1 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4 * varasm.c (hash_section): Return an unchangeble hash value
5 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6 Return !aarch64_can_use_per_function_literal_pools_p ().
8 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12 rather than address of cmode as last argument to prepare_cmp_insn.
14 2018-03-15 Julia Koval <julia.koval@intel.com>
16 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17 F_AVX512VNNI, F_AVX512BITALG): New.
19 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
22 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
23 insn for floating-point loads and stores.
25 2018-03-14 Carl Love <cel@us.ibm.com>
27 * config/rs6000/rs6000-c.c: Add macro definitions for
28 ALTIVEC_BUILTIN_VEC_PERMXOR.
29 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
30 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
31 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
32 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
34 * config/doc/extend.texi: Add prototypes for vec_permxor.
36 2018-03-14 David Malcolm <dmalcolm@redhat.com>
39 * diagnostic-show-locus.c (class layout_point): Convert m_line
40 from int to linenum_type.
41 (line_span::comparator): Use linenum "compare" function when
42 comparing line numbers.
43 (test_line_span): New function.
44 (layout_range::contains_point): Convert param "row" from int to
46 (layout_range::intersects_line_p): Likewise.
47 (layout::will_show_line_p): Likewise.
48 (layout::print_source_line): Likewise.
49 (layout::should_print_annotation_line_p): Likewise.
50 (layout::print_annotation_line): Likewise.
51 (layout::print_leading_fixits): Likewise.
52 (layout::annotation_line_showed_range_p): Likewise.
53 (struct line_corrections): Likewise for field m_row.
54 (line_corrections::line_corrections): Likewise for param "row".
55 (layout::print_trailing_fixits): Likewise.
56 (layout::get_state_at_point): Likewise.
57 (layout::get_x_bound_for_row): Likewise.
58 (layout::print_line): Likewise.
59 (diagnostic_show_locus): Likewise for locals "last_line" and
61 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
62 * input.c (selftest::test_linenum_comparisons): New function.
63 (selftest::input_c_tests): Call it.
64 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
65 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
66 * selftest.h (ASSERT_GT): New macro.
67 (ASSERT_GT_AT): New macro.
68 (ASSERT_LT): New macro.
69 (ASSERT_LT_AT): New macro.
71 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
73 PR rtl-optimization/84780
74 * combine.c (distribute_links): Don't make a link based on pc_rtx.
76 2018-03-14 Martin Liska <mliska@suse.cz>
78 * tree.c (record_node_allocation_statistics): Use
80 (get_stats_node_kind): New function extracted from
81 record_node_allocation_statistics.
82 (free_node): Use get_stats_node_kind.
84 2018-03-14 Richard Biener <rguenther@suse.de>
86 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
87 that the value-set of ANTIC_IN doesn't grow.
90 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
92 (BB_VISITED_WITH_VISITED_SUCCS): New define.
93 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
95 2018-03-14 Julia Koval <julia.koval@intel.com>
97 * config.gcc (icelake-client, icelake-server): New.
99 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
100 (initial_ix86_arch_features): Ditto.
101 (PTA_SKYLAKE): Add SGX.
102 (PTA_ICELAKE): Remove.
103 (PTA_ICELAKE_CLIENT): New.
104 (PTA_ICELAKE_SERVER): New.
105 (ix86_option_override_internal): Split up icelake on icelake client and
107 (get_builtin_code_for_version): Ditto.
108 (fold_builtin_cpu): Ditto.
109 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
110 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
111 * config/i386/i386.h (processor_type): Ditto.
112 * doc/invoke.texi: Ditto.
114 2018-03-14 Jakub Jelinek <jakub@redhat.com>
117 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
118 INTEGER_CST offset, add it together with bitpos / 8 and
119 sign extend based on POINTER_SIZE.
123 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
126 * config/i386/constraints.md (Yc): New register constraint.
127 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
128 Use Yc constraint for alternative 2 of operand 0. Remove
129 preferred_for_speed attribute.
131 2018-03-14 Richard Biener <rguenther@suse.de>
133 PR tree-optimization/84830
134 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
135 with the old one to avoid oscillations.
137 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
140 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
142 (assign_by_spills): Return a flag of reload assignment failure.
143 Do not process the reload assignment failures. Do not spill other
144 reload pseudos if they has the same reg class. Update n if
146 (lra_assign): Add a return arg. Set up from the result of
147 assign_by_spills call.
148 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
149 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
150 usage_insns if it is not NULL.
151 (spill_hard_reg_in_range): New function.
152 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
153 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
155 (lra_assign): Change prototype.
156 * lra.c (lra): Add code to deal with fails by splitting hard reg
159 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
161 * config/riscv/riscv.opt (mrelax): New option.
162 * config/riscv/riscv.c (riscv_file_start): Emit ".option
163 "norelax" when riscv_mrelax is disabled.
164 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
166 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
169 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
170 reassociation for int modes.
172 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
174 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
175 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
177 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
178 * config/aarch64/aarch64-sve.md
179 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
180 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
181 (*extend<mode><Vwide>2): Rename to...
182 (aarch64_sve_extend<mode><Vwide>2): ...this.
183 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
184 renaming the old pattern to...
185 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
187 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
188 define_expand, renaming the old pattern to...
189 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
190 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
191 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
192 account when deciding which SVE instruction the optab should use.
193 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
195 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
197 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
198 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
199 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
200 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
201 (tlsdesc_small_<mode>): Turn a define_expand and use
202 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
203 (tlsdesc_small_advsimd_<mode>): ...this.
204 (tlsdesc_small_sve_<mode>): New pattern.
206 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
208 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
209 (UNSPEC_UMUL_HIGHPART): New constants.
210 (MUL_HIGHPART): New int iteraor.
211 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
212 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
214 (*<su>mul<mode>3_highpart): New define_insn.
216 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
219 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
222 2018-03-13 Martin Liska <mliska@suse.cz>
225 * (sem_item_optimizer::sem_item_optimizer): Initialize new
227 (sem_item_optimizer::~sem_item_optimizer): Release it.
228 (sem_item_optimizer::merge_classes): Register variable aliases.
229 (sem_item_optimizer::fixup_pt_set): New function.
230 (sem_item_optimizer::fixup_points_to_sets): Likewise.
231 * ipa-icf.h: Declare new variables and functions.
233 2018-03-13 Jakub Jelinek <jakub@redhat.com>
236 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
237 integer_pow2p@2 and test integer_pow2p in condition.
238 (A < 0 ? C : 0): Similarly for @1.
241 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
242 characters starting at p contain '\0' character, don't look beyond
246 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
247 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
250 * reg-stack.c (change_stack): Change update_end var from int to
251 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
252 also call set_block_for_insn on the newly added insns and rescan.
255 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
259 * tree.c (stabilize_reference_1): Return save_expr (e) for
260 STATEMENT_LIST even if it doesn't have side-effects.
262 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
264 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
266 2018-03-12 Renlin Li <renlin.li@arm.com>
268 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
269 aarch64_output_scalar_simd_mov_immediate.
271 2018-03-12 Martin Sebor <msebor@redhat.com>
273 PR tree-optimization/83456
274 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
275 for perfectly overlapping calls to memcpy.
276 (gimple_fold_builtin_memory_chk): Same.
277 (gimple_fold_builtin_strcpy): Handle no-warning.
278 (gimple_fold_builtin_stxcpy_chk): Same.
279 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
281 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
283 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
284 parameter. Use it for SFmode.
285 (rs6000_function_arg_advance_1): Adjust.
286 (rs6000_function_arg): Adjust.
287 (rs6000_gimplify_va_arg): Pass false for that new parameter.
289 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
291 PR rtl-optimization/84169
292 PR rtl-optimization/84780
293 * combine.c (can_combine_p): Check for a 2-insn combination whether
294 the destination register is used between the two insns, too.
296 2018-03-12 Richard Biener <rguenther@suse.de>
298 PR tree-optimization/84803
299 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
300 for refs DR analysis didn't process.
302 2018-03-12 Richard Biener <rguenther@suse.de>
304 PR tree-optimization/84777
305 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
306 force-vectorize loops ignore whether we are optimizing for size.
308 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
310 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
311 (TARGET_MD_ASM_ADJUST): Define.
313 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
314 Kito Cheng <kito.cheng@gmail.com>
315 Chung-Ju Wu <jasonwucj@gmail.com>
317 * config/nds32/nds32.c (nds32_compute_stack_frame,
318 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
319 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
320 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
321 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
322 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
323 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
324 * config/nds32/nds32.md (prologue, epilogue): Use macro
325 NDS32_V3PUSH_AVAILABLE_P to do checking.
327 2018-03-11 Jakub Jelinek <jakub@redhat.com>
330 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
331 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
332 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
333 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
334 addition of most attributes on !orig_type_die or the attribute not
335 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
337 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
338 Chung-Ju Wu <jasonwucj@gmail.com>
340 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
342 * config/nds32/nds32.opt (mvh): New option.
344 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
345 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
349 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
350 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
353 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
354 Chung-Ju Wu <jasonwucj@gmail.com>
356 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
358 * config/nds32/nds32-multiple.md (strlensi): New pattern.
359 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
361 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
362 Kito Cheng <kito.cheng@gmail.com>
363 Chung-Ju Wu <jasonwucj@gmail.com>
365 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
366 UNSPEC_FFMISM and UNSPEC_FLMISM.
367 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
368 for ffb, ffmism and flmism.
369 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
370 (unspec_ffmism): Ditto.
371 (unspec_flmism): Ditto.
372 (nds32_expand_builtin_impl): Check if string extension is available.
373 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
374 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
376 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
379 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
382 * lra-assigns.c (assign_by_spills): Return a flag of reload
383 assignment failure. Do not process the reload assignment
384 failures. Do not spill other reload pseudos if they has the same
386 (lra_assign): Add a return arg. Set up from the result of
387 assign_by_spills call.
388 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
389 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
390 usage_insns if it is not NULL.
391 (spill_hard_reg_in_range): New function.
392 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
393 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
395 (lra_assign): Change prototype.
396 * lra.c (lra): Add code to deal with fails by splitting hard reg
399 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
402 * config/i386/i386.opt: Replace Enforcment with Enforcement.
404 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
407 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
408 (dw_val_node): Add val_symbolic_view.
409 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
410 (symview_upper_bound): New.
411 (new_line_info_table): Initialize symviews_since_reset.
412 (dwarf2out_source_line): Count symviews_since_reset and set
414 (dw_val_equal_p): Handle symview.
415 (add_AT_symview): New.
416 (print_dw_val): Handle symview.
417 (attr_checksum, attr_checksum_ordered): Likewise.
418 (same_dw_val_p, size_of_die): Likewise.
419 (value_format, output_die): Likewise.
420 (add_high_low_attributes): Use add_AT_symview for entry_view.
421 (dwarf2out_finish): Reset symview_upper_bound, clear
424 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
427 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
428 Add strict argument and use it.
429 (rs6000_split_multireg_move): Update for new strict argument.
430 (mem_operand_gpr): Disallow all non-offsettable addresses.
431 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
433 2018-03-09 Jakub Jelinek <jakub@redhat.com>
436 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
437 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
438 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
441 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
442 decl, use remap_type if we want to use the type.
444 2018-03-09 Martin Sebor <msebor@redhat.com>
446 PR tree-optimization/84526
447 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
449 (builtin_access::generic_overlap): Be prepared to handle non-array
452 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
454 PR rtl-optimization/84682
455 * lra-constraints.c (process_address_1): Check is_address flag
456 for address constraints.
457 (process_alt_operands): Likewise.
458 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
459 preprocess_constraints.
460 * recog.h (preprocess_constraints): Add oploc parameter.
462 * recog.c (preprocess_constraints): Test address_operand for
463 CT_ADDRESS constraints.
465 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
468 * lra-assigns.c (assign_by_spills): Return a flag of reload
469 assignment failure. Do not process the reload assignment
470 failures. Do not spill other reload pseudos if they has the same
472 (lra_assign): Add a return arg. Set up from the result of
473 assign_by_spills call.
474 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
475 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
476 usage_insns if it is not NULL.
477 (spill_hard_reg_in_range): New function.
478 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
479 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
481 (lra_assign): Change prototype.
482 * lra.c (lra): Add code to deal with fails by splitting hard reg
485 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
488 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
489 Accept complain bool parameter. Only emit errors if it is true.
490 (arm_parse_cpu_option_name): Likewise.
491 (arm_target_thumb_only): Adjust callers of the above.
492 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
493 prototype to take a default true bool parameter.
494 (arm_parse_arch_option_name): Likewise.
496 2018-03-09 David Malcolm <dmalcolm@redhat.com>
497 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
501 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
502 * configure: Regenerate.
503 * configure.ac ("linker --version-script option"): New.
504 ("linker soname option"): New.
506 2018-03-09 Richard Biener <rguenther@suse.de>
508 PR tree-optimization/84775
509 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
510 immediate uses of predicate stmts and mark them modified.
513 PR tree-optimization/84178
514 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
516 (version_loop_for_if_conversion): Delay update_ssa call.
517 (tree_if_conversion): Delay update_ssa until after predicate
520 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
523 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
524 when the function accesses prior frames.
526 2018-03-08 Jakub Jelinek <jakub@redhat.com>
529 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
530 gen_llsym, otherwise call maybe_gen_llsym.
533 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
534 has ',' character inside of it.
536 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
539 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
540 as clobbering CC_REGNUM.
542 2018-03-08 Richard Biener <rguenther@suse.de>
545 * tree-scalar-evolution.c: Include tree-into-ssa.h.
546 (follow_copies_to_constant): Do not follow SSA names registered
549 2018-03-08 Richard Biener <rguenther@suse.de>
551 PR tree-optimization/84178
552 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
554 (version_loop_for_if_conversion): Delay update_ssa call.
555 (tree_if_conversion): Delay update_ssa until after predicate
558 2018-03-08 David Malcolm <dmalcolm@redhat.com>
560 PR tree-optimization/84178
561 * tree-if-conv.c (release_bb_predicate): Remove the
562 the assertion that the stmts have NULL use_ops.
563 Discard the statements, asserting that they haven't
564 yet been added to a BB.
566 2018-03-08 Richard Biener <rguenther@suse.de>
568 PR tree-optimization/84746
569 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
570 (phi_translate): Pass in destination ANTIC_OUT set.
571 (phi_translate_1): Likewise. For a simplified result lookup
572 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
573 (phi_translate_set): Adjust.
574 (do_pre_regular_insertion): Likewise.
575 (do_pre_partial_partial_insertion): Likewise.
577 2018-03-08 Martin Liska <mliska@suse.cz>
579 PR gcov-profile/84735
580 * doc/gcov.texi: Document usage of profile files.
581 * gcov-io.h: Document changes in the format.
583 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
587 * dwarf2out.c (struct dw_line_info_table): Update comments for
589 (FORCE_RESET_NEXT_VIEW): New.
590 (FORCE_RESETTING_VIEW_P): New.
591 (RESETTING_VIEW_P): Check for -1 too.
592 (ZERO_VIEW_P): Likewise.
593 (new_line_info_table): Force-reset next view.
594 (dwarf2out_begin_function): Likewise.
595 (dwarf2out_source_line): Simplify zero_view_p initialization.
596 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
597 view directly. Omit view when omitting .loc at line 0.
599 2018-03-08 Jakub Jelinek <jakub@redhat.com>
601 PR tree-optimization/84740
602 * tree-switch-conversion.c (process_switch): Call build_constructors
603 only if info.phi_count is non-zero.
605 PR tree-optimization/84739
606 * tree-tailcall.c (find_tail_calls): Check call arguments against
607 DECL_ARGUMENTS (current_function_decl) rather than
608 DECL_ARGUMENTS (func) when checking for tail recursion.
610 2018-03-07 Jakub Jelinek <jakub@redhat.com>
612 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
613 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
614 Volker Reichelt's entry and add entries for people that perform
615 GCC fuzzy testing and report numerous bugs.
617 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
620 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
621 readonly data in sdata, if that is disabled.
622 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
623 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
624 -mreadonly-in-sdata option.
626 2018-03-07 Martin Sebor <msebor@redhat.com>
628 PR tree-optimization/84468
629 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
630 basic block when looking for nul assignment.
632 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
635 * except.h (output_function_exception_table): Adjust prototype.
636 * except.c (output_function_exception_table): Remove FNNAME parameter
637 and add SECTION parameter. Ouput one part of the table at a time.
638 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
639 the first part of the exception table and emit unwind directives.
640 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
641 (i386_pe_seh_cold_init): Likewise.
642 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
643 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
644 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
645 (ix86_output_call_insn): Emit a nop in one more case for SEH.
646 * config/i386/winnt.c: Include except.h.
647 (struct seh_frame_state): Add reg_offset, after_prologue and
648 in_cold_section fields.
649 (i386_pe_seh_end_prologue): Set seh->after_prologue.
650 (i386_pe_seh_cold_init): New function.
651 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
652 to seh->in_cold_section.
653 (seh_emit_push): Record the offset of the push.
654 (seh_emit_save): Record the offet of the save.
655 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
656 Test seh->after_prologue to disregard the epilogue.
657 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
658 (i386_pe_end_cold_function): New function.
660 2018-03-07 Jakub Jelinek <jakub@redhat.com>
663 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
664 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
667 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
669 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
670 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
673 * multiple_target.c: Include tree-inline.h and intl.h.
674 (expand_target_clones): Diagnose and fail if node->definition and
675 !tree_versionable_function_p (node->decl).
677 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
679 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
681 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
682 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
683 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
685 2018-03-06 Jakub Jelinek <jakub@redhat.com>
688 * combine.c (try_combine): Use reg_or_subregno instead of handling
689 just paradoxical SUBREGs and REGs.
691 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
693 * config/arc/arc.c (arc_finalize_pic): Remove function.
694 (arc_must_save_register): We use single base PIC register, remove
695 checks to save/restore the PIC register.
696 (arc_expand_prologue): Likewise.
697 * config/arc/arc-protos.h (arc_set_default_type_attributes):
699 (arc_verify_short): Likewise.
700 (arc_attr_type): Likewise.
701 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
702 (walk_stores): Likewise.
703 (arc_address_cost): Make it static.
704 (arc_verify_short): Likewise.
705 (branch_dest): Likewise.
706 (arc_attr_type): Likewise.
707 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
708 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
709 (arc_final_prescan_insn): Remove inserting the nops due to
710 hardware hazards. It is done in reorg step.
711 (insn_length_variant_t): Remove.
712 (insn_length_parameters_t): Likewise.
713 (arc_insn_length_parameters): Likewise.
714 (arc_get_insn_variants): Likewise.
715 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
717 2018-03-06 Jakub Jelinek <jakub@redhat.com>
720 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
723 PR tree-optimization/84687
724 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
726 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
728 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
730 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
731 Rename to ppc_speculation_barrier.
732 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
733 __builtin_ppc_speculation_barrier.
735 2018-03-05 Jakub Jelinek <jakub@redhat.com>
738 * combine.c (combine_simplify_rtx): Don't try to simplify if
739 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
742 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
744 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
745 to 32 bytes when compiling for POWER9.
747 2018-03-05 Jakub Jelinek <jakub@redhat.com>
750 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
751 regparm >= 3 with no arg reg available also for calls with
752 flag_force_indirect_call. Pass decl to ix86_function_regparm.
755 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
757 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
759 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
762 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
764 2018-03-05 Richard Biener <rguenther@suse.de>
766 PR tree-optimization/84486
767 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
768 When inserting a __builtin_assume_aligned call set the LHS
769 SSA name alignment info accordingly.
771 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
773 PR tree-optimization/84114
774 * config/aarch64/aarch64.c (aarch64_reassociation_width)
775 Avoid reassociation of FLOAT_MODE addition.
777 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
779 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
780 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
781 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
782 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
783 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
784 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
787 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
788 __builtin_ia32_wbinvd): New builtins.
789 (SPECIAL_ARGS2): New.
790 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
791 (SPECIAL_ARGS2): New.
792 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
793 (ix86_valid_target_attribute_inner_p): Ditto.
794 (ix86_init_mmx_sse_builtins): Add special_args2.
795 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
796 TARGET_WBNOINVD_P): New.
797 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
798 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
799 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
800 * config/i386/immintrin.h (_wbinvd): New intrinsic.
801 * config/i386/pconfigintrin.h: New file.
802 * config/i386/wbnoinvdintrin.h: Ditto.
803 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
804 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
806 2018-03-05 Richard Biener <rguenther@suse.de>
808 PR tree-optimization/84670
809 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
811 (BB_VISITED_WITH_VISITED_SUCCS): New define.
812 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
813 (compute_antic_aux): Only assert the number of values in ANTIC_IN
814 doesn't grow if all successors (recursively) were visited at least
817 2018-03-05 Richard Biener <rguenther@suse.de>
819 PR tree-optimization/84650
820 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
821 if executed in the loop pipeline.
823 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
825 * doc/configfiles.texi (Configuration Files): Move info about
826 conditionalizing $target-protos.h to...
827 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
828 differs from $target-protos.h.
830 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
831 Chung-Ju Wu <jasonwucj@gmail.com>
833 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
834 * config/nds32/nds32-multiple.md (setmemsi): Define.
835 * config/nds32/nds32-memory-manipulation.c
836 (nds32_gen_dup_4_byte_to_word_value): New.
837 (emit_setmem_word_loop): New.
838 (emit_setmem_byte_loop): New.
839 (nds32_expand_setmem_loop): New.
840 (nds32_expand_setmem_loop_v3m): New.
841 (nds32_expand_setmem_unroll): New.
842 (nds32_expand_setmem): New.
844 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
845 Chung-Ju Wu <jasonwucj@gmail.com>
847 * config/nds32/nds32-memory-manipulation.c
848 (nds32_emit_load_store): New.
849 (nds32_emit_post_inc_load_store): New.
850 (nds32_emit_mem_move): New.
851 (nds32_emit_mem_move_block): New.
852 (nds32_expand_movmemsi_loop_unknown_size): New.
853 (nds32_expand_movmemsi_loop_known_size): New.
854 (nds32_expand_movmemsi_loop): New.
855 (nds32_expand_movmemsi_unroll): New.
856 (nds32_expand_movmemqi): Rename ...
857 (nds32_expand_movmemsi): ... to this.
858 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
859 (movmemsi): ... to this.
860 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
861 (nds32_expand_movmemsi): ... to this.
863 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
864 Monk Chiang <sh.chiang04@gmail.com>
865 Chung-Ju Wu <jasonwucj@gmail.com>
867 * config/nds32/nds32-protos.h
868 (nds32_expand_load_multiple): New arguments.
869 (nds32_expand_store_multiple): Ditto.
870 (nds32_valid_multiple_load_store): Rename ...
871 (nds32_valid_multiple_load_store_p): ... to this.
872 * config/nds32/nds32-memory-manipulation.c
873 (nds32_expand_load_multiple): Refine implementation.
874 (nds32_expand_store_multiple): Ditto.
875 * config/nds32/nds32-multiple.md
876 (load_multiple): Update nds32_expand_load_multiple interface.
877 (store_multiple): Update nds32_expand_store_multiple interface.
878 * config/nds32/nds32-predicates.c
879 (nds32_valid_multiple_load_store): Rename ...
880 (nds32_valid_multiple_load_store_p): ... to this and refine
882 * config/nds32/predicates.md
883 (nds32_load_multiple_and_update_address_operation): New predicate.
884 (nds32_store_multiple_and_update_address_operation): New predicate.
886 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
887 Chung-Ju Wu <jasonwucj@gmail.com>
889 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
890 (combo): New attribute.
891 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
893 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
895 * config/nds32/nds32.opt: Change -mcmodel= default value.
897 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
898 Monk Chiang <sh.chiang04@gmail.com>
899 Chung-Ju Wu <jasonwucj@gmail.com>
901 * config/nds32/constants.md (unspec_element): New enum.
902 * config/nds32/constraints.md (Umw): New constraint.
903 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
904 * config/nds32/nds32-intrinsic.md: Likewise.
905 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
906 (nds32_valid_smw_lwm_base_p): New.
907 (nds32_output_smw_single_word): New.
908 (nds32_output_lmw_single_word): New.
909 (nds32_expand_unaligned_load): New.
910 (nds32_expand_unaligned_store): New.
911 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
912 (nds32_output_smw_single_word): Declare.
913 (nds32_output_lmw_single_word): Declare.
914 (nds32_expand_unaligned_load): Declare.
915 (nds32_expand_unaligned_store): Declare.
916 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
917 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
918 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
919 NDS32_BUILTIN_UASTORE_DW.
920 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
923 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
924 Kito Cheng <kito.cheng@gmail.com>
925 Chung-Ju Wu <jasonwucj@gmail.com>
927 * config/nds32/nds32-intrinsic.c
928 (nds32_expand_builtin_null_ftype_reg): Delete.
929 (nds32_expand_builtin_reg_ftype_imm): Ditto.
930 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
931 (nds32_read_argument): New.
932 (nds32_legitimize_target): Ditto.
933 (nds32_legitimize_argument): Ditto.
934 (nds32_check_constant_argument): Ditto.
935 (nds32_expand_unop_builtin): Ditto.
936 (nds32_expand_unopimm_builtin): Ditto.
937 (nds32_expand_binop_builtin): Ditto.
938 (nds32_builtin_decl_impl): Ditto.
939 (builtin_description): Ditto.
940 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
941 (nds32_init_builtins_impl): Ditto.
942 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
943 (nds32_builtin_decl): New.
944 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
945 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
947 2018-03-02 Jeff Law <law@redhat.com>
949 * reorg.c (stop_search_p): Handle DEBUG_INSN.
950 (redundant_insn, fill_simple_delay_slots): Likewise.
951 (fill_slots_from_thread): Likewise.
952 * resource.c (mark_referenced_resources): Likewise.
953 (mark_set_resources, find_dead_or_set_registers): Likewise.
955 2018-03-02 Jakub Jelinek <jakub@redhat.com>
957 * substring-locations.h (format_warning_va): Formatting fix for
959 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
961 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
962 * substring-locations.c: Include intl.h.
963 (format_warning_va): Turned into small wrapper around
964 format_warning_n_va, renamed to ...
965 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
966 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
968 (format_warning_at_substring_n): New function.
969 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
970 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
971 format_warning_at_substring with just a shorter name instead of
972 const function pointer.
973 (fmtwarn_n): New function.
974 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
975 appropriate, get rid of all the fmtstr temporaries, move conditionals
976 with G_() wrapped string literals directly into fmtwarn arguments,
977 cast dir.len to (int), formatting fixes.
979 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
981 * doc/invoke.texi: Remove "Cilk Plus" references.
983 2018-03-02 Jakub Jelinek <jakub@redhat.com>
984 Richard Biener <rguenther@suse.de>
987 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
988 for error or warning attributes if CALL_FROM_THUNK_P is set.
991 2018-03-02 Jakub Jelinek <jakub@redhat.com>
994 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
995 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
998 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
999 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1001 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1002 instead of -1U in last predictors element's probability member.
1004 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1007 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1008 arguments if they are comparable.
1010 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1012 PR tree-optimization/84634
1013 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1014 masks and masked_loop_p with a single loop_masks, making sure it's
1015 null for bb vectorization.
1017 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1019 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1020 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1021 loop->force_vectorize to check whether there is no alias.
1023 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1026 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1028 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1030 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1031 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1032 instead of a loop around prev_real_insn.
1033 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1037 * config/i386/i386.c (ix86_print_operand): Use conditional
1038 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1041 2018-03-02 Richard Biener <rguenther@suse.de>
1043 PR tree-optimization/84427
1044 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1045 (bitmap_set_subtract_values): Rewrite to handle multiple
1048 (prune_clobbered_mems): Likewise.
1049 (phi_translate): Take edge instead of pred/phiblock.
1050 (phi_translate_1): Likewise.
1051 (phi_translate_set): Likewise. Insert all translated
1052 exprs for a value into the set, keeping possibly multiple
1053 expressions per value.
1054 (compute_antic_aux): Adjust for phi_translate changes.
1055 When intersecting union the expressions and prune those
1056 not in the final value set, keeping possibly multiple
1057 expressions per value. Do not use value-insertion
1058 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1059 all expressions. Add verification that the value-sets
1060 only shrink during iteration.
1061 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1062 (do_pre_regular_insertion): Likewise.
1063 (do_pre_partial_partial_insertion): Likewise.
1065 2018-03-02 Richard Biener <rguenther@suse.de>
1068 * config/darwin.c (saved_debug_info_level): New static global.
1069 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1070 (darwin_asm_lto_end): Restore debug info generation settings.
1072 2018-03-01 Martin Liska <mliska@suse.cz>
1075 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1078 2018-03-01 Richard Biener <rguenther@suse.de>
1081 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1082 type annotation with LTO when debug was disabled at compile-time.
1084 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1086 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1088 (mips_final_postscan_insn): Likewise.
1090 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1092 PR rtl-optimization/84528
1093 * alias.c (init_alias_target): Add commentary.
1094 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1095 a unique base value if the frame pointer is not eliminated
1096 to the stack pointer.
1098 2018-03-01 Tom de Vries <tom@codesourcery.com>
1100 PR rtl-optimization/83327
1101 * lra-int.h (hard_regs_spilled_into): Declare.
1102 * lra.c (hard_regs_spilled_into): Define.
1103 (init_reg_info): Init hard_regs_spilled_into.
1104 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1105 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1106 (process_bb_lives): Handle hard_regs_spilled_into.
1107 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1108 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1110 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1112 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1113 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1114 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1115 * config/rs6000/aix72.h: New file.
1117 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1119 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1120 instead of warning_at with conditional singular and plural messages
1124 * stor-layout.c (update_alignment_for_field): For
1125 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1126 && !DECL_PACKED (field), do the alignment update, just use
1127 only desired_align instead of MAX (type_align, desired_align)
1129 (place_field): Don't do known_align < desired_align handling
1130 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1131 is non-NULL, instead do it after rli->prev_field handling and
1132 only if not within a bitfield word. For DECL_PACKED (field)
1133 use type_align of BITS_PER_UNIT.
1135 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1137 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1138 superfluous parentheses and trailing spaces.
1140 2018-02-28 Richard Biener <rguenther@suse.de>
1142 PR tree-optimization/84584
1143 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1144 SCoPs with fake exit edge.
1146 2018-02-28 Martin Liska <mliska@suse.cz>
1149 * timevar.c (timer::print): Fix format to properly print 100%
1152 2018-02-28 Richard Biener <rguenther@suse.de>
1155 * genmatch.c (capture_info::walk_match): Do not mark
1156 captured expressions without operands as expr_p given
1157 they act more like predicates and should be subject to
1158 "lost tail" side-effect preserving.
1160 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1162 PR rtl-optimization/81611
1163 * auto-inc-dec.c (attempt_change): Move dead note from
1164 mem_insn if it's the next use of regno
1165 (find_address): Take address use of reg holding
1166 non-incremented value. Add parm to limit search to the named
1168 (merge_in_block): Attempt to use a mem insn that is the next
1169 use of the original regno.
1171 2018-02-27 Martin Sebor <msebor@redhat.com>
1174 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1175 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1177 2018-02-27 Martin Sebor <msebor@redhat.com>
1179 PR translation/84207
1180 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1181 n argument to unsigned HOST_WIDE_INT.
1182 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1183 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1184 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1185 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1187 2018-02-27 Richard Biener <rguenther@suse.de>
1189 PR tree-optimization/84512
1190 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1191 Do not use the estimate returned from record_stmt_cost for
1192 the scalar iteration cost but sum properly using add_stmt_cost.
1194 2018-02-27 Richard Biener <rguenther@suse.de>
1196 PR tree-optimization/84466
1197 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1198 Adjust last change to less strictly validate use operands.
1200 2018-02-27 Martin Liska <mliska@suse.cz>
1202 PR gcov-profile/84548
1203 * gcov.c (process_file): Allow partial overlap and consider it
1204 also as group functions.
1205 (output_lines): Properly calculate range of lines for a group.
1207 2018-02-27 Martin Liska <mliska@suse.cz>
1209 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1210 'ggc' suffixes. Change first column width.
1211 (timer::print): Fix formatting of the column.
1213 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1215 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1216 preserve inline entry blocks for the sake of debug inline
1217 entry point markers alone.
1218 (remove_unused_locals): Suggest in comments a better place to
1219 force the preservation of inline entry blocks that are
1220 otherwise unused, but do not preserve them.
1222 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1224 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1226 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1229 * config/i386/constraints.md (Bs): Replace
1230 ix86_indirect_branch_register with
1231 TARGET_INDIRECT_BRANCH_REGISTER.
1233 * config/i386/i386.md (indirect_jump): Likewise.
1234 (tablejump): Likewise.
1235 (*sibcall_memory): Likewise.
1236 (*sibcall_value_memory): Likewise.
1237 Peepholes of indirect call and jump via memory: Likewise.
1238 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1239 (*sibcall_value_GOT_32): Likewise.
1240 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1241 (GOT_memory_operand): Likewise.
1242 (call_insn_operand): Likewise.
1243 (sibcall_insn_operand): Likewise.
1244 (GOT32_symbol_operand): Likewise.
1245 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1247 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1249 PR rtl-optimization/83496
1250 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1251 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1252 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1253 redundant insn, if any.
1254 (relax_delay_slots): Likewise.
1255 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1257 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1259 PR tree-optimization/83965
1260 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1261 that grouped statements are part of a reduction chain. Return
1262 true if the statement is not marked as a reduction itself but
1264 (vect_recog_dot_prod_pattern): Don't check whether the statement
1265 is part of a group here.
1266 (vect_recog_sad_pattern): Likewise.
1267 (vect_recog_widen_sum_pattern): Likewise.
1269 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1272 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1274 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1277 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1279 (ix86_output_indirect_function_return): New prototype.
1280 (ix86_split_simple_return_pop_internal): Likewise.
1281 * config/i386/i386.c (indirect_return_via_cx): New.
1282 (indirect_return_via_cx_bnd): Likewise.
1283 (indirect_thunk_name): Handle return va CX_REG.
1284 (output_indirect_thunk_function): Create alias for
1285 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1286 (ix86_output_indirect_jmp): Remove the bool argument.
1287 (ix86_output_indirect_function_return): New function.
1288 (ix86_split_simple_return_pop_internal): Likewise.
1289 * config/i386/i386.md (*indirect_jump): Don't pass false
1290 to ix86_output_indirect_jmp.
1291 (*tablejump_1): Likewise.
1292 (simple_return_pop_internal): Change it to define_insn_and_split.
1293 Call ix86_split_simple_return_pop_internal to split it for
1295 (simple_return_indirect_internal): Call
1296 ix86_output_indirect_function_return instead of
1297 ix86_output_indirect_jmp.
1299 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1302 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1303 memset and value initialization afterwards.
1305 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1307 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1309 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1312 * common/config/aarch64/aarch64-common.c
1313 (aarch_option_optimization_table[]): Switch
1314 off fomit-frame-pointer
1316 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1317 Chung-Ju Wu <jasonwucj@gmail.com>
1319 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1321 (store_multiple): Ditto.
1323 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1325 * config.gcc: Add --with-cpu support for nds32 target.
1326 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1327 * config/nds32/nds32.opt: Add -mcpu= option.
1329 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1331 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1332 isel=yes): Warn for these deprecated options.
1334 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1336 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1337 ISA_2_5_MASKS_EMBEDDED.
1339 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1341 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1342 p->max as pointers rather than using iterative_hash_expr.
1344 2018-02-23 Carl Love <cel@us.ibm.com>
1346 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1347 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1349 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1350 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1351 P8V_BUILTIN_VEC_VUNSIGNED2.
1353 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1356 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1357 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1359 * lra-constraints.c (curr_insn_transform): Set up
1360 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1362 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1363 * lra-eliminations.c (spill_pseudos): Ditto.
1364 (process_insn_for_elimination): Ditto.
1365 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1366 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1367 LRA_NON_CLOBBERED_ALT.
1369 2018-02-22 Martin Sebor <msebor@redhat.com>
1371 PR tree-optimization/84480
1372 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1373 to maybe_diag_stxncpy_trunc. Call it.
1374 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1375 from gimple_fold_builtin_strcpy. Print inlining stack.
1376 (handle_builtin_stxncpy): Print inlining stack.
1377 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1379 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1382 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1383 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1384 and -fcheck-pointer-bounds are used together.
1385 (indirect_thunk_prefix): New enum.
1386 (indirect_thunk_need_prefix): New function.
1387 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1388 "_nt" instead of "_bnd" for NOTRACK prefix.
1389 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1390 (output_indirect_thunk_function): Likewise.
1392 (ix86_code_end): Update output_indirect_thunk_function calls.
1393 (ix86_output_indirect_branch_via_reg): Replace
1394 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1395 (ix86_output_indirect_branch_via_push): Likewise.
1396 (ix86_output_function_return): Likewise.
1397 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1398 incompatible with -fcf-protection=branch and
1399 -fcheck-pointer-bounds.
1401 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1404 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1405 Change gcc_assert call to output_operand_lossage.
1407 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1409 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1411 2018-02-22 DJ Delorie <dj@redhat.com>
1412 Sebastian Perta <sebastian.perta@renesas.com>
1413 Oleg Endo <olegendo@gcc.gnu.org>
1415 * config/rx/rx.c (rx_rtx_costs): New function.
1416 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1418 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1420 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1422 2018-02-22 Martin Liska <mliska@suse.cz>
1425 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1426 Add "native" as a possible value.
1428 2018-02-22 Martin Liska <mliska@suse.cz>
1431 * config/i386/i386.c (ix86_option_override_internal):
1432 Add "native" as a possible value for -march and -mtune.
1434 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1437 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1438 to all type variants.
1440 PR tree-optimization/84503
1441 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1442 width as info->bitpos + info->bitsize - start.
1443 (merged_store_group::merge_overlapping): Simplify width computation.
1444 (check_no_overlap): New function.
1445 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1446 start + width and last_order of the group, fail if check_no_overlap
1448 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1449 to group if check_no_overlap fails.
1451 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1453 * config/rs6000/altivec.md: Delete contraint arguments to
1454 define_expand, define_split, and define_peephole2, and in
1455 define_insn_and_split if always unused.
1456 * config/rs6000/darwin.md: Ditto.
1457 * config/rs6000/dfp.md: Ditto.
1458 * config/rs6000/rs6000.md: Ditto.
1459 * config/rs6000/sync.md: Ditto.
1460 * config/rs6000/vector.md: Ditto.
1461 * config/rs6000/vsx.md: Ditto.
1463 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1465 * config/rs6000/altivec.md: Write output control strings as braced
1466 blocks instead of double-quoted strings.
1467 * config/rs6000/darwin.md: Ditto.
1468 * config/rs6000/rs6000.md: Ditto.
1469 * config/rs6000/vector.md: Ditto.
1470 * config/rs6000/vsx.md: Ditto.
1472 2018-02-21 Jason Merrill <jason@redhat.com>
1474 PR c++/84314 - ICE with templates and fastcall attribute.
1475 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1477 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1479 * ipa-cp.c (determine_versionability): Fix comment typos.
1481 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1484 * ipa-cp.c (determine_versionability): Do not version functions caling
1487 2018-02-21 Martin Liska <mliska@suse.cz>
1490 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1491 Add "native" as a possible value.
1492 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1493 the macro when native cpu detection is available.
1495 2018-02-21 Martin Liska <mliska@suse.cz>
1498 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1499 Add "native" as a possible value.
1500 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1501 when native cpu detection is available.
1503 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1504 Martin Sebor <msebor@redhat.com>
1506 PR tree-optimization/84478
1507 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1509 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1510 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1511 support which is conservatively correct, for 2 only stay conservative
1512 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1513 argument to the 2 argument get_range_strlen, adjust 6 arg
1514 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1516 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1517 (gimple_fold_builtin_strlen): Pass true as last argument to
1520 2018-02-20 Martin Sebor <msebor@redhat.com>
1523 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1524 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1525 (builtin_memref::builtin_memref): Factor out parts into
1526 set_base_and_offset and call it.
1528 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1531 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1532 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1533 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1534 search at the associated MODE_INT.
1536 2018-02-20 Jeff Law <law@redhat.com>
1539 PR tree-optimization/81592
1541 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1542 for range data rather than using global data.
1543 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1544 range data rather than using global data.
1545 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1546 pass it to children as needed.
1547 (struct directive::fmtresult): Similarly.
1548 (struct directive::set_width): Similarly.
1549 (struct directive::set_precision): Similarly.
1550 (format_integer, format_directive, parse_directive): Similarly.
1551 (format_none): Accept unnamed vr_values parameter.
1552 (format_percent, format_floating, format_character): Similarly.
1553 (format_string, format_plain): Similarly.
1554 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1555 the EVRP range analyzer for range data rather than using global data.
1556 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1557 gimple-ssa-evrp-analyze.h
1558 (class sprintf_dom_walker): Add after_dom_children member function.
1559 Add evrp_range_analyzer member.
1560 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1561 range analyzer as needed.
1562 (sprintf_dom_walker::after_dom_children): New member function.
1563 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1565 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1566 (evrp_range_analyzer::pop_to_marker): Likewise.
1568 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1570 PR tree-optimization/84419
1571 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1572 with the required type if its current type is compatible but
1575 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1578 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1579 after vectorization.
1581 2018-02-20 Martin Liska <mliska@suse.cz>
1584 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1585 possible values if we don't have a hint.
1587 2018-02-20 Martin Liska <mliska@suse.cz>
1591 * final.c (shorten_branches): Build align_tab array with one
1593 * opts.c (finish_options): Add alignment option limit check.
1594 (MAX_CODE_ALIGN): Likewise.
1595 (MAX_CODE_ALIGN_VALUE): Likewise.
1596 * doc/invoke.texi: Document maximum allowed option value for
1597 all -falign-* options.
1599 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1602 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1603 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1604 * var-tracking.c (emit_note_insn_var_location): Remove all references
1605 to NOTE_INSN_CALL_ARG_LOCATION.
1606 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1607 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1608 Use copy_rtx_if_shared.
1609 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1610 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1611 (dwarf2out_var_location): Remove handling of
1612 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1614 * final.c (final_scan_insn): Remove all references to
1615 NOTE_INSN_CALL_ARG_LOCATION.
1616 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1617 before dumping final insns.
1618 * except.c (emit_note_eh_region_end): Remove all references to
1619 NOTE_INSN_CALL_ARG_LOCATION.
1620 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1621 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1622 * config/arc/arc.c (hwloop_optimize): Likewise.
1623 * config/arm/arm.c (create_fix_barrier): Likewise.
1624 * config/s390/s390.c (s390_chunkify_start): Likewise.
1625 * config/sh/sh.c (find_barrier): Likewise.
1626 * config/i386/i386.c (rest_of_insert_endbranch,
1627 ix86_seh_fixup_eh_fallthru): Likewise.
1628 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1629 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1630 * config/frv/frv.c (frv_function_prologue): Likewise.
1631 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1633 (note_outside_basic_block_p): Remove all references to
1634 NOTE_INSN_CALL_ARG_LOCATION.
1635 * gengtype.c (adjust_field_rtx_def): Likewise.
1636 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1638 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1639 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1642 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1645 PR tree-optimization/84452
1646 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1647 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1650 2018-02-19 Martin Liska <mliska@suse.cz>
1653 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1655 2018-02-19 Martin Liska <mliska@suse.cz>
1656 Richard Sandiford <richard.sandiford@linaro.org>
1658 PR tree-optimization/82491
1659 * gimple-fold.c (get_base_constructor): Make earlier bail out
1662 2018-02-19 Carl Love <cel@us.ibm.com>
1664 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1665 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1667 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1668 P8V_BUILTIN_VEC_NEG.
1670 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1672 * config/rl78/rl78.md (movdf): New define expand.
1674 2018-02-19 Martin Liska <mliska@suse.cz>
1677 * doc/invoke.texi: Fix typo.
1678 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1680 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1682 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1683 handle rs6000_single_float and rs6000_double_float specially for
1686 2018-02-16 Jeff Law <law@redhat.com>
1688 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1691 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1694 * ipa-inline.c (inline_small_functions): Fix a typo.
1696 2018-02-16 Nathan Sidwell <nathan@acm.org>
1698 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1700 2018-02-16 Carl Love <cel@us.ibm.com>
1702 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1703 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1704 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1705 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1706 expansion to P8V_BUILTIN_VEC_FLOAT2.
1708 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1710 PR rtl-optimization/70023
1711 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1712 src_regno into account.
1714 2018-02-16 Carl Love <cel@us.ibm.com>
1716 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1717 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1718 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1719 * config/rs6000/rs6000.c: Remove case statements for
1720 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1721 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1722 and P9V_BUILTIN_VEC_VINSERT4B.
1723 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1724 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1725 * config/rs6000/vsx.md:
1726 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1729 2018-02-16 Carl Love <cel@us.ibm.com>
1731 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1733 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1735 * config/rs6000/rs6000-c.c: Add the definitions for
1736 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1737 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1738 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1739 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1740 definition for insert4b and define insn *insert3b_internal.
1741 * doc/extend.texi: Add documentation for vec_extract4b.
1743 2018-02-16 Nathan Sidwell <nathan@acm.org>
1745 * doc/extend.texi (Backwards Compatibility): Mention friend
1746 injection. Note for-scope is deprecated.
1747 * doc/invoke.texi (-ffriend-injection): Deprecate.
1749 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1751 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1752 that moved to I2, also allow destinations that are a paradoxical
1753 subreg (instead of a normal reg).
1755 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1758 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1761 2018-02-16 Richard Biener <rguenther@suse.de>
1763 PR tree-optimization/84037
1764 PR tree-optimization/84016
1766 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1767 Adjust vec_construct for the fact we need additional higher latency
1768 128bit inserts for AVX256 and AVX512 vector builds.
1769 (ix86_add_stmt_cost): Scale vector construction cost for
1772 2018-02-16 Richard Biener <rguenther@suse.de>
1774 PR tree-optimization/84417
1775 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1776 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1777 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1779 2018-02-16 Martin Liska <mliska@suse.cz>
1782 * internal-fn.def (ASAN_CHECK): Set proper flags.
1783 (ASAN_MARK): Likewise.
1785 2018-02-16 Julia Koval <julia.koval@intel.com>
1787 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1788 from PTA_CANNONLAKE.
1790 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1793 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1794 Use ++iter rather than iter++ for std::list iterators.
1795 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1796 defer deleting them until all nodes in the forest are processed. Do
1797 free even leaf nodes. Change to_process into auto_vec.
1800 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1801 * vec.h (vec_default_construct): Use memset instead of placement new
1802 if BROKEN_VALUE_INITIALIZATION is defined.
1803 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1804 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1807 PR rtl-optimization/83723
1808 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1809 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1810 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1812 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1814 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1816 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1818 PR rtl-optimization/81443
1819 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1820 from inner REGs to paradoxical SUBREGs.
1822 2018-02-16 Richard Biener <rguenther@suse.de>
1824 PR tree-optimization/84399
1825 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1826 For operands we can analyze at their definition make sure we can
1827 analyze them at each use as well.
1829 2018-02-16 Richard Biener <rguenther@suse.de>
1831 PR tree-optimization/84190
1832 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1833 volatile accesses if the decl isn't volatile.
1835 2018-02-15 Jason Merrill <jason@redhat.com>
1837 PR c++/84314 - ICE with templates and fastcall attribute.
1838 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1839 TYPE_CANONICAL on an existing type.
1841 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1843 PR tree-optimization/84383
1844 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1845 dstoff nor call operand_equal_p if dstbase is NULL.
1847 PR tree-optimization/84334
1848 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1849 also a CONSTANT_CLASS_P, punt.
1851 2018-02-14 Jim Wilson <jimw@sifive.com>
1853 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1854 first SMALL_OPERAND check. New local min_second_step. Move assert
1855 to where locals are set. Add TARGET_RVC support.
1856 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1858 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1860 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1862 2018-02-14 Martin Sebor <msebor@redhat.com>
1864 PR tree-optimization/83698
1865 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1866 arrays constrain the offset range to their bounds.
1867 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1868 (builtin_access::overlap): Avoid setting the size of overlap if it's
1870 (maybe_diag_overlap): Also consider arrays when deciding what values
1871 of offsets to include in diagnostics.
1873 2018-02-14 Martin Sebor <msebor@redhat.com>
1876 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1877 that correspond to the kind of a declaration.
1879 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1882 * config/pa/pa.md: Load address of PIC label using the linkage table
1883 if the label is nonlocal.
1885 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1887 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1888 warning message if user requests -maltivec=be.
1889 * doc/invoke.texi: Document deprecation of -maltivec=be.
1891 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1894 * config/rs6000/rs6000-c.c: Update definitions for
1895 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1896 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1898 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1901 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1902 add _get_ssp intrinsics. Remove argument from
1903 __builtin_ia32_rdssp[d|q].
1904 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1905 * config/i386/i386-builtin.def: Remove argument from
1906 __builtin_ia32_rdssp[d|q].
1907 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1908 ix86_expand_special_args_builtin for _rdssp[d|q].
1909 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1910 Clear register before usage.
1911 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1912 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1914 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1916 PR tree-optimization/84357
1917 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1918 operand 1 of an ARRAY_REF too.
1920 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1923 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1924 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1926 (set_of_reg): New struct.
1927 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1928 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1929 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1931 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1932 Split into bitclr, bitset, bitinvert patterns if appropriate.
1933 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1934 use rx_fuse_in_memory_bitop.
1935 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1936 to named insn, correct maximum insn length.
1938 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1941 * machmode.def: Define a complex mode for PARTIAL_INT.
1942 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1944 * doc/rtl.texi: Document CSPImode.
1945 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1947 (msp430_hard_regno_nregs_with_padding): Likewise.
1949 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1952 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1954 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1956 PR rtl-optimization/84169
1957 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1958 we generated a parallel as new i3 and we split that to new i2 and i3
1959 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1960 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1961 those to i2, not i1. Partially rewrite this scan code.
1963 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1966 * stor-layout.c (place_field): For variable length fields, adjust
1967 offset_align afterwards not just based on the field's alignment,
1968 but also on the size.
1971 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1972 of exps and logs in the use_exp2 case.
1974 2018-02-13 Jeff Law <law@redhat.com>
1976 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1979 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1982 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1986 * common.opt (gas-loc-support, gas-locview-support): New.
1987 (ginline-points, ginternal-reset-location-views): New.
1988 * doc/invoke.texi: Document them. Use @itemx where intended.
1989 (gvariable-location-views): Adjust.
1990 * target.def (reset_location_view): New.
1991 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1992 (TARGET_RESET_LOCATION_VIEW): New.
1993 * doc/tm.texi: Rebuilt.
1994 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1995 (dwarf2out_default_as_locview_support): New.
1996 (output_asm_line_debug_info): Use option variables.
1997 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1998 (output_loc_list): Likewise.
1999 (add_high_low_attributes): Check option variables.
2000 Don't output entry view attribute in strict mode.
2001 (gen_inlined_subroutine_die): Check option variables.
2002 (dwarf2out_inline_entry): Likewise.
2003 (init_sections_and_labels): Likewise.
2004 (dwarf2out_early_finish): Likewise.
2005 (maybe_reset_location_view): New, from...
2006 (dwarf2out_var_location): ... here. Call it.
2007 * debug.h (dwarf2out_default_as_loc_support): Declare.
2008 (dwarf2out_default_as_locview_support): Declare.
2009 * hooks.c (hook_int_rtx_insn_0): New.
2010 * hooks.h (hook_int_rtx_insn_0): Declare.
2011 * toplev.c (process_options): Take -gas-loc-support and
2012 -gas-locview-support from dwarf2out. Enable
2013 -gvariable-location-views by default only with locview
2014 assembler support. Enable -ginternal-reset-location-views by
2015 default only if the target defines the corresponding hook.
2016 Enable -ginline-points by default if location views are
2017 enabled; force it disabled if statement frontiers are
2019 * tree-inline.c (expand_call_inline): Check option variables.
2020 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2022 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2024 PR tree-optimization/84321
2025 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2026 handling. Also check whether the anti-range contains any values
2027 that satisfy the mask; switch to a VR_RANGE if not.
2029 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2032 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2034 2018-02-13 Martin Jambor <mjambor@suse.cz>
2037 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2038 of call statements, also set location of a load to a temporary.
2040 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2042 * config/rl78/rl78.c (add_vector_labels): New function.
2043 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2044 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2045 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2046 which checks that no arguments are passed.
2047 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2048 * doc/extend.texi: Documentation for the new attribute.
2050 2018-02-13 Andreas Schwab <schwab@suse.de>
2052 * config/riscv/linux.h (CPP_SPEC): Define.
2054 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2057 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2058 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2059 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2060 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2061 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2062 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2063 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2064 temporarily for AES and PCLMUL builtins.
2066 PR tree-optimization/84339
2067 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2068 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2072 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2073 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2074 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2076 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2078 * omp-simd-clone.h: New file.
2079 * omp-simd-clone.c: Include omp-simd-clone.h.
2080 (expand_simd_clones): No longer static.
2081 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2082 cgraph.h and omp-simd-clone.h.
2083 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2084 (vect_recog_widen_shift_pattern): Formatting fix.
2085 (vect_pattern_recog_1): Don't check optab for calls.
2088 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2089 operands[2] into a REG before using gen_lowpart on it.
2091 2018-02-12 Jeff Law <law@redhat.com>
2094 * config/sh/sh.c (find_barrier): Consider a sibling call
2097 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2098 successfully back substituting a reg.
2100 2018-02-12 Richard Biener <rguenther@suse.de>
2102 PR tree-optimization/84037
2103 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2104 parameter, move visited init to caller.
2105 (vect_slp_analyze_operations): Separate cost from validity
2106 check, initialize visited once for all instances.
2107 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2109 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2111 (vect_model_promotion_demotion_cost): Likewise.
2112 (vectorizable_bswap): Guard cost modeling with !slp_node
2113 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2115 (vectorizable_call): Likewise.
2116 (vectorizable_conversion): Likewise.
2117 (vectorizable_assignment): Likewise.
2118 (vectorizable_shift): Likewise.
2119 (vectorizable_operation): Likewise.
2120 (vectorizable_store): Likewise.
2121 (vectorizable_load): Likewise.
2122 (vectorizable_condition): Likewise.
2123 (vectorizable_comparison): Likewise.
2125 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2128 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2129 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2130 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2132 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2135 * params.def (inline-min-speedup): Increase from 8 to 15.
2136 (max-inline-insns-auto): Decrease from 40 to 30.
2137 * ipa-split.c (consider_split): Add some buffer for function to
2138 be considered inlining candidate.
2139 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2142 2018-02-12 Richard Biener <rguenther@suse.de>
2144 PR tree-optimization/84037
2145 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2146 matched stmts if we cannot swap the non-matched ones.
2148 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2150 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2151 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2152 _mm_maskz_scalef_round_ss): New intrinsics.
2153 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2154 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2155 __builtin_ia32_scalefss_round): Remove.
2156 (__builtin_ia32_scalefsd_mask_round,
2157 __builtin_ia32_scalefss_mask_round): New intrinsics.
2158 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2159 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2160 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2161 "<round_constraint>")): Changed to ...
2162 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2163 "<round_scalar_constraint>")): ... this.
2164 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2165 %0, %1, %2<round_op3>}"): Changed to ...
2166 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2167 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2168 %2<round_scalar_mask_op3>}"): ... this.
2169 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2171 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2173 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2174 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2175 (_mm_maskz_sqrt_round_ss): New intrinsics.
2176 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2177 (__builtin_ia32_sqrtsd_mask_round)
2178 (__builtin_ia32_sqrtss_mask_round): New builtins.
2179 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2180 (__builtin_ia32_sqrtss_round): Remove.
2181 (__builtin_ia32_sqrtsd_mask_round)
2182 (__builtin_ia32_sqrtss_mask_round): New builtins.
2183 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2184 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2185 ((match_operand:VF_128 1 "vector_operand"
2186 "xBm,<round_constraint>")): Changed to ...
2187 ((match_operand:VF_128 1 "vector_operand"
2188 "xBm,<round_scalar_constraint>")): ... this.
2189 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2190 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2191 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2192 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2193 %<iptr>1<round_scalar_mask_op3>}): ... this.
2194 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2195 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2197 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2200 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2201 Cast vec_cmpeq result to correct type.
2202 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2203 Cast vec_cmpgt result to correct type.
2205 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2207 * final.c (final_scan_insn_1): Renamed from...
2208 (final_scan_insn): ... this. New wrapper, to recover
2209 seen from the outermost call in recursive ones.
2210 * config/sparc/sparc.c (output_return): Drop seen from call.
2211 (output_sibcall): Likewise.
2212 * config/visium/visium.c (output_branch): Likewise.
2214 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2216 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2219 2018-02-10 Alan Modra <amodra@gmail.com>
2222 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2223 Specify LR as an input.
2225 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2228 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2229 remove_member_access_dummy_vars): New functions.
2230 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2231 lower_omp_1, execute_lower_omp): Use them.
2233 PR rtl-optimization/84308
2234 * shrink-wrap.c (spread_components): Release todo vector.
2236 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2238 PR rtl-optimization/57193
2239 * ira-color.c (struct allocno_color_data): Add member
2240 conflict_allocno_hard_prefs.
2241 (update_conflict_allocno_hard_prefs): New.
2242 (bucket_allocno_compare_func): Add a preference based on
2243 conflict_allocno_hard_prefs.
2244 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2245 (color_allocnos): Remove a dead code. Initiate
2246 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2248 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2251 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2252 constraint from =wa to wa. Avoid a subreg on the output operand,
2253 instead use a pseudo and subreg it in a move.
2254 (p9_xxbrd_<mode>): Changed to ...
2255 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2256 (p9_xxbrd_v2df): New expander.
2257 (p9_xxbrw_<mode>): Changed to ...
2258 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2259 (p9_xxbrw_v4sf): New expander.
2261 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2263 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2265 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2268 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2269 multiply in 32-bit mode.
2270 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2271 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2274 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2276 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2277 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2278 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2279 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2281 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2284 * dwarf2out.c (is_trivial_indirect_ref): New function.
2285 (dwarf2out_late_global_decl): Do not generate a location
2286 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2287 and that are not defined in the current unit.
2289 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2291 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2292 instead of a libcall for UNORDERED.
2294 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2297 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2298 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2300 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2303 * config/s390/s390.c (s390_set_current_function): Invoke
2304 s390_indirect_branch_settings also if fndecl didn't change.
2306 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2308 * config/rs6000/rs6000.md (blockage): Set length to zero.
2310 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2312 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2314 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2317 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2318 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2322 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2323 PARALLEL incoming that failed vt_get_decl_and_offset check.
2326 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2327 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2329 (get_variable_section): For decls in named .bss* sections pass true as
2330 second argument to bss_initializer_p.
2332 2018-02-09 Marek Polacek <polacek@redhat.com>
2333 Jakub Jelinek <jakub@redhat.com>
2336 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2337 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2338 Sync some changes from cxx_fold_indirect_ref.
2340 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2342 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2344 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2345 (BLOCK_INLINE_ENTRY_LABEL): New.
2346 (dwarf2out_var_location): Disregard inline entry markers.
2347 (inline_entry_data): New struct.
2348 (inline_entry_data_hasher): New hashtable type.
2349 (inline_entry_data_hasher::hash): New.
2350 (inline_entry_data_hasher::equal): New.
2351 (inline_entry_data_table): New variable.
2352 (add_high_low_attributes): Add DW_AT_entry_pc and
2353 DW_AT_GNU_entry_view attributes if a pending entry is found
2354 in inline_entry_data_table. Add old entry_pc attribute only
2355 if debug nonbinding markers are disabled.
2356 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2357 markers are enabled.
2358 (block_within_block_p, dwarf2out_inline_entry): New.
2359 (dwarf2out_finish): Check that no entries remained in
2360 inline_entry_data_table.
2361 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2362 (final_scan_insn, notice_source_line): Likewise.
2363 (rest_of_clean_state): Skip inline entry markers.
2364 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2366 * gimple.c (gimple_build_debug_inline_entry): New.
2367 * gimple.h (enum gimple_debug_subcode): Add
2368 GIMPLE_DEBUG_INLINE_ENTRY.
2369 (gimple_build_debug_inline_entry): Declare.
2370 (gimple_debug_inline_entry_p): New.
2371 (gimple_debug_nonbind_marker_p): Adjust.
2372 * insn-notes.def (INLINE_ENTRY): New.
2373 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2374 inline entry marker notes.
2375 (print_insn): Likewise.
2376 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2377 (INSN_DEBUG_MARKER_KIND): Likewise.
2378 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2379 * tree-inline.c (expand_call_inline): Build and insert
2380 debug_inline_entry stmt.
2381 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2382 inline entry blocks early, if nonbind markers are enabled.
2383 (dump_scope_block): Dump fragment info.
2384 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2385 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2386 (gimple_build_debug_inline_entry): New.
2387 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2388 Enable/disable inline entry points too.
2389 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2390 (DEBUG_INSN): Describe inline entry markers.
2392 * common.opt (gvariable-location-views): New.
2393 (gvariable-location-views=incompat5): New.
2394 * config.in: Rebuilt.
2395 * configure: Rebuilt.
2396 * configure.ac: Test assembler for view support.
2397 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2398 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2399 * dwarf2out.c (var_loc_view): New typedef.
2400 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2401 (dwarf2out_locviews_in_attribute): New.
2402 (dwarf2out_locviews_in_loclist): New.
2403 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2404 (enum dw_line_info_opcode): Add LI_adv_address.
2405 (struct dw_line_info_table): Add view.
2406 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2407 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2408 (zero_view_p): New variable.
2409 (ZERO_VIEW_P): New macro.
2410 (output_asm_line_debug_info): New.
2411 (struct var_loc_node): Add view.
2412 (add_AT_view_list, AT_loc_list): New.
2413 (add_var_loc_to_decl): Add view param. Test it against last.
2414 (new_loc_list): Add view params. Record them.
2415 (AT_loc_list_ptr): Handle loc and view lists.
2416 (view_list_to_loc_list_val_node): New.
2417 (print_dw_val): Handle dw_val_class_view_list.
2418 (size_of_die): Likewise.
2419 (value_format): Likewise.
2420 (loc_list_has_views): New.
2421 (gen_llsym): Set vl_symbol too.
2422 (maybe_gen_llsym, skip_loc_list_entry): New.
2423 (dwarf2out_maybe_output_loclist_view_pair): New.
2424 (output_loc_list): Output view list or entries too.
2425 (output_view_list_offset): New.
2426 (output_die): Handle dw_val_class_view_list.
2427 (output_dwarf_version): New.
2428 (output_compilation_unit_header): Use it.
2429 (output_skeleton_debug_sections): Likewise.
2430 (output_rnglists, output_line_info): Likewise.
2431 (output_pubnames, output_aranges): Update version comments.
2432 (output_one_line_info_table): Output view numbers in asm comments.
2433 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2434 Call maybe_gen_llsym.
2435 (loc_list_from_tree_1): Adjust.
2436 (add_AT_location_description): Create view list attribute if
2437 needed, check it's absent otherwise.
2438 (convert_cfa_to_fb_loc_list): Adjust.
2439 (maybe_emit_file): Call output_asm_line_debug_info for test.
2440 (dwarf2out_var_location): Reset views as needed. Precompute
2441 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2442 attribute. Set view.
2443 (new_line_info_table): Reset next view.
2444 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2445 (dwarf2out_source_line): Likewise. Output view resets and labels to
2446 the assembler, or select appropriate line info opcodes.
2447 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2448 (optimize_string_length): Catch it. Adjust.
2449 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2450 dw_val_class_view_list, and remove it if no longer needed.
2451 (hash_loc_list): Hash view numbers.
2452 (loc_list_hasher::equal): Compare them.
2453 (optimize_location_lists): Check whether a view list symbol is
2454 needed, and whether the locview attribute is present, and
2455 whether they match. Remove the locview attribute if no longer
2457 (index_location_lists): Call skip_loc_list_entry for test.
2458 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2459 Use output_dwarf_version.
2460 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2461 (struct dw_val_node): Add val_view_list.
2462 * final.c (SEEN_NEXT_VIEW): New.
2463 (set_next_view_needed): New.
2464 (clear_next_view_needed): New.
2465 (maybe_output_next_view): New.
2466 (final_start_function): Rename to...
2467 (final_start_function_1): ... this. Take pointer to FIRST,
2468 add SEEN parameter. Emit param bindings in the initial view.
2469 (final_start_function): Reintroduce SEEN-less interface.
2470 (final): Rename to...
2471 (final_1): ... this. Take SEEN parameter. Output final pending
2472 next view at the end.
2473 (final): Reintroduce seen-less interface.
2474 (final_scan_insn): Output pending next view before switching
2475 sections or ending a block. Mark the next view as needed when
2476 outputting variable locations. Notify debug backend of section
2477 changes, and of location view changes.
2478 (rest_of_handle_final): Adjust.
2479 * toplev.c (process_options): Autodetect value for debug variable
2480 location views option. Warn on incompat5 without -gdwarf-5.
2481 * doc/invoke.texi (gvariable-location-views): New.
2482 (gvariable-location-views=incompat5): New.
2483 (gno-variable-location-views): New.
2485 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2487 PR tree-optimization/84136
2488 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2489 that the result of find_edge is non-NULL.
2491 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2494 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2495 storing integer register in SImode. Fix cost of 256 and 512
2496 byte aligned SSE register store.
2498 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2500 * config/i386/i386.c (ix86_multiplication_cost): Fix
2501 multiplication cost for TARGET_AVX512DQ.
2503 2018-02-08 Marek Polacek <polacek@redhat.com>
2505 PR tree-optimization/84238
2506 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2509 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2511 PR tree-optimization/84265
2512 * tree-vect-stmts.c (vectorizable_store): Don't treat
2513 VMAT_CONTIGUOUS accesses as grouped.
2514 (vectorizable_load): Likewise.
2516 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2518 PR tree-optimization/81635
2519 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2520 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2521 (test_round_for_mask): New functions.
2522 (wide_int_cc_tests): Call test_round_for_mask.
2523 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2524 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2525 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2526 range returned by get_range_info.
2528 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2531 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2532 * symtab.c: Include builtins.h
2533 (symtab_node::output_to_lto_symbol_table_p): Move here
2534 from lto-streamer-out.c:output_symbol_p.
2535 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2536 (output_symbol_p): Move all logic to symtab.c
2537 (produce_symtab): Update.
2539 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2541 * config/s390/s390-opts.h (enum indirect_branch): Define.
2542 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2543 (s390_indirect_branch_via_thunk)
2544 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2545 (enum s390_indirect_branch_type): Define.
2546 * config/s390/s390.c (struct s390_frame_layout, struct
2547 machine_function): Remove.
2548 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2549 (indirect_branch_table_label_no, indirect_branch_table_name):
2551 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2552 (enum s390_indirect_branch_option): Define.
2553 (s390_return_addr_from_memory): New function.
2554 (s390_handle_string_attribute): New function.
2555 (s390_attribute_table): Add new attribute handler.
2556 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2557 (s390_indirect_branch_via_thunk): New function.
2558 (s390_indirect_branch_via_inline_thunk): New function.
2559 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2560 sibling call optimization for non z10 compiles.
2561 (s390_emit_call): Force indirect branch target to be a single
2562 register. Add r1 clobber for non-z10 compiles.
2563 (s390_emit_epilogue): Emit return jump via return_use expander.
2564 (s390_reorg): Handle JUMP_INSNs as execute targets.
2565 (s390_option_override_internal): Perform validity checks for the
2566 new command line options.
2567 (s390_indirect_branch_attrvalue): New function.
2568 (s390_indirect_branch_settings): New function.
2569 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2570 (s390_output_indirect_thunk_function): New function.
2571 (s390_code_end): Implement target hook.
2572 (s390_case_values_threshold): Implement target hook.
2573 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2575 * config/s390/s390.h (struct s390_frame_layout)
2576 (struct machine_function): Move here from s390.c.
2577 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2578 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2579 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2580 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2581 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2582 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2583 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2584 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2585 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2586 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2587 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2588 (mnemonic attribute): Add values which aren't recognized
2590 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2591 pattern for branch conversion. Fix mnemonic attribute.
2592 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2593 indirect branch via thunk if requested.
2594 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2595 ("*indirect_jump"): Disable for branch conversion using out of
2597 ("indirect_jump_via_thunk<mode>_z10")
2598 ("indirect_jump_via_thunk<mode>")
2599 ("indirect_jump_via_inlinethunk<mode>_z10")
2600 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2601 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2602 ("casesi_jump_via_inlinethunk<mode>_z10")
2603 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2604 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2605 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2606 ("*indirect2_jump"): Disable for branch conversion.
2607 ("casesi_jump"): Turn into expander and expand patterns for branch
2609 ("return_use"): New expander.
2610 ("*return"): Emit return via thunk and rename it to ...
2611 ("*return<mode>"): ... this one.
2612 * config/s390/s390.opt: Add new options and and enum for the
2615 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2617 * lra-constraints.c (match_reload): Unconditionally use
2618 gen_lowpart_SUBREG, rather than selecting between that
2619 and equivalent gen_rtx_SUBREG code.
2621 2018-02-08 Richard Biener <rguenther@suse.de>
2623 PR tree-optimization/84233
2624 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2625 changed flag instead of boguously re-using phi_inserted.
2627 2018-02-08 Martin Jambor <mjambor@suse.cz>
2629 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2630 static local variables.
2632 2018-02-08 Richard Biener <rguenther@suse.de>
2634 PR tree-optimization/84278
2635 * tree-vect-stmts.c (vectorizable_store): When looking for
2636 smaller vector types to perform grouped strided loads/stores
2637 make sure the mode is supported by the target.
2638 (vectorizable_load): Likewise.
2640 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2642 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2643 Increase LDP/STP opportunities by adding adjacent callee-saves.
2645 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2647 PR rtl-optimization/84068
2648 PR rtl-optimization/83459
2649 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2651 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2653 PR tree-optimization/84224
2654 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2655 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2658 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2661 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2662 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2663 count, remove one USE.
2665 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2667 * doc/install.texi (Configuration): Document the
2668 --with-long-double-format={ibm,ieee} PowerPC configuration
2672 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2673 Convert from define_expand to be define_insn_and_split. Rework
2674 float/double/_Float128 conversions to QI/HI/SImode to work with
2675 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2676 conversions to QI/HImode types did a store and then a load to
2677 truncate the value. For conversions to VSX registers, don't split
2678 the insn, instead emit the code directly. Use the code iterator
2679 any_fix to combine signed and unsigned conversions.
2680 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2681 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2682 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2683 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2684 (fix_<mode>di2_hw): Likewise.
2685 (fixuns_<mode>di2_hw): Likewise.
2686 (fix_<mode>si2_hw): Likewise.
2687 (fixuns_<mode>si2_hw): Likewise.
2688 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2689 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2690 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2691 fix<uns>_trunc<SFDF:mode>si2_p8.
2692 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2694 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2695 (fix<uns>_<mode>_mem): Likewise.
2696 (fctiw<u>z_<mode>_mem): Likewise.
2697 (fix<uns>_<mode>_mem): Likewise.
2698 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2699 the register allocator from doing a direct move to the GPRs to do
2700 a store, and instead use the ISA 3.0 store byte/half-word from
2701 vector register instruction. For IEEE 128-bit floating point,
2702 also optimize stores of 32-bit ints.
2703 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2705 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2707 * genextract.c (push_pathstr_operand): New function to support
2709 (walk_rtx): Call push_pathstr_operand.
2710 (print_path): Support [a-zA-Z].
2712 2018-02-07 Richard Biener <rguenther@suse.de>
2714 PR tree-optimization/84037
2715 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2716 (cse_and_gimplify_to_preheader): Declare.
2717 (vect_get_place_in_interleaving_chain): Likewise.
2718 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2720 (_loop_vec_info::~_loop_vec_info): Delete it.
2721 (cse_and_gimplify_to_preheader): New function.
2722 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2723 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2724 (vectorizable_load): Likewise. For grouped stores always base
2725 the IV on the first element.
2726 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2727 condition before gimplifying.
2729 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2731 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2732 *DIV_EXPR and *MOD_EXPR.
2734 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2737 * config/i386/i386.c (ix86_option_override_internal): Mask out
2738 the CF_SET bit when checking -fcf-protection.
2740 2018-02-07 Tom de Vries <tom@codesourcery.com>
2743 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2746 2018-02-07 Richard Biener <rguenther@suse.de>
2748 PR tree-optimization/84204
2749 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2752 PR tree-optimization/84205
2753 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2754 special-case isl_ast_op_zdiv_r.
2756 PR tree-optimization/84223
2757 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2758 Only add conditions from within the region.
2759 (gather_bbs::after_dom_children): Adjust.
2761 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2764 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2765 * config/avr/avr.md: Only post-reload split REG-REG moves if
2766 either register is GENERAL_REG_P.
2768 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2770 PR tree-optimization/84235
2771 * tree-ssa-scopedtables.c
2772 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2773 if the subtraction is performed in floating point type where NaNs are
2774 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2775 build 1. Formatting fix.
2777 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2780 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2781 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2782 and skip it regardless of bb boundaries. Use CALL_P macro,
2783 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2784 unnecessarily, formatting fix.
2786 2018-02-06 Michael Collison <michael.collison@arm.com>
2788 * config/arm/thumb2.md:
2789 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2790 (*thumb_mov_notscc): Ditto.
2792 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2795 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2796 unsigned_fix, not "s".
2798 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2800 * configure.ac (gcc_fn_eh_frame_ro): New function.
2801 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2802 correct .eh_frame permissions.
2803 * configure: Regenerate.
2805 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2807 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2810 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2812 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2813 Display warning message for -mno-speculate-indirect-jumps.
2815 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2817 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2819 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2821 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2823 PR tree-optimization/84225
2824 * tree-eh.c (find_trapping_overflow): Only call
2825 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2827 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2830 * config/i386/i386.c: Reimplement the check of possible options
2831 -mibt/-mshstk conbination. Change error messages.
2832 * doc/invoke.texi: Fix a typo: remove extra '='.
2834 2018-02-06 Marek Polacek <polacek@redhat.com>
2836 PR tree-optimization/84228
2837 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2839 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2842 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2843 emitted arch directives.
2844 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2845 __ARM_FEATURE_COPROC before changing architectures.
2847 2018-02-06 Richard Biener <rguenther@suse.de>
2849 * config/i386/i386.c (print_reg): Fix typo.
2850 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2852 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2854 * configure: Regenerate.
2856 2018-02-05 Martin Sebor <msebor@redhat.com>
2858 PR tree-optimization/83369
2859 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2862 2018-02-05 Martin Liska <mliska@suse.cz>
2864 * doc/invoke.texi: Cherry-pick upstream r323995.
2866 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2868 * ira.c (ira_init_register_move_cost): Adjust comment.
2870 2018-02-05 Martin Liska <mliska@suse.cz>
2872 PR gcov-profile/84137
2873 * doc/gcov.texi: Fix typo in documentation.
2875 2018-02-05 Martin Liska <mliska@suse.cz>
2877 PR gcov-profile/83879
2878 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2879 using dlopen functionality.
2881 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2883 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2884 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2885 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2886 _mm_maskz_range_round_ss): New intrinsics.
2887 (__builtin_ia32_rangesd128_round)
2888 (__builtin_ia32_rangess128_round): Remove.
2889 (__builtin_ia32_rangesd128_mask_round,
2890 __builtin_ia32_rangess128_mask_round): New builtins.
2891 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2892 __builtin_ia32_rangess128_round): Remove.
2893 (__builtin_ia32_rangesd128_mask_round,
2894 __builtin_ia32_rangess128_mask_round): New builtins.
2895 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2896 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2897 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2898 "<round_saeonly_constraint>")): Changed to ...
2899 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2900 "<round_saeonly_scalar_constraint>")): ... this.
2901 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2902 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2903 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2904 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2905 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2907 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2909 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2911 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2912 Remove all values except native, 8540 and 8548.
2914 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2916 * config/i386/i386.c (ix86_output_function_return): Pass
2917 INVALID_REGNUM, instead of -1, as invalid register number to
2918 indirect_thunk_name and output_indirect_thunk.
2920 2018-02-02 Julia Koval <julia.koval@intel.com>
2922 * config.gcc: Add -march=icelake.
2923 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2924 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2925 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2926 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2927 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2928 (processor_target_table): Add icelake.
2929 (ix86_option_override_internal): Handle new PTAs.
2930 (get_builtin_code_for_version): Handle icelake.
2931 (M_INTEL_COREI7_ICELAKE): New.
2932 (fold_builtin_cpu): Handle icelake.
2933 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2934 * doc/invoke.texi: Add -march=icelake.
2936 2018-02-02 Julia Koval <julia.koval@intel.com>
2938 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2939 to wide_int_bitmask.
2940 * wide-int-bitmask.h: New.
2942 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2945 * config/i386/i386.md: Replace Pmode with word_mode in
2946 builtin_setjmp_setup and builtin_longjmp to support x32.
2948 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2952 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2954 (rs6000_supported_cpu_names): New static variable.
2955 (linux_cpu_translation_table): Likewise.
2956 (elf_platform) <cpu>: Define new static variable and use it.
2957 Translate kernel AT_PLATFORM name to canonical name if needed.
2958 Error if platform name is unknown.
2960 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2963 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2965 2018-02-01 Jeff Law <law@redhat.com>
2968 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2969 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2970 the scratch if RELEASE_VIA_POP is false.
2971 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2972 If we have to save a temporary register, decrement SIZE appropriately.
2973 Pass new arguments to release_scratch_register_on_entry.
2974 (ix86_adjust_stack_and_probe): Likewise.
2975 (ix86_emit_probe_stack_range): Pass new arguments to
2976 release_scratch_register_on_entry.
2978 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2980 PR rtl-optimization/84157
2981 * combine.c (change_zero_ext): Use REG_P predicate in
2982 front of HARD_REGISTER_P predicate.
2984 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2986 * config/avr/avr.c (avr_option_override): Move disabling of
2987 -fdelete-null-pointer-checks to...
2988 * common/config/avr/avr-common.c (avr_option_optimization_table):
2991 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2993 PR tree-optimization/81635
2994 * tree-data-ref.c (split_constant_offset_1): For types that
2995 wrap on overflow, try to use range info to prove that wrapping
2998 2018-02-01 Renlin Li <renlin.li@arm.com>
3001 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3003 (aarch64_register_move_cost): Likewise.
3004 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3006 (REG_CLASS_NAMES): Likewise.
3007 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3008 TAILCALL_ADDR_REGS. Remove IP registers.
3009 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3011 2018-02-01 Richard Biener <rguenther@suse.de>
3013 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3014 for specifying RPO order and allow NULL for that.
3015 * domwalk.c (dom_walker::dom_walker): Likewise.
3016 (dom_walker::walk): Handle NULL RPO order.
3017 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3019 (rewrite_update_dom_walker): Likewise.
3020 (mark_def_dom_walker): Likewise.
3022 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3024 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3025 (aarch64_maybe_expand_sve_subreg_move): Declare.
3026 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3027 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3029 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3030 that are semantically a reverse operation.
3031 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3032 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3033 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3035 (aarch64_can_change_mode_class): For big-endian, forbid changes
3036 between two SVE modes if they have different element sizes.
3038 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3040 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3041 the TImode handling for big-endian targets.
3043 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3045 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3046 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3048 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3049 Remove BSWAP handing for big-endian targets and use the form of
3050 LD1RQ appropariate for the mode.
3052 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3054 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3055 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3058 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3061 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3062 check for operands that need to go through aarch64_sve_reload_be.
3064 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3066 PR tree-optimization/81661
3067 PR tree-optimization/84117
3068 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3069 * tree-eh.c: Include gimplify.h.
3070 (find_trapping_overflow, replace_trapping_overflow,
3071 rewrite_to_non_trapping_overflow): New functions.
3072 * tree-vect-loop.c: Include tree-eh.h.
3073 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3074 * tree-data-ref.c: Include tree-eh.h.
3075 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3077 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3079 PR rtl-optimization/84123
3080 * combine.c (change_zero_ext): Check if hard register satisfies
3081 can_change_dest_mode before calling gen_lowpart_SUBREG.
3083 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3086 * ira.c (ira_init_register_move_cost): Remove assert.
3088 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3090 PR rtl-optimization/84071
3091 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3092 * doc/tm.texi: Regenerate.
3094 2018-01-31 Richard Biener <rguenther@suse.de>
3096 PR tree-optimization/84132
3097 * tree-data-ref.c (analyze_miv_subscript): Properly
3098 check whether evolution_function_is_affine_multivariate_p
3099 before calling gcd_of_steps_may_divide_p.
3101 2018-01-31 Julia Koval <julia.koval@intel.com>
3104 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3105 * config/i386/i386.md (rdpid_rex64) New.
3106 (rdpid): Make 32bit only.
3108 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3111 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3112 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3114 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3117 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3119 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3121 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3123 PR rtl-optimization/84071
3124 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3125 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3127 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3129 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3130 (arc_attribute_table): Add 'aux' attribute.
3131 (arc_in_small_data_p): Consider aux like variables.
3132 (arc_is_aux_reg_p): New function.
3133 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3134 (arc_get_aux_arg): New function.
3135 (prepare_move_operands): Handle aux-register access.
3136 (arc_handle_aux_attribute): New function.
3137 * doc/extend.texi (ARC Variable attributes): Add subsection.
3139 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3141 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3142 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3143 (arc_attribute_table): Add 'uncached' attribute.
3144 (arc_print_operand): Print '.di' flag for uncached memory
3146 (arc_in_small_data_p): Do not consider for small data the uncached
3148 (arc_is_uncached_mem_p): New function.
3149 * config/arc/predicates.md (compact_store_memory_operand): Check
3150 for uncached memory accesses.
3151 (nonvol_nonimm_operand): Likewise.
3152 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3154 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3157 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3158 falign-loops=): Add Optimization flag.
3160 2018-01-30 Jeff Law <law@redhat.com>
3163 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3164 INT_REGISTERS_SAVED. Check it prior to calling
3165 get_scratch_register_on_entry.
3166 (ix86_adjust_stack_and_probe): Similarly.
3167 (ix86_emit_probe_stack_range): Similarly.
3168 (ix86_expand_prologue): Corresponding changes.
3170 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3173 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3174 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3176 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3179 * lra-constraints.c (curr_insn_transform): Process AND in the
3182 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3184 PR rtl-optimization/83986
3185 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3186 dependence against last_pending_memory_flush in addition to
3189 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3191 PR tree-optimization/81611
3192 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3195 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3198 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3201 2018-01-30 Richard Biener <rguenther@suse.de>
3202 Jakub Jelinek <jakub@redhat.com>
3204 PR tree-optimization/84111
3205 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3206 inner loops added during recursion, as they don't have up-to-date
3209 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3212 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3213 (can_inline_edge_by_limits_p): ... here.
3214 (can_early_inline_edge_p, check_callers,
3215 update_caller_keys, update_callee_keys, recursive_inlining,
3216 add_new_edges_to_heap, speculation_useful_p,
3217 inline_small_functions,
3218 inline_small_functions, flatten_function,
3219 inline_to_all_callers_1): Update.
3221 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3223 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3226 2018-01-30 Richard Biener <rguenther@suse.de>
3228 PR tree-optimization/83008
3229 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3230 invariant and constant vector uses in stmts when they need
3233 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3236 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3237 * configure: Regenerate.
3239 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3241 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3243 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3244 Use gen_rtx_REG rather than gen_lowpart.
3246 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3248 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3249 rather than 0 when creating partial subregs.
3251 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3253 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3256 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3259 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3260 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3261 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3262 flags. This restores the settings used before the 2017-07-24.
3263 Turning off pre increment/decrement/modify allows IVOPTS to
3264 optimize DF/SF loops where the index is an int.
3266 2018-01-29 Richard Biener <rguenther@suse.de>
3267 Kelvin Nilsen <kelvin@gcc.gnu.org>
3270 * tree-vect-stmts.c (vectorizable_call): Don't call
3271 targetm.vectorize_builtin_md_vectorized_function if callee is
3274 2018-01-22 Carl Love <cel@us.ibm.com>
3276 * doc/extend.tex: Fix typo in second arg in
3277 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3279 2018-01-29 Richard Biener <rguenther@suse.de>
3281 PR tree-optimization/84086
3282 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3283 (flush_ssaname_freelist): When SSA names were released reset
3284 the SCEV hash table.
3286 2018-01-29 Richard Biener <rguenther@suse.de>
3288 PR tree-optimization/84057
3289 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3290 removed paths when removing edges.
3292 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3294 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3295 -mfunction-return=@var{choice}.
3297 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3300 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3302 (layout::print_source_line): Likewise.
3303 (test_get_line_width_without_trailing_whitespace): Add test cases.
3305 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3308 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3311 2018-01-26 Jim Wilson <jimw@sifive.com>
3313 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3315 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3318 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3320 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3321 and CMP + SUB-immediate -> SUBS.
3323 2018-01-26 Martin Sebor <msebor@redhat.com>
3325 PR tree-optimization/83896
3326 * tree-ssa-strlen.c (get_string_len): Rename...
3327 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3328 Avoid assuming length is constant.
3329 (handle_char_store): Use HOST_WIDE_INT for string length.
3331 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3334 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3335 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3337 2018-01-26 Richard Biener <rguenther@suse.de>
3339 PR rtl-optimization/84003
3340 * dse.c (record_store): Only record redundant stores when
3341 the earlier store aliases at least all accesses the later one does.
3343 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3345 PR rtl-optimization/83985
3346 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3347 REG_CFA_RESTORE insns.
3348 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3349 REG_CFA_RESTORE insns here.
3352 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3353 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3355 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3357 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3358 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3359 (arc_init): Likewise.
3360 (arc_override_options): Likewise.
3361 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3363 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3365 * config/arc/arc.h (TARGET_DBNZ): Define.
3366 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3367 properly set the tune attribute.
3368 (dbnz): Use TARGET_DBNZ guard.
3369 * config/arc/arc.opt (mtune): Add core3 option.
3371 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3373 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3374 recognize new pic like addresses.
3375 (arc_delegitimize_address): Clean up.
3377 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3379 * config/arc/arc-arches.def: Option mrf16 valid for all
3381 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3382 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3383 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3384 * config/arc/arc-tables.opt: Regenerate.
3385 * config/arc/arc.c (arc_conditional_register_usage): Handle
3386 reduced register file case.
3387 (arc_file_start): Set must have build attributes.
3388 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3390 * config/arc/arc.opt (mrf16): Add new option.
3391 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3392 * config/arc/genmultilib.awk: Handle new mrf16 option.
3393 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3394 * config/arc/t-multilib: Regenerate.
3395 * doc/invoke.texi (ARC Options): Document mrf16 option.
3397 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3399 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3400 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3401 (arc_attribute_table): Add 'secure_call' attribute.
3402 (arc_print_operand): Print secure call operand.
3403 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3405 (arc_is_secure_call_p): New function. * config/arc/arc.md
3406 (call_i): Add support for sjli instruction.
3407 (call_value_i): Likewise.
3408 * config/arc/constraints.md (Csc): New constraint.
3410 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3411 John Eric Martin <John.Martin@emmicro-us.com>
3413 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3414 * config/arc/arc.c (_arc_jli_section): New struct.
3415 (arc_jli_section): New type.
3416 (rc_jli_sections): New static variable.
3417 (arc_handle_jli_attribute): New function.
3418 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3419 (arc_file_end): New function.
3420 (TARGET_ASM_FILE_END): Define.
3421 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3422 (arc_add_jli_section): New function.
3423 (jli_call_scan): Likewise.
3424 (arc_reorg): Call jli_call_scan.
3425 (arc_output_addsi): Remove 'S' from printing asm operand.
3426 (arc_is_jli_call_p): New function.
3427 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3429 (movhi_insn): Likewise.
3430 (movsi_insn): Likewise.
3431 (movsi_set_cc_insn): Likewise.
3432 (loadqi_update): Likewise.
3433 (load_zeroextendqisi_update): Likewise.
3434 (load_signextendqisi_update): Likewise.
3435 (loadhi_update): Likewise.
3436 (load_zeroextendhisi_update): Likewise.
3437 (load_signextendhisi_update): Likewise.
3438 (loadsi_update): Likewise.
3439 (loadsf_update): Likewise.
3440 (movsicc_insn): Likewise.
3441 (bset_insn): Likewise.
3442 (bxor_insn): Likewise.
3443 (bclr_insn): Likewise.
3444 (bmsk_insn): Likewise.
3445 (bicsi3_insn): Likewise.
3446 (cmpsi_cc_c_insn): Likewise.
3447 (movsi_ne): Likewise.
3448 (movsi_cond_exec): Likewise.
3449 (clrsbsi2): Likewise.
3460 (call_i): Remove 'S' asm letter, add jli instruction.
3461 (call_value_i): Likewise.
3462 * config/arc/arc.op (mjli-always): New option.
3463 * config/arc/constraints.md (Cji): New constraint.
3464 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3466 (subsf3_fpx): Likewise.
3467 (mulsf3_fpx): Likewise.
3468 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3470 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3472 * doc/invoke.texi (ARC): Document mjli-always option.
3474 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3476 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3477 avoid addition with 0 and use incw and decw where possible.
3479 2018-01-26 Richard Biener <rguenther@suse.de>
3481 PR tree-optimization/81082
3482 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3483 association if it requires casting to unsigned.
3484 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3485 from fold_plusminus_mult_expr to catch important cases late when
3486 range info is available.
3488 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3490 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3491 * configure.ac (hidden_linkonce): New test.
3492 * configure: Regenerate.
3493 * config.in: Regenerate.
3495 2018-01-26 Julia Koval <julia.koval@intel.com>
3497 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3498 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3499 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3500 _mm_mask_bitshuffle_epi64_mask): Fix type.
3501 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3502 USI_FTYPE_V4DI_V4DI_USI): Remove.
3503 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3504 __builtin_ia32_vpshufbitqmb256_mask,
3505 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3506 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3507 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3509 2018-01-26 Alan Modra <amodra@gmail.com>
3512 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3513 UNSPEC_VBPERMQ. Sort other unspecs.
3515 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3517 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3519 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3522 * predict.c (drop_profile): Do not push/pop cfun; update also
3524 (handle_missing_profiles): Fix logic looking for zero profiles.
3526 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3529 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3530 on functions with #pragma omp declare simd or functions with simd
3532 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3533 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3534 Remove trailing \n from warning_at calls.
3536 2018-01-25 Tom de Vries <tom@codesourcery.com>
3539 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3540 for neutered workers.
3542 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3545 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3546 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3548 2018-01-24 Jeff Law <law@redhat.com>
3551 * i386.c (get_probe_interval): Move to earlier point.
3552 (ix86_compute_frame_layout): If -fstack-clash-protection and
3553 the frame is larger than the probe interval, then use pushes
3554 to save registers rather than reg->mem moves.
3555 (ix86_expand_prologue): Remove conditional for int_registers_saved
3558 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3561 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3562 min/max for never referenced object.
3564 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3567 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3569 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3570 attributes from DECL_ATTRIBUTES (decl) without affecting
3571 DECL_ATTRIBUTES (current_function_decl).
3572 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3573 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3575 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3577 PR tree-optimization/83979
3578 * fold-const.c (fold_comparison): Use constant_boolean_node
3579 instead of boolean_{true,false}_node.
