1 2018-07-20 Martin Sebor <msebor@redhat.com>
4 * builtins.c (expand_builtin_alloca): Adjust.
5 * calls.c (alloc_max_size): Simplify.
6 * cgraphunit.c (cgraph_node::expand): Adjust.
7 * common.opt (larger_than_size, warn_frame_larger_than): Remove
9 (frame_larger_than_size): Same.
10 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
13 * doc/invoke.texi (GCC Command Options): Document option arguments.
14 Explain byte-size arguments and suffixes.
15 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
16 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
17 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
18 * doc/options.texi (UInteger): Expand.
19 (Host_Wide_Int, ByteSize): Document new properties.
20 * final.c (final_start_function_1): Include sizes in an error message.
21 * function.c (frame_offset_overflow): Same.
22 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
23 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
24 Diagnose unbounded alloca calls only for limits of less than
26 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
27 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
29 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
30 only for limits of less than PTRDIFF_MAX.
31 * langhooks-def.h (lhd_handle_option): Change function argument
33 * langhooks.c (lhd_handle_option): Same.
34 * langhooks.h (handle_option): Same.
35 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
37 (var_type, var_type_struct): Same.
38 (var_set): Handle ByteSize flag.
39 * optc-gen.awk: Add comments to output to ease debugging. Make
40 use of HOST_WIDE_INT where appropriate.
41 * opts-gen-save.awk: Use %lx to format unsigned long.
42 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
43 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
44 arguments. Parse bytes-size suffixes.
45 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
46 (enum_value_to_arg): Same.
47 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
48 (handle_option): Adjust.
49 (generate_option): Change function argument to HOST_WIDE_INT.
50 (cmdline_handle_error): Adjust.
51 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
52 (set_option): Change function argument to HOST_WIDE_INT.
53 (option_enabled): Handle cl_host_wide_int.
54 (get_option_state): Handle CLVC_SIZE.
55 (control_warning_option): Same.
56 * opