ditto.
6385 (vfmlalq_low_u32): Ditto.
6386 (vfmlslq_low_u32): Ditto.
6387 (vfmlal_high_u32): Ditto.
6388 (vfmlsl_high_u32): Ditto.
6389 (vfmlalq_high_u32): Ditto.
6390 (vfmlslq_high_u32): Ditto.
6391 (vfmlal_lane_low_u32): Ditto.
6392 (vfmlsl_lane_low_u32): Ditto.
6393 (vfmlal_laneq_low_u32): Ditto.
6394 (vfmlsl_laneq_low_u32): Ditto.
6395 (vfmlalq_lane_low_u32): Ditto.
6396 (vfmlslq_lane_low_u32): Ditto.
6397 (vfmlalq_laneq_low_u32): Ditto.
6398 (vfmlslq_laneq_low_u32): Ditto.
6399 (vfmlal_lane_high_u32): Ditto.
6400 (vfmlsl_lane_high_u32): Ditto.
6401 (vfmlal_laneq_high_u32): Ditto.
6402 (vfmlsl_laneq_high_u32): Ditto.
6403 (vfmlalq_lane_high_u32): Ditto.
6404 (vfmlslq_lane_high_u32): Ditto.
6405 (vfmlalq_laneq_high_u32): Ditto.
6406 (vfmlslq_laneq_high_u32): Ditto.
6407 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6408 (AARCH64_FL_FOR_ARCH8_4): New.
6409 (AARCH64_ISA_F16FML): New ISA flag.
6410 (TARGET_F16FML): New feature flag for fp16fml.
6411 (doc/invoke.texi): Document new fp16fml option.
6413 2018-01-10 Michael Collison <michael.collison@arm.com>
6415 * config/aarch64/aarch64-builtins.c:
6416 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6417 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6418 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6419 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6420 (AARCH64_ISA_SHA3): New ISA flag.
6421 (TARGET_SHA3): New feature flag for sha3.
6422 * config/aarch64/iterators.md (sha512_op): New int attribute.
6423 (CRYPTO_SHA512): New int iterator.
6424 (UNSPEC_SHA512H): New unspec.
6425 (UNSPEC_SHA512H2): Ditto.
6426 (UNSPEC_SHA512SU0): Ditto.
6427 (UNSPEC_SHA512SU1): Ditto.
6428 * config/aarch64/aarch64-simd-builtins.def
6429 (aarch64_crypto_sha512hqv2di): New builtin.
6430 (aarch64_crypto_sha512h2qv2di): Ditto.
6431 (aarch64_crypto_sha512su0qv2di): Ditto.
6432 (aarch64_crypto_sha512su1qv2di): Ditto.
6433 (aarch64_eor3qv8hi): Ditto.
6434 (aarch64_rax1qv2di): Ditto.
6435 (aarch64_xarqv2di): Ditto.
6436 (aarch64_bcaxqv8hi): Ditto.
6437 * config/aarch64/aarch64-simd.md:
6438 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6439 (aarch64_crypto_sha512su0qv2di): Ditto.
6440 (aarch64_crypto_sha512su1qv2di): Ditto.
6441 (aarch64_eor3qv8hi): Ditto.
6442 (aarch64_rax1qv2di): Ditto.
6443 (aarch64_xarqv2di): Ditto.
6444 (aarch64_bcaxqv8hi): Ditto.
6445 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6446 (vsha512h2q_u64): Ditto.
6447 (vsha512su0q_u64): Ditto.
6448 (vsha512su1q_u64): Ditto.
6449 (veor3q_u16): Ditto.
6450 (vrax1q_u64): Ditto.
6452 (vbcaxq_u16): Ditto.
6453 * config/arm/types.md (crypto_sha512): New type attribute.
6454 (crypto_sha3): Ditto.
6455 (doc/invoke.texi): Document new sha3 option.
6457 2018-01-10 Michael Collison <michael.collison@arm.com>
6459 * config/aarch64/aarch64-builtins.c:
6460 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6461 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6462 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6463 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6464 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6465 (AARCH64_ISA_SM4): New ISA flag.
6466 (TARGET_SM4): New feature flag for sm4.
6467 * config/aarch64/aarch64-simd-builtins.def
6468 (aarch64_sm3ss1qv4si): Ditto.
6469 (aarch64_sm3tt1aq4si): Ditto.
6470 (aarch64_sm3tt1bq4si): Ditto.
6471 (aarch64_sm3tt2aq4si): Ditto.
6472 (aarch64_sm3tt2bq4si): Ditto.
6473 (aarch64_sm3partw1qv4si): Ditto.
6474 (aarch64_sm3partw2qv4si): Ditto.
6475 (aarch64_sm4eqv4si): Ditto.
6476 (aarch64_sm4ekeyqv4si): Ditto.
6477 * config/aarch64/aarch64-simd.md:
6478 (aarch64_sm3ss1qv4si): Ditto.
6479 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6480 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6481 (aarch64_sm4eqv4si): Ditto.
6482 (aarch64_sm4ekeyqv4si): Ditto.
6483 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6484 (sm3part_op): Ditto.
6485 (CRYPTO_SM3TT): Ditto.
6486 (CRYPTO_SM3PART): Ditto.
6487 (UNSPEC_SM3SS1): New unspec.
6488 (UNSPEC_SM3TT1A): Ditto.
6489 (UNSPEC_SM3TT1B): Ditto.
6490 (UNSPEC_SM3TT2A): Ditto.
6491 (UNSPEC_SM3TT2B): Ditto.
6492 (UNSPEC_SM3PARTW1): Ditto.
6493 (UNSPEC_SM3PARTW2): Ditto.
6494 (UNSPEC_SM4E): Ditto.
6495 (UNSPEC_SM4EKEY): Ditto.
6496 * config/aarch64/constraints.md (Ui2): New constraint.
6497 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6498 * config/arm/types.md (crypto_sm3): New type attribute.
6499 (crypto_sm4): Ditto.
6500 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6501 (vsm3tt1aq_u32): Ditto.
6502 (vsm3tt1bq_u32): Ditto.
6503 (vsm3tt2aq_u32): Ditto.
6504 (vsm3tt2bq_u32): Ditto.
6505 (vsm3partw1q_u32): Ditto.
6506 (vsm3partw2q_u32): Ditto.
6507 (vsm4eq_u32): Ditto.
6508 (vsm4ekeyq_u32): Ditto.
6509 (doc/invoke.texi): Document new sm4 option.
6511 2018-01-10 Michael Collison <michael.collison@arm.com>
6513 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6514 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6515 (AARCH64_FL_FOR_ARCH8_4): New.
6516 (AARCH64_FL_V8_4): New flag.
6517 (doc/invoke.texi): Document new armv8.4-a option.
6519 2018-01-10 Michael Collison <michael.collison@arm.com>
6521 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6522 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6523 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6524 * config/aarch64/aarch64-option-extension.def: Add
6525 AARCH64_OPT_EXTENSION of 'sha2'.
6526 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6527 (crypto): Disable sha2 and aes if crypto disabled.
6528 (crypto): Enable aes and sha2 if enabled.
6529 (simd): Disable sha2 and aes if simd disabled.
6530 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6532 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6533 (TARGET_SHA2): New feature flag for sha2.
6534 (TARGET_AES): New feature flag for aes.
6535 * config/aarch64/aarch64-simd.md:
6536 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6537 conditional on TARGET_AES.
6538 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6539 (aarch64_crypto_sha1hsi): Make pattern conditional
6541 (aarch64_crypto_sha1hv4si): Ditto.
6542 (aarch64_be_crypto_sha1hv4si): Ditto.
6543 (aarch64_crypto_sha1su1v4si): Ditto.
6544 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6545 (aarch64_crypto_sha1su0v4si): Ditto.
6546 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6547 (aarch64_crypto_sha256su0v4si): Ditto.
6548 (aarch64_crypto_sha256su1v4si): Ditto.
6549 (doc/invoke.texi): Document new aes and sha2 options.
6551 2018-01-10 Martin Sebor <msebor@redhat.com>
6553 PR tree-optimization/83781
6554 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6557 2018-01-11 Martin Sebor <msebor@gmail.com>
6558 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6560 PR tree-optimization/83501
6561 PR tree-optimization/81703
6563 * tree-ssa-strlen.c (get_string_cst): Rename...
6564 (get_string_len): ...to this. Handle global constants.
6565 (handle_char_store): Adjust.
6567 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6568 Jim Wilson <jimw@sifive.com>
6570 * config/riscv/riscv-protos.h (riscv_output_return): New.
6571 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6572 (riscv_attribute_table, riscv_output_return),
6573 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6574 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6575 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6577 (riscv_expand_prologue): Add early return for naked function.
6578 (riscv_expand_epilogue): Likewise.
6579 (riscv_function_ok_for_sibcall): Return false for naked function.
6580 (riscv_set_current_function): New.
6581 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6582 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6583 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6584 * doc/extend.texi (RISC-V Function Attributes): New.
6586 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6588 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6589 check for 128-bit long double before checking TCmode.
6590 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6591 128-bit long doubles before checking TFmode or TCmode.
6592 (FLOAT128_IBM_P): Likewise.
6594 2018-01-10 Martin Sebor <msebor@redhat.com>
6596 PR tree-optimization/83671
6597 * builtins.c (c_strlen): Unconditionally return zero for the empty
6599 Use -Warray-bounds for warnings.
6600 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6601 for non-constant array indices with COMPONENT_REF, arrays of
6602 arrays, and pointers to arrays.
6603 (gimple_fold_builtin_strlen): Determine and set length range for
6604 non-constant character arrays.
6606 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6609 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6612 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6614 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6616 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6619 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6620 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6621 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6622 indexed_or_indirect_operand predicate.
6623 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6624 (*vsx_le_perm_load_v8hi): Likewise.
6625 (*vsx_le_perm_load_v16qi): Likewise.
6626 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6627 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6628 (*vsx_le_perm_store_v8hi): Likewise.
6629 (*vsx_le_perm_store_v16qi): Likewise.
6630 (eight unnamed splitters): Likewise.
6632 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6634 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6635 * config/rs6000/emmintrin.h: Likewise.
6636 * config/rs6000/mmintrin.h: Likewise.
6637 * config/rs6000/xmmintrin.h: Likewise.
6639 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6642 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6644 * tree.c (tree_nop_conversion): Return true for location wrapper
6646 (maybe_wrap_with_location): New function.
6647 (selftest::check_strip_nops): New function.
6648 (selftest::test_location_wrappers): New function.
6649 (selftest::tree_c_tests): Call it.
6650 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6651 (maybe_wrap_with_location): New decl.
6652 (EXPR_LOCATION_WRAPPER_P): New macro.
6653 (location_wrapper_p): New inline function.
6654 (tree_strip_any_location_wrapper): New inline function.
6656 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6659 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6660 stack_realign_offset for the largest alignment of stack slot
6662 (ix86_find_max_used_stack_alignment): New function.
6663 (ix86_finalize_stack_frame_flags): Use it. Set
6664 max_used_stack_alignment if we don't realign stack.
6665 * config/i386/i386.h (machine_function): Add
6666 max_used_stack_alignment.
6668 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6670 * config/arm/arm.opt (-mbranch-cost): New option.
6671 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6674 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6677 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6678 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6680 2018-01-10 Richard Biener <rguenther@suse.de>
6683 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6684 early out so it also covers the case where we have a non-NULL
6687 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6689 PR tree-optimization/83753
6690 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6691 for non-strided grouped accesses if the number of elements is 1.
6693 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6696 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6697 * i386.h (TARGET_USE_GATHER): Define.
6698 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6700 2018-01-10 Martin Liska <mliska@suse.cz>
6703 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6704 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6706 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6707 CLEANUP_NO_PARTITIONING is not set.
6709 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6711 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6712 for vectors, as a partial revert of r254296.
6713 * rtl.h (const_vec_p): Delete.
6714 (const_vec_duplicate_p): Don't test for vector CONSTs.
6715 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6716 * expmed.c (make_tree): Likewise.
6719 * common.md (E, F): Use CONSTANT_P instead of checking for
6721 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6722 checking for CONST_VECTOR.
6724 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6727 * predict.c (force_edge_cold): Handle in more sane way edges
6730 2018-01-09 Carl Love <cel@us.ibm.com>
6732 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6734 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6735 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6736 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6737 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6738 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6739 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6740 * config/rs6000/rs6000-protos.h: Add extern defition for
6741 rs6000_generate_float2_double_code.
6742 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6744 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6745 (float2_v2df): Add define_expand.
6747 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6750 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6751 op_mode in the force_to_mode call.
6753 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6755 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6756 instead of checking each element individually.
6757 (aarch64_evpc_uzp): Likewise.
6758 (aarch64_evpc_zip): Likewise.
6759 (aarch64_evpc_ext): Likewise.
6760 (aarch64_evpc_rev): Likewise.
6761 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6762 instead of checking each element individually. Return true without
6764 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6765 whether all selected elements come from the same input, instead of
6766 checking each element individually. Remove calls to gen_rtx_REG,
6767 start_sequence and end_sequence and instead assert that no rtl is
6770 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6772 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6773 order of HIGH and CONST checks.
6775 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6777 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6778 if the destination isn't an SSA_NAME.
6780 2018-01-09 Richard Biener <rguenther@suse.de>
6782 PR tree-optimization/83668
6783 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6785 (canonicalize_loop_form): ... here, renamed from ...
6786 (canonicalize_loop_closed_ssa_form): ... this and amended to
6787 swap successor edges for loop exit blocks to make us use
6788 the RPO order we need for initial schedule generation.
6790 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6792 PR tree-optimization/64811
6793 * match.pd: When optimizing comparisons with Inf, avoid
6794 introducing or losing exceptions from comparisons with NaN.
6796 2018-01-09 Martin Liska <mliska@suse.cz>
6799 * asan.c (shadow_mem_size): Add gcc_assert.
6801 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6803 Don't save registers in main().
6806 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6807 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6808 * config/avr/avr.c (avr_set_current_function): Don't error if
6809 naked, OS_task or OS_main are specified at the same time.
6810 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6812 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6814 * common/config/avr/avr-common.c (avr_option_optimization_table):
6815 Switch on -mmain-is-OS_task for optimizing compilations.
6817 2018-01-09 Richard Biener <rguenther@suse.de>
6819 PR tree-optimization/83572
6820 * graphite.c: Include cfganal.h.
6821 (graphite_transform_loops): Connect infinite loops to exit
6822 and remove fake edges at the end.
6824 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6826 * ipa-inline.c (edge_badness): Revert accidental checkin.
6828 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6831 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6832 symbols; not inline clones.
6834 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6837 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6838 hard registers. Formatting fixes.
6840 PR preprocessor/83722
6841 * gcc.c (try_generate_repro): Pass
6842 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6843 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6846 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6847 Kito Cheng <kito.cheng@gmail.com>
6849 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6850 (riscv_leaf_function_p): Delete.
6851 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6853 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6855 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6857 (do_ifelse): New function.
6858 (do_isel): New function.
6859 (do_sub3): New function.
6860 (do_add3): New function.
6861 (do_load_mask_compare): New function.
6862 (do_overlap_load_compare): New function.
6863 (expand_compare_loop): New function.
6864 (expand_block_compare): Call expand_compare_loop() when appropriate.
6865 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6867 (-mblock-compare-inline-loop-limit): New option.
6869 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6872 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6873 Reverse order of second and third operands in first alternative.
6874 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6875 of first and second elements in UNSPEC_VPERMR vector.
6876 (altivec_expand_vec_perm_le): Likewise.
6878 2018-01-08 Jeff Law <law@redhat.com>
6880 PR rtl-optimizatin/81308
6881 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6882 (process_switch): If group_case_labels makes a change, then set
6884 (pass_convert_switch::execute): If a switch is converted, then
6885 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6887 PR rtl-optimization/81308
6888 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6891 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6893 PR target/83663 - Revert r255946
6894 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6895 generation for cases where splatting a value is not useful.
6896 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6897 across a vec_duplicate and a paradoxical subreg forming a vector
6898 mode to a vec_concat.
6900 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6902 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6903 -march=armv8.3-a variants.
6904 * config/arm/t-multilib: Likewise.
6905 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6907 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6909 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6911 (cceq_ior_compare_complement): Give it a name so I can use it, and
6912 change boolean_or_operator predicate to boolean_operator so it can
6913 be used to generate a crand.
6914 (eqne): New code iterator.
6915 (bd/bd_neg): New code_attrs.
6916 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6917 a single define_insn.
6918 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6919 decrement (bdnzt/bdnzf/bdzt/bdzf).
6920 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6921 with the new names of the branch decrement patterns, and added the
6922 names of the branch decrement conditional patterns.
6924 2018-01-08 Richard Biener <rguenther@suse.de>
6926 PR tree-optimization/83563
6927 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6930 2018-01-08 Richard Biener <rguenther@suse.de>
6933 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6935 2018-01-08 Richard Biener <rguenther@suse.de>
6937 PR tree-optimization/83685
6938 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6939 references to abnormals.
6941 2018-01-08 Richard Biener <rguenther@suse.de>
6944 * dwarf2out.c (output_indirect_strings): Handle empty
6945 skeleton_debug_str_hash.
6946 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6948 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6950 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6951 (emit_store_direct): Likewise.
6952 (arc_trampoline_adjust_address): Likewise.
6953 (arc_asm_trampoline_template): New function.
6954 (arc_initialize_trampoline): Use asm_trampoline_template.
6955 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6956 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6957 * config/arc/arc.md (flush_icache): Delete pattern.
6959 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6961 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6962 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6965 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6968 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6969 by not USED_FOR_TARGET.
6970 (make_pass_resolve_sw_modes): Likewise.
6972 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6974 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6977 2018-01-08 Richard Biener <rguenther@suse.de>
6980 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6982 2018-01-08 Richard Biener <rguenther@suse.de>
6985 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6987 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6990 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6991 basic blocks with a small number of successors.
6992 (convert_control_dep_chain_into_preds): Improve handling of
6994 (dump_predicates): Split apart into...
6995 (dump_pred_chain): ...here...
6996 (dump_pred_info): ...and here.