3581 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3583 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3586 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3588 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3589 Simplify the clause that sets the length attribute.
3590 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3591 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3592 clause that sets the length attribute.
3593 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3595 2018-01-24 Tom de Vries <tom@codesourcery.com>
3598 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3599 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3600 Add strict parameter.
3601 (prevent_branch_around_nothing): Insert dummy insn between branch to
3602 label and label with no ptx insn inbetween.
3603 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3605 2018-01-24 Tom de Vries <tom@codesourcery.com>
3608 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3609 for neutered threads in warp.
3610 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3612 2018-01-24 Richard Biener <rguenther@suse.de>
3614 PR tree-optimization/83176
3615 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3618 2018-01-24 Richard Biener <rguenther@suse.de>
3620 PR tree-optimization/82819
3621 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3622 code generating pluses that are no-ops in the target precision.
3624 2018-01-24 Richard Biener <rguenther@suse.de>
3627 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3629 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3631 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3632 to merge probabilities.
3633 * predict.c (probably_never_executed): Also mark as cold functions
3634 with global 0 profile and guessed local profile.
3635 * profile-count.c (profile_probability::combine_with_count): New
3637 * profile-count.h (profile_probability::operator*,
3638 profile_probability::operator*=, profile_probability::operator/,
3639 profile_probability::operator/=): Reduce precision to adjusted
3640 and set value to guessed on contradictory divisions.
3641 (profile_probability::combine_with_freq): Remove.
3642 (profile_probability::combine_wiht_count): Declare.
3643 (profile_count::force_nonzero):: Set to adjusted.
3644 (profile_count::probability_in):: Set quality to adjusted.
3645 * tree-ssa-tail-merge.c (replace_block_by): Use
3648 2018-01-23 Andrew Waterman <andrew@sifive.com>
3649 Jim Wilson <jimw@sifive.com>
3651 * config/riscv/riscv.c (riscv_stack_boundary): New.
3652 (riscv_option_override): Set riscv_stack_boundary. Handle
3653 riscv_preferred_stack_boundary_arg.
3654 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3655 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3656 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3657 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3658 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3659 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3661 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3664 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3665 of struct ix86_frame.
3666 (ix86_expand_epilogue): Likewise. Add a local variable for
3667 the reg_save_offset field in struct ix86_frame.
3669 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3671 PR tree-optimization/82604
3672 * tree-loop-distribution.c (enum partition_kind): New enum item
3673 PKIND_PARTIAL_MEMSET.
3674 (partition_builtin_p): Support above new enum item.
3675 (generate_code_for_partition): Ditto.
3676 (compute_access_range): Differentiate cases that equality can be
3677 proven at all loops, the innermost loops or no loops.
3678 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3679 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3680 (finalize_partitions, distribute_loop): Don't fuse partition of
3681 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3682 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3685 2018-01-23 Martin Liska <mliska@suse.cz>
3687 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3688 order to ignore the predictor.
3689 (PRED_POLYMORPHIC_CALL): Likewise.
3690 (PRED_RECURSIVE_CALL): Likewise.
3692 2018-01-23 Martin Liska <mliska@suse.cz>
3694 * tree-profile.c (tree_profiling): Print function header to
3695 aware reader which function we are working on.
3696 * value-prof.c (gimple_find_values_to_profile): Do not print
3697 not interesting value histograms.
3699 2018-01-23 Martin Liska <mliska@suse.cz>
3701 * profile-count.h (enum profile_quality): Add
3702 profile_uninitialized as the first value. Do not number values
3703 as they are zero based.
3704 (profile_count::verify): Update sanity check.
3705 (profile_probability::verify): Likewise.
3707 2018-01-23 Nathan Sidwell <nathan@acm.org>
3709 * doc/invoke.texi (ffor-scope): Deprecate.
3711 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3713 PR tree-optimization/83510
3714 * domwalk.c (set_all_edges_as_executable): New function.
3715 (dom_walker::dom_walker): Convert bool param
3716 "skip_unreachable_blocks" to enum reachability. Move setup of
3717 edge flags to set_all_edges_as_executable and only do it when
3718 reachability is REACHABLE_BLOCKS.
3719 * domwalk.h (enum dom_walker::reachability): New enum.
3720 (dom_walker::dom_walker): Convert bool param
3721 "skip_unreachable_blocks" to enum reachability.
3722 (set_all_edges_as_executable): New decl.
3723 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3724 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3726 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3727 but converting true to REACHABLE_BLOCKS.
3728 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3730 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3731 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3732 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3734 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3735 if check_all_array_refs will be called.
3737 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3739 * tree.c (selftest::test_location_wrappers): Add more test
3742 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3744 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3745 (selftest::test_bit_in_range): Likewise.
3747 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3750 * doc/sourcebuild.texi (vect_float): Say that the selector
3751 only describes the situation when -funsafe-math-optimizations is on.
3752 (vect_float_strict): Document.
3754 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3756 PR tree-optimization/83965
3757 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3758 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3759 instead of checking only for a reduction.
3760 (vect_recog_widen_sum_pattern): Likewise.
3762 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3764 * predict.c (probably_never_executed): Only use precise profile info.
3765 (compute_function_frequency): Skip after inlining hack since we now
3766 have quality checking.
3768 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3770 * profile-count.h (profile_probability::very_unlikely,
3771 profile_probability::unlikely, profile_probability::even): Set
3772 precision to guessed.
3774 2018-01-23 Richard Biener <rguenther@suse.de>
3776 PR tree-optimization/83963
3777 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3778 Properly terminate dominator walk when crossing the exit edge not
3779 when visiting its source block.
3781 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3784 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3785 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3787 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3789 PR tree-optimization/83957
3790 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3791 semicolon after for body surrounded by braces.
3793 PR tree-optimization/83081
3794 * profile-count.h (profile_probability::split): New method.
3795 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3796 Use profile_probability::split.
3797 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3798 when splitting a single conditional jump into 2.
3800 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3802 PR tree-optimization/69452
3803 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3806 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3808 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3809 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3810 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3812 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3814 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3815 * config/rl78/rl78.md (movdi): New define_expand.
3816 * config/rl78/rl78.c (rl78_split_movdi): New function.
3818 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3821 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3823 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3824 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3825 128-bit to produce an UNSPEC move to get the double word with the
3826 signbit and then a shift directly to do signbit.
3827 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3828 implementation with a new version that just does either a direct
3829 move or a regular move. Move memory interface to separate insns.
3830 Move insns so they are next to the expander.
3831 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3832 with signbit move. Split big and little endian case.
3833 (signbit<mode>2_dm_mem_le): Likewise.
3834 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3835 (signbit<mode>2_dm2): Likewise.
3837 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3839 * config/rl78/rl78.md (anddi3): New define_expand.
3841 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3843 * config/rl78/rl78.md (umindi3): New define_expand.
3845 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3847 * config/rl78/rl78.md (smindi3): New define_expand.
3849 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3851 * config/rl78/rl78.md (smaxdi3): New define_expand.
3853 2018-01-22 Carl Love <cel@us.ibm.com>
3855 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3856 LVX_V1TI): Add macro expansion.
3857 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3858 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3859 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3860 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3861 Change check to determine if the instruction is a byte reversing
3862 entry. Fix typo in comment.
3863 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3864 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3865 Add def_builtin calls for new builtins.
3866 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3867 Add define_insn expansion.
3869 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3871 * config/rl78/rl78.md (umaxdi3): New define_expand.
3873 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3875 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3876 for non-QImode registers.
3878 2018-01-22 Richard Biener <rguenther@suse.de>
3880 PR tree-optimization/83963
3881 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3882 including the loop exit block.
3883 (scop_detection::merge_sese): Likewise.
3884 (scop_detection::add_scop): Do it here instead.
3886 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3888 * doc/sourcebuild.texi (arm_softfloat): Document.
3890 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3893 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3894 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3895 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3897 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3898 David Edelsohn <dje.gcc@gmail.com>
3901 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3902 Change "crset eq" to "crset 2".
3903 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3904 (*call_indirect_aix<mode>_nospec): Likewise.
3905 (*call_value_indirect_aix<mode>_nospec): Likewise.
3906 (*call_indirect_elfv2<mode>_nospec): Likewise.
3907 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3908 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3909 change assembly output from . to $.
3910 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3911 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3912 (*tablejump<mode>_internal1_nospec): Likewise.
3914 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3917 * config/sh/sh_optimize_sett_clrt.cc:
3918 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3920 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3922 PR tree-optimization/83940
3923 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3924 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3925 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3926 use it to pass back the definition type.
3927 (vect_check_store_rhs): Likewise rhs_dt_out.
3928 (vect_build_gather_load_calls): Add a mask_dt argument and use
3929 it instead of a call to vect_is_simple_use.
3930 (vectorizable_store): Update calls to vect_check_load_store_mask
3931 and vect_check_store_rhs. Use the dt returned by the latter instead
3932 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3933 instead of calls to vect_is_simple_use. Pass the scalar rather
3934 than the vector operand to vect_is_simple_use when handling
3935 second and subsequent copies of an rhs value.
3936 (vectorizable_load): Update calls to vect_check_load_store_mask
3937 and vect_build_gather_load_calls. Use the cached mask_dt and
3938 gs_info.offset_dt instead of calls to vect_is_simple_use.
3940 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3943 * tree-emutls.c: Include gimplify.h.
3944 (lower_emutls_2): New function.
3945 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3946 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3947 it before further processing.
3950 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3951 UINTVAL (trueop1) instead of INTVAL (op1).
3953 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3957 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3958 INCOMING_FRAME_SP_OFFSET if not defined.
3959 (scan_trace): Add ENTRY argument. If true and
3960 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3961 emit a note to adjust the CFA offset.
3962 (create_cfi_notes): Adjust scan_trace callers.
3963 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3964 INCOMING_FRAME_SP_OFFSET in the CIE.
3965 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3966 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3968 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3969 * doc/tm.texi: Regenerated.
3971 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3973 PR rtl-optimization/83147
3974 * lra-constraints.c (remove_inheritance_pseudos): Use
3975 lra_substitute_pseudo_within_insn.
3977 2018-01-19 Tom de Vries <tom@codesourcery.com>
3978 Cesar Philippidis <cesar@codesourcery.com>
3981 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3983 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3986 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3987 spaces for function labels.
3989 2018-01-19 Martin Liska <mliska@suse.cz>
3991 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3992 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3993 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3994 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3995 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3996 (PRED_CONST_RETURN): Change from 69 to 65.
3997 (PRED_NULL_RETURN): Change from 91 to 71.
3998 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3999 (PRED_LOOP_GUARD): Change from 66 to 73.
4001 2018-01-19 Martin Liska <mliska@suse.cz>
4003 * predict.c (predict_insn_def): Add new assert.
4004 (struct branch_predictor): Change type to signed integer.
4005 (test_prediction_value_range): Amend test to cover
4007 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4008 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4009 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4010 (PRED_LOOP_IV_COMPARE): Likewise.
4011 * predict.h (PROB_UNINITIALIZED): Define new constant.
4013 2018-01-19 Martin Liska <mliska@suse.cz>
4015 * predict.c (dump_prediction): Add new format for
4016 analyze_brprob.py script which is enabled with -details
4018 * profile-count.h (precise_p): New function.
4020 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4022 PR tree-optimization/83922
4023 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4024 there are no statements that need masking.
4025 (vect_active_double_reduction_p): New function.
4026 (vect_analyze_loop_operations): Use it when handling phis that
4027 are not in the loop header.
4029 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4031 PR tree-optimization/83914
4032 * tree-vect-loop.c (vectorizable_induction): Don't convert
4033 init_expr or apply the peeling adjustment for inductions
4034 that are nested within the vectorized loop.
4036 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4038 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4041 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4045 * function.h (gimplify_parameters): Add gimple_seq * argument.
4046 * function.c: Include gimple.h and options.h.
4047 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4048 for the added local temporaries if needed.
4049 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4050 if there are any parameter cleanups, wrap whole body into a
4051 try/finally with the cleanups.
4053 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4056 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4057 Use GET_MODE_CLASS for scalar floating point.
4059 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4063 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4064 Fix call of call_cgraph_insertion_hooks.
4066 2018-01-18 Martin Sebor <msebor@redhat.com>
4068 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4070 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4073 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4076 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4079 * common.opt: (-ffile-prefix-map): New option.
4080 * opts.c (common_handle_option): Defer it.
4081 * opts-global.c (handle_common_deferred_options): Handle it.
4082 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4083 * file-prefix-map.h: New file.
4084 (remap_debug_filename, add_debug_prefix_map): ...here.
4085 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4086 * final.c (debug_prefix_map, add_debug_prefix_map
4087 remap_debug_filename): Move to...
4088 * file-prefix-map.c: New file.
4089 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4090 generalize, get rid of alloca(), use strrchr() instead of strchr().
4091 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4092 Implement in terms of add_prefix_map().
4093 (remap_macro_filename, remap_debug_filename): Implement in term of
4095 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4096 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4097 * dbxout.c: Include file-prefix-map.h.
4098 * varasm.c: Likewise.
4099 * vmsdbgout.c: Likewise.
4100 * xcoffout.c: Likewise.
4101 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4102 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4103 * doc/invoke.texi (-ffile-prefix-map): Document.
4104 (-fdebug-prefix-map): Update description.
4106 2018-01-18 Martin Liska <mliska@suse.cz>
4108 * config/i386/i386.c (indirect_thunk_name): Document that also
4110 (output_indirect_thunk): Document why both instructions
4111 (pause and lfence) are generated.
4113 2018-01-18 Richard Biener <rguenther@suse.de>
4115 PR tree-optimization/83887
4116 * graphite-scop-detection.c
4117 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4118 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4119 (scop_detection::merge_sese): Re-implement with a flood-fill
4120 algorithm that properly finds a SESE region if it exists.
4122 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4125 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4126 pointer_diff optimizations use view_convert instead of convert.
4128 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4130 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4131 Generate different code for -mno-speculate-indirect-jumps.
4132 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4133 (*call_indirect_aix<mode>): Disable for
4134 -mno-speculate-indirect-jumps.
4135 (*call_indirect_aix<mode>_nospec): New define_insn.
4136 (*call_value_indirect_aix<mode>): Disable for
4137 -mno-speculate-indirect-jumps.
4138 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4139 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4140 -mno-speculate-indirect-jumps.
4141 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4143 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4145 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4146 long double type, set the flags for noting the default long double
4147 type, even if we don't pass or return a long double type.
4149 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4152 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4155 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4157 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4158 support for merge[hl].
4159 (fold_mergehl_helper): New helper function.
4160 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4161 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4162 (altivec_vmrglw_direct): Add xxmrglw insn.
4164 2018-01-17 Andrew Waterman <andrew@sifive.com>
4166 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4167 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4169 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4172 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4173 call the lto_location_cache before reading the
4174 DECL_SOURCE_LOCATION of the types.
4176 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4177 Richard Sandiford <richard.sandiford@linaro.org>
4179 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4180 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4181 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4182 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4183 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4185 * config/aarch64/constraints.md (aarch64_movti_operand):
4187 * config/aarch64/predicates.md (Uti): Add new constraint.
4189 2018-01-17 Carl Love <cel@us.ibm.com>
4191 * config/rs6000/vsx.md (define_expand xl_len_r,
4192 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4193 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4195 (define_expand, define_insn): Move the shift left from the
4196 define_insn to the define_expand for lxvl and stxvl instructions.
4197 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4198 and XL_LEN_R definitions to PURE.
4200 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4202 * config/i386/i386.c (indirect_thunk_name): Declare regno
4203 as unsigned int. Compare regno with INVALID_REGNUM.
4204 (output_indirect_thunk): Ditto.
4205 (output_indirect_thunk_function): Ditto.
4206 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4207 in the call to output_indirect_thunk_function.
4209 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4212 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4213 rather than the size of inner_type to determine the stack slot size
4214 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4216 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4219 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4222 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4224 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4225 endian Linux systems to optionally enable multilibs for selecting
4226 the long double type if the user configured an explicit type.
4227 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4228 have no long double multilibs if not defined.
4229 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4230 warn if the user used -mabi={ieee,ibm}longdouble and we built
4231 multilibs for long double.
4232 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4233 appropriate multilib option.
4234 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4236 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4237 for building long double multilibs.
4238 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4240 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4242 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4245 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4247 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4250 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4253 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4256 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4258 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4259 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4261 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4263 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4264 different rtl trees depending on TARGET_64BIT.
4265 (rs6000_gen_lvx): Likewise.
4267 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4269 * config/visium/visium.md (nop): Tweak comment.
4270 (hazard_nop): Likewise.
4272 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4274 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4275 -mspeculate-indirect-jumps.
4276 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4277 for -mno-speculate-indirect-jumps.
4278 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4279 (*call_value_indirect_elfv2<mode>): Disable for
4280 -mno-speculate-indirect-jumps.
4281 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4282 (indirect_jump): Emit different RTL for
4283 -mno-speculate-indirect-jumps.
4284 (*indirect_jump<mode>): Disable for
4285 -mno-speculate-indirect-jumps.
4286 (*indirect_jump<mode>_nospec): New define_insn.
4287 (tablejump): Emit different RTL for
4288 -mno-speculate-indirect-jumps.
4289 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4290 (tablejumpsi_nospec): New define_expand.
4291 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4292 (tablejumpdi_nospec): New define_expand.
4293 (*tablejump<mode>_internal1): Disable for
4294 -mno-speculate-indirect-jumps.
4295 (*tablejump<mode>_internal1_nospec): New define_insn.
4296 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4299 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4301 * caller-save.c (insert_save): Drop unnecessary parameter. All
4304 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4305 Richard Biener <rguenth@suse.de>
4308 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4309 return early, inline manually is_gimple_sizepos. Make sure if we
4310 call gimplify_expr we don't end up with a gimple constant.
4311 * tree.c (variably_modified_type_p): Don't return true for
4312 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4313 * gimplify.h (is_gimple_sizepos): Remove.
4315 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4317 PR tree-optimization/83857
4318 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4319 vectorizable_live_operation for pure SLP statements.
4320 (vectorizable_live_operation): Handle PHIs.
4322 2018-01-16 Richard Biener <rguenther@suse.de>
4324 PR tree-optimization/83867
4325 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4326 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4328 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4331 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4332 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4333 If off is not INTEGER_CST, issue a may not be aligned warning
4334 rather than isn't aligned. Use isn%'t rather than isn't.
4335 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4337 <case MULT_EXPR>: Improve the case when bottom and one of the
4338 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4339 operand, in that case check if the other operand is multiple of
4340 bottom divided by the INTEGER_CST operand.
4342 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4345 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4346 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4347 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4348 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4349 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4350 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4351 (pa_function_arg_size): New function.
4353 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4355 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4356 in a separate statement.
4358 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4360 PR tree-optimization/83847
4361 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4362 group gathers and scatters.
4364 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4366 PR rtl-optimization/86620
4367 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4369 PR rtl-optimization/83213
4370 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4371 to last if both are JUMP_INSNs.
4373 PR tree-optimization/83843
4374 * gimple-ssa-store-merging.c
4375 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4376 store_immediate_info for bswap/nop orig_stores.
4378 2018-01-15 Andrew Waterman <andrew@sifive.com>
4380 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4382 <UDIV>: Increase cost if !TARGET_DIV.
4384 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4386 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4387 (define_attr "cr_logical_3op"): New.
4388 (cceq_ior_compare): Adjust.
4389 (cceq_ior_compare_complement): Adjust.
4390 (*cceq_rev_compare): Adjust.
4391 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4392 (is_cracked_insn): Adjust.
4393 (insn_must_be_first_in_group): Adjust.
4394 * config/rs6000/40x.md: Adjust.
4395 * config/rs6000/440.md: Adjust.
4396 * config/rs6000/476.md: Adjust.
4397 * config/rs6000/601.md: Adjust.
4398 * config/rs6000/603.md: Adjust.
4399 * config/rs6000/6xx.md: Adjust.
4400 * config/rs6000/7450.md: Adjust.
4401 * config/rs6000/7xx.md: Adjust.
4402 * config/rs6000/8540.md: Adjust.
4403 * config/rs6000/cell.md: Adjust.
4404 * config/rs6000/e300c2c3.md: Adjust.
4405 * config/rs6000/e500mc.md: Adjust.
4406 * config/rs6000/e500mc64.md: Adjust.
4407 * config/rs6000/e5500.md: Adjust.
4408 * config/rs6000/e6500.md: Adjust.
4409 * config/rs6000/mpc.md: Adjust.
4410 * config/rs6000/power4.md: Adjust.
4411 * config/rs6000/power5.md: Adjust.
4412 * config/rs6000/power6.md: Adjust.
4413 * config/rs6000/power7.md: Adjust.
4414 * config/rs6000/power8.md: Adjust.
4415 * config/rs6000/power9.md: Adjust.
4416 * config/rs6000/rs64.md: Adjust.
4417 * config/rs6000/titan.md: Adjust.
4419 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4421 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4422 ix86_indirect_branch_register logic.
4424 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4426 * config/i386/constraints.md (Bs): Update
4427 ix86_indirect_branch_register check. Don't check
4428 ix86_indirect_branch_register with GOT_memory_operand.
4430 * config/i386/predicates.md (GOT_memory_operand): Don't check
4431 ix86_indirect_branch_register here.
4432 (GOT32_symbol_operand): Likewise.
4434 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4436 * config/i386/predicates.md (constant_call_address_operand):
4437 Rewrite ix86_indirect_branch_register logic.
4438 (sibcall_insn_operand): Likewise.
4440 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4442 * config/i386/constraints.md (Bs): Replace
4443 ix86_indirect_branch_thunk_register with
4444 ix86_indirect_branch_register.
4446 * config/i386/i386.md (indirect_jump): Likewise.
4447 (tablejump): Likewise.
4448 (*sibcall_memory): Likewise.
4449 (*sibcall_value_memory): Likewise.
4450 Peepholes of indirect call and jump via memory: Likewise.
4451 * config/i386/i386.opt: Likewise.
4452 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4453 (GOT_memory_operand): Likewise.
4454 (call_insn_operand): Likewise.
4455 (sibcall_insn_operand): Likewise.
4456 (GOT32_symbol_operand): Likewise.
4458 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4461 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4462 type rather than type addr's type points to.
4463 (expand_omp_atomic_mutex): Likewise.
4464 (expand_omp_atomic): Likewise.
4466 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4469 * config/i386/i386.c (output_indirect_thunk_function): Use
4470 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4471 for __x86_return_thunk.
4473 2018-01-15 Richard Biener <rguenther@suse.de>
4476 * expmed.c (extract_bit_field_1): Fix typo.
4478 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4481 * config/arm/iterators.md (VF): New mode iterator.
4482 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4483 Remove integer-related logic from pattern.
4484 (neon_vabd<mode>_3): Likewise.
4486 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4489 * common.opt (fstrict-overflow): No longer an alias.
4490 (fwrapv-pointer): New option.
4491 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4492 also for pointer types based on flag_wrapv_pointer.
4493 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4494 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4495 opts->x_flag_wrapv got set.
4496 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4497 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4498 POINTER_TYPE_OVERFLOW_UNDEFINED.
4499 * match.pd: Likewise in address comparison pattern.
4500 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4502 2018-01-15 Richard Biener <rguenther@suse.de>
4505 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4506 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4507 Reset type names to their identifier if their TYPE_DECL doesn't
4508 have linkage (and thus is used for ODR and devirt).
4509 (save_debug_info_for_decl): Remove.
4510 (save_debug_info_for_type): Likewise.
4511 (add_tree_to_fld_list): Adjust.
4512 * tree-pretty-print.c (dump_generic_node): Make dumping of
4513 type names more robust.
4515 2018-01-15 Richard Biener <rguenther@suse.de>
4517 * BASE-VER: Bump to 8.0.1.
4519 2018-01-14 Martin Sebor <msebor@redhat.com>
4522 * builtins.c (check_access): Avoid warning when the no-warning bit
4525 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4527 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4528 * ira-color (allocno_hard_regs_compare): Likewise.
4530 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4533 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4534 Use .pushsection/.popsection.
4536 2018-01-14 Martin Sebor <msebor@redhat.com>
4539 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4541 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4543 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4544 entry from extra_headers.
4545 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4546 extra_headers, make the list bitwise identical to the i?86-*-* one.
4548 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4550 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4551 -mcmodel=large with -mindirect-branch=thunk,
4552 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4553 -mfunction-return=thunk-extern.
4554 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4555 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4556 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4558 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4560 * config/i386/i386.c (print_reg): Print the name of the full
4561 integer register without '%'.
4562 (ix86_print_operand): Handle 'V'.
4563 * doc/extend.texi: Document 'V' modifier.
4565 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4567 * config/i386/constraints.md (Bs): Disallow memory operand for
4568 -mindirect-branch-register.
4570 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4571 (GOT_memory_operand): Likewise.
4572 (call_insn_operand): Likewise.
4573 (sibcall_insn_operand): Likewise.
4574 (GOT32_symbol_operand): Likewise.
4575 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4576 for -mindirect-branch-register.
4577 (tablejump): Likewise.
4578 (*sibcall_memory): Likewise.
4579 (*sibcall_value_memory): Likewise.
4580 Disallow peepholes of indirect call and jump via memory for
4581 -mindirect-branch-register.
4582 (*call_pop): Replace m with Bw.
4583 (*call_value_pop): Likewise.
4584 (*sibcall_pop_memory): Replace m with Bs.
4585 * config/i386/i386.opt (mindirect-branch-register): New option.
4586 * doc/invoke.texi: Document -mindirect-branch-register option.
4588 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4590 * config/i386/i386-protos.h (ix86_output_function_return): New.
4591 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4592 set function_return_type.
4593 (indirect_thunk_name): Add ret_p to indicate thunk for function
4595 (output_indirect_thunk_function): Pass false to
4596 indirect_thunk_name.
4597 (ix86_output_indirect_branch_via_reg): Likewise.
4598 (ix86_output_indirect_branch_via_push): Likewise.
4599 (output_indirect_thunk_function): Create alias for function
4600 return thunk if regno < 0.
4601 (ix86_output_function_return): New function.
4602 (ix86_handle_fndecl_attribute): Handle function_return.
4603 (ix86_attribute_table): Add function_return.
4604 * config/i386/i386.h (machine_function): Add
4605 function_return_type.
4606 * config/i386/i386.md (simple_return_internal): Use
4607 ix86_output_function_return.
4608 (simple_return_internal_long): Likewise.
4609 * config/i386/i386.opt (mfunction-return=): New option.
4610 (indirect_branch): Mention -mfunction-return=.
4611 * doc/extend.texi: Document function_return function attribute.
4612 * doc/invoke.texi: Document -mfunction-return= option.
4614 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4616 * config/i386/i386-opts.h (indirect_branch): New.
4617 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4618 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4619 with local indirect jump when converting indirect call and jump.
4620 (ix86_set_indirect_branch_type): New.
4621 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4622 (indirectlabelno): New.
4623 (indirect_thunk_needed): Likewise.
4624 (indirect_thunk_bnd_needed): Likewise.
4625 (indirect_thunks_used): Likewise.
4626 (indirect_thunks_bnd_used): Likewise.
4627 (INDIRECT_LABEL): Likewise.
4628 (indirect_thunk_name): Likewise.
4629 (output_indirect_thunk): Likewise.
4630 (output_indirect_thunk_function): Likewise.
4631 (ix86_output_indirect_branch_via_reg): Likewise.
4632 (ix86_output_indirect_branch_via_push): Likewise.
4633 (ix86_output_indirect_branch): Likewise.
4634 (ix86_output_indirect_jmp): Likewise.
4635 (ix86_code_end): Call output_indirect_thunk_function if needed.
4636 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4638 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4639 (ix86_attribute_table): Add indirect_branch.
4640 * config/i386/i386.h (machine_function): Add indirect_branch_type
4641 and has_local_indirect_jump.
4642 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4644 (tablejump): Likewise.
4645 (*indirect_jump): Use ix86_output_indirect_jmp.
4646 (*tablejump_1): Likewise.
4647 (simple_return_indirect_internal): Likewise.
4648 * config/i386/i386.opt (mindirect-branch=): New option.
4649 (indirect_branch): New.
4652 (thunk-inline): Likewise.
4653 (thunk-extern): Likewise.
4654 * doc/extend.texi: Document indirect_branch function attribute.
4655 * doc/invoke.texi: Document -mindirect-branch= option.
4657 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4660 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4662 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4664 * ipa-inline.c (want_inline_small_function_p): Return false if
4665 inlining has already failed with CIF_FINAL_ERROR.
4666 (update_caller_keys): Call want_inline_small_function_p before
4668 (update_callee_keys): Likewise.
4670 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4672 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4674 (rs6000_quadword_masked_address_p): Likewise.
4675 (quad_aligned_load_p): Likewise.
4676 (quad_aligned_store_p): Likewise.
4677 (const_load_sequence_p): Add comment to describe the outer-most loop.
4678 (mimic_memory_attributes_and_flags): New function.
4679 (rs6000_gen_stvx): Likewise.
4680 (replace_swapped_aligned_store): Likewise.
4681 (rs6000_gen_lvx): Likewise.
4682 (replace_swapped_aligned_load): Likewise.
4683 (replace_swapped_load_constant): Capitalize argument name in
4684 comment describing this function.
4685 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4686 and stores that access quad-word aligned addresses and replace
4687 with stvx or lvx instructions when appropriate.
4688 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4689 New function prototype.
4690 (rs6000_quadword_masked_address_p): Likewise.
4691 (rs6000_gen_lvx): Likewise.
4692 (rs6000_gen_stvx): Likewise.
4693 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4694 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4695 when memory address is aligned.
4696 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4697 this split to select lvx instruction when memory address is aligned.
4698 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4699 instruction when memory address is aligned.
4700 (*vsx_le_perm_load_v16qi): Likewise.
4701 (four unnamed splitters): Modify to select the stvx instruction
4702 when memory is aligned.
4704 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4706 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4707 which appears in the queue multiple times.
4709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4710 Alan Hayward <alan.hayward@arm.com>
4711 David Sherwood <david.sherwood@arm.com>
4713 * tree-vectorizer.h (vec_lower_bound): New structure.
4714 (_loop_vec_info): Add check_nonzero and lower_bounds.
4715 (LOOP_VINFO_CHECK_NONZERO): New macro.
4716 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4717 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4718 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4719 fields. Make seg_len the distance travelled, not including the
4721 (dr_direction_indicator): Declare.
4722 (dr_zero_step_indicator): Likewise.
4723 (dr_known_forward_stride_p): Likewise.
4724 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4726 (runtime_alias_check_p): Allow runtime alias checks with
4728 (operator ==): Compare access_size and align.