6997 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6998 (can_chain_union_be_invalidated_p): Improve check for invalidation
7000 (uninit_uses_cannot_happen): Avoid unnecessary if
7001 convert_control_dep_chain_into_preds yielded nothing.
7003 2018-01-06 Martin Sebor <msebor@redhat.com>
7005 PR tree-optimization/83640
7006 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7007 subtracting negative offset from size.
7008 (builtin_access::overlap): Adjust offset bounds of the access to fall
7009 within the size of the object if possible.
7011 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7013 PR rtl-optimization/83699
7014 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7015 extract_bit_field_as_subreg to cases in which the extracted
7016 value is also a vector.
7018 * lra-constraints.c (process_alt_operands): Test for the equivalence
7019 substitutions when detecting a possible reload cycle.
7021 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7024 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7025 by default if flag_selective_schedling{,2}. Formatting fixes.
7027 PR rtl-optimization/83682
7028 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7029 if it has non-VECTOR_MODE element mode.
7030 (vec_duplicate_p): Likewise.
7033 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7034 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7036 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7039 * config/i386/i386-builtin.def
7040 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7041 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7042 Require also OPTION_MASK_ISA_AVX512F in addition to
7043 OPTION_MASK_ISA_GFNI.
7044 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7045 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7046 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7047 to OPTION_MASK_ISA_GFNI.
7048 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7049 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7050 OPTION_MASK_ISA_AVX512BW.
7051 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7052 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7053 addition to OPTION_MASK_ISA_GFNI.
7054 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7055 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7056 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7057 to OPTION_MASK_ISA_GFNI.
7058 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7059 a requirement for all ISAs rather than any of them with a few
7061 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7063 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7064 bitmasks to be enabled with 3 exceptions, instead of requiring any
7065 enabled ISA with lots of exceptions.
7066 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7067 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7068 Change avx512bw in isa attribute to avx512f.
7069 * config/i386/sgxintrin.h: Add license boilerplate.
7070 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7071 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7072 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7073 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7075 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7076 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7077 temporarily sse2 rather than sse if not enabled already.
7080 * config/i386/sse.md (VI248_VLBW): Rename to ...
7081 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7082 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7083 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7084 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7085 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7086 mode iterator instead of VI248_VLBW.
7088 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7090 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7091 (record_modified): Skip clobbers; add debug output.
7092 (param_change_prob): Use sreal frequencies.
7094 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7096 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7097 punt for user-aligned variables.
7099 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7101 * tree-chrec.c (chrec_contains_symbols): Return true for
7104 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7107 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7108 of (x|y) == x for BICS pattern.
7110 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7112 PR tree-optimization/83605
7113 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7114 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7117 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7119 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7120 * config/epiphany/rtems.h: New file.
7122 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7123 Uros Bizjak <ubizjak@gmail.com>
7126 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7127 QIreg_operand instead of register_operand predicate.
7128 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7129 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7130 comments instead of -fmitigate[-_]rop.
7132 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7135 * cgraphunit.c (symbol_table::compile): Switch to text_section
7136 before calling assembly_start debug hook.
7137 * run-rtl-passes.c (run_rtl_passes): Likewise.
7140 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7142 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7143 range_int_cst_p rather than !symbolic_range_p before calling
7144 extract_range_from_multiplicative_op_1.
7146 2018-01-04 Jeff Law <law@redhat.com>
7148 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7149 redundant test in assertion.
7151 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7153 * doc/rtl.texi: Document machine_mode wrapper classes.
7155 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7157 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7160 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7162 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7163 the VEC_PERM_EXPR fold to fail.
7165 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7168 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7169 to switched_sections.
7171 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7174 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7177 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7180 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7181 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7183 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7186 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7187 is BLKmode and bitpos not zero or mode change is needed.
7189 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7192 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7195 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7198 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7199 instead of MULT rtx. Update all corresponding splitters.
7201 (*ssub<modesuffix>): Ditto.
7203 (*cmp_sadd_di): Update split patterns.
7204 (*cmp_sadd_si): Ditto.
7205 (*cmp_sadd_sidi): Ditto.
7206 (*cmp_ssub_di): Ditto.
7207 (*cmp_ssub_si): Ditto.
7208 (*cmp_ssub_sidi): Ditto.
7209 * config/alpha/predicates.md (const23_operand): New predicate.
7210 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7211 Look for ASHIFT, not MULT inner operand.
7212 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7214 2018-01-04 Martin Liska <mliska@suse.cz>
7216 PR gcov-profile/83669
7217 * gcov.c (output_intermediate_file): Add version to intermediate
7219 * doc/gcov.texi: Document new field 'version' in intermediate
7220 file format. Fix location of '-k' option of gcov command.
7222 2018-01-04 Martin Liska <mliska@suse.cz>
7225 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7227 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7229 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7231 2018-01-03 Martin Sebor <msebor@redhat.com>
7233 PR tree-optimization/83655
7234 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7235 checking calls with invalid arguments.
7237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7239 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7240 (vectorizable_mask_load_store): Delete.
7241 (vectorizable_call): Return false for masked loads and stores.
7242 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7243 instead of gimple_assign_rhs1.
7244 (vectorizable_load): Handle IFN_MASK_LOAD.
7245 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7249 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7251 (vectorizable_mask_load_store): ...here.
7252 (vectorizable_load): ...and here.
7254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7256 * tree-vect-stmts.c (vect_build_all_ones_mask)
7257 (vect_build_zero_merge_argument): New functions, split out from...
7258 (vectorizable_load): ...here.
7260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7262 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7264 (vectorizable_mask_load_store): ...here.
7265 (vectorizable_store): ...and here.
7267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7269 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7271 (vectorizable_mask_load_store): ...here.
7273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7275 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7276 (vect_model_store_cost): Take a vec_load_store_type instead of a
7278 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7279 (vect_model_store_cost): Take a vec_load_store_type instead of a
7281 (vectorizable_mask_load_store): Update accordingly.
7282 (vectorizable_store): Likewise.
7283 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7287 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7288 IFN_MASK_LOAD calls here rather than...
7289 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7292 Alan Hayward <alan.hayward@arm.com>
7293 David Sherwood <david.sherwood@arm.com>
7295 * expmed.c (extract_bit_field_1): For vector extracts,
7296 fall back to extract_bit_field_as_subreg if vec_extract
7299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7300 Alan Hayward <alan.hayward@arm.com>
7301 David Sherwood <david.sherwood@arm.com>
7303 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7304 they are variable or constant sized.
7305 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7306 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7313 handling COND_EXPRs with boolean comparisons, try to find a better
7314 basis for the mask type than the boolean itself.
7316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7318 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7319 is calculated and how it can be overridden.
7320 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7321 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7323 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7327 Alan Hayward <alan.hayward@arm.com>
7328 David Sherwood <david.sherwood@arm.com>
7330 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7331 Remove the mode argument.
7332 (aarch64_simd_valid_immediate): Remove the mode and inverse
7334 * config/aarch64/iterators.md (bitsize): New iterator.
7335 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7336 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7337 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7338 aarch64_simd_valid_immediate.
7339 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7340 (aarch64_reg_or_bic_imm): Likewise.
7341 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7342 with an insn_type enum and msl with a modifier_type enum.
7343 Replace element_width with a scalar_mode. Change the shift
7344 to unsigned int. Add constructors for scalar_float_mode and
7345 scalar_int_mode elements.
7346 (aarch64_vect_float_const_representable_p): Delete.
7347 (aarch64_can_const_movi_rtx_p)
7348 (aarch64_simd_scalar_immediate_valid_for_move)
7349 (aarch64_simd_make_constant): Update call to
7350 aarch64_simd_valid_immediate.
7351 (aarch64_advsimd_valid_immediate_hs): New function.
7352 (aarch64_advsimd_valid_immediate): Likewise.
7353 (aarch64_simd_valid_immediate): Remove mode and inverse
7354 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7355 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7356 and aarch64_float_const_representable_p on the result.
7357 (aarch64_output_simd_mov_immediate): Remove mode argument.
7358 Update call to aarch64_simd_valid_immediate and use of
7359 simd_immediate_info.
7360 (aarch64_output_scalar_simd_mov_immediate): Update call
7363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7364 Alan Hayward <alan.hayward@arm.com>
7365 David Sherwood <david.sherwood@arm.com>
7367 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7368 (mode_nunits): Likewise CONST_MODE_NUNITS.
7369 * machmode.def (ADJUST_NUNITS): Document.
7370 * genmodes.c (mode_data::need_nunits_adj): New field.
7371 (blank_mode): Update accordingly.
7372 (adj_nunits): New variable.
7373 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7375 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7376 listed in adj_nunits.
7377 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7378 listed in adj_nunits. Don't emit case statements for such modes.
7379 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7380 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7381 nothing if adj_nunits is nonnull.
7382 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7383 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7384 (emit_mode_fbit): Update use of print_maybe_const_decl.
7385 (emit_move_size): Likewise. Treat the array as non-const
7387 (emit_mode_adjustments): Handle adj_nunits.
7389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7391 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7392 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7393 (VECTOR_MODES): Use it.
7394 (make_vector_modes): Take the prefix as an argument.