4729 (prune_runtime_alias_test_list): Rework for new distinction between
4730 the access_size and seg_len.
4731 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4733 (get_segment_min_max): New function.
4734 (create_intersect_range_checks): Use it.
4735 (dr_step_indicator): New function.
4736 (dr_direction_indicator): Likewise.
4737 (dr_zero_step_indicator): Likewise.
4738 (dr_known_forward_stride_p): Likewise.
4739 * tree-loop-distribution.c (data_ref_segment_size): Return
4740 DR_STEP * (niters - 1).
4741 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4743 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4744 (vect_preserves_scalar_order_p): New function, split out from...
4745 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4746 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4747 (vect_vfa_access_size): New function.
4748 (vect_vfa_align): Likewise.
4749 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4750 (dump_lower_bound): New function.
4751 (vect_check_lower_bound): Likewise.
4752 (vect_small_gap_p): Likewise.
4753 (vectorizable_with_step_bound_p): Likewise.
4754 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4755 depencies if the vectorization factor is 1. Convert the checks
4756 for nonzero steps into checks on the bounds of DR_STEP. Try using
4757 a bunds check for variable steps if the minimum required step is
4758 relatively small. Update calls to the dr_with_seg_len
4759 constructor and to vect_compile_time_alias.
4760 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4762 (vect_loop_versioning): Call it.
4763 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4765 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4768 Alan Hayward <alan.hayward@arm.com>
4769 David Sherwood <david.sherwood@arm.com>
4771 * doc/sourcebuild.texi (vect_scatter_store): Document.
4772 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4774 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4776 * genopinit.c (main): Add supports_vec_scatter_store and
4777 supports_vec_scatter_store_cached to target_optabs.
4778 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4779 IFN_MASK_SCATTER_STORE.
4780 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4782 * internal-fn.h (internal_store_fn_p): Declare.
4783 (internal_fn_stored_value_index): Likewise.
4784 * internal-fn.c (scatter_store_direct): New macro.
4785 (expand_scatter_store_optab_fn): New function.
4786 (direct_scatter_store_optab_supported_p): New macro.
4787 (internal_store_fn_p): New function.
4788 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4789 IFN_MASK_SCATTER_STORE.
4790 (internal_fn_mask_index): Likewise.
4791 (internal_fn_stored_value_index): New function.
4792 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4794 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4795 * optabs-query.c (supports_vec_scatter_store_p): New function.
4796 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4797 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4798 true for scatter stores.
4799 (vect_gather_scatter_fn_p): Handle scatter stores too.
4800 (vect_check_gather_scatter): Consider using scatter stores if
4801 supports_vec_scatter_store_p.
4802 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4804 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4805 internal_fn_stored_value_index.
4806 (check_load_store_masking): Handle scatter stores too.
4807 (vect_get_store_rhs): Make public.
4808 (vectorizable_call): Use internal_store_fn_p.
4809 (vectorizable_store): Handle scatter store internal functions.
4810 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4811 when deciding whether the end of the group has been reached.
4812 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4813 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4814 (mask_scatter_store<mode>): New insns.
4816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4817 Alan Hayward <alan.hayward@arm.com>
4818 David Sherwood <david.sherwood@arm.com>
4820 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4821 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4822 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4824 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4825 Use vect_truncate_gather_scatter_offset if we can't treat the
4826 operation as a normal gather load or scatter store.
4827 (get_group_load_store_type): Take the gather_scatter_info
4828 as argument. Try using a gather load or scatter store for
4829 single-element groups.
4830 (get_load_store_type): Update calls to get_group_load_store_type
4831 and vect_use_strided_gather_scatters_p.
4833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4834 Alan Hayward <alan.hayward@arm.com>
4835 David Sherwood <david.sherwood@arm.com>
4837 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4838 optional tree argument.
4839 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4841 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4842 but continue to use the current value as a fallback.
4843 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4844 to compare the updates.
4845 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4846 (get_load_store_type): Use it when handling a strided access.
4847 (vect_get_strided_load_store_ops): New function.
4848 (vect_get_data_ptr_increment): Likewise.
4849 (vectorizable_load): Handle strided gather loads. Always pass
4850 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4853 Alan Hayward <alan.hayward@arm.com>
4854 David Sherwood <david.sherwood@arm.com>
4856 * doc/md.texi (gather_load@var{m}): Document.
4857 (mask_gather_load@var{m}): Likewise.
4858 * genopinit.c (main): Add supports_vec_gather_load and
4859 supports_vec_gather_load_cached to target_optabs.
4860 * optabs-tree.c (init_tree_optimization_optabs): Use
4861 ggc_cleared_alloc to allocate target_optabs.
4862 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4863 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4865 * internal-fn.h (internal_load_fn_p): Declare.
4866 (internal_gather_scatter_fn_p): Likewise.
4867 (internal_fn_mask_index): Likewise.
4868 (internal_gather_scatter_fn_supported_p): Likewise.
4869 * internal-fn.c (gather_load_direct): New macro.
4870 (expand_gather_load_optab_fn): New function.
4871 (direct_gather_load_optab_supported_p): New macro.
4872 (direct_internal_fn_optab): New function.
4873 (internal_load_fn_p): Likewise.
4874 (internal_gather_scatter_fn_p): Likewise.
4875 (internal_fn_mask_index): Likewise.
4876 (internal_gather_scatter_fn_supported_p): Likewise.
4877 * optabs-query.c (supports_at_least_one_mode_p): New function.
4878 (supports_vec_gather_load_p): Likewise.
4879 * optabs-query.h (supports_vec_gather_load_p): Declare.
4880 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4881 and memory_type field.
4882 (NUM_PATTERNS): Bump to 15.
4883 * tree-vect-data-refs.c: Include internal-fn.h.
4884 (vect_gather_scatter_fn_p): New function.
4885 (vect_describe_gather_scatter_call): Likewise.
4886 (vect_check_gather_scatter): Try using internal functions for
4887 gather loads. Recognize existing calls to a gather load function.
4888 (vect_analyze_data_refs): Consider using gather loads if
4889 supports_vec_gather_load_p.
4890 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4891 (vect_get_gather_scatter_offset_type): Likewise.
4892 (vect_convert_mask_for_vectype): Likewise.
4893 (vect_add_conversion_to_patterm): Likewise.
4894 (vect_try_gather_scatter_pattern): Likewise.
4895 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4896 (vect_vect_recog_func_ptrs): Add it.
4897 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4898 internal_fn_mask_index and internal_gather_scatter_fn_p.
4899 (check_load_store_masking): Take the gather_scatter_info as an
4900 argument and handle gather loads.
4901 (vect_get_gather_scatter_ops): New function.
4902 (vectorizable_call): Check internal_load_fn_p.
4903 (vectorizable_load): Likewise. Handle gather load internal
4905 (vectorizable_store): Update call to check_load_store_masking.
4906 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4907 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4908 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4909 (aarch64_gather_scale_operand_d): New predicates.
4910 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4911 (mask_gather_load<mode>): New insns.
4913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4914 Alan Hayward <alan.hayward@arm.com>
4915 David Sherwood <david.sherwood@arm.com>
4917 * optabs.def (fold_left_plus_optab): New optab.
4918 * doc/md.texi (fold_left_plus_@var{m}): Document.
4919 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4920 * internal-fn.c (fold_left_direct): Define.
4921 (expand_fold_left_optab_fn): Likewise.
4922 (direct_fold_left_optab_supported_p): Likewise.
4923 * fold-const-call.c (fold_const_fold_left): New function.
4924 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4925 * tree-parloops.c (valid_reduction_p): New function.
4926 (gather_scalar_reductions): Use it.
4927 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4928 (vect_finish_replace_stmt): Declare.
4929 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4930 (needs_fold_left_reduction_p): New function, split out from...
4931 (vect_is_simple_reduction): ...here. Accept reductions that
4932 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4933 (vect_force_simple_reduction): Also store the reduction type in
4934 the assignment's STMT_VINFO_REDUC_TYPE.
4935 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4936 (merge_with_identity): New function.
4937 (vect_expand_fold_left): Likewise.
4938 (vectorize_fold_left_reduction): Likewise.
4939 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4940 scalar phi in place for it. Check for target support and reject
4941 cases that would reassociate the operation. Defer the transform
4942 phase to vectorize_fold_left_reduction.
4943 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4944 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4945 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4949 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4950 call to ifc_temp_var.
4952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4953 Alan Hayward <alan.hayward@arm.com>
4954 David Sherwood <david.sherwood@arm.com>
4956 * target.def (legitimize_address_displacement): Take the original
4957 offset as a poly_int.
4958 * targhooks.h (default_legitimize_address_displacement): Update
4960 * targhooks.c (default_legitimize_address_displacement): Likewise.
4961 * doc/tm.texi: Regenerate.
4962 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4963 as an argument, moving assert of ad->disp == ad->disp_term to...
4964 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4965 Try calling targetm.legitimize_address_displacement before expanding
4966 the address rather than afterwards, and adjust for the new interface.
4967 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4968 Match the new hook interface. Handle SVE addresses.
4969 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4974 * Makefile.in (OBJS): Add early-remat.o.
4975 * target.def (select_early_remat_modes): New hook.
4976 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4977 * doc/tm.texi: Regenerate.
4978 * targhooks.h (default_select_early_remat_modes): Declare.
4979 * targhooks.c (default_select_early_remat_modes): New function.
4980 * timevar.def (TV_EARLY_REMAT): New timevar.
4981 * passes.def (pass_early_remat): New pass.
4982 * tree-pass.h (make_pass_early_remat): Declare.
4983 * early-remat.c: New file.
4984 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4986 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4989 Alan Hayward <alan.hayward@arm.com>
4990 David Sherwood <david.sherwood@arm.com>
4992 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4993 vfm1 with a bound_epilog parameter.
4994 (vect_do_peeling): Update calls accordingly, and move the prologue
4995 call earlier in the function. Treat the base bound_epilog as 0 for
4996 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4997 this base when peeling for gaps.
4998 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4999 with fully-masked loops.
5000 (vect_estimate_min_profitable_iters): Handle the single peeled
5001 iteration in that case.
5003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5004 Alan Hayward <alan.hayward@arm.com>
5005 David Sherwood <david.sherwood@arm.com>
5007 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5008 single-element interleaving even if the size is not a power of 2.
5009 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5010 accesses for single-element interleaving if the group size is
5013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5014 Alan Hayward <alan.hayward@arm.com>
5015 David Sherwood <david.sherwood@arm.com>
5017 * doc/md.texi (fold_extract_last_@var{m}): Document.
5018 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5019 * optabs.def (fold_extract_last_optab): New optab.
5020 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5021 * internal-fn.c (fold_extract_direct): New macro.
5022 (expand_fold_extract_optab_fn): Likewise.
5023 (direct_fold_extract_optab_supported_p): Likewise.
5024 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5025 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5026 EXTRACT_LAST_REDUCTION.
5027 (get_initial_def_for_reduction): Do not create an initial vector
5028 for EXTRACT_LAST_REDUCTION reductions.
5029 (vectorizable_reduction): Leave the scalar phi in place for
5030 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5031 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5032 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5033 transform phase to vectorizable_condition.
5034 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5036 (vect_finish_stmt_generation): ...here.
5037 (vect_finish_replace_stmt): New function.
5038 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5039 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5041 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5044 Alan Hayward <alan.hayward@arm.com>
5045 David Sherwood <david.sherwood@arm.com>
5047 * doc/md.texi (extract_last_@var{m}): Document.
5048 * optabs.def (extract_last_optab): New optab.
5049 * internal-fn.def (EXTRACT_LAST): New internal function.
5050 * internal-fn.c (cond_unary_direct): New macro.
5051 (expand_cond_unary_optab_fn): Likewise.
5052 (direct_cond_unary_optab_supported_p): Likewise.
5053 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5054 loops using EXTRACT_LAST.
5055 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5056 (extract_last_<mode>): ...this optab.
5057 (vec_extract<mode><Vel>): Update accordingly.
5059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5060 Alan Hayward <alan.hayward@arm.com>
5061 David Sherwood <david.sherwood@arm.com>
5063 * target.def (empty_mask_is_expensive): New hook.
5064 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5065 * doc/tm.texi: Regenerate.
5066 * targhooks.h (default_empty_mask_is_expensive): Declare.
5067 * targhooks.c (default_empty_mask_is_expensive): New function.
5068 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5069 if the target says that empty masks are expensive.
5070 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5072 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5074 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5075 Alan Hayward <alan.hayward@arm.com>
5076 David Sherwood <david.sherwood@arm.com>
5078 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5079 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5080 (vect_use_loop_mask_for_alignment_p): New function.
5081 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5082 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5083 niters_skip argument. Make sure that the first niters_skip elements
5084 of the first iteration are inactive.
5085 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5086 Update call to vect_set_loop_masks_directly.
5087 (get_misalign_in_elems): New function, split out from...
5088 (vect_gen_prolog_loop_niters): ...here.
5089 (vect_update_init_of_dr): Take a code argument that specifies whether
5090 the adjustment should be added or subtracted.
5091 (vect_update_init_of_drs): Likewise.
5092 (vect_prepare_for_masked_peels): New function.
5093 (vect_do_peeling): Skip prologue peeling if we're using a mask
5094 instead. Update call to vect_update_inits_of_drs.
5095 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5097 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5098 alignment. Do not include the number of peeled iterations in
5099 the minimum threshold in that case.
5100 (vectorizable_induction): Adjust the start value down by
5101 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5102 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5103 Take the number of skipped iterations into account when calculating
5105 * tree-vect-stmts.c (vect_gen_while_not): New function.
5107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5108 Alan Hayward <alan.hayward@arm.com>
5109 David Sherwood <david.sherwood@arm.com>
5111 * doc/sourcebuild.texi (vect_fully_masked): Document.
5112 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5114 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5116 (vect_analyze_loop_2): ...here. Don't check the vectorization
5117 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5125 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5126 (dump_groups): Update accordingly.
5127 (iv_use::mem_type): New member variable.
5128 (address_p): New function.
5129 (record_use): Add a mem_type argument and initialize the new
5131 (record_group_use): Add a mem_type argument. Use address_p.
5132 Remove obsolete null checks of base_object. Update call to record_use.
5133 (find_interesting_uses_op): Update call to record_group_use.
5134 (find_interesting_uses_cond): Likewise.
5135 (find_interesting_uses_address): Likewise.
5136 (get_mem_type_for_internal_fn): New function.
5137 (find_address_like_use): Likewise.
5138 (find_interesting_uses_stmt): Try find_address_like_use before
5139 calling find_interesting_uses_op.
5140 (addr_offset_valid_p): Use the iv mem_type field as the type
5141 of the addressed memory.
5142 (add_autoinc_candidates): Likewise.
5143 (get_address_cost): Likewise.
5144 (split_small_address_groups_p): Use address_p.
5145 (split_address_groups): Likewise.
5146 (add_iv_candidate_for_use): Likewise.
5147 (autoinc_possible_for_pair): Likewise.
5148 (rewrite_groups): Likewise.
5149 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5150 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5151 (get_alias_ptr_type_for_ptr_address): New function.
5152 (rewrite_use_address): Rewrite address uses in calls that were
5153 identified by find_address_like_use.
5155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5156 Alan Hayward <alan.hayward@arm.com>
5157 David Sherwood <david.sherwood@arm.com>
5159 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5161 * gimple-expr.h (is_gimple_addressable: Likewise.
5162 * gimple-expr.c (is_gimple_address): Likewise.
5163 * internal-fn.c (expand_call_mem_ref): New function.
5164 (expand_mask_load_optab_fn): Use it.
5165 (expand_mask_store_optab_fn): Likewise.
5167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5168 Alan Hayward <alan.hayward@arm.com>
5169 David Sherwood <david.sherwood@arm.com>
5171 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5172 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5173 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5174 (cond_umax@var{mode}): Document.
5175 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5176 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5177 (cond_umin_optab, cond_umax_optab): New optabs.
5178 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5179 (COND_IOR, COND_XOR): New internal functions.
5180 * internal-fn.h (get_conditional_internal_fn): Declare.
5181 * internal-fn.c (cond_binary_direct): New macro.
5182 (expand_cond_binary_optab_fn): Likewise.
5183 (direct_cond_binary_optab_supported_p): Likewise.
5184 (get_conditional_internal_fn): New function.
5185 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5186 Cope with reduction statements that are vectorized as calls rather
5188 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5189 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5190 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5191 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5192 (UNSPEC_COND_EOR): New unspecs.
5193 (optab): Add mappings for them.
5194 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5195 (sve_int_op, sve_fp_op): New int attributes.
5197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5198 Alan Hayward <alan.hayward@arm.com>
5199 David Sherwood <david.sherwood@arm.com>
5201 * optabs.def (while_ult_optab): New optab.
5202 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5203 * internal-fn.def (WHILE_ULT): New internal function.
5204 * internal-fn.h (direct_internal_fn_supported_p): New override
5205 that takes two types as argument.
5206 * internal-fn.c (while_direct): New macro.
5207 (expand_while_optab_fn): New function.
5208 (convert_optab_supported_p): Likewise.
5209 (direct_while_optab_supported_p): New macro.
5210 * wide-int.h (wi::udiv_ceil): New function.
5211 * tree-vectorizer.h (rgroup_masks): New structure.
5212 (vec_loop_masks): New typedef.
5213 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5215 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5216 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5217 (vect_max_vf): New function.
5218 (slpeel_make_loop_iterate_ntimes): Delete.
5219 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5220 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5221 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5222 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5223 internal-fn.h, stor-layout.h and optabs-query.h.
5224 (vect_set_loop_mask): New function.
5225 (add_preheader_seq): Likewise.
5226 (add_header_seq): Likewise.
5227 (interleave_supported_p): Likewise.
5228 (vect_maybe_permute_loop_masks): Likewise.
5229 (vect_set_loop_masks_directly): Likewise.
5230 (vect_set_loop_condition_masked): Likewise.
5231 (vect_set_loop_condition_unmasked): New function, split out from
5232 slpeel_make_loop_iterate_ntimes.
5233 (slpeel_make_loop_iterate_ntimes): Rename to..
5234 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5235 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5236 (vect_do_peeling): Update call accordingly.
5237 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5239 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5240 mask_compare_type, can_fully_mask_p and fully_masked_p.
5241 (release_vec_loop_masks): New function.
5242 (_loop_vec_info): Use it to free the loop masks.
5243 (can_produce_all_loop_masks_p): New function.
5244 (vect_get_max_nscalars_per_iter): Likewise.
5245 (vect_verify_full_masking): Likewise.
5246 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5247 retries, and free the mask rgroups before retrying. Check loop-wide
5248 reasons for disallowing fully-masked loops. Make the final decision
5249 about whether use a fully-masked loop or not.
5250 (vect_estimate_min_profitable_iters): Do not assume that peeling
5251 for the number of iterations will be needed for fully-masked loops.
5252 (vectorizable_reduction): Disable fully-masked loops.
5253 (vectorizable_live_operation): Likewise.
5254 (vect_halve_mask_nunits): New function.
5255 (vect_double_mask_nunits): Likewise.
5256 (vect_record_loop_mask): Likewise.
5257 (vect_get_loop_mask): Likewise.
5258 (vect_transform_loop): Handle the case in which the final loop
5259 iteration might handle a partial vector. Call vect_set_loop_condition
5260 instead of slpeel_make_loop_iterate_ntimes.
5261 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5262 (check_load_store_masking): New function.
5263 (prepare_load_store_mask): Likewise.
5264 (vectorizable_store): Handle fully-masked loops.
5265 (vectorizable_load): Likewise.
5266 (supportable_widening_operation): Use vect_halve_mask_nunits for
5268 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5269 (vect_gen_while): New function.
5270 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5271 (aarch64_uqdec<mode>): New insn.
5273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5274 Alan Hayward <alan.hayward@arm.com>
5275 David Sherwood <david.sherwood@arm.com>
5277 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5278 (reduc_xor_scal_optab): New optabs.
5279 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5280 (reduc_xor_scal_@var{m}): Document.
5281 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5282 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5284 * fold-const-call.c (fold_const_call): Handle them.
5285 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5286 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5287 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5288 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5289 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5290 (UNSPEC_XORV): New unspecs.
5291 (optab): Add entries for them.
5292 (BITWISEV): New int iterator.
5293 (bit_reduc_op): New int attributes.
5295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5296 Alan Hayward <alan.hayward@arm.com>
5297 David Sherwood <david.sherwood@arm.com>
5299 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5300 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5301 * optabs.def (vec_shl_insert_optab): New optab.
5302 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5303 (duplicate_and_interleave): Likewise.
5304 * tree-vect-loop.c: Include internal-fn.h.
5305 (neutral_op_for_slp_reduction): New function, split out from
5306 get_initial_defs_for_reduction.
5307 (get_initial_def_for_reduction): Handle option 2 for variable-length
5308 vectors by loading the neutral value into a vector and then shifting
5309 the initial value into element 0.
5310 (get_initial_defs_for_reduction): Replace the code argument with
5311 the neutral value calculated by neutral_op_for_slp_reduction.
5312 Use gimple_build_vector for constant-length vectors.
5313 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5314 but the first group_size elements have a neutral value.
5315 Use duplicate_and_interleave otherwise.
5316 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5317 Update call to get_initial_defs_for_reduction. Handle SLP
5318 reductions for variable-length vectors by creating one vector
5319 result for each scalar result, with the elements associated
5320 with other scalar results stubbed out with the neutral value.
5321 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5322 Require IFN_VEC_SHL_INSERT for double reductions on
5323 variable-length vectors, or SLP reductions that have
5324 a neutral value. Require can_duplicate_and_interleave_p
5325 support for variable-length unchained SLP reductions if there
5326 is no neutral value, such as for MIN/MAX reductions. Also require
5327 the number of vector elements to be a multiple of the number of
5328 SLP statements when doing variable-length unchained SLP reductions.
5329 Update call to vect_create_epilog_for_reduction.
5330 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5331 and remove initial values.
5332 (duplicate_and_interleave): Make public.
5333 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5334 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5337 Alan Hayward <alan.hayward@arm.com>
5338 David Sherwood <david.sherwood@arm.com>
5340 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5341 (can_duplicate_and_interleave_p): New function.
5342 (vect_get_and_check_slp_defs): Take the vector of statements
5343 rather than just the current one. Remove excess parentheses.
5344 Restriction rejectinon of vect_constant_def and vect_external_def
5345 for variable-length vectors to boolean types, or types for which
5346 can_duplicate_and_interleave_p is false.
5347 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5348 (duplicate_and_interleave): New function.
5349 (vect_get_constant_vectors): Use gimple_build_vector for
5350 constant-length vectors and suitable variable-length constant
5351 vectors. Use duplicate_and_interleave for other variable-length
5352 vectors. Don't defer the update when inserting new statements.
5354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5355 Alan Hayward <alan.hayward@arm.com>
5356 David Sherwood <david.sherwood@arm.com>
5358 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5359 min_profitable_iters doesn't go negative.
5361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5362 Alan Hayward <alan.hayward@arm.com>
5363 David Sherwood <david.sherwood@arm.com>
5365 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5366 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5367 * optabs.def (vec_mask_load_lanes_optab): New optab.
5368 (vec_mask_store_lanes_optab): Likewise.
5369 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5370 (MASK_STORE_LANES): Likewise.
5371 * internal-fn.c (mask_load_lanes_direct): New macro.
5372 (mask_store_lanes_direct): Likewise.
5373 (expand_mask_load_optab_fn): Handle masked operations.
5374 (expand_mask_load_lanes_optab_fn): New macro.
5375 (expand_mask_store_optab_fn): Handle masked operations.
5376 (expand_mask_store_lanes_optab_fn): New macro.
5377 (direct_mask_load_lanes_optab_supported_p): Likewise.
5378 (direct_mask_store_lanes_optab_supported_p): Likewise.
5379 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5381 (vect_load_lanes_supported): Likewise.
5382 * tree-vect-data-refs.c (strip_conversion): New function.
5383 (can_group_stmts_p): Likewise.
5384 (vect_analyze_data_ref_accesses): Use it instead of checking
5385 for a pair of assignments.
5386 (vect_store_lanes_supported): Take a masked_p parameter.
5387 (vect_load_lanes_supported): Likewise.
5388 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5389 vect_store_lanes_supported and vect_load_lanes_supported.
5390 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5391 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5392 parameter. Don't allow gaps for masked accesses.
5393 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5394 and vect_load_lanes_supported.
5395 (get_load_store_type): Take a masked_p parameter and update
5396 call to get_group_load_store_type.
5397 (vectorizable_store): Update call to get_load_store_type.
5398 Handle IFN_MASK_STORE_LANES.
5399 (vectorizable_load): Update call to get_load_store_type.
5400 Handle IFN_MASK_LOAD_LANES.
5402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5403 Alan Hayward <alan.hayward@arm.com>
5404 David Sherwood <david.sherwood@arm.com>
5406 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5408 * config/aarch64/aarch64-protos.h
5409 (aarch64_sve_struct_memory_operand_p): Declare.
5410 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5411 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5412 (VPRED, vpred): Handle SVE structure modes.
5413 * config/aarch64/constraints.md (Utx): New constraint.
5414 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5415 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5416 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5417 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5418 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5419 structure modes. Split into pieces after RA.
5420 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5421 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5423 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5424 SVE structure modes.
5425 (aarch64_classify_address): Likewise.
5426 (sizetochar): Move earlier in file.
5427 (aarch64_print_operand): Handle SVE register lists.
5428 (aarch64_array_mode): New function.
5429 (aarch64_sve_struct_memory_operand_p): Likewise.
5430 (TARGET_ARRAY_MODE): Redefine.
5432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5433 Alan Hayward <alan.hayward@arm.com>
5434 David Sherwood <david.sherwood@arm.com>
5436 * target.def (array_mode): New target hook.
5437 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5438 * doc/tm.texi: Regenerate.
5439 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5440 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5441 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5443 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5447 Alan Hayward <alan.hayward@arm.com>
5448 David Sherwood <david.sherwood@arm.com>
5450 * fold-const.c (fold_binary_loc): Check the argument types
5451 rather than the result type when testing for a vector operation.
5453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5455 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5456 * doc/tm.texi: Regenerate.
5458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5459 Alan Hayward <alan.hayward@arm.com>
5460 David Sherwood <david.sherwood@arm.com>
5462 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5463 (sve): Document new AArch64 extension.
5464 * doc/md.texi (w): Extend the description of the AArch64
5465 constraint to include SVE vectors.
5466 (Upl, Upa): Document new AArch64 predicate constraints.
5467 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5469 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5470 (msve-vector-bits=): New option.
5471 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5472 SVE when these are disabled.
5473 (sve): New extension.
5474 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5475 modes. Adjust their number of units based on aarch64_sve_vg.
5476 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5477 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5478 aarch64_addr_query_type.
5479 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5480 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5481 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5482 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5483 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5484 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5485 (aarch64_simd_imm_zero_p): Delete.
5486 (aarch64_check_zero_based_sve_index_immediate): Declare.
5487 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5488 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5489 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5490 (aarch64_sve_float_mul_immediate_p): Likewise.
5491 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5493 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5494 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5495 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5496 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5497 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5498 (aarch64_regmode_natural_size): Likewise.
5499 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5500 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5502 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5503 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5504 for VG and the SVE predicate registers.
5505 (V_ALIASES): Add a "z"-prefixed alias.
5506 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5507 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5508 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5509 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5510 (REG_CLASS_NAMES): Add entries for them.
5511 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5512 and the predicate registers.
5513 (aarch64_sve_vg): Declare.
5514 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5515 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5516 (REGMODE_NATURAL_SIZE): Define.
5517 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5519 * config/aarch64/aarch64.c: Include cfgrtl.h.
5520 (simd_immediate_info): Add a constructor for series vectors,
5521 and an associated step field.
5522 (aarch64_sve_vg): New variable.
5523 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5524 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5525 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5526 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5527 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5528 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5529 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5530 (aarch64_get_mask_mode): New functions.
5531 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5532 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5533 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5534 predicate modes and predicate registers. Explicitly restrict
5535 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5536 to store a vector mode if it is recognized by
5537 aarch64_classify_vector_mode.
5538 (aarch64_regmode_natural_size): New function.
5539 (aarch64_hard_regno_caller_save_mode): Return the original mode
5541 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5542 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5543 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5544 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5546 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5547 does not overlap dest if the function is frame-related. Handle
5549 (aarch64_split_add_offset): New function.
5550 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5551 them aarch64_add_offset.
5552 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5553 and update call to aarch64_sub_sp.
5554 (aarch64_add_cfa_expression): New function.
5555 (aarch64_expand_prologue): Pass extra temporary registers to the
5556 functions above. Handle the case in which we need to emit new
5557 DW_CFA_expressions for registers that were originally saved
5558 relative to the stack pointer, but now have to be expressed
5559 relative to the frame pointer.
5560 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5562 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5563 IP0 and IP1 values for SVE frames.
5564 (aarch64_expand_vec_series): New function.
5565 (aarch64_expand_sve_widened_duplicate): Likewise.
5566 (aarch64_expand_sve_const_vector): Likewise.
5567 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5568 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5569 into the register, rather than emitting a SET directly.
5570 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5571 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5572 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5573 (offset_9bit_signed_scaled_p): New functions.
5574 (aarch64_replicate_bitmask_imm): New function.
5575 (aarch64_bitmask_imm): Use it.
5576 (aarch64_cannot_force_const_mem): Reject expressions involving
5577 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5578 (aarch64_classify_index): Handle SVE indices, by requiring
5579 a plain register index with a scale that matches the element size.
5580 (aarch64_classify_address): Handle SVE addresses. Assert that
5581 the mode of the address is VOIDmode or an integer mode.
5582 Update call to aarch64_classify_symbol.
5583 (aarch64_classify_symbolic_expression): Update call to
5584 aarch64_classify_symbol.
5585 (aarch64_const_vec_all_in_range_p): New function.
5586 (aarch64_print_vector_float_operand): Likewise.
5587 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5588 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5589 and the FP immediates 1.0 and 0.5.
5590 (aarch64_print_address_internal): Handle SVE addresses.
5591 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5592 (aarch64_regno_regclass): Handle predicate registers.
5593 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5595 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5596 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5597 (aarch64_convert_sve_vector_bits): New function.
5598 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5599 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5601 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5602 Handle SVE vector and predicate modes. Accept VL-based constants
5603 that need only one temporary register, and VL offsets that require
5604 no temporary registers.
5605 (aarch64_conditional_register_usage): Mark the predicate registers
5606 as fixed if SVE isn't available.