7396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7397 Alan Hayward <alan.hayward@arm.com>
7398 David Sherwood <david.sherwood@arm.com>
7400 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7401 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7402 for MODE_VECTOR_BOOL.
7403 * machmode.def (VECTOR_BOOL_MODE): Document.
7404 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7405 (make_vector_bool_mode): New function.
7406 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7408 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7409 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7411 * stor-layout.c (int_mode_for_mode): Likewise.
7412 * tree.c (build_vector_type_for_mode): Likewise.
7413 * varasm.c (output_constant_pool_2): Likewise.
7414 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7415 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7416 for MODE_VECTOR_BOOL.
7417 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7418 of mode class checks.
7419 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7420 instead of a list of mode class checks.
7421 (expand_vector_scalar_condition): Likewise.
7422 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7425 Alan Hayward <alan.hayward@arm.com>
7426 David Sherwood <david.sherwood@arm.com>
7428 * machmode.h (mode_size): Change from unsigned short to
7430 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7431 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7432 or if measurement_type is not polynomial.
7433 (fixed_size_mode::includes_p): Check for constant-sized modes.
7434 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7435 return a poly_uint16 rather than an unsigned short.
7436 (emit_mode_size): Change the type of mode_size from unsigned short
7437 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7438 (emit_mode_adjustments): Cope with polynomial vector sizes.
7439 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7441 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7443 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7444 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7445 * caller-save.c (setup_save_areas): Likewise.
7446 (replace_reg_with_saved_mem): Likewise.
7447 * calls.c (emit_library_call_value_1): Likewise.
7448 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7449 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7450 (gen_lowpart_for_combine): Likewise.
7451 * convert.c (convert_to_integer_1): Likewise.
7452 * cse.c (equiv_constant, cse_insn): Likewise.
7453 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7454 (cselib_subst_to_values): Likewise.
7455 * dce.c (word_dce_process_block): Likewise.
7456 * df-problems.c (df_word_lr_mark_ref): Likewise.
7457 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7458 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7459 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7460 (rtl_for_decl_location): Likewise.
7461 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7462 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7463 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7464 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7465 (expand_expr_real_1): Likewise.
7466 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7467 (pad_below): Likewise.
7468 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7469 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7470 * ira.c (get_subreg_tracking_sizes): Likewise.
7471 * ira-build.c (ira_create_allocno_objects): Likewise.
7472 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7473 (ira_sort_regnos_for_alter_reg): Likewise.
7474 * ira-costs.c (record_operand_costs): Likewise.
7475 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7476 (resolve_simple_move): Likewise.
7477 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7478 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7479 (lra_constraints): Likewise.
7480 (CONST_POOL_OK_P): Reject variable-sized modes.
7481 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7482 (add_pseudo_to_slot, lra_spill): Likewise.
7483 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7484 * optabs-query.c (get_best_extraction_insn): Likewise.
7485 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7486 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7487 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7488 * recog.c (offsettable_address_addr_space_p): Likewise.
7489 * regcprop.c (maybe_mode_change): Likewise.
7490 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7491 * regrename.c (build_def_use): Likewise.
7492 * regstat.c (dump_reg_info): Likewise.
7493 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7494 (find_reloads, find_reloads_subreg_address): Likewise.
7495 * reload1.c (eliminate_regs_1): Likewise.
7496 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7497 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7498 (simplify_binary_operation_1, simplify_subreg): Likewise.
7499 * targhooks.c (default_function_arg_padding): Likewise.
7500 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7501 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7502 (verify_gimple_assign_ternary): Likewise.
7503 * tree-inline.c (estimate_move_cost): Likewise.
7504 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7505 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7506 (get_address_cost_ainc): Likewise.
7507 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7508 (vect_supportable_dr_alignment): Likewise.
7509 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7510 (vectorizable_reduction): Likewise.
7511 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7512 (vectorizable_operation, vectorizable_load): Likewise.
7513 * tree.c (build_same_sized_truth_vector_type): Likewise.
7514 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7515 * var-tracking.c (emit_note_insn_var_location): Likewise.
7516 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7517 (ADDR_VEC_ALIGN): Likewise.
7519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7520 Alan Hayward <alan.hayward@arm.com>
7521 David Sherwood <david.sherwood@arm.com>
7523 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7525 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7526 or if measurement_type is polynomial.
7527 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7528 * combine.c (make_extraction): Likewise.
7529 * dse.c (find_shift_sequence): Likewise.
7530 * dwarf2out.c (mem_loc_descriptor): Likewise.
7531 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7532 (extract_bit_field, extract_low_bits): Likewise.
7533 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7534 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7535 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7536 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7537 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7538 * reload.c (find_reloads): Likewise.
7539 * reload1.c (alter_reg): Likewise.
7540 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7541 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7542 * tree-if-conv.c (predicate_mem_writes): Likewise.
7543 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7544 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7545 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7546 * valtrack.c (dead_debug_insert_temp): Likewise.
7547 * varasm.c (mergeable_constant_section): Likewise.
7548 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7551 Alan Hayward <alan.hayward@arm.com>
7552 David Sherwood <david.sherwood@arm.com>
7554 * expr.c (expand_assignment): Cope with polynomial mode sizes
7555 when assigning to a CONCAT.
7557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7558 Alan Hayward <alan.hayward@arm.com>
7559 David Sherwood <david.sherwood@arm.com>
7561 * machmode.h (mode_precision): Change from unsigned short to
7563 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7565 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7566 or if measurement_type is not polynomial.
7567 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7568 in which the mode is already known to be a scalar_int_mode.
7569 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7570 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7572 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7573 for GET_MODE_PRECISION.
7574 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7575 for GET_MODE_PRECISION.
7576 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7578 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7579 (expand_field_assignment, make_extraction): Likewise.
7580 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7581 (get_last_value): Likewise.
7582 * convert.c (convert_to_integer_1): Likewise.
7583 * cse.c (cse_insn): Likewise.
7584 * expr.c (expand_expr_real_1): Likewise.
7585 * lra-constraints.c (simplify_operand_subreg): Likewise.
7586 * optabs-query.c (can_atomic_load_p): Likewise.
7587 * optabs.c (expand_atomic_load): Likewise.
7588 (expand_atomic_store): Likewise.
7589 * ree.c (combine_reaching_defs): Likewise.
7590 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7591 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7592 * tree.h (type_has_mode_precision_p): Likewise.
7593 * ubsan.c (instrument_si_overflow): Likewise.
7595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7596 Alan Hayward <alan.hayward@arm.com>
7597 David Sherwood <david.sherwood@arm.com>
7599 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7600 polynomial numbers of units.
7601 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7602 (valid_vector_subparts_p): New function.
7603 (build_vector_type): Remove temporary shim and take the number
7604 of units as a poly_uint64 rather than an int.
7605 (build_opaque_vector_type): Take the number of units as a
7606 poly_uint64 rather than an int.
7607 * tree.c (build_vector_from_ctor): Handle polynomial
7608 TYPE_VECTOR_SUBPARTS.
7609 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7610 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7611 (build_vector_from_val): If the number of units is variable,
7612 use build_vec_duplicate_cst for constant operands and
7613 VEC_DUPLICATE_EXPR otherwise.
7614 (make_vector_type): Remove temporary is_constant ().
7615 (build_vector_type, build_opaque_vector_type): Take the number of
7616 units as a poly_uint64 rather than an int.
7617 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7619 * cfgexpand.c (expand_debug_expr): Likewise.
7620 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7621 (store_constructor, expand_expr_real_1): Likewise.
7622 (const_scalar_mask_from_tree): Likewise.
7623 * fold-const-call.c (fold_const_reduction): Likewise.
7624 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7625 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7626 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7627 (fold_relational_const): Likewise.
7628 (native_interpret_vector): Likewise. Change the size from an
7629 int to an unsigned int.
7630 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7631 TYPE_VECTOR_SUBPARTS.
7632 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7633 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7634 duplicating a non-constant operand into a variable-length vector.
7635 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7636 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7637 * ipa-icf.c (sem_variable::equals): Likewise.
7638 * match.pd: Likewise.
7639 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7640 * print-tree.c (print_node): Likewise.
7641 * stor-layout.c (layout_type): Likewise.
7642 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7643 * tree-cfg.c (verify_gimple_comparison): Likewise.
7644 (verify_gimple_assign_binary): Likewise.
7645 (verify_gimple_assign_ternary): Likewise.
7646 (verify_gimple_assign_single): Likewise.
7647 * tree-pretty-print.c (dump_generic_node): Likewise.
7648 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7649 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7650 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7651 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7652 (vect_shift_permute_load_chain): Likewise.
7653 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7654 (expand_vector_condition, optimize_vector_constructor): Likewise.
7655 (lower_vec_perm, get_compute_type): Likewise.
7656 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7657 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7658 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7659 (vect_recog_mask_conversion_pattern): Likewise.
7660 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7661 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7662 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7663 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7664 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7665 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7666 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7667 (supportable_widening_operation): Likewise.
7668 (supportable_narrowing_operation): Likewise.
7669 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7671 * varasm.c (output_constant): Likewise.
7673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7674 Alan Hayward <alan.hayward@arm.com>
7675 David Sherwood <david.sherwood@arm.com>
7677 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7678 so that both the length == 3 and length != 3 cases set up their
7679 own permute vectors. Add comments explaining why we know the
7680 number of elements is constant.