5607 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5608 Return true for SVE vector and predicate modes.
5609 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5610 rather than an unsigned int. Handle SVE modes.
5611 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5613 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5615 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5616 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5617 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5618 (aarch64_sve_float_mul_immediate_p): New functions.
5619 (aarch64_sve_valid_immediate): New function.
5620 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5621 Explicitly reject structure modes. Check for INDEX constants.
5622 Handle PTRUE and PFALSE constants.
5623 (aarch64_check_zero_based_sve_index_immediate): New function.
5624 (aarch64_simd_imm_zero_p): Delete.
5625 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5626 vector modes. Accept constants in the range of CNT[BHWD].
5627 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5628 ask for an Advanced SIMD mode.
5629 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5630 (aarch64_simd_vector_alignment): Handle SVE predicates.
5631 (aarch64_vectorize_preferred_vector_alignment): New function.
5632 (aarch64_simd_vector_alignment_reachable): Use it instead of
5634 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5635 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5637 (MAX_VECT_LEN): Delete.
5638 (expand_vec_perm_d): Add a vec_flags field.
5639 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5640 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5641 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5643 (aarch64_evpc_rev): Rename to...
5644 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5645 (aarch64_evpc_rev_global): New function.
5646 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5647 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5649 (aarch64_evpc_sve_tbl): New function.
5650 (aarch64_expand_vec_perm_const_1): Update after rename of
5651 aarch64_evpc_rev. Handle SVE permutes too, trying
5652 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5653 than aarch64_evpc_tbl.
5654 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5655 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5656 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5657 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5658 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5659 (aarch64_expand_sve_vcond): New functions.
5660 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5661 of aarch64_vector_mode_p.
5662 (aarch64_dwarf_poly_indeterminate_value): New function.
5663 (aarch64_compute_pressure_classes): Likewise.
5664 (aarch64_can_change_mode_class): Likewise.
5665 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5666 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5667 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5668 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5669 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5670 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5671 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5672 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5674 (Dn, Dl, Dr): Accept const as well as const_vector.
5675 (Dz): Likewise. Compare against CONST0_RTX.
5676 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5677 of "vector" where appropriate.
5678 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5679 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5680 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5681 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5682 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5683 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5684 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5685 (v_int_equiv): Extend to SVE modes.
5686 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5688 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5689 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5690 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5691 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5692 (SVE_COND_FP_CMP): New int iterators.
5693 (perm_hilo): Handle the new unpack unspecs.
5694 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5696 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5697 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5698 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5699 (aarch64_equality_operator, aarch64_constant_vector_operand)
5700 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5701 (aarch64_sve_nonimmediate_operand): Likewise.
5702 (aarch64_sve_general_operand): Likewise.
5703 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5704 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5705 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5706 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5707 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5708 (aarch64_sve_float_arith_immediate): Likewise.
5709 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5710 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5711 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5712 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5713 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5714 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5715 (aarch64_sve_float_arith_operand): Likewise.
5716 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5717 (aarch64_sve_float_mul_operand): Likewise.
5718 (aarch64_sve_vec_perm_operand): Likewise.
5719 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5720 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5721 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5722 as well as const_vector.
5723 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5724 in file. Use CONST0_RTX and CONSTM1_RTX.
5725 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5726 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5727 Use aarch64_simd_imm_zero.
5728 * config/aarch64/aarch64-sve.md: New file.
5729 * config/aarch64/aarch64.md: Include it.
5730 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5731 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5732 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5733 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5734 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5735 (sve): New attribute.
5736 (enabled): Disable instructions with the sve attribute unless
5738 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5739 aarch64_expand_mov_immediate.
5740 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5741 CNT[BHSD] immediates.
5742 (movti): Split CONST_POLY_INT moves into two halves.
5743 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5744 Split additions that need a temporary here if the destination
5745 is the stack pointer.
5746 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5747 (*add<mode>3_poly_1): New instruction.
5748 (set_clobber_cc): New expander.
5750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5752 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5753 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5754 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5755 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5756 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5757 Change innermode from fixed_mode_size to machine_mode.
5758 (simplify_subreg): Update call accordingly. Handle a constant-sized
5759 subreg of a variable-length CONST_VECTOR.
5761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5762 Alan Hayward <alan.hayward@arm.com>
5763 David Sherwood <david.sherwood@arm.com>
5765 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5766 (add_offset_to_base): New function, split out from...
5767 (create_mem_ref): ...here. When handling a scale other than 1,
5768 check first whether the address is valid without the offset.
5769 Add it into the base if so, leaving the index and scale as-is.
5771 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5774 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5775 fold_for_warn before checking if arg2 is INTEGER_CST.
5777 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5779 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5780 (store_multiple_operation): Delete.
5781 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5782 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5783 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5784 guarded by TARGET_STRING.
5785 (rs6000_output_load_multiple): Delete.
5786 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5787 OPTION_MASK_STRING / TARGET_STRING handling.
5788 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5789 (const rs6000_opt_masks) <"string">: Change mask to 0.
5790 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5791 (MASK_STRING): Delete.
5792 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5794 (load_multiple): Delete.
5801 (store_multiple): Delete.
5808 (movmemsi_8reg): Delete.
5809 (corresponding unnamed define_insn): Delete.
5810 (movmemsi_6reg): Delete.
5811 (corresponding unnamed define_insn): Delete.
5812 (movmemsi_4reg): Delete.
5813 (corresponding unnamed define_insn): Delete.
5814 (movmemsi_2reg): Delete.
5815 (corresponding unnamed define_insn): Delete.
5816 (movmemsi_1reg): Delete.
5817 (corresponding unnamed define_insn): Delete.
5818 * config/rs6000/rs6000.opt (mno-string): New.
5819 (mstring): Replace by deprecation warning stub.
5820 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5822 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5824 * regrename.c (regrename_do_replace): If replacing the same
5825 reg multiple times, try to reuse last created gen_raw_REG.
5828 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5829 main to workaround a bug in GDB.
5831 2018-01-12 Tom de Vries <tom@codesourcery.com>
5834 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5836 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5838 PR rtl-optimization/80481
5839 * ira-color.c (get_cap_member): New function.
5840 (allocnos_conflict_by_live_ranges_p): Use it.
5841 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5842 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5844 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5847 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5848 (*saddl_se_1): Ditto.
5850 (*ssubl_se_1): Ditto.
5852 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5854 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5855 rather than wi::to_widest for DR_INITs.
5856 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5857 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5858 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5860 (vect_analyze_group_access_1): Note that here.
5862 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5864 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5865 polynomial type sizes.
5867 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5869 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5870 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5871 (gimple_add_tmp_var): Likewise.
5873 2018-01-12 Martin Liska <mliska@suse.cz>
5875 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5876 (gimple_alloc_sizes): Likewise.
5877 (dump_gimple_statistics): Use PRIu64 in printf format.
5878 * gimple.h: Change uint64_t to int.
5880 2018-01-12 Martin Liska <mliska@suse.cz>
5882 * tree-core.h: Use uint64_t instead of int.
5883 * tree.c (tree_node_counts): Likewise.
5884 (tree_node_sizes): Likewise.
5885 (dump_tree_statistics): Use PRIu64 in printf format.
5887 2018-01-12 Martin Liska <mliska@suse.cz>
5889 * Makefile.in: As qsort_chk is implemented in vec.c, add
5890 vec.o to linkage of gencfn-macros.
5891 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5892 passing the info to record_node_allocation_statistics.
5893 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5895 * ggc-common.c (struct ggc_usage): Add operator== and use
5896 it in operator< and compare function.
5897 * mem-stats.h (struct mem_usage): Likewise.
5898 * vec.c (struct vec_usage): Remove operator< and compare
5899 function. Can be simply inherited.
5901 2018-01-12 Martin Jambor <mjambor@suse.cz>
5904 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5905 * tree-ssa-math-opts.c: Include domwalk.h.
5906 (convert_mult_to_fma_1): New function.
5907 (fma_transformation_info): New type.
5908 (fma_deferring_state): Likewise.
5909 (cancel_fma_deferring): New function.
5910 (result_of_phi): Likewise.
5911 (last_fma_candidate_feeds_initial_phi): Likewise.
5912 (convert_mult_to_fma): Added deferring logic, split actual
5913 transformation to convert_mult_to_fma_1.
5914 (math_opts_dom_walker): New type.
5915 (math_opts_dom_walker::after_dom_children): New method, body moved
5916 here from pass_optimize_widening_mul::execute, added deferring logic
5918 (pass_optimize_widening_mul::execute): Moved most of code to
5919 math_opts_dom_walker::after_dom_children.
5920 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5921 * config/i386/i386.c (ix86_option_override_internal): Added
5922 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5924 2018-01-12 Richard Biener <rguenther@suse.de>
5927 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5928 inline instance vars.
5930 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5933 * config/rx/rx.c (rx_is_restricted_memory_address):
5936 2018-01-12 Richard Biener <rguenther@suse.de>
5938 PR tree-optimization/80846
5939 * target.def (split_reduction): New target hook.
5940 * targhooks.c (default_split_reduction): New function.
5941 * targhooks.h (default_split_reduction): Declare.
5942 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5943 target requests first reduce vectors by combining low and high
5945 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5946 (get_vectype_for_scalar_type_and_size): Export.
5947 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5948 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5949 * doc/tm.texi: Regenerate.
5950 * config/i386/i386.c (ix86_split_reduction): Implement
5951 TARGET_VECTORIZE_SPLIT_REDUCTION.
5953 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5956 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5957 in PIC mode except for TARGET_VXWORKS_RTP.
5958 * config/sparc/sparc.c: Include cfgrtl.h.
5959 (TARGET_INIT_PIC_REG): Define.
5960 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5961 (sparc_pic_register_p): New predicate.
5962 (sparc_legitimate_address_p): Use it.
5963 (sparc_legitimize_pic_address): Likewise.
5964 (sparc_delegitimize_address): Likewise.
5965 (sparc_mode_dependent_address_p): Likewise.
5966 (gen_load_pcrel_sym): Remove 4th parameter.
5967 (load_got_register): Adjust call to above. Remove obsolete stuff.
5968 (sparc_expand_prologue): Do not call load_got_register here.
5969 (sparc_flat_expand_prologue): Likewise.
5970 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5971 (sparc_use_pseudo_pic_reg): New function.
5972 (sparc_init_pic_reg): Likewise.
5973 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5974 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5976 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5978 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5979 Add item for branch_cost.
5981 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5983 PR rtl-optimization/83565
5984 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5985 not extend the result to a larger mode for rotate operations.
5986 (num_sign_bit_copies1): Likewise.
5988 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5991 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5993 Use values-Xc.o for -pedantic.
5994 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5996 2018-01-12 Martin Liska <mliska@suse.cz>
5999 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6001 (possible_polymorphic_call_targets): Use it.
6002 (ipa_devirt): Likewise.
6004 2018-01-12 Martin Liska <mliska@suse.cz>
6006 * profile-count.h (enum profile_quality): Use 0 as invalid
6007 enum value of profile_quality.
6009 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6011 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6012 -mext-string options.
6014 2018-01-12 Richard Biener <rguenther@suse.de>
6016 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6017 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6018 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6020 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6022 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6024 * configure.ac (--with-long-double-format): Add support for the
6025 configuration option to change the default long double format on
6027 * config.gcc (powerpc*-linux*-*): Likewise.
6028 * configure: Regenerate.
6029 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6030 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6031 used without modification.
6033 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6035 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6036 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6037 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6038 MISC_BUILTIN_SPEC_BARRIER.
6039 (rs6000_init_builtins): Likewise.
6040 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6042 (speculation_barrier): New define_insn.
6043 * doc/extend.texi: Document __builtin_speculation_barrier.
6045 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6048 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6049 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6050 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6052 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6053 integral modes instead of "ss" and "sd".
6054 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6055 vectors with 32-bit and 64-bit elements.
6056 (vecdupssescalarmodesuffix): New mode attribute.
6057 (vec_dup<mode>): Use it.
6059 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6062 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6063 frame if argument is passed on stack.
6065 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6068 * ree.c (combine_reaching_defs): Optimize also
6069 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6070 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6072 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6075 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6077 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6080 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6081 after they are computed.
6083 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6085 PR tree-optimization/83695
6086 * gimple-loop-linterchange.cc
6087 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6088 reset cached scev information after interchange.
6089 (pass_linterchange::execute): Remove call to scev_reset_htab.
6091 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6093 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6094 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6095 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6096 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6097 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6098 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6099 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6100 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6101 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6102 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6103 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6104 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6105 (V_lane_reg): Likewise.
6106 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6108 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6109 (vfmal_lane_low<mode>_intrinsic,
6110 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6111 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6112 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6113 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6114 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6115 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6117 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6119 * config/arm/arm-cpus.in (fp16fml): New feature.
6120 (ALL_SIMD): Add fp16fml.
6121 (armv8.2-a): Add fp16fml as an option.
6122 (armv8.3-a): Likewise.
6123 (armv8.4-a): Add fp16fml as part of fp16.
6124 * config/arm/arm.h (TARGET_FP16FML): Define.
6125 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6127 * config/arm/arm-modes.def (V2HF): Define.
6128 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6129 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6130 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6131 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6132 vfmsl_low, vfmsl_high): New set of builtins.
6133 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6134 (vfml_op): New code attribute.
6135 (VFMLHALVES): New int iterator.
6136 (VFML, VFMLSEL): New mode attributes.
6137 (V_reg): Define mapping for V2HF.
6138 (V_hi, V_lo): New mode attributes.
6139 (VF_constraint): Likewise.
6140 (vfml_half, vfml_half_selector): New int attributes.
6141 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6143 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6144 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6146 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6147 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6148 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6149 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6151 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6152 Document new effective target and option set.
6154 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6156 * config/arm/arm-cpus.in (armv8_4): New feature.
6157 (ARMv8_4a): New fgroup.
6158 (armv8.4-a): New arch.
6159 * config/arm/arm-tables.opt: Regenerate.
6160 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6161 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6162 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6163 Add matching rules for -march=armv8.4-a and extensions.
6164 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6166 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6169 * config/rx/rx.md (BW): New mode attribute.
6170 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6172 2018-01-11 Richard Biener <rguenther@suse.de>
6174 PR tree-optimization/83435
6175 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6176 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6177 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6179 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6180 Alan Hayward <alan.hayward@arm.com>
6181 David Sherwood <david.sherwood@arm.com>
6183 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6185 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6186 (aarch64_print_address_internal): Use it to check for a zero offset.
6188 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6189 Alan Hayward <alan.hayward@arm.com>
6190 David Sherwood <david.sherwood@arm.com>
6192 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6193 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6194 Return a poly_int64 rather than a HOST_WIDE_INT.
6195 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6196 rather than a HOST_WIDE_INT.
6197 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6198 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6199 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6200 final_offset from HOST_WIDE_INT to poly_int64.
6201 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6202 to_constant when getting the number of units in an Advanced SIMD
6204 (aarch64_builtin_vectorized_function): Check for a constant number
6206 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6208 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6209 attribute instead of GET_MODE_NUNITS.
6210 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6211 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6212 GET_MODE_SIZE for fixed-size registers.
6213 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6214 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6215 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6216 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6217 (aarch64_print_operand, aarch64_print_address_internal)
6218 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6219 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6220 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6221 Handle polynomial GET_MODE_SIZE.
6222 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6223 wider than SImode without modification.
6224 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6225 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6226 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6227 passing and returning SVE modes.
6228 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6229 rather than GEN_INT.
6230 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6231 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6232 (aarch64_allocate_and_probe_stack_space): Likewise.
6233 (aarch64_layout_frame): Cope with polynomial offsets.
6234 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6235 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6237 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6238 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6239 poly_int64 rather than a HOST_WIDE_INT.
6240 (aarch64_get_separate_components, aarch64_process_components)
6241 (aarch64_expand_prologue, aarch64_expand_epilogue)
6242 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6243 (aarch64_anchor_offset): New function, split out from...
6244 (aarch64_legitimize_address): ...here.
6245 (aarch64_builtin_vectorization_cost): Handle polynomial
6246 TYPE_VECTOR_SUBPARTS.
6247 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6249 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6250 number of elements from the PARALLEL rather than the mode.
6251 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6252 rather than GET_MODE_BITSIZE.
6253 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6254 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6255 (aarch64_expand_vec_perm_const_1): Handle polynomial
6256 d->perm.length () and d->perm elements.
6257 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6258 Apply to_constant to d->perm elements.
6259 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6260 polynomial CONST_VECTOR_NUNITS.
6261 (aarch64_move_pointer): Take amount as a poly_int64 rather
6263 (aarch64_progress_pointer): Avoid temporary variable.
6264 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6265 the mode attribute instead of GET_MODE.
6267 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6268 Alan Hayward <alan.hayward@arm.com>
6269 David Sherwood <david.sherwood@arm.com>
6271 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6272 x exists before using it.
6273 (aarch64_add_constant_internal): Rename to...
6274 (aarch64_add_offset_1): ...this. Replace regnum with separate
6275 src and dest rtxes. Handle the case in which they're different,
6276 including when the offset is zero. Replace scratchreg with an rtx.
6277 Use 2 additions if there is no spare register into which we can
6278 move a 16-bit constant.
6279 (aarch64_add_constant): Delete.
6280 (aarch64_add_offset): Replace reg with separate src and dest
6281 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6282 Use aarch64_add_offset_1.
6283 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6284 an rtx rather than an int. Take the delta as a poly_int64
6285 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6286 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6287 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6288 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6289 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6291 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6292 aarch64_add_constant.
6294 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6296 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6297 Use scalar_float_mode.
6299 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6301 * config/aarch64/aarch64-simd.md
6302 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6303 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6304 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6305 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6306 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6307 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6308 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6309 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6310 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6311 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6313 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6316 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6317 targ_options->x_arm_arch_string is non NULL.
6319 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6321 * config/aarch64/aarch64.h
6322 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6324 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6327 * expmed.c (emit_store_flag_force): Swap if const op0
6328 and change VOIDmode to mode of op0.
6330 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6332 PR rtl-optimization/83761
6333 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6334 than bytes to mode_for_size.
6336 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6339 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6340 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6343 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6346 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6347 when in layout mode.
6348 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6349 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6352 2018-01-10 Michael Collison <michael.collison@arm.com>
6354 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6355 * config/aarch64/aarch64-option-extension.def: Add
6356 AARCH64_OPT_EXTENSION of 'fp16fml'.
6357 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6358 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6359 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6360 * config/aarch64/constraints.md (Ui7): New constraint.
6361 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6362 (VFMLA_SEL_W): Ditto.
6365 (VFMLA16_LOW): New int iterator.
6366 (VFMLA16_HIGH): Ditto.
6367 (UNSPEC_FMLAL): New unspec.
6368 (UNSPEC_FMLSL): Ditto.
6369 (UNSPEC_FMLAL2): Ditto.
6370 (UNSPEC_FMLSL2): Ditto.
6371 (f16mac): New code attribute.
6372 * config/aarch64/aarch64-simd-builtins.def
6373 (aarch64_fmlal_lowv2sf): Ditto.
6374 (aarch64_fmlsl_lowv2sf): Ditto.
6375 (aarch64_fmlalq_lowv4sf): Ditto.
6376 (aarch64_fmlslq_lowv4sf): Ditto.
6377 (aarch64_fmlal_highv2sf): Ditto.
6378 (aarch64_fmlsl_highv2sf): Ditto.
6379 (aarch64_fmlalq_highv4sf): Ditto.
6380 (aarch64_fmlslq_highv4sf): Ditto.
6381 (aarch64_fmlal_lane_lowv2sf): Ditto.
6382 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6383 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6384 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6385 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6386 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6387 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6388 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6389 (aarch64_fmlal_lane_highv2sf): Ditto.
6390 (aarch64_fmlsl_lane_highv2sf): Ditto.
6391 (aarch64_fmlal_laneq_highv2sf): Ditto.
6392 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6393 (aarch64_fmlalq_lane_highv4sf): Ditto.
6394 (aarch64_fmlsl_lane_highv4sf): Ditto.
6395 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6396 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6397 * config/aarch64/aarch64-simd.md:
6398 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6399 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6400 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6401 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6402 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6403 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6404 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6405 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6406 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6407 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6408 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6409 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6410 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6411 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6412 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6413 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6414 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6415 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6416 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6417 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6418 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6419 (vfmlsl_low_u32): Ditto.
6420 (vfmlalq_low_u32): Ditto.
6421 (vfmlslq_low_u32): Ditto.
6422 (vfmlal_high_u32): Ditto.
6423 (vfmlsl_high_u32): Ditto.
6424 (vfmlalq_high_u32): Ditto.
6425 (vfmlslq_high_u32): Ditto.
6426 (vfmlal_lane_low_u32): Ditto.
6427 (vfmlsl_lane_low_u32): Ditto.
6428 (vfmlal_laneq_low_u32): Ditto.
6429 (vfmlsl_laneq_low_u32): Ditto.
6430 (vfmlalq_lane_low_u32): Ditto.
6431 (vfmlslq_lane_low_u32): Ditto.
6432 (vfmlalq_laneq_low_u32): Ditto.
6433 (vfmlslq_laneq_low_u32): Ditto.
6434 (vfmlal_lane_high_u32): Ditto.
6435 (vfmlsl_lane_high_u32): Ditto.
6436 (vfmlal_laneq_high_u32): Ditto.
6437 (vfmlsl_laneq_high_u32): Ditto.
6438 (vfmlalq_lane_high_u32): Ditto.
6439 (vfmlslq_lane_high_u32): Ditto.
6440 (vfmlalq_laneq_high_u32): Ditto.
6441 (vfmlslq_laneq_high_u32): Ditto.
6442 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6443 (AARCH64_FL_FOR_ARCH8_4): New.
6444 (AARCH64_ISA_F16FML): New ISA flag.
6445 (TARGET_F16FML): New feature flag for fp16fml.
6446 (doc/invoke.texi): Document new fp16fml option.
6448 2018-01-10 Michael Collison <michael.collison@arm.com>
6450 * config/aarch64/aarch64-builtins.c:
6451 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6452 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6453 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6454 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6455 (AARCH64_ISA_SHA3): New ISA flag.
6456 (TARGET_SHA3): New feature flag for sha3.
6457 * config/aarch64/iterators.md (sha512_op): New int attribute.
6458 (CRYPTO_SHA512): New int iterator.
6459 (UNSPEC_SHA512H): New unspec.
6460 (UNSPEC_SHA512H2): Ditto.
6461 (UNSPEC_SHA512SU0): Ditto.
6462 (UNSPEC_SHA512SU1): Ditto.
6463 * config/aarch64/aarch64-simd-builtins.def
6464 (aarch64_crypto_sha512hqv2di): New builtin.
6465 (aarch64_crypto_sha512h2qv2di): Ditto.
6466 (aarch64_crypto_sha512su0qv2di): Ditto.
6467 (aarch64_crypto_sha512su1qv2di): Ditto.
6468 (aarch64_eor3qv8hi): Ditto.
6469 (aarch64_rax1qv2di): Ditto.
6470 (aarch64_xarqv2di): Ditto.
6471 (aarch64_bcaxqv8hi): Ditto.
6472 * config/aarch64/aarch64-simd.md:
6473 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6474 (aarch64_crypto_sha512su0qv2di): Ditto.
6475 (aarch64_crypto_sha512su1qv2di): Ditto.
6476 (aarch64_eor3qv8hi): Ditto.
6477 (aarch64_rax1qv2di): Ditto.
6478 (aarch64_xarqv2di): Ditto.
6479 (aarch64_bcaxqv8hi): Ditto.
6480 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6481 (vsha512h2q_u64): Ditto.
6482 (vsha512su0q_u64): Ditto.
6483 (vsha512su1q_u64): Ditto.
6484 (veor3q_u16): Ditto.
6485 (vrax1q_u64): Ditto.
6487 (vbcaxq_u16): Ditto.
6488 * config/arm/types.md (crypto_sha512): New type attribute.
6489 (crypto_sha3): Ditto.
6490 (doc/invoke.texi): Document new sha3 option.
6492 2018-01-10 Michael Collison <michael.collison@arm.com>
6494 * config/aarch64/aarch64-builtins.c:
6495 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6496 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6497 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6498 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6499 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6500 (AARCH64_ISA_SM4): New ISA flag.
6501 (TARGET_SM4): New feature flag for sm4.
6502 * config/aarch64/aarch64-simd-builtins.def
6503 (aarch64_sm3ss1qv4si): Ditto.
6504 (aarch64_sm3tt1aq4si): Ditto.
6505 (aarch64_sm3tt1bq4si): Ditto.
6506 (aarch64_sm3tt2aq4si): Ditto.
6507 (aarch64_sm3tt2bq4si): Ditto.
6508 (aarch64_sm3partw1qv4si): Ditto.
6509 (aarch64_sm3partw2qv4si): Ditto.
6510 (aarch64_sm4eqv4si): Ditto.
6511 (aarch64_sm4ekeyqv4si): Ditto.
6512 * config/aarch64/aarch64-simd.md:
6513 (aarch64_sm3ss1qv4si): Ditto.
6514 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6515 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6516 (aarch64_sm4eqv4si): Ditto.
6517 (aarch64_sm4ekeyqv4si): Ditto.
6518 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6519 (sm3part_op): Ditto.
6520 (CRYPTO_SM3TT): Ditto.
6521 (CRYPTO_SM3PART): Ditto.
6522 (UNSPEC_SM3SS1): New unspec.
6523 (UNSPEC_SM3TT1A): Ditto.
6524 (UNSPEC_SM3TT1B): Ditto.
6525 (UNSPEC_SM3TT2A): Ditto.
6526 (UNSPEC_SM3TT2B): Ditto.
6527 (UNSPEC_SM3PARTW1): Ditto.
6528 (UNSPEC_SM3PARTW2): Ditto.
6529 (UNSPEC_SM4E): Ditto.
6530 (UNSPEC_SM4EKEY): Ditto.
6531 * config/aarch64/constraints.md (Ui2): New constraint.
6532 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6533 * config/arm/types.md (crypto_sm3): New type attribute.
6534 (crypto_sm4): Ditto.
6535 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6536 (vsm3tt1aq_u32): Ditto.
6537 (vsm3tt1bq_u32): Ditto.
6538 (vsm3tt2aq_u32): Ditto.
6539 (vsm3tt2bq_u32): Ditto.
6540 (vsm3partw1q_u32): Ditto.
6541 (vsm3partw2q_u32): Ditto.
6542 (vsm4eq_u32): Ditto.
6543 (vsm4ekeyq_u32): Ditto.
6544 (doc/invoke.texi): Document new sm4 option.
6546 2018-01-10 Michael Collison <michael.collison@arm.com>
6548 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6549 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6550 (AARCH64_FL_FOR_ARCH8_4): New.
6551 (AARCH64_FL_V8_4): New flag.
6552 (doc/invoke.texi): Document new armv8.4-a option.
6554 2018-01-10 Michael Collison <michael.collison@arm.com>
6556 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6557 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6558 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6559 * config/aarch64/aarch64-option-extension.def: Add
6560 AARCH64_OPT_EXTENSION of 'sha2'.
6561 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6562 (crypto): Disable sha2 and aes if crypto disabled.
6563 (crypto): Enable aes and sha2 if enabled.
6564 (simd): Disable sha2 and aes if simd disabled.
6565 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6567 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6568 (TARGET_SHA2): New feature flag for sha2.
6569 (TARGET_AES): New feature flag for aes.
6570 * config/aarch64/aarch64-simd.md:
6571 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6572 conditional on TARGET_AES.
6573 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6574 (aarch64_crypto_sha1hsi): Make pattern conditional
6576 (aarch64_crypto_sha1hv4si): Ditto.
6577 (aarch64_be_crypto_sha1hv4si): Ditto.
6578 (aarch64_crypto_sha1su1v4si): Ditto.
6579 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6580 (aarch64_crypto_sha1su0v4si): Ditto.
6581 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6582 (aarch64_crypto_sha256su0v4si): Ditto.
6583 (aarch64_crypto_sha256su1v4si): Ditto.
6584 (doc/invoke.texi): Document new aes and sha2 options.
6586 2018-01-10 Martin Sebor <msebor@redhat.com>
6588 PR tree-optimization/83781
6589 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6592 2018-01-11 Martin Sebor <msebor@gmail.com>
6593 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6595 PR tree-optimization/83501
6596 PR tree-optimization/81703
6598 * tree-ssa-strlen.c (get_string_cst): Rename...
6599 (get_string_len): ...to this. Handle global constants.
6600 (handle_char_store): Adjust.
6602 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6603 Jim Wilson <jimw@sifive.com>
6605 * config/riscv/riscv-protos.h (riscv_output_return): New.
6606 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6607 (riscv_attribute_table, riscv_output_return),
6608 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6609 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6610 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6612 (riscv_expand_prologue): Add early return for naked function.
6613 (riscv_expand_epilogue): Likewise.
6614 (riscv_function_ok_for_sibcall): Return false for naked function.
6615 (riscv_set_current_function): New.
6616 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6617 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6618 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6619 * doc/extend.texi (RISC-V Function Attributes): New.
6621 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6623 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6624 check for 128-bit long double before checking TCmode.
6625 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6626 128-bit long doubles before checking TFmode or TCmode.
6627 (FLOAT128_IBM_P): Likewise.
6629 2018-01-10 Martin Sebor <msebor@redhat.com>
6631 PR tree-optimization/83671
6632 * builtins.c (c_strlen): Unconditionally return zero for the empty
6634 Use -Warray-bounds for warnings.
6635 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6636 for non-constant array indices with COMPONENT_REF, arrays of
6637 arrays, and pointers to arrays.
6638 (gimple_fold_builtin_strlen): Determine and set length range for
6639 non-constant character arrays.
6641 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6644 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6647 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6649 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6651 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6654 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6655 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6656 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6657 indexed_or_indirect_operand predicate.
6658 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6659 (*vsx_le_perm_load_v8hi): Likewise.
6660 (*vsx_le_perm_load_v16qi): Likewise.
6661 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6662 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6663 (*vsx_le_perm_store_v8hi): Likewise.
6664 (*vsx_le_perm_store_v16qi): Likewise.
6665 (eight unnamed splitters): Likewise.
6667 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6669 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6670 * config/rs6000/emmintrin.h: Likewise.
6671 * config/rs6000/mmintrin.h: Likewise.
6672 * config/rs6000/xmmintrin.h: Likewise.
6674 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6677 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6679 * tree.c (tree_nop_conversion): Return true for location wrapper
6681 (maybe_wrap_with_location): New function.
6682 (selftest::check_strip_nops): New function.