7681 (vect_permute_load_chain): Likewise.
7683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7684 Alan Hayward <alan.hayward@arm.com>
7685 David Sherwood <david.sherwood@arm.com>
7687 * machmode.h (mode_nunits): Change from unsigned char to
7689 (ONLY_FIXED_SIZE_MODES): New macro.
7690 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7691 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7692 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7694 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7695 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7696 or if measurement_type is not polynomial.
7697 * genmodes.c (ZERO_COEFFS): New macro.
7698 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7700 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7701 Use ZERO_COEFFS when emitting initializers.
7702 * data-streamer.h (bp_pack_poly_value): New function.
7703 (bp_unpack_poly_value): Likewise.
7704 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7705 for GET_MODE_NUNITS.
7706 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7707 for GET_MODE_NUNITS.
7708 * tree.c (make_vector_type): Remove temporary shim and make
7709 the real function take the number of units as a poly_uint64
7711 (build_vector_type_for_mode): Handle polynomial nunits.
7712 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7713 * emit-rtl.c (const_vec_series_p_1): Likewise.
7714 (gen_rtx_CONST_VECTOR): Likewise.
7715 * fold-const.c (test_vec_duplicate_folding): Likewise.
7716 * genrecog.c (validate_pattern): Likewise.
7717 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7718 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7719 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7720 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7721 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7722 * rtlanal.c (subreg_get_info): Likewise.
7723 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7724 (vect_grouped_load_supported): Likewise.
7725 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7726 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7727 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7728 (simplify_const_unary_operation, simplify_binary_operation_1)
7729 (simplify_const_binary_operation, simplify_ternary_operation)
7730 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7731 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7732 instead of CONST_VECTOR_NUNITS.
7733 * varasm.c (output_constant_pool_2): Likewise.
7734 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7735 explicit-encoded elements in the XVEC for variable-length vectors.
7737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7739 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7742 Alan Hayward <alan.hayward@arm.com>
7743 David Sherwood <david.sherwood@arm.com>
7745 * coretypes.h (fixed_size_mode): Declare.
7746 (fixed_size_mode_pod): New typedef.
7747 * builtins.h (target_builtins::x_apply_args_mode)
7748 (target_builtins::x_apply_result_mode): Change type to
7749 fixed_size_mode_pod.
7750 * builtins.c (apply_args_size, apply_result_size, result_vector)
7751 (expand_builtin_apply_args_1, expand_builtin_apply)
7752 (expand_builtin_return): Update accordingly.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7756 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7757 * cselib.c (cselib_hash_rtx): Likewise.
7758 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7759 CONST_VECTOR encoding.
7761 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7762 Jeff Law <law@redhat.com>
7765 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7766 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7767 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7768 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7771 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7772 explicitly probe *sp in a noreturn function if there were any callee
7773 register saves or frame pointer is needed.
7775 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7778 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7779 BLKmode for ternary, binary or unary expressions.
7782 * var-tracking.c (delete_vta_debug_insn): New inline function.
7783 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7784 insns from get_insns () to NULL instead of each bb separately.
7785 Use delete_vta_debug_insn. No longer static.
7786 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7787 delete_vta_debug_insns callers.
7788 * rtl.h (delete_vta_debug_insns): Declare.
7789 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7790 instead of variable_tracking_main.
7792 2018-01-03 Martin Sebor <msebor@redhat.com>
7794 PR tree-optimization/83603
7795 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7796 arguments past the endof the argument list in functions declared
7797 without a prototype.
7798 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7799 Avoid checking when arguments are null.
7801 2018-01-03 Martin Sebor <msebor@redhat.com>
7804 * doc/extend.texi (attribute const): Fix a typo.
7805 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7806 issuing -Wsuggest-attribute for void functions.
7808 2018-01-03 Martin Sebor <msebor@redhat.com>
7810 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7811 offset_int::from instead of wide_int::to_shwi.
7812 (maybe_diag_overlap): Remove assertion.
7813 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7814 * gimple-ssa-sprintf.c (format_directive): Same.
7815 (parse_directive): Same.
7816 (sprintf_dom_walker::compute_format_length): Same.
7817 (try_substitute_return_value): Same.
7819 2018-01-03 Jeff Law <law@redhat.com>
7822 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7823 non-constant residual for zero at runtime and avoid probing in
7824 that case. Reorganize code for trailing problem to mirror handling
7827 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7829 PR tree-optimization/83501
7830 * tree-ssa-strlen.c (get_string_cst): New.
7831 (handle_char_store): Call get_string_cst.
7833 2018-01-03 Martin Liska <mliska@suse.cz>
7835 PR tree-optimization/83593
7836 * tree-ssa-strlen.c: Include tree-cfg.h.
7837 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7838 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7839 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7841 (strlen_dom_walker::before_dom_children): Call
7842 gimple_purge_dead_eh_edges. Dump tranformation with details
7844 (strlen_dom_walker::before_dom_children): Update call by adding
7845 new argument cleanup_eh.
7846 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7848 2018-01-03 Martin Liska <mliska@suse.cz>
7851 * cif-code.def (VARIADIC_THUNK): New enum value.
7852 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7855 2018-01-03 Jan Beulich <jbeulich@suse.com>
7857 * sse.md (mov<mode>_internal): Tighten condition for when to use
7858 vmovdqu<ssescalarsize> for TI and OI modes.
7860 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7862 Update copyright years.
7864 2018-01-03 Martin Liska <mliska@suse.cz>
7867 * ipa-visibility.c (function_and_variable_visibility): Skip
7868 functions with noipa attribure.
7870 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7872 * gcc.c (process_command): Update copyright notice dates.
7873 * gcov-dump.c (print_version): Ditto.
7874 * gcov.c (print_version): Ditto.
7875 * gcov-tool.c (print_version): Ditto.
7876 * gengtype.c (create_file): Ditto.
7877 * doc/cpp.texi: Bump @copying's copyright year.
7878 * doc/cppinternals.texi: Ditto.
7879 * doc/gcc.texi: Ditto.
7880 * doc/gccint.texi: Ditto.
7881 * doc/gcov.texi: Ditto.
7882 * doc/install.texi: Ditto.
7883 * doc/invoke.texi: Ditto.
7885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7887 * vector-builder.h (vector_builder::m_full_nelts): Change from
7888 unsigned int to poly_uint64.
7889 (vector_builder::full_nelts): Update prototype accordingly.
7890 (vector_builder::new_vector): Likewise.
7891 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7892 (vector_builder::operator ==): Likewise.
7893 (vector_builder::finalize): Likewise.
7894 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7895 Take the number of elements as a poly_uint64 rather than an
7897 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7898 from unsigned int to poly_uint64.
7899 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7900 (vec_perm_indices::new_vector): Likewise.
7901 (vec_perm_indices::length): Likewise.
7902 (vec_perm_indices::nelts_per_input): Likewise.
7903 (vec_perm_indices::input_nelts): Likewise.
7904 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7905 number of elements per input as a poly_uint64 rather than an
7906 unsigned int. Use the original encoding for variable-length
7907 vectors, rather than clamping each individual element.
7908 For the second and subsequent elements in each pattern,
7909 clamp the step and base before clamping their sum.
7910 (vec_perm_indices::series_p): Handle polynomial element counts.
7911 (vec_perm_indices::all_in_range_p): Likewise.
7912 (vec_perm_indices_to_tree): Likewise.
7913 (vec_perm_indices_to_rtx): Likewise.
7914 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7915 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7916 (tree_vector_builder::new_binary_operation): Handle polynomial
7917 element counts. Return false if we need to know the number
7918 of elements at compile time.
7919 * fold-const.c (fold_vec_perm): Punt if the number of elements
7920 isn't known at compile time.
7922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7924 * vec-perm-indices.h (vec_perm_builder): Change element type
7925 from HOST_WIDE_INT to poly_int64.
7926 (vec_perm_indices::element_type): Update accordingly.
7927 (vec_perm_indices::clamp): Handle polynomial element_types.
7928 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7929 (vec_perm_indices::all_in_range_p): Likewise.
7930 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7932 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7933 polynomial vec_perm_indices element types.
7934 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7935 * fold-const.c (fold_vec_perm): Likewise.
7936 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7937 * tree-vect-generic.c (lower_vec_perm): Likewise.
7938 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7939 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7940 element type to HOST_WIDE_INT.
7942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7943 Alan Hayward <alan.hayward@arm.com>
7944 David Sherwood <david.sherwood@arm.com>
7946 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7947 rather than an int. Use plus_constant.
7948 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7949 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7952 Alan Hayward <alan.hayward@arm.com>
7953 David Sherwood <david.sherwood@arm.com>
7955 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7956 a HOST_WIDE_INT to a poly_int64.
7958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7959 Alan Hayward <alan.hayward@arm.com>
7960 David Sherwood <david.sherwood@arm.com>
7962 * calls.c (load_register_parameters): Cope with polynomial
7963 mode sizes. Require a constant size for BLKmode parameters
7964 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7965 forces a parameter to be padded at the lsb end in order to
7966 fill a complete number of words, require the parameter size
7967 to be ordered wrt UNITS_PER_WORD.