6683 (selftest::test_location_wrappers): New function.
6684 (selftest::tree_c_tests): Call it.
6685 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6686 (maybe_wrap_with_location): New decl.
6687 (EXPR_LOCATION_WRAPPER_P): New macro.
6688 (location_wrapper_p): New inline function.
6689 (tree_strip_any_location_wrapper): New inline function.
6691 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6694 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6695 stack_realign_offset for the largest alignment of stack slot
6697 (ix86_find_max_used_stack_alignment): New function.
6698 (ix86_finalize_stack_frame_flags): Use it. Set
6699 max_used_stack_alignment if we don't realign stack.
6700 * config/i386/i386.h (machine_function): Add
6701 max_used_stack_alignment.
6703 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6705 * config/arm/arm.opt (-mbranch-cost): New option.
6706 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6709 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6712 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6713 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6715 2018-01-10 Richard Biener <rguenther@suse.de>
6718 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6719 early out so it also covers the case where we have a non-NULL
6722 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6724 PR tree-optimization/83753
6725 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6726 for non-strided grouped accesses if the number of elements is 1.
6728 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6731 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6732 * i386.h (TARGET_USE_GATHER): Define.
6733 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6735 2018-01-10 Martin Liska <mliska@suse.cz>
6738 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6739 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6741 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6742 CLEANUP_NO_PARTITIONING is not set.
6744 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6746 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6747 for vectors, as a partial revert of r254296.
6748 * rtl.h (const_vec_p): Delete.
6749 (const_vec_duplicate_p): Don't test for vector CONSTs.
6750 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6751 * expmed.c (make_tree): Likewise.
6754 * common.md (E, F): Use CONSTANT_P instead of checking for
6756 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6757 checking for CONST_VECTOR.
6759 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6762 * predict.c (force_edge_cold): Handle in more sane way edges
6765 2018-01-09 Carl Love <cel@us.ibm.com>
6767 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6769 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6770 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6771 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6772 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6773 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6774 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6775 * config/rs6000/rs6000-protos.h: Add extern defition for
6776 rs6000_generate_float2_double_code.
6777 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6779 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6780 (float2_v2df): Add define_expand.
6782 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6785 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6786 op_mode in the force_to_mode call.
6788 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6790 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6791 instead of checking each element individually.
6792 (aarch64_evpc_uzp): Likewise.
6793 (aarch64_evpc_zip): Likewise.
6794 (aarch64_evpc_ext): Likewise.
6795 (aarch64_evpc_rev): Likewise.
6796 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6797 instead of checking each element individually. Return true without
6799 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6800 whether all selected elements come from the same input, instead of
6801 checking each element individually. Remove calls to gen_rtx_REG,
6802 start_sequence and end_sequence and instead assert that no rtl is
6805 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6807 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6808 order of HIGH and CONST checks.
6810 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6812 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6813 if the destination isn't an SSA_NAME.
6815 2018-01-09 Richard Biener <rguenther@suse.de>
6817 PR tree-optimization/83668
6818 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6820 (canonicalize_loop_form): ... here, renamed from ...
6821 (canonicalize_loop_closed_ssa_form): ... this and amended to
6822 swap successor edges for loop exit blocks to make us use
6823 the RPO order we need for initial schedule generation.
6825 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6827 PR tree-optimization/64811
6828 * match.pd: When optimizing comparisons with Inf, avoid
6829 introducing or losing exceptions from comparisons with NaN.
6831 2018-01-09 Martin Liska <mliska@suse.cz>
6834 * asan.c (shadow_mem_size): Add gcc_assert.
6836 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6838 Don't save registers in main().
6841 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6842 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6843 * config/avr/avr.c (avr_set_current_function): Don't error if
6844 naked, OS_task or OS_main are specified at the same time.
6845 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6847 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6849 * common/config/avr/avr-common.c (avr_option_optimization_table):
6850 Switch on -mmain-is-OS_task for optimizing compilations.
6852 2018-01-09 Richard Biener <rguenther@suse.de>
6854 PR tree-optimization/83572
6855 * graphite.c: Include cfganal.h.
6856 (graphite_transform_loops): Connect infinite loops to exit
6857 and remove fake edges at the end.
6859 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6861 * ipa-inline.c (edge_badness): Revert accidental checkin.
6863 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6866 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6867 symbols; not inline clones.
6869 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6872 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6873 hard registers. Formatting fixes.
6875 PR preprocessor/83722
6876 * gcc.c (try_generate_repro): Pass
6877 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6878 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6881 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6882 Kito Cheng <kito.cheng@gmail.com>
6884 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6885 (riscv_leaf_function_p): Delete.
6886 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6888 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6890 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6892 (do_ifelse): New function.
6893 (do_isel): New function.
6894 (do_sub3): New function.
6895 (do_add3): New function.
6896 (do_load_mask_compare): New function.
6897 (do_overlap_load_compare): New function.
6898 (expand_compare_loop): New function.
6899 (expand_block_compare): Call expand_compare_loop() when appropriate.
6900 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6902 (-mblock-compare-inline-loop-limit): New option.
6904 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6907 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6908 Reverse order of second and third operands in first alternative.
6909 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6910 of first and second elements in UNSPEC_VPERMR vector.
6911 (altivec_expand_vec_perm_le): Likewise.
6913 2018-01-08 Jeff Law <law@redhat.com>
6915 PR rtl-optimizatin/81308
6916 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6917 (process_switch): If group_case_labels makes a change, then set
6919 (pass_convert_switch::execute): If a switch is converted, then
6920 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6922 PR rtl-optimization/81308
6923 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6926 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6928 PR target/83663 - Revert r255946
6929 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6930 generation for cases where splatting a value is not useful.
6931 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6932 across a vec_duplicate and a paradoxical subreg forming a vector
6933 mode to a vec_concat.
6935 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6937 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6938 -march=armv8.3-a variants.
6939 * config/arm/t-multilib: Likewise.
6940 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6942 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6944 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6946 (cceq_ior_compare_complement): Give it a name so I can use it, and
6947 change boolean_or_operator predicate to boolean_operator so it can
6948 be used to generate a crand.
6949 (eqne): New code iterator.
6950 (bd/bd_neg): New code_attrs.
6951 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6952 a single define_insn.
6953 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6954 decrement (bdnzt/bdnzf/bdzt/bdzf).
6955 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6956 with the new names of the branch decrement patterns, and added the
6957 names of the branch decrement conditional patterns.
6959 2018-01-08 Richard Biener <rguenther@suse.de>
6961 PR tree-optimization/83563
6962 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6965 2018-01-08 Richard Biener <rguenther@suse.de>
6968 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6970 2018-01-08 Richard Biener <rguenther@suse.de>
6972 PR tree-optimization/83685
6973 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6974 references to abnormals.
6976 2018-01-08 Richard Biener <rguenther@suse.de>
6979 * dwarf2out.c (output_indirect_strings): Handle empty
6980 skeleton_debug_str_hash.
6981 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6983 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6985 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6986 (emit_store_direct): Likewise.
6987 (arc_trampoline_adjust_address): Likewise.
6988 (arc_asm_trampoline_template): New function.
6989 (arc_initialize_trampoline): Use asm_trampoline_template.
6990 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6991 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6992 * config/arc/arc.md (flush_icache): Delete pattern.
6994 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6996 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6997 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7000 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7003 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7004 by not USED_FOR_TARGET.
7005 (make_pass_resolve_sw_modes): Likewise.
7007 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7009 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7012 2018-01-08 Richard Biener <rguenther@suse.de>
7015 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7017 2018-01-08 Richard Biener <rguenther@suse.de>
7020 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7022 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7025 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7026 basic blocks with a small number of successors.
7027 (convert_control_dep_chain_into_preds): Improve handling of
7029 (dump_predicates): Split apart into...
7030 (dump_pred_chain): ...here...
7031 (dump_pred_info): ...and here.
7032 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7033 (can_chain_union_be_invalidated_p): Improve check for invalidation
7035 (uninit_uses_cannot_happen): Avoid unnecessary if
7036 convert_control_dep_chain_into_preds yielded nothing.
7038 2018-01-06 Martin Sebor <msebor@redhat.com>
7040 PR tree-optimization/83640
7041 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7042 subtracting negative offset from size.
7043 (builtin_access::overlap): Adjust offset bounds of the access to fall
7044 within the size of the object if possible.
7046 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7048 PR rtl-optimization/83699
7049 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7050 extract_bit_field_as_subreg to cases in which the extracted
7051 value is also a vector.
7053 * lra-constraints.c (process_alt_operands): Test for the equivalence
7054 substitutions when detecting a possible reload cycle.
7056 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7059 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7060 by default if flag_selective_schedling{,2}. Formatting fixes.
7062 PR rtl-optimization/83682
7063 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7064 if it has non-VECTOR_MODE element mode.
7065 (vec_duplicate_p): Likewise.
7068 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7069 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7071 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7074 * config/i386/i386-builtin.def
7075 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7076 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7077 Require also OPTION_MASK_ISA_AVX512F in addition to
7078 OPTION_MASK_ISA_GFNI.
7079 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7080 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7081 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7082 to OPTION_MASK_ISA_GFNI.
7083 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7084 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7085 OPTION_MASK_ISA_AVX512BW.
7086 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7087 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7088 addition to OPTION_MASK_ISA_GFNI.
7089 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7090 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7091 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7092 to OPTION_MASK_ISA_GFNI.
7093 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7094 a requirement for all ISAs rather than any of them with a few
7096 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7098 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7099 bitmasks to be enabled with 3 exceptions, instead of requiring any
7100 enabled ISA with lots of exceptions.
7101 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7102 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7103 Change avx512bw in isa attribute to avx512f.
7104 * config/i386/sgxintrin.h: Add license boilerplate.
7105 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7106 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7107 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7108 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7110 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7111 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7112 temporarily sse2 rather than sse if not enabled already.
7115 * config/i386/sse.md (VI248_VLBW): Rename to ...
7116 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7117 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7118 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7119 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7120 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7121 mode iterator instead of VI248_VLBW.
7123 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7125 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7126 (record_modified): Skip clobbers; add debug output.
7127 (param_change_prob): Use sreal frequencies.
7129 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7131 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7132 punt for user-aligned variables.
7134 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7136 * tree-chrec.c (chrec_contains_symbols): Return true for
7139 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7142 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7143 of (x|y) == x for BICS pattern.
7145 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7147 PR tree-optimization/83605
7148 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7149 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7152 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7154 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7155 * config/epiphany/rtems.h: New file.
7157 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7158 Uros Bizjak <ubizjak@gmail.com>
7161 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7162 QIreg_operand instead of register_operand predicate.
7163 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7164 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7165 comments instead of -fmitigate[-_]rop.
7167 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7170 * cgraphunit.c (symbol_table::compile): Switch to text_section
7171 before calling assembly_start debug hook.
7172 * run-rtl-passes.c (run_rtl_passes): Likewise.
7175 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7177 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7178 range_int_cst_p rather than !symbolic_range_p before calling
7179 extract_range_from_multiplicative_op_1.
7181 2018-01-04 Jeff Law <law@redhat.com>
7183 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7184 redundant test in assertion.
7186 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7188 * doc/rtl.texi: Document machine_mode wrapper classes.
7190 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7192 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7195 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7197 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7198 the VEC_PERM_EXPR fold to fail.
7200 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7203 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7204 to switched_sections.
7206 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7209 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7212 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7215 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7216 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7218 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7221 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7222 is BLKmode and bitpos not zero or mode change is needed.
7224 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7227 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7230 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7233 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7234 instead of MULT rtx. Update all corresponding splitters.
7236 (*ssub<modesuffix>): Ditto.
7238 (*cmp_sadd_di): Update split patterns.
7239 (*cmp_sadd_si): Ditto.
7240 (*cmp_sadd_sidi): Ditto.
7241 (*cmp_ssub_di): Ditto.
7242 (*cmp_ssub_si): Ditto.
7243 (*cmp_ssub_sidi): Ditto.
7244 * config/alpha/predicates.md (const23_operand): New predicate.
7245 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7246 Look for ASHIFT, not MULT inner operand.
7247 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7249 2018-01-04 Martin Liska <mliska@suse.cz>
7251 PR gcov-profile/83669
7252 * gcov.c (output_intermediate_file): Add version to intermediate
7254 * doc/gcov.texi: Document new field 'version' in intermediate
7255 file format. Fix location of '-k' option of gcov command.
7257 2018-01-04 Martin Liska <mliska@suse.cz>
7260 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7262 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7264 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7266 2018-01-03 Martin Sebor <msebor@redhat.com>
7268 PR tree-optimization/83655
7269 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7270 checking calls with invalid arguments.
7272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7274 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7275 (vectorizable_mask_load_store): Delete.
7276 (vectorizable_call): Return false for masked loads and stores.
7277 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7278 instead of gimple_assign_rhs1.
7279 (vectorizable_load): Handle IFN_MASK_LOAD.
7280 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7284 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7286 (vectorizable_mask_load_store): ...here.
7287 (vectorizable_load): ...and here.
7289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7291 * tree-vect-stmts.c (vect_build_all_ones_mask)
7292 (vect_build_zero_merge_argument): New functions, split out from...
7293 (vectorizable_load): ...here.
7295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7297 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7299 (vectorizable_mask_load_store): ...here.
7300 (vectorizable_store): ...and here.
7302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7304 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7306 (vectorizable_mask_load_store): ...here.
7308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7310 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7311 (vect_model_store_cost): Take a vec_load_store_type instead of a
7313 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7314 (vect_model_store_cost): Take a vec_load_store_type instead of a
7316 (vectorizable_mask_load_store): Update accordingly.
7317 (vectorizable_store): Likewise.
7318 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7322 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7323 IFN_MASK_LOAD calls here rather than...
7324 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
7331 fall back to extract_bit_field_as_subreg if vec_extract
7334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7335 Alan Hayward <alan.hayward@arm.com>
7336 David Sherwood <david.sherwood@arm.com>
7338 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7339 they are variable or constant sized.
7340 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7341 slots for constant-sized data.
7343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7344 Alan Hayward <alan.hayward@arm.com>
7345 David Sherwood <david.sherwood@arm.com>
7347 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7348 handling COND_EXPRs with boolean comparisons, try to find a better
7349 basis for the mask type than the boolean itself.
7351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7353 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7354 is calculated and how it can be overridden.
7355 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7356 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7358 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7362 Alan Hayward <alan.hayward@arm.com>
7363 David Sherwood <david.sherwood@arm.com>
7365 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7366 Remove the mode argument.
7367 (aarch64_simd_valid_immediate): Remove the mode and inverse
7369 * config/aarch64/iterators.md (bitsize): New iterator.
7370 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7371 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7372 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7373 aarch64_simd_valid_immediate.
7374 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7375 (aarch64_reg_or_bic_imm): Likewise.
7376 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7377 with an insn_type enum and msl with a modifier_type enum.
7378 Replace element_width with a scalar_mode. Change the shift
7379 to unsigned int. Add constructors for scalar_float_mode and
7380 scalar_int_mode elements.
7381 (aarch64_vect_float_const_representable_p): Delete.
7382 (aarch64_can_const_movi_rtx_p)
7383 (aarch64_simd_scalar_immediate_valid_for_move)
7384 (aarch64_simd_make_constant): Update call to
7385 aarch64_simd_valid_immediate.
7386 (aarch64_advsimd_valid_immediate_hs): New function.
7387 (aarch64_advsimd_valid_immediate): Likewise.
7388 (aarch64_simd_valid_immediate): Remove mode and inverse
7389 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7390 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7391 and aarch64_float_const_representable_p on the result.
7392 (aarch64_output_simd_mov_immediate): Remove mode argument.
7393 Update call to aarch64_simd_valid_immediate and use of
7394 simd_immediate_info.
7395 (aarch64_output_scalar_simd_mov_immediate): Update call
7398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7399 Alan Hayward <alan.hayward@arm.com>
7400 David Sherwood <david.sherwood@arm.com>
7402 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7403 (mode_nunits): Likewise CONST_MODE_NUNITS.
7404 * machmode.def (ADJUST_NUNITS): Document.
7405 * genmodes.c (mode_data::need_nunits_adj): New field.
7406 (blank_mode): Update accordingly.
7407 (adj_nunits): New variable.
7408 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7410 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7411 listed in adj_nunits.
7412 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7413 listed in adj_nunits. Don't emit case statements for such modes.
7414 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7415 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7416 nothing if adj_nunits is nonnull.
7417 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7418 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7419 (emit_mode_fbit): Update use of print_maybe_const_decl.
7420 (emit_move_size): Likewise. Treat the array as non-const
7422 (emit_mode_adjustments): Handle adj_nunits.
7424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7426 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7427 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7428 (VECTOR_MODES): Use it.
7429 (make_vector_modes): Take the prefix as an argument.
7431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7432 Alan Hayward <alan.hayward@arm.com>
7433 David Sherwood <david.sherwood@arm.com>
7435 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7436 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7437 for MODE_VECTOR_BOOL.
7438 * machmode.def (VECTOR_BOOL_MODE): Document.
7439 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7440 (make_vector_bool_mode): New function.
7441 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7443 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7444 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7446 * stor-layout.c (int_mode_for_mode): Likewise.
7447 * tree.c (build_vector_type_for_mode): Likewise.
7448 * varasm.c (output_constant_pool_2): Likewise.
7449 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7450 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7451 for MODE_VECTOR_BOOL.
7452 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7453 of mode class checks.
7454 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7455 instead of a list of mode class checks.
7456 (expand_vector_scalar_condition): Likewise.
7457 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7460 Alan Hayward <alan.hayward@arm.com>
7461 David Sherwood <david.sherwood@arm.com>
7463 * machmode.h (mode_size): Change from unsigned short to
7465 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7466 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7467 or if measurement_type is not polynomial.
7468 (fixed_size_mode::includes_p): Check for constant-sized modes.
7469 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7470 return a poly_uint16 rather than an unsigned short.
7471 (emit_mode_size): Change the type of mode_size from unsigned short
7472 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7473 (emit_mode_adjustments): Cope with polynomial vector sizes.
7474 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7476 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7478 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7479 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7480 * caller-save.c (setup_save_areas): Likewise.
7481 (replace_reg_with_saved_mem): Likewise.
7482 * calls.c (emit_library_call_value_1): Likewise.
7483 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7484 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7485 (gen_lowpart_for_combine): Likewise.
7486 * convert.c (convert_to_integer_1): Likewise.
7487 * cse.c (equiv_constant, cse_insn): Likewise.
7488 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7489 (cselib_subst_to_values): Likewise.
7490 * dce.c (word_dce_process_block): Likewise.
7491 * df-problems.c (df_word_lr_mark_ref): Likewise.
7492 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7493 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7494 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7495 (rtl_for_decl_location): Likewise.
7496 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7497 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7498 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7499 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7500 (expand_expr_real_1): Likewise.
7501 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7502 (pad_below): Likewise.
7503 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7504 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7505 * ira.c (get_subreg_tracking_sizes): Likewise.
7506 * ira-build.c (ira_create_allocno_objects): Likewise.
7507 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7508 (ira_sort_regnos_for_alter_reg): Likewise.
7509 * ira-costs.c (record_operand_costs): Likewise.
7510 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7511 (resolve_simple_move): Likewise.
7512 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7513 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7514 (lra_constraints): Likewise.
7515 (CONST_POOL_OK_P): Reject variable-sized modes.
7516 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7517 (add_pseudo_to_slot, lra_spill): Likewise.
7518 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7519 * optabs-query.c (get_best_extraction_insn): Likewise.
7520 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7521 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7522 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7523 * recog.c (offsettable_address_addr_space_p): Likewise.
7524 * regcprop.c (maybe_mode_change): Likewise.
7525 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7526 * regrename.c (build_def_use): Likewise.
7527 * regstat.c (dump_reg_info): Likewise.
7528 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7529 (find_reloads, find_reloads_subreg_address): Likewise.
7530 * reload1.c (eliminate_regs_1): Likewise.
7531 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7532 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7533 (simplify_binary_operation_1, simplify_subreg): Likewise.
7534 * targhooks.c (default_function_arg_padding): Likewise.
7535 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7536 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7537 (verify_gimple_assign_ternary): Likewise.
7538 * tree-inline.c (estimate_move_cost): Likewise.
7539 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7540 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7541 (get_address_cost_ainc): Likewise.
7542 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7543 (vect_supportable_dr_alignment): Likewise.
7544 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7545 (vectorizable_reduction): Likewise.
7546 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7547 (vectorizable_operation, vectorizable_load): Likewise.
7548 * tree.c (build_same_sized_truth_vector_type): Likewise.
7549 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7550 * var-tracking.c (emit_note_insn_var_location): Likewise.
7551 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7552 (ADDR_VEC_ALIGN): Likewise.
7554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7555 Alan Hayward <alan.hayward@arm.com>
7556 David Sherwood <david.sherwood@arm.com>
7558 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7560 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7561 or if measurement_type is polynomial.
7562 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7563 * combine.c (make_extraction): Likewise.
7564 * dse.c (find_shift_sequence): Likewise.
7565 * dwarf2out.c (mem_loc_descriptor): Likewise.
7566 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7567 (extract_bit_field, extract_low_bits): Likewise.
7568 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7569 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7570 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7571 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7572 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7573 * reload.c (find_reloads): Likewise.
7574 * reload1.c (alter_reg): Likewise.
7575 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7576 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7577 * tree-if-conv.c (predicate_mem_writes): Likewise.
7578 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7579 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7580 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7581 * valtrack.c (dead_debug_insert_temp): Likewise.
7582 * varasm.c (mergeable_constant_section): Likewise.
7583 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7586 Alan Hayward <alan.hayward@arm.com>
7587 David Sherwood <david.sherwood@arm.com>
7589 * expr.c (expand_assignment): Cope with polynomial mode sizes
7590 when assigning to a CONCAT.
7592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * machmode.h (mode_precision): Change from unsigned short to
7598 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7600 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7601 or if measurement_type is not polynomial.
7602 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7603 in which the mode is already known to be a scalar_int_mode.
7604 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7605 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7607 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7608 for GET_MODE_PRECISION.
7609 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7610 for GET_MODE_PRECISION.
7611 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7613 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7614 (expand_field_assignment, make_extraction): Likewise.
7615 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7616 (get_last_value): Likewise.
7617 * convert.c (convert_to_integer_1): Likewise.
7618 * cse.c (cse_insn): Likewise.
7619 * expr.c (expand_expr_real_1): Likewise.
7620 * lra-constraints.c (simplify_operand_subreg): Likewise.
7621 * optabs-query.c (can_atomic_load_p): Likewise.
7622 * optabs.c (expand_atomic_load): Likewise.
7623 (expand_atomic_store): Likewise.
7624 * ree.c (combine_reaching_defs): Likewise.
7625 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7626 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7627 * tree.h (type_has_mode_precision_p): Likewise.
7628 * ubsan.c (instrument_si_overflow): Likewise.
7630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7631 Alan Hayward <alan.hayward@arm.com>
7632 David Sherwood <david.sherwood@arm.com>
7634 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7635 polynomial numbers of units.
7636 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7637 (valid_vector_subparts_p): New function.
7638 (build_vector_type): Remove temporary shim and take the number
7639 of units as a poly_uint64 rather than an int.
7640 (build_opaque_vector_type): Take the number of units as a
7641 poly_uint64 rather than an int.
7642 * tree.c (build_vector_from_ctor): Handle polynomial
7643 TYPE_VECTOR_SUBPARTS.
7644 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7645 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7646 (build_vector_from_val): If the number of units is variable,
7647 use build_vec_duplicate_cst for constant operands and
7648 VEC_DUPLICATE_EXPR otherwise.
7649 (make_vector_type): Remove temporary is_constant ().
7650 (build_vector_type, build_opaque_vector_type): Take the number of
7651 units as a poly_uint64 rather than an int.
7652 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7654 * cfgexpand.c (expand_debug_expr): Likewise.
7655 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7656 (store_constructor, expand_expr_real_1): Likewise.
7657 (const_scalar_mask_from_tree): Likewise.
7658 * fold-const-call.c (fold_const_reduction): Likewise.
7659 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7660 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7661 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7662 (fold_relational_const): Likewise.
7663 (native_interpret_vector): Likewise. Change the size from an
7664 int to an unsigned int.
7665 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7666 TYPE_VECTOR_SUBPARTS.
7667 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7668 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7669 duplicating a non-constant operand into a variable-length vector.
7670 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7671 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7672 * ipa-icf.c (sem_variable::equals): Likewise.
7673 * match.pd: Likewise.
7674 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7675 * print-tree.c (print_node): Likewise.
7676 * stor-layout.c (layout_type): Likewise.
7677 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7678 * tree-cfg.c (verify_gimple_comparison): Likewise.
7679 (verify_gimple_assign_binary): Likewise.
7680 (verify_gimple_assign_ternary): Likewise.
7681 (verify_gimple_assign_single): Likewise.
7682 * tree-pretty-print.c (dump_generic_node): Likewise.
7683 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7684 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7685 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7686 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7687 (vect_shift_permute_load_chain): Likewise.
7688 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7689 (expand_vector_condition, optimize_vector_constructor): Likewise.
7690 (lower_vec_perm, get_compute_type): Likewise.
7691 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7692 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7693 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7694 (vect_recog_mask_conversion_pattern): Likewise.
7695 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7696 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7697 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7698 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7699 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7700 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7701 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7702 (supportable_widening_operation): Likewise.
7703 (supportable_narrowing_operation): Likewise.
7704 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7706 * varasm.c (output_constant): Likewise.
7708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7709 Alan Hayward <alan.hayward@arm.com>
7710 David Sherwood <david.sherwood@arm.com>
7712 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7713 so that both the length == 3 and length != 3 cases set up their
7714 own permute vectors. Add comments explaining why we know the
7715 number of elements is constant.
7716 (vect_permute_load_chain): Likewise.
7718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7719 Alan Hayward <alan.hayward@arm.com>
7720 David Sherwood <david.sherwood@arm.com>
7722 * machmode.h (mode_nunits): Change from unsigned char to
7724 (ONLY_FIXED_SIZE_MODES): New macro.
7725 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7726 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7727 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7729 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7730 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7731 or if measurement_type is not polynomial.
7732 * genmodes.c (ZERO_COEFFS): New macro.
7733 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7735 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7736 Use ZERO_COEFFS when emitting initializers.
7737 * data-streamer.h (bp_pack_poly_value): New function.
7738 (bp_unpack_poly_value): Likewise.
7739 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7740 for GET_MODE_NUNITS.
7741 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7742 for GET_MODE_NUNITS.
7743 * tree.c (make_vector_type): Remove temporary shim and make
7744 the real function take the number of units as a poly_uint64
7746 (build_vector_type_for_mode): Handle polynomial nunits.
7747 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7748 * emit-rtl.c (const_vec_series_p_1): Likewise.
7749 (gen_rtx_CONST_VECTOR): Likewise.
7750 * fold-const.c (test_vec_duplicate_folding): Likewise.
7751 * genrecog.c (validate_pattern): Likewise.
7752 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7753 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7754 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7755 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7756 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7757 * rtlanal.c (subreg_get_info): Likewise.
7758 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7759 (vect_grouped_load_supported): Likewise.
7760 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7761 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7762 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7763 (simplify_const_unary_operation, simplify_binary_operation_1)
7764 (simplify_const_binary_operation, simplify_ternary_operation)
7765 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7766 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7767 instead of CONST_VECTOR_NUNITS.
7768 * varasm.c (output_constant_pool_2): Likewise.
7769 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7770 explicit-encoded elements in the XVEC for variable-length vectors.
7772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7774 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7777 Alan Hayward <alan.hayward@arm.com>
7778 David Sherwood <david.sherwood@arm.com>
7780 * coretypes.h (fixed_size_mode): Declare.
7781 (fixed_size_mode_pod): New typedef.
7782 * builtins.h (target_builtins::x_apply_args_mode)
7783 (target_builtins::x_apply_result_mode): Change type to
7784 fixed_size_mode_pod.
7785 * builtins.c (apply_args_size, apply_result_size, result_vector)
7786 (expand_builtin_apply_args_1, expand_builtin_apply)
7787 (expand_builtin_return): Update accordingly.
7789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7791 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7792 * cselib.c (cselib_hash_rtx): Likewise.
7793 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7794 CONST_VECTOR encoding.
7796 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7797 Jeff Law <law@redhat.com>
7800 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7801 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7802 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7803 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7806 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7807 explicitly probe *sp in a noreturn function if there were any callee
7808 register saves or frame pointer is needed.
7810 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7813 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7814 BLKmode for ternary, binary or unary expressions.
7817 * var-tracking.c (delete_vta_debug_insn): New inline function.
7818 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7819 insns from get_insns () to NULL instead of each bb separately.
7820 Use delete_vta_debug_insn. No longer static.
7821 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7822 delete_vta_debug_insns callers.
7823 * rtl.h (delete_vta_debug_insns): Declare.
7824 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7825 instead of variable_tracking_main.
7827 2018-01-03 Martin Sebor <msebor@redhat.com>
7829 PR tree-optimization/83603
7830 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7831 arguments past the endof the argument list in functions declared
7832 without a prototype.
7833 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7834 Avoid checking when arguments are null.
7836 2018-01-03 Martin Sebor <msebor@redhat.com>
7839 * doc/extend.texi (attribute const): Fix a typo.
7840 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7841 issuing -Wsuggest-attribute for void functions.
7843 2018-01-03 Martin Sebor <msebor@redhat.com>
7845 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7846 offset_int::from instead of wide_int::to_shwi.
7847 (maybe_diag_overlap): Remove assertion.
7848 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7849 * gimple-ssa-sprintf.c (format_directive): Same.
7850 (parse_directive): Same.
7851 (sprintf_dom_walker::compute_format_length): Same.
7852 (try_substitute_return_value): Same.
7854 2018-01-03 Jeff Law <law@redhat.com>
7857 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7858 non-constant residual for zero at runtime and avoid probing in
7859 that case. Reorganize code for trailing problem to mirror handling
7862 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7864 PR tree-optimization/83501
7865 * tree-ssa-strlen.c (get_string_cst): New.
7866 (handle_char_store): Call get_string_cst.
7868 2018-01-03 Martin Liska <mliska@suse.cz>
7870 PR tree-optimization/83593
7871 * tree-ssa-strlen.c: Include tree-cfg.h.
7872 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7873 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7874 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7876 (strlen_dom_walker::before_dom_children): Call
7877 gimple_purge_dead_eh_edges. Dump tranformation with details
7879 (strlen_dom_walker::before_dom_children): Update call by adding
7880 new argument cleanup_eh.