7969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7970 Alan Hayward <alan.hayward@arm.com>
7971 David Sherwood <david.sherwood@arm.com>
7973 * reload1.c (spill_stack_slot_width): Change element type
7974 from unsigned int to poly_uint64_pod.
7975 (alter_reg): Treat mode sizes as polynomial.
7977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7978 Alan Hayward <alan.hayward@arm.com>
7979 David Sherwood <david.sherwood@arm.com>
7981 * reload.c (complex_word_subreg_p): New function.
7982 (reload_inner_reg_of_subreg, push_reload): Use it.
7984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7985 Alan Hayward <alan.hayward@arm.com>
7986 David Sherwood <david.sherwood@arm.com>
7988 * lra-constraints.c (process_alt_operands): Reject matched
7989 operands whose sizes aren't ordered.
7990 (match_reload): Refer to this check here.
7992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7993 Alan Hayward <alan.hayward@arm.com>
7994 David Sherwood <david.sherwood@arm.com>
7996 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7997 that the mode size is in the set {1, 2, 4, 8, 16}.
7999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8000 Alan Hayward <alan.hayward@arm.com>
8001 David Sherwood <david.sherwood@arm.com>
8003 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8004 Use plus_constant instead of gen_rtx_PLUS.
8006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8007 Alan Hayward <alan.hayward@arm.com>
8008 David Sherwood <david.sherwood@arm.com>
8010 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8011 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8012 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8013 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8014 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8015 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8016 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8017 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8018 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8019 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8021 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8022 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8023 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8024 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8025 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8026 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8027 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8028 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8029 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8030 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8032 * expr.c (emit_move_resolve_push): Treat the input and result
8033 of PUSH_ROUNDING as a poly_int64.
8034 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8035 (emit_push_insn): Likewise.
8036 * lra-eliminations.c (mark_not_eliminable): Likewise.
8037 * recog.c (push_operand): Likewise.
8038 * reload1.c (elimination_effects): Likewise.
8039 * rtlanal.c (nonzero_bits1): Likewise.
8040 * calls.c (store_one_arg): Likewise. Require the padding to be
8041 known at compile time.
8043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8044 Alan Hayward <alan.hayward@arm.com>
8045 David Sherwood <david.sherwood@arm.com>
8047 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8048 Use plus_constant instead of gen_rtx_PLUS.
8050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8051 Alan Hayward <alan.hayward@arm.com>
8052 David Sherwood <david.sherwood@arm.com>
8054 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8058 Alan Hayward <alan.hayward@arm.com>
8059 David Sherwood <david.sherwood@arm.com>
8061 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8062 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8063 via stack temporaries. Treat the mode size as polynomial too.
8065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8066 Alan Hayward <alan.hayward@arm.com>
8067 David Sherwood <david.sherwood@arm.com>
8069 * expr.c (expand_expr_real_2): When handling conversions involving
8070 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8071 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8072 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8081 Alan Hayward <alan.hayward@arm.com>
8082 David Sherwood <david.sherwood@arm.com>
8084 * combine.c (can_change_dest_mode): Handle polynomial
8085 REGMODE_NATURAL_SIZE.
8086 * expmed.c (store_bit_field_1): Likewise.
8087 * expr.c (store_constructor): Likewise.
8088 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8089 and polynomial REGMODE_NATURAL_SIZE.
8090 (gen_lowpart_common): Likewise.
8091 * reginfo.c (record_subregs_of_mode): Likewise.
8092 * rtlanal.c (read_modify_subreg_p): Likewise.
8094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8095 Alan Hayward <alan.hayward@arm.com>
8096 David Sherwood <david.sherwood@arm.com>
8098 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8099 numbers of elements.
8101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8102 Alan Hayward <alan.hayward@arm.com>
8103 David Sherwood <david.sherwood@arm.com>
8105 * match.pd: Cope with polynomial numbers of vector elements.
8107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8108 Alan Hayward <alan.hayward@arm.com>
8109 David Sherwood <david.sherwood@arm.com>
8111 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8112 in a POINTER_PLUS_EXPR.
8114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8115 Alan Hayward <alan.hayward@arm.com>
8116 David Sherwood <david.sherwood@arm.com>
8118 * omp-simd-clone.c (simd_clone_subparts): New function.
8119 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8120 (ipa_simd_modify_function_body): Likewise.
8122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8123 Alan Hayward <alan.hayward@arm.com>
8124 David Sherwood <david.sherwood@arm.com>
8126 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8127 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8128 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8129 (expand_vector_condition, vector_element): Likewise.
8130 (subparts_gt): New function.
8131 (get_compute_type): Use subparts_gt.
8132 (count_type_subparts): Delete.
8133 (expand_vector_operations_1): Use subparts_gt instead of
8134 count_type_subparts.
8136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8137 Alan Hayward <alan.hayward@arm.com>
8138 David Sherwood <david.sherwood@arm.com>
8140 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8141 (vect_compile_time_alias): ...this new function. Do the calculation
8142 on poly_ints rather than trees.
8143 (vect_prune_runtime_alias_test_list): Update call accordingly.
8145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8146 Alan Hayward <alan.hayward@arm.com>
8147 David Sherwood <david.sherwood@arm.com>
8149 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8151 (vect_schedule_slp_instance): Likewise.
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8154 Alan Hayward <alan.hayward@arm.com>
8155 David Sherwood <david.sherwood@arm.com>
8157 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8158 constant and extern definitions for variable-length vectors.
8159 (vect_get_constant_vectors): Note that the number of units
8160 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
8167 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
8175 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8178 Alan Hayward <alan.hayward@arm.com>
8179 David Sherwood <david.sherwood@arm.com>
8181 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8182 vectors as polynomial. Use build_index_vector for
8185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8186 Alan Hayward <alan.hayward@arm.com>
8187 David Sherwood <david.sherwood@arm.com>
8189 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8190 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8191 for variable-length vectors.
8192 (vectorizable_mask_load_store): Treat the number of units as
8193 polynomial, asserting that it is constant if the condition has
8194 already been enforced.
8195 (vectorizable_store, vectorizable_load): Likewise.
8197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8198 Alan Hayward <alan.hayward@arm.com>
8199 David Sherwood <david.sherwood@arm.com>
8201 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8202 of units as polynomial. Punt if we can't tell at compile time
8203 which vector contains the final result.
8205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8206 Alan Hayward <alan.hayward@arm.com>
8207 David Sherwood <david.sherwood@arm.com>
8209 * tree-vect-loop.c (vectorizable_induction): Treat the number
8210 of units as polynomial. Punt on SLP inductions. Use an integer
8211 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8212 cast of such a series for variable-length floating-point
8215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8216 Alan Hayward <alan.hayward@arm.com>
8217 David Sherwood <david.sherwood@arm.com>
8219 * tree.h (build_index_vector): Declare.
8220 * tree.c (build_index_vector): New function.
8221 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8222 of units as polynomial, forcibly converting it to a constant if
8223 vectorizable_reduction has already enforced the condition.
8224 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8225 to create a {1,2,3,...} vector.
8226 (vectorizable_reduction): Treat the number of units as polynomial.
8227 Choose vectype_in based on the largest scalar element size rather
8228 than the smallest number of units. Enforce the restrictions
8231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8232 Alan Hayward <alan.hayward@arm.com>
8233 David Sherwood <david.sherwood@arm.com>
8235 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8236 number of units as polynomial.
8238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8239 Alan Hayward <alan.hayward@arm.com>
8240 David Sherwood <david.sherwood@arm.com>
8242 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8243 * target.def (autovectorize_vector_sizes): Return the vector sizes
8244 by pointer, using vector_sizes rather than a bitmask.
8245 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8246 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8247 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8249 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8250 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8251 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8252 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8253 * omp-general.c (omp_max_vf): Likewise.
8254 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8255 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8256 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8257 * tree-vect-slp.c (vect_slp_bb): Likewise.
8258 * doc/tm.texi: Regenerate.
8259 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8261 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8262 the vector size as a poly_uint64 rather than an unsigned int.
8263 (current_vector_size): Change from an unsigned int to a poly_uint64.
8264 (get_vectype_for_scalar_type): Update accordingly.
8265 * tree.h (build_truth_vector_type): Take the size and number of
8266 units as a poly_uint64 rather than an unsigned int.
8267 (build_vector_type): Add a temporary overload that takes
8268 the number of units as a poly_uint64 rather than an unsigned int.
8269 * tree.c (make_vector_type): Likewise.
8270 (build_truth_vector_type): Take the number of units as a poly_uint64
8271 rather than an unsigned int.
8273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8274 Alan Hayward <alan.hayward@arm.com>
8275 David Sherwood <david.sherwood@arm.com>
8277 * target.def (get_mask_mode): Take the number of units and length
8278 as poly_uint64s rather than unsigned ints.
8279 * targhooks.h (default_get_mask_mode): Update accordingly.
8280 * targhooks.c (default_get_mask_mode): Likewise.
8281 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8282 * doc/tm.texi: Regenerate.
8284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8285 Alan Hayward <alan.hayward@arm.com>
8286 David Sherwood <david.sherwood@arm.com>
8288 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8289 * omp-general.c (omp_max_vf): Likewise.