7881 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7883 2018-01-03 Martin Liska <mliska@suse.cz>
7886 * cif-code.def (VARIADIC_THUNK): New enum value.
7887 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7890 2018-01-03 Jan Beulich <jbeulich@suse.com>
7892 * sse.md (mov<mode>_internal): Tighten condition for when to use
7893 vmovdqu<ssescalarsize> for TI and OI modes.
7895 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7897 Update copyright years.
7899 2018-01-03 Martin Liska <mliska@suse.cz>
7902 * ipa-visibility.c (function_and_variable_visibility): Skip
7903 functions with noipa attribure.
7905 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7907 * gcc.c (process_command): Update copyright notice dates.
7908 * gcov-dump.c (print_version): Ditto.
7909 * gcov.c (print_version): Ditto.
7910 * gcov-tool.c (print_version): Ditto.
7911 * gengtype.c (create_file): Ditto.
7912 * doc/cpp.texi: Bump @copying's copyright year.
7913 * doc/cppinternals.texi: Ditto.
7914 * doc/gcc.texi: Ditto.
7915 * doc/gccint.texi: Ditto.
7916 * doc/gcov.texi: Ditto.
7917 * doc/install.texi: Ditto.
7918 * doc/invoke.texi: Ditto.
7920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7922 * vector-builder.h (vector_builder::m_full_nelts): Change from
7923 unsigned int to poly_uint64.
7924 (vector_builder::full_nelts): Update prototype accordingly.
7925 (vector_builder::new_vector): Likewise.
7926 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7927 (vector_builder::operator ==): Likewise.
7928 (vector_builder::finalize): Likewise.
7929 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7930 Take the number of elements as a poly_uint64 rather than an
7932 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7933 from unsigned int to poly_uint64.
7934 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7935 (vec_perm_indices::new_vector): Likewise.
7936 (vec_perm_indices::length): Likewise.
7937 (vec_perm_indices::nelts_per_input): Likewise.
7938 (vec_perm_indices::input_nelts): Likewise.
7939 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7940 number of elements per input as a poly_uint64 rather than an
7941 unsigned int. Use the original encoding for variable-length
7942 vectors, rather than clamping each individual element.
7943 For the second and subsequent elements in each pattern,
7944 clamp the step and base before clamping their sum.
7945 (vec_perm_indices::series_p): Handle polynomial element counts.
7946 (vec_perm_indices::all_in_range_p): Likewise.
7947 (vec_perm_indices_to_tree): Likewise.
7948 (vec_perm_indices_to_rtx): Likewise.
7949 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7950 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7951 (tree_vector_builder::new_binary_operation): Handle polynomial
7952 element counts. Return false if we need to know the number
7953 of elements at compile time.
7954 * fold-const.c (fold_vec_perm): Punt if the number of elements
7955 isn't known at compile time.
7957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7959 * vec-perm-indices.h (vec_perm_builder): Change element type
7960 from HOST_WIDE_INT to poly_int64.
7961 (vec_perm_indices::element_type): Update accordingly.
7962 (vec_perm_indices::clamp): Handle polynomial element_types.
7963 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7964 (vec_perm_indices::all_in_range_p): Likewise.
7965 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7967 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7968 polynomial vec_perm_indices element types.
7969 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7970 * fold-const.c (fold_vec_perm): Likewise.
7971 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7972 * tree-vect-generic.c (lower_vec_perm): Likewise.
7973 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7974 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7975 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7982 rather than an int. Use plus_constant.
7983 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7984 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 Alan Hayward <alan.hayward@arm.com>
7988 David Sherwood <david.sherwood@arm.com>
7990 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7991 a HOST_WIDE_INT to a poly_int64.
7993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7994 Alan Hayward <alan.hayward@arm.com>
7995 David Sherwood <david.sherwood@arm.com>
7997 * calls.c (load_register_parameters): Cope with polynomial
7998 mode sizes. Require a constant size for BLKmode parameters
7999 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8000 forces a parameter to be padded at the lsb end in order to
8001 fill a complete number of words, require the parameter size
8002 to be ordered wrt UNITS_PER_WORD.
8004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8005 Alan Hayward <alan.hayward@arm.com>
8006 David Sherwood <david.sherwood@arm.com>
8008 * reload1.c (spill_stack_slot_width): Change element type
8009 from unsigned int to poly_uint64_pod.
8010 (alter_reg): Treat mode sizes as polynomial.
8012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8013 Alan Hayward <alan.hayward@arm.com>
8014 David Sherwood <david.sherwood@arm.com>
8016 * reload.c (complex_word_subreg_p): New function.
8017 (reload_inner_reg_of_subreg, push_reload): Use it.
8019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8020 Alan Hayward <alan.hayward@arm.com>
8021 David Sherwood <david.sherwood@arm.com>
8023 * lra-constraints.c (process_alt_operands): Reject matched
8024 operands whose sizes aren't ordered.
8025 (match_reload): Refer to this check here.
8027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8028 Alan Hayward <alan.hayward@arm.com>
8029 David Sherwood <david.sherwood@arm.com>
8031 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8032 that the mode size is in the set {1, 2, 4, 8, 16}.
8034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8035 Alan Hayward <alan.hayward@arm.com>
8036 David Sherwood <david.sherwood@arm.com>
8038 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8039 Use plus_constant instead of gen_rtx_PLUS.
8041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8042 Alan Hayward <alan.hayward@arm.com>
8043 David Sherwood <david.sherwood@arm.com>
8045 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8046 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8047 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8048 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8049 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8050 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8051 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8052 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8053 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8054 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8056 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8057 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8058 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8059 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8060 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8061 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8062 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8063 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8064 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8065 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8067 * expr.c (emit_move_resolve_push): Treat the input and result
8068 of PUSH_ROUNDING as a poly_int64.
8069 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8070 (emit_push_insn): Likewise.
8071 * lra-eliminations.c (mark_not_eliminable): Likewise.
8072 * recog.c (push_operand): Likewise.
8073 * reload1.c (elimination_effects): Likewise.
8074 * rtlanal.c (nonzero_bits1): Likewise.
8075 * calls.c (store_one_arg): Likewise. Require the padding to be
8076 known at compile time.
8078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8079 Alan Hayward <alan.hayward@arm.com>
8080 David Sherwood <david.sherwood@arm.com>
8082 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8083 Use plus_constant instead of gen_rtx_PLUS.
8085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8086 Alan Hayward <alan.hayward@arm.com>
8087 David Sherwood <david.sherwood@arm.com>
8089 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8093 Alan Hayward <alan.hayward@arm.com>
8094 David Sherwood <david.sherwood@arm.com>
8096 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8097 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8098 via stack temporaries. Treat the mode size as polynomial too.
8100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8101 Alan Hayward <alan.hayward@arm.com>
8102 David Sherwood <david.sherwood@arm.com>
8104 * expr.c (expand_expr_real_2): When handling conversions involving
8105 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8106 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8107 as a poly_uint64 too.
8109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8110 Alan Hayward <alan.hayward@arm.com>
8111 David Sherwood <david.sherwood@arm.com>
8113 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8116 Alan Hayward <alan.hayward@arm.com>
8117 David Sherwood <david.sherwood@arm.com>
8119 * combine.c (can_change_dest_mode): Handle polynomial
8120 REGMODE_NATURAL_SIZE.
8121 * expmed.c (store_bit_field_1): Likewise.
8122 * expr.c (store_constructor): Likewise.
8123 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8124 and polynomial REGMODE_NATURAL_SIZE.
8125 (gen_lowpart_common): Likewise.
8126 * reginfo.c (record_subregs_of_mode): Likewise.
8127 * rtlanal.c (read_modify_subreg_p): Likewise.
8129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8130 Alan Hayward <alan.hayward@arm.com>
8131 David Sherwood <david.sherwood@arm.com>
8133 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8134 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
8142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8143 Alan Hayward <alan.hayward@arm.com>
8144 David Sherwood <david.sherwood@arm.com>
8146 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8147 in a POINTER_PLUS_EXPR.
8149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8150 Alan Hayward <alan.hayward@arm.com>
8151 David Sherwood <david.sherwood@arm.com>
8153 * omp-simd-clone.c (simd_clone_subparts): New function.
8154 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8155 (ipa_simd_modify_function_body): Likewise.
8157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8158 Alan Hayward <alan.hayward@arm.com>
8159 David Sherwood <david.sherwood@arm.com>
8161 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8162 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8163 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8164 (expand_vector_condition, vector_element): Likewise.
8165 (subparts_gt): New function.
8166 (get_compute_type): Use subparts_gt.
8167 (count_type_subparts): Delete.
8168 (expand_vector_operations_1): Use subparts_gt instead of
8169 count_type_subparts.
8171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8172 Alan Hayward <alan.hayward@arm.com>
8173 David Sherwood <david.sherwood@arm.com>
8175 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8176 (vect_compile_time_alias): ...this new function. Do the calculation
8177 on poly_ints rather than trees.
8178 (vect_prune_runtime_alias_test_list): Update call accordingly.
8180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8181 Alan Hayward <alan.hayward@arm.com>
8182 David Sherwood <david.sherwood@arm.com>
8184 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8186 (vect_schedule_slp_instance): Likewise.
8188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8189 Alan Hayward <alan.hayward@arm.com>
8190 David Sherwood <david.sherwood@arm.com>
8192 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8193 constant and extern definitions for variable-length vectors.
8194 (vect_get_constant_vectors): Note that the number of units
8195 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
8202 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
8210 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8213 Alan Hayward <alan.hayward@arm.com>
8214 David Sherwood <david.sherwood@arm.com>
8216 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8217 vectors as polynomial. Use build_index_vector for
8220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8221 Alan Hayward <alan.hayward@arm.com>
8222 David Sherwood <david.sherwood@arm.com>
8224 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8225 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8226 for variable-length vectors.
8227 (vectorizable_mask_load_store): Treat the number of units as
8228 polynomial, asserting that it is constant if the condition has
8229 already been enforced.
8230 (vectorizable_store, vectorizable_load): Likewise.
8232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8233 Alan Hayward <alan.hayward@arm.com>
8234 David Sherwood <david.sherwood@arm.com>
8236 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8237 of units as polynomial. Punt if we can't tell at compile time
8238 which vector contains the final result.
8240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8241 Alan Hayward <alan.hayward@arm.com>
8242 David Sherwood <david.sherwood@arm.com>
8244 * tree-vect-loop.c (vectorizable_induction): Treat the number
8245 of units as polynomial. Punt on SLP inductions. Use an integer
8246 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8247 cast of such a series for variable-length floating-point
8250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8251 Alan Hayward <alan.hayward@arm.com>
8252 David Sherwood <david.sherwood@arm.com>
8254 * tree.h (build_index_vector): Declare.
8255 * tree.c (build_index_vector): New function.
8256 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8257 of units as polynomial, forcibly converting it to a constant if
8258 vectorizable_reduction has already enforced the condition.
8259 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8260 to create a {1,2,3,...} vector.
8261 (vectorizable_reduction): Treat the number of units as polynomial.
8262 Choose vectype_in based on the largest scalar element size rather
8263 than the smallest number of units. Enforce the restrictions
8266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8267 Alan Hayward <alan.hayward@arm.com>
8268 David Sherwood <david.sherwood@arm.com>
8270 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8271 number of units as polynomial.
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.h (vector_sizes, auto_vector_sizes): New typedefs.
8278 * target.def (autovectorize_vector_sizes): Return the vector sizes
8279 by pointer, using vector_sizes rather than a bitmask.
8280 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8281 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8282 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8284 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8285 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8286 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8287 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8288 * omp-general.c (omp_max_vf): Likewise.
8289 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8290 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8291 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8292 * tree-vect-slp.c (vect_slp_bb): Likewise.
8293 * doc/tm.texi: Regenerate.
8294 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8296 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8297 the vector size as a poly_uint64 rather than an unsigned int.
8298 (current_vector_size): Change from an unsigned int to a poly_uint64.
8299 (get_vectype_for_scalar_type): Update accordingly.
8300 * tree.h (build_truth_vector_type): Take the size and number of
8301 units as a poly_uint64 rather than an unsigned int.
8302 (build_vector_type): Add a temporary overload that takes
8303 the number of units as a poly_uint64 rather than an unsigned int.
8304 * tree.c (make_vector_type): Likewise.
8305 (build_truth_vector_type): Take the number of units as a poly_uint64
8306 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
8313 as poly_uint64s rather than unsigned ints.
8314 * targhooks.h (default_get_mask_mode): Update accordingly.
8315 * targhooks.c (default_get_mask_mode): Likewise.
8316 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8317 * doc/tm.texi: Regenerate.
8319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8320 Alan Hayward <alan.hayward@arm.com>
8321 David Sherwood <david.sherwood@arm.com>
8323 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8324 * omp-general.c (omp_max_vf): Likewise.
8325 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8326 (expand_omp_simd): Handle polynomial safelen.
8327 * omp-low.c (omplow_simd_context): Add a default constructor.
8328 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8329 (lower_rec_simd_input_clauses): Update accordingly.
8330 (lower_rec_input_clauses): Likewise.
8332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8333 Alan Hayward <alan.hayward@arm.com>
8334 David Sherwood <david.sherwood@arm.com>
8336 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8337 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8338 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8339 (vect_analyze_slp_cost): Likewise.
8340 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8341 (vect_model_load_cost): Likewise.
8343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8344 Alan Hayward <alan.hayward@arm.com>
8345 David Sherwood <david.sherwood@arm.com>
8347 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8348 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8349 from an unsigned int * to a poly_uint64_pod *.
8350 (calculate_unrolling_factor): New function.
8351 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8354 Alan Hayward <alan.hayward@arm.com>
8355 David Sherwood <david.sherwood@arm.com>
8357 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8358 from an unsigned int to a poly_uint64.
8359 (_loop_vec_info::slp_unrolling_factor): Likewise.
8360 (_loop_vec_info::vectorization_factor): Change from an int
8362 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8363 (vect_get_num_vectors): New function.
8364 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8365 (vect_get_num_copies): Use vect_get_num_vectors.
8366 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8367 to an unsigned int *.
8368 (vect_analyze_data_refs): Change min_vf from an int * to a
8370 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8371 than an unsigned HOST_WIDE_INT.
8372 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8373 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8374 to an unsigned int *.
8375 (vect_analyze_data_ref_dependences): Likewise.
8376 (vect_compute_data_ref_alignment): Handle polynomial vf.
8377 (vect_enhance_data_refs_alignment): Likewise.
8378 (vect_prune_runtime_alias_test_list): Likewise.
8379 (vect_shift_permute_load_chain): Likewise.
8380 (vect_supportable_dr_alignment): Likewise.
8381 (dependence_distance_ge_vf): Take the vectorization factor as a
8382 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8383 (vect_analyze_data_refs): Change min_vf from an int * to a
8385 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8386 vfm1 as a poly_uint64 rather than an int. Make the same change
8387 for the returned bound_scalar.
8388 (vect_gen_vector_loop_niters): Handle polynomial vf.
8389 (vect_do_peeling): Likewise. Update call to
8390 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8391 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8393 * tree-vect-loop.c (vect_determine_vectorization_factor)
8394 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8395 (vect_get_known_peeling_cost): Likewise.
8396 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8397 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8398 (vect_transform_loop): Likewise. Use the lowest possible VF when
8399 updating the upper bounds of the loop.
8400 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8402 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8403 polynomial unroll factors.
8404 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8405 (vect_make_slp_decision): Likewise.
8406 (vect_supported_load_permutation_p): Likewise, and polynomial
8408 (vect_analyze_slp_cost): Handle polynomial vf.
8409 (vect_slp_analyze_node_operations): Likewise.
8410 (vect_slp_analyze_bb_1): Likewise.
8411 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8412 than an unsigned HOST_WIDE_INT.
8413 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8414 (vectorizable_load): Handle polynomial vf.
8415 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8417 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8420 Alan Hayward <alan.hayward@arm.com>
8421 David Sherwood <david.sherwood@arm.com>
8423 * match.pd: Handle bit operations involving three constants
8424 and try to fold one pair.
8426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8428 * tree-vect-loop-manip.c: Include gimple-fold.h.
8429 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8430 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8431 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8432 Add a path that uses a step of VF instead of 1, but disable it
8434 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8435 and niters_no_overflow parameters. Update calls to
8436 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8437 Create a new SSA name if the latter choses to use a ste other
8438 than zero, and return it via niters_vector_mult_vf_var.
8439 * tree-vect-loop.c (vect_transform_loop): Update calls to
8440 vect_do_peeling, vect_gen_vector_loop_niters and
8441 slpeel_make_loop_iterate_ntimes.
8442 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8443 (vect_gen_vector_loop_niters): Update declarations after above changes.
8445 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8447 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8448 128-bit round to integer instructions.
8449 (ceil<mode>2): Likewise.
8450 (btrunc<mode>2): Likewise.
8451 (round<mode>2): Likewise.
8453 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8455 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8456 unaligned VSX load/store on P8/P9.
8457 (expand_block_clear): Allow the use of unaligned VSX
8458 load/store on P8/P9.
8460 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8462 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8464 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8465 swap associated with both a load and a store.
8467 2018-01-02 Andrew Waterman <andrew@sifive.com>
8469 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8470 * config/riscv/riscv.md (clear_cache): Use it.
8472 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8474 * web.c: Remove out-of-date comment.
8476 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8478 * expr.c (fixup_args_size_notes): Check that any existing
8479 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8480 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8481 (emit_single_push_insn): ...here.
8483 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8485 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8486 (const_vector_encoded_nelts): New function.
8487 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8488 (const_vector_int_elt, const_vector_elt): Declare.
8489 * emit-rtl.c (const_vector_int_elt_1): New function.
8490 (const_vector_elt): Likewise.
8491 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8492 of CONST_VECTOR_ELT.
8494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8496 * expr.c: Include rtx-vector-builder.h.
8497 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8498 directly on the tree encoding.
8499 (const_vector_from_tree): Likewise.
8500 * optabs.c: Include rtx-vector-builder.h.
8501 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8502 sequence of "u" values.
8503 * vec-perm-indices.c: Include rtx-vector-builder.h.
8504 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8505 directly on the vec_perm_indices encoding.
8507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8509 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8510 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8511 * rtx-vector-builder.h: New file.
8512 * rtx-vector-builder.c: Likewise.
8513 * rtl.h (rtx_def::u2): Add a const_vector field.
8514 (CONST_VECTOR_NPATTERNS): New macro.
8515 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8516 (CONST_VECTOR_DUPLICATE_P): Likewise.
8517 (CONST_VECTOR_STEPPED_P): Likewise.
8518 (CONST_VECTOR_ENCODED_ELT): Likewise.
8519 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8520 (unwrap_const_vec_duplicate): Likewise.
8521 (const_vec_series_p): Check for a non-duplicated vector encoding.
8522 Say that the function only returns true for integer vectors.
8523 * emit-rtl.c: Include rtx-vector-builder.h.
8524 (gen_const_vec_duplicate_1): Delete.
8525 (gen_const_vector): Call gen_const_vec_duplicate instead of
8526 gen_const_vec_duplicate_1.
8527 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8528 (gen_const_vec_duplicate): Use rtx_vector_builder.
8529 (gen_const_vec_series): Likewise.
8530 (gen_rtx_CONST_VECTOR): Likewise.
8531 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8532 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8533 Build a new vector rather than modifying a CONST_VECTOR in-place.
8534 (handle_special_swappables): Update call accordingly.
8535 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8536 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8537 Build a new vector rather than modifying a CONST_VECTOR in-place.
8538 (handle_special_swappables): Update call accordingly.
8540 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8542 * simplify-rtx.c (simplify_const_binary_operation): Use
8543 CONST_VECTOR_ELT instead of XVECEXP.
8545 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8547 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8548 the selector elements to be different from the data elements
8549 if the selector is a VECTOR_CST.
8550 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8551 ssizetype for the selector.
8553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8555 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8556 before testing each element individually.
8557 * tree-vect-generic.c (lower_vec_perm): Likewise.
8559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8561 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8562 * selftest-run-tests.c (selftest::run_tests): Call it.
8563 * vector-builder.h (vector_builder::operator ==): New function.
8564 (vector_builder::operator !=): Likewise.
8565 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8566 (vec_perm_indices::all_from_input_p): New function.
8567 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8568 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8569 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8570 instead of reading the VECTOR_CST directly. Detect whether both
8571 vector inputs are the same before constructing the vec_perm_indices,
8572 and update the number of inputs argument accordingly. Use the
8573 utility functions added above. Only construct sel2 if we need to.
8575 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8577 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8578 the broadcast of the low byte.
8579 (expand_mult_highpart): Use an explicit encoding for the permutes.
8580 * optabs-query.c (can_mult_highpart_p): Likewise.
8581 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8582 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8583 (vectorizable_bswap): Likewise.
8584 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8585 explicit encoding for the power-of-2 permutes.
8586 (vect_permute_store_chain): Likewise.
8587 (vect_grouped_load_supported): Likewise.
8588 (vect_permute_load_chain): Likewise.
8590 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8592 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8593 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8594 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8595 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8596 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8597 (vect_gen_perm_mask_any): Likewise.
8599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8601 * int-vector-builder.h: New file.
8602 * vec-perm-indices.h: Include int-vector-builder.h.
8603 (vec_perm_indices): Redefine as an int_vector_builder.
8604 (auto_vec_perm_indices): Delete.
8605 (vec_perm_builder): Redefine as a stand-alone class.
8606 (vec_perm_indices::vec_perm_indices): New function.
8607 (vec_perm_indices::clamp): Likewise.
8608 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8609 (vec_perm_indices::new_vector): New function.
8610 (vec_perm_indices::new_expanded_vector): Update for new
8611 vec_perm_indices class.
8612 (vec_perm_indices::rotate_inputs): New function.
8613 (vec_perm_indices::all_in_range_p): Operate directly on the
8614 encoded form, without computing elided elements.
8615 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8616 encoding. Update for new vec_perm_indices class.
8617 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8618 the given vec_perm_builder.
8619 (expand_vec_perm_var): Update vec_perm_builder constructor.
8620 (expand_mult_highpart): Use vec_perm_builder instead of
8621 auto_vec_perm_indices.
8622 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8623 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8624 or double series encoding as appropriate.
8625 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8626 vec_perm_indices instead of auto_vec_perm_indices.
8627 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8628 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8629 (vect_permute_store_chain): Likewise.
8630 (vect_grouped_load_supported): Likewise.
8631 (vect_permute_load_chain): Likewise.
8632 (vect_shift_permute_load_chain): Likewise.
8633 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8634 (vect_transform_slp_perm_load): Likewise.
8635 (vect_schedule_slp_instance): Likewise.
8636 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8637 (vectorizable_mask_load_store): Likewise.
8638 (vectorizable_bswap): Likewise.
8639 (vectorizable_store): Likewise.
8640 (vectorizable_load): Likewise.
8641 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8642 vec_perm_indices instead of auto_vec_perm_indices. Use
8643 tree_to_vec_perm_builder to read the vector from a tree.
8644 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8645 vec_perm_builder instead of a vec_perm_indices.
8646 (have_whole_vector_shift): Use vec_perm_builder and
8647 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8648 truncation to calc_vec_perm_mask_for_shift.
8649 (vect_create_epilog_for_reduction): Likewise.
8650 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8651 from auto_vec_perm_indices to vec_perm_indices.
8652 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8653 instead of changing individual elements.
8654 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8655 the vector in d.perm.
8656 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8657 from auto_vec_perm_indices to vec_perm_indices.
8658 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8659 instead of changing individual elements.
8660 (arm_vectorize_vec_perm_const): Use new_vector to install
8661 the vector in d.perm.
8662 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8663 Update vec_perm_builder constructor.
8664 (rs6000_expand_interleave): Likewise.
8665 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8666 (rs6000_expand_interleave): Likewise.
8668 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8670 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8671 to qimode could truncate the indices.
8672 * optabs.c (expand_vec_perm_var): Likewise.
8674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8676 * Makefile.in (OBJS): Add vec-perm-indices.o.
8677 * vec-perm-indices.h: New file.
8678 * vec-perm-indices.c: Likewise.
8679 * target.h (vec_perm_indices): Replace with a forward class
8681 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8682 * optabs.h: Include vec-perm-indices.h.
8683 (expand_vec_perm): Delete.
8684 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8685 (expand_vec_perm_const): Declare.
8686 * target.def (vec_perm_const_ok): Replace with...
8687 (vec_perm_const): ...this new hook.
8688 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8689 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8690 * doc/tm.texi: Regenerate.
8691 * optabs.def (vec_perm_const): Delete.
8692 * doc/md.texi (vec_perm_const): Likewise.
8693 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8694 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8695 expand_vec_perm for constant permutation vectors. Assert that
8696 the mode of variable permutation vectors is the integer equivalent
8697 of the mode that is being permuted.
8698 * optabs-query.h (selector_fits_mode_p): Declare.
8699 * optabs-query.c: Include vec-perm-indices.h.
8700 (selector_fits_mode_p): New function.
8701 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8702 is defined, instead of checking whether the vec_perm_const_optab
8703 exists. Use targetm.vectorize.vec_perm_const instead of
8704 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8705 fit in the vector mode before using a variable permute.
8706 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8707 vec_perm_indices instead of an rtx.
8708 (expand_vec_perm): Replace with...
8709 (expand_vec_perm_const): ...this new function. Take the selector
8710 as a vec_perm_indices rather than an rtx. Also take the mode of
8711 the selector. Update call to shift_amt_for_vec_perm_mask.
8712 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8713 Use vec_perm_indices::new_expanded_vector to expand the original
8714 selector into bytes. Check whether the indices fit in the vector
8715 mode before using a variable permute.
8716 (expand_vec_perm_var): Make global.
8717 (expand_mult_highpart): Use expand_vec_perm_const.
8718 * fold-const.c: Includes vec-perm-indices.h.
8719 * tree-ssa-forwprop.c: Likewise.
8720 * tree-vect-data-refs.c: Likewise.
8721 * tree-vect-generic.c: Likewise.
8722 * tree-vect-loop.c: Likewise.
8723 * tree-vect-slp.c: Likewise.
8724 * tree-vect-stmts.c: Likewise.
8725 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8727 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8728 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8729 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8730 (aarch64_vectorize_vec_perm_const): ...this new function.
8731 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8732 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8733 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8734 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8735 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8736 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8737 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8739 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8740 check for NEON modes.
8741 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8742 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8743 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8744 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8746 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8747 the old VEC_PERM_CONST conditions.
8748 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8749 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8750 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8751 (ia64_vectorize_vec_perm_const_ok): Merge into...
8752 (ia64_vectorize_vec_perm_const): ...this new function.
8753 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8754 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8755 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8756 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8757 * config/mips/mips.c (mips_expand_vec_perm_const)
8758 (mips_vectorize_vec_perm_const_ok): Merge into...
8759 (mips_vectorize_vec_perm_const): ...this new function.
8760 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8761 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8762 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8763 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8764 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8765 (rs6000_expand_vec_perm_const): Delete.
8766 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8768 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8769 (altivec_expand_vec_perm_const_le): Take each operand individually.
8770 Operate on constant selectors rather than rtxes.
8771 (altivec_expand_vec_perm_const): Likewise. Update call to
8772 altivec_expand_vec_perm_const_le.
8773 (rs6000_expand_vec_perm_const): Delete.
8774 (rs6000_vectorize_vec_perm_const_ok): Delete.
8775 (rs6000_vectorize_vec_perm_const): New function.
8776 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8777 an element count and rtx array.
8778 (rs6000_expand_extract_even): Update call accordingly.
8779 (rs6000_expand_interleave): Likewise.
8780 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8781 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8782 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8783 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8784 (rs6000_expand_vec_perm_const): Delete.
8785 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8786 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8787 (altivec_expand_vec_perm_const_le): Take each operand individually.
8788 Operate on constant selectors rather than rtxes.
8789 (altivec_expand_vec_perm_const): Likewise. Update call to
8790 altivec_expand_vec_perm_const_le.
8791 (rs6000_expand_vec_perm_const): Delete.
8792 (rs6000_vectorize_vec_perm_const_ok): Delete.
8793 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8794 reference to the SPE evmerge intructions.
8795 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8796 an element count and rtx array.
8797 (rs6000_expand_extract_even): Update call accordingly.
8798 (rs6000_expand_interleave): Likewise.
8799 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8800 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8802 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8806 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8807 vector mode and that that mode matches the mode of the data
8809 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8810 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8811 directly using expand_vec_perm_1 when forcing selectors into
8813 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8817 * optabs-query.h (can_vec_perm_p): Delete.
8818 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8819 * optabs-query.c (can_vec_perm_p): Split into...
8820 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8821 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8822 particular selector is valid.
8823 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8824 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8825 (vect_grouped_load_supported): Likewise.
8826 (vect_shift_permute_load_chain): Likewise.
8827 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8828 (vect_transform_slp_perm_load): Likewise.
8829 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8830 (vectorizable_bswap): Likewise.
8831 (vect_gen_perm_mask_checked): Likewise.
8832 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8833 implementations of variable permutation vectors into account
8834 when deciding which selector to use.
8835 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8836 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8837 with a false third argument.
8838 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8839 to test whether the constant selector is valid and can_vec_perm_var_p
8840 to test whether a variable selector is valid.
8842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8844 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8845 * optabs-query.c (can_vec_perm_p): Likewise.
8846 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8847 instead of vec_perm_indices.
8848 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8849 (vect_gen_perm_mask_checked): Likewise,
8850 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8851 (vect_gen_perm_mask_checked): Likewise,
8853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8855 * optabs-query.h (qimode_for_vec_perm): Declare.
8856 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8857 (qimode_for_vec_perm): ...this new function.
8858 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8860 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8862 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8863 does not have a conditional at the top.
8865 2018-01-02 Richard Biener <rguenther@suse.de>
8867 * ipa-inline.c (big_speedup_p): Fix expression.
8869 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8872 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8875 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8879 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8880 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8881 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8882 cond_taken_branch_cost 3->4.
8884 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8886 PR tree-optimization/83581
8887 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8888 TODO_cleanup_cfg if any changes have been made.
8891 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8892 convert_modes if target mode has the right side, but different mode
8896 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8897 last argument when extracting from CONCAT. If either from_real or
8898 from_imag is NULL, use expansion through memory. If result is not
8899 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8900 the parts directly to inner mode, if even that fails, use expansion
8904 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8905 check for bswap in mode rather than HImode and use that in expand_unop
8908 Copyright (C) 2018 Free Software Foundation, Inc.
8910 Copying and distribution of this file, with or without modification,
8911 are permitted in any medium without royalty provided the copyright
8912 notice and this notice are preserved.