8290 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8291 (expand_omp_simd): Handle polynomial safelen.
8292 * omp-low.c (omplow_simd_context): Add a default constructor.
8293 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8294 (lower_rec_simd_input_clauses): Update accordingly.
8295 (lower_rec_input_clauses): Likewise.
8297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8298 Alan Hayward <alan.hayward@arm.com>
8299 David Sherwood <david.sherwood@arm.com>
8301 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8302 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8303 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8304 (vect_analyze_slp_cost): Likewise.
8305 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8306 (vect_model_load_cost): Likewise.
8308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8309 Alan Hayward <alan.hayward@arm.com>
8310 David Sherwood <david.sherwood@arm.com>
8312 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8313 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8314 from an unsigned int * to a poly_uint64_pod *.
8315 (calculate_unrolling_factor): New function.
8316 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8319 Alan Hayward <alan.hayward@arm.com>
8320 David Sherwood <david.sherwood@arm.com>
8322 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8323 from an unsigned int to a poly_uint64.
8324 (_loop_vec_info::slp_unrolling_factor): Likewise.
8325 (_loop_vec_info::vectorization_factor): Change from an int
8327 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8328 (vect_get_num_vectors): New function.
8329 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8330 (vect_get_num_copies): Use vect_get_num_vectors.
8331 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8332 to an unsigned int *.
8333 (vect_analyze_data_refs): Change min_vf from an int * to a
8335 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8336 than an unsigned HOST_WIDE_INT.
8337 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8338 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8339 to an unsigned int *.
8340 (vect_analyze_data_ref_dependences): Likewise.
8341 (vect_compute_data_ref_alignment): Handle polynomial vf.
8342 (vect_enhance_data_refs_alignment): Likewise.
8343 (vect_prune_runtime_alias_test_list): Likewise.
8344 (vect_shift_permute_load_chain): Likewise.
8345 (vect_supportable_dr_alignment): Likewise.
8346 (dependence_distance_ge_vf): Take the vectorization factor as a
8347 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8348 (vect_analyze_data_refs): Change min_vf from an int * to a
8350 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8351 vfm1 as a poly_uint64 rather than an int. Make the same change
8352 for the returned bound_scalar.
8353 (vect_gen_vector_loop_niters): Handle polynomial vf.
8354 (vect_do_peeling): Likewise. Update call to
8355 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8356 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8358 * tree-vect-loop.c (vect_determine_vectorization_factor)
8359 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8360 (vect_get_known_peeling_cost): Likewise.
8361 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8362 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8363 (vect_transform_loop): Likewise. Use the lowest possible VF when
8364 updating the upper bounds of the loop.
8365 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8367 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8368 polynomial unroll factors.
8369 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8370 (vect_make_slp_decision): Likewise.
8371 (vect_supported_load_permutation_p): Likewise, and polynomial
8373 (vect_analyze_slp_cost): Handle polynomial vf.
8374 (vect_slp_analyze_node_operations): Likewise.
8375 (vect_slp_analyze_bb_1): Likewise.
8376 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8377 than an unsigned HOST_WIDE_INT.
8378 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8379 (vectorizable_load): Handle polynomial vf.
8380 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8382 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8385 Alan Hayward <alan.hayward@arm.com>
8386 David Sherwood <david.sherwood@arm.com>
8388 * match.pd: Handle bit operations involving three constants
8389 and try to fold one pair.
8391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8393 * tree-vect-loop-manip.c: Include gimple-fold.h.
8394 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8395 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8396 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8397 Add a path that uses a step of VF instead of 1, but disable it
8399 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8400 and niters_no_overflow parameters. Update calls to
8401 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8402 Create a new SSA name if the latter choses to use a ste other
8403 than zero, and return it via niters_vector_mult_vf_var.
8404 * tree-vect-loop.c (vect_transform_loop): Update calls to
8405 vect_do_peeling, vect_gen_vector_loop_niters and
8406 slpeel_make_loop_iterate_ntimes.
8407 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8408 (vect_gen_vector_loop_niters): Update declarations after above changes.
8410 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8412 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8413 128-bit round to integer instructions.
8414 (ceil<mode>2): Likewise.
8415 (btrunc<mode>2): Likewise.
8416 (round<mode>2): Likewise.
8418 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8420 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8421 unaligned VSX load/store on P8/P9.
8422 (expand_block_clear): Allow the use of unaligned VSX
8423 load/store on P8/P9.
8425 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8427 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8429 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8430 swap associated with both a load and a store.
8432 2018-01-02 Andrew Waterman <andrew@sifive.com>
8434 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8435 * config/riscv/riscv.md (clear_cache): Use it.
8437 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8439 * web.c: Remove out-of-date comment.
8441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8443 * expr.c (fixup_args_size_notes): Check that any existing
8444 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8445 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8446 (emit_single_push_insn): ...here.
8448 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8450 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8451 (const_vector_encoded_nelts): New function.
8452 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8453 (const_vector_int_elt, const_vector_elt): Declare.
8454 * emit-rtl.c (const_vector_int_elt_1): New function.
8455 (const_vector_elt): Likewise.
8456 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8457 of CONST_VECTOR_ELT.
8459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8461 * expr.c: Include rtx-vector-builder.h.
8462 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8463 directly on the tree encoding.
8464 (const_vector_from_tree): Likewise.
8465 * optabs.c: Include rtx-vector-builder.h.
8466 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8467 sequence of "u" values.
8468 * vec-perm-indices.c: Include rtx-vector-builder.h.
8469 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8470 directly on the vec_perm_indices encoding.
8472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8474 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8475 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8476 * rtx-vector-builder.h: New file.
8477 * rtx-vector-builder.c: Likewise.
8478 * rtl.h (rtx_def::u2): Add a const_vector field.
8479 (CONST_VECTOR_NPATTERNS): New macro.
8480 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8481 (CONST_VECTOR_DUPLICATE_P): Likewise.
8482 (CONST_VECTOR_STEPPED_P): Likewise.
8483 (CONST_VECTOR_ENCODED_ELT): Likewise.
8484 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8485 (unwrap_const_vec_duplicate): Likewise.
8486 (const_vec_series_p): Check for a non-duplicated vector encoding.
8487 Say that the function only returns true for integer vectors.
8488 * emit-rtl.c: Include rtx-vector-builder.h.
8489 (gen_const_vec_duplicate_1): Delete.
8490 (gen_const_vector): Call gen_const_vec_duplicate instead of
8491 gen_const_vec_duplicate_1.
8492 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8493 (gen_const_vec_duplicate): Use rtx_vector_builder.
8494 (gen_const_vec_series): Likewise.
8495 (gen_rtx_CONST_VECTOR): Likewise.
8496 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8497 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8498 Build a new vector rather than modifying a CONST_VECTOR in-place.
8499 (handle_special_swappables): Update call accordingly.
8500 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8501 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8502 Build a new vector rather than modifying a CONST_VECTOR in-place.
8503 (handle_special_swappables): Update call accordingly.
8505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8507 * simplify-rtx.c (simplify_const_binary_operation): Use
8508 CONST_VECTOR_ELT instead of XVECEXP.
8510 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8512 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8513 the selector elements to be different from the data elements
8514 if the selector is a VECTOR_CST.
8515 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8516 ssizetype for the selector.
8518 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8520 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8521 before testing each element individually.
8522 * tree-vect-generic.c (lower_vec_perm): Likewise.
8524 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8526 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8527 * selftest-run-tests.c (selftest::run_tests): Call it.
8528 * vector-builder.h (vector_builder::operator ==): New function.
8529 (vector_builder::operator !=): Likewise.
8530 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8531 (vec_perm_indices::all_from_input_p): New function.
8532 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8533 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8534 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8535 instead of reading the VECTOR_CST directly. Detect whether both
8536 vector inputs are the same before constructing the vec_perm_indices,
8537 and update the number of inputs argument accordingly. Use the
8538 utility functions added above. Only construct sel2 if we need to.
8540 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8542 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8543 the broadcast of the low byte.
8544 (expand_mult_highpart): Use an explicit encoding for the permutes.
8545 * optabs-query.c (can_mult_highpart_p): Likewise.
8546 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8547 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8548 (vectorizable_bswap): Likewise.
8549 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8550 explicit encoding for the power-of-2 permutes.
8551 (vect_permute_store_chain): Likewise.
8552 (vect_grouped_load_supported): Likewise.
8553 (vect_permute_load_chain): Likewise.
8555 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8557 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8558 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8559 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8560 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8561 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8562 (vect_gen_perm_mask_any): Likewise.
8564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8566 * int-vector-builder.h: New file.
8567 * vec-perm-indices.h: Include int-vector-builder.h.
8568 (vec_perm_indices): Redefine as an int_vector_builder.
8569 (auto_vec_perm_indices): Delete.
8570 (vec_perm_builder): Redefine as a stand-alone class.
8571 (vec_perm_indices::vec_perm_indices): New function.
8572 (vec_perm_indices::clamp): Likewise.
8573 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8574 (vec_perm_indices::new_vector): New function.
8575 (vec_perm_indices::new_expanded_vector): Update for new
8576 vec_perm_indices class.
8577 (vec_perm_indices::rotate_inputs): New function.
8578 (vec_perm_indices::all_in_range_p): Operate directly on the
8579 encoded form, without computing elided elements.
8580 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8581 encoding. Update for new vec_perm_indices class.
8582 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8583 the given vec_perm_builder.
8584 (expand_vec_perm_var): Update vec_perm_builder constructor.
8585 (expand_mult_highpart): Use vec_perm_builder instead of
8586 auto_vec_perm_indices.
8587 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8588 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8589 or double series encoding as appropriate.
8590 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8591 vec_perm_indices instead of auto_vec_perm_indices.
8592 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8593 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8594 (vect_permute_store_chain): Likewise.
8595 (vect_grouped_load_supported): Likewise.
8596 (vect_permute_load_chain): Likewise.
8597 (vect_shift_permute_load_chain): Likewise.
8598 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8599 (vect_transform_slp_perm_load): Likewise.
8600 (vect_schedule_slp_instance): Likewise.
8601 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8602 (vectorizable_mask_load_store): Likewise.
8603 (vectorizable_bswap): Likewise.
8604 (vectorizable_store): Likewise.
8605 (vectorizable_load): Likewise.
8606 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8607 vec_perm_indices instead of auto_vec_perm_indices. Use
8608 tree_to_vec_perm_builder to read the vector from a tree.
8609 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8610 vec_perm_builder instead of a vec_perm_indices.
8611 (have_whole_vector_shift): Use vec_perm_builder and
8612 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8613 truncation to calc_vec_perm_mask_for_shift.
8614 (vect_create_epilog_for_reduction): Likewise.
8615 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8616 from auto_vec_perm_indices to vec_perm_indices.
8617 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8618 instead of changing individual elements.
8619 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8620 the vector in d.perm.
8621 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8622 from auto_vec_perm_indices to vec_perm_indices.
8623 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8624 instead of changing individual elements.
8625 (arm_vectorize_vec_perm_const): Use new_vector to install
8626 the vector in d.perm.
8627 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8628 Update vec_perm_builder constructor.
8629 (rs6000_expand_interleave): Likewise.
8630 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8631 (rs6000_expand_interleave): Likewise.
8633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8635 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8636 to qimode could truncate the indices.
8637 * optabs.c (expand_vec_perm_var): Likewise.
8639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8641 * Makefile.in (OBJS): Add vec-perm-indices.o.
8642 * vec-perm-indices.h: New file.
8643 * vec-perm-indices.c: Likewise.
8644 * target.h (vec_perm_indices): Replace with a forward class
8646 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8647 * optabs.h: Include vec-perm-indices.h.
8648 (expand_vec_perm): Delete.
8649 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8650 (expand_vec_perm_const): Declare.
8651 * target.def (vec_perm_const_ok): Replace with...
8652 (vec_perm_const): ...this new hook.
8653 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8654 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8655 * doc/tm.texi: Regenerate.
8656 * optabs.def (vec_perm_const): Delete.
8657 * doc/md.texi (vec_perm_const): Likewise.
8658 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8659 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8660 expand_vec_perm for constant permutation vectors. Assert that
8661 the mode of variable permutation vectors is the integer equivalent
8662 of the mode that is being permuted.
8663 * optabs-query.h (selector_fits_mode_p): Declare.
8664 * optabs-query.c: Include vec-perm-indices.h.
8665 (selector_fits_mode_p): New function.
8666 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8667 is defined, instead of checking whether the vec_perm_const_optab
8668 exists. Use targetm.vectorize.vec_perm_const instead of
8669 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8670 fit in the vector mode before using a variable permute.
8671 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8672 vec_perm_indices instead of an rtx.
8673 (expand_vec_perm): Replace with...
8674 (expand_vec_perm_const): ...this new function. Take the selector
8675 as a vec_perm_indices rather than an rtx. Also take the mode of
8676 the selector. Update call to shift_amt_for_vec_perm_mask.
8677 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8678 Use vec_perm_indices::new_expanded_vector to expand the original
8679 selector into bytes. Check whether the indices fit in the vector
8680 mode before using a variable permute.
8681 (expand_vec_perm_var): Make global.
8682 (expand_mult_highpart): Use expand_vec_perm_const.
8683 * fold-const.c: Includes vec-perm-indices.h.
8684 * tree-ssa-forwprop.c: Likewise.
8685 * tree-vect-data-refs.c: Likewise.
8686 * tree-vect-generic.c: Likewise.
8687 * tree-vect-loop.c: Likewise.
8688 * tree-vect-slp.c: Likewise.
8689 * tree-vect-stmts.c: Likewise.
8690 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8692 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8693 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8694 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8695 (aarch64_vectorize_vec_perm_const): ...this new function.
8696 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8697 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8698 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8699 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8700 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8701 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8702 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8704 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8705 check for NEON modes.
8706 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8707 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8708 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8709 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8711 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8712 the old VEC_PERM_CONST conditions.
8713 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8714 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8715 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8716 (ia64_vectorize_vec_perm_const_ok): Merge into...
8717 (ia64_vectorize_vec_perm_const): ...this new function.
8718 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8719 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8720 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8721 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8722 * config/mips/mips.c (mips_expand_vec_perm_const)
8723 (mips_vectorize_vec_perm_const_ok): Merge into...
8724 (mips_vectorize_vec_perm_const): ...this new function.
8725 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8726 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8727 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8728 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8729 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8730 (rs6000_expand_vec_perm_const): Delete.
8731 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8733 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8734 (altivec_expand_vec_perm_const_le): Take each operand individually.
8735 Operate on constant selectors rather than rtxes.
8736 (altivec_expand_vec_perm_const): Likewise. Update call to
8737 altivec_expand_vec_perm_const_le.
8738 (rs6000_expand_vec_perm_const): Delete.
8739 (rs6000_vectorize_vec_perm_const_ok): Delete.
8740 (rs6000_vectorize_vec_perm_const): New function.
8741 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8742 an element count and rtx array.
8743 (rs6000_expand_extract_even): Update call accordingly.
8744 (rs6000_expand_interleave): Likewise.
8745 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8746 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8747 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8748 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8749 (rs6000_expand_vec_perm_const): Delete.
8750 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8751 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8752 (altivec_expand_vec_perm_const_le): Take each operand individually.
8753 Operate on constant selectors rather than rtxes.
8754 (altivec_expand_vec_perm_const): Likewise. Update call to
8755 altivec_expand_vec_perm_const_le.
8756 (rs6000_expand_vec_perm_const): Delete.
8757 (rs6000_vectorize_vec_perm_const_ok): Delete.
8758 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8759 reference to the SPE evmerge intructions.
8760 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8761 an element count and rtx array.
8762 (rs6000_expand_extract_even): Update call accordingly.
8763 (rs6000_expand_interleave): Likewise.
8764 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8765 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8767 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8769 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8771 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8772 vector mode and that that mode matches the mode of the data
8774 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8775 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8776 directly using expand_vec_perm_1 when forcing selectors into
8778 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8780 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8782 * optabs-query.h (can_vec_perm_p): Delete.
8783 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8784 * optabs-query.c (can_vec_perm_p): Split into...
8785 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8786 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8787 particular selector is valid.
8788 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8789 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8790 (vect_grouped_load_supported): Likewise.
8791 (vect_shift_permute_load_chain): Likewise.
8792 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8793 (vect_transform_slp_perm_load): Likewise.
8794 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8795 (vectorizable_bswap): Likewise.
8796 (vect_gen_perm_mask_checked): Likewise.
8797 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8798 implementations of variable permutation vectors into account
8799 when deciding which selector to use.
8800 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8801 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8802 with a false third argument.
8803 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8804 to test whether the constant selector is valid and can_vec_perm_var_p
8805 to test whether a variable selector is valid.
8807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8809 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8810 * optabs-query.c (can_vec_perm_p): Likewise.
8811 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8812 instead of vec_perm_indices.
8813 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8814 (vect_gen_perm_mask_checked): Likewise,
8815 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8816 (vect_gen_perm_mask_checked): Likewise,
8818 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8820 * optabs-query.h (qimode_for_vec_perm): Declare.
8821 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8822 (qimode_for_vec_perm): ...this new function.
8823 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8825 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8827 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8828 does not have a conditional at the top.
8830 2018-01-02 Richard Biener <rguenther@suse.de>
8832 * ipa-inline.c (big_speedup_p): Fix expression.
8834 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8837 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8840 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8844 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8845 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8846 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8847 cond_taken_branch_cost 3->4.
8849 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8851 PR tree-optimization/83581
8852 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8853 TODO_cleanup_cfg if any changes have been made.
8856 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8857 convert_modes if target mode has the right side, but different mode
8861 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8862 last argument when extracting from CONCAT. If either from_real or
8863 from_imag is NULL, use expansion through memory. If result is not
8864 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8865 the parts directly to inner mode, if even that fails, use expansion
8869 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8870 check for bswap in mode rather than HImode and use that in expand_unop
8873 Copyright (C) 2018 Free Software Foundation, Inc.
8875 Copying and distribution of this file, with or without modification,
8876 are permitted in any medium without royalty provided the copyright
8877 notice and this notice are preserved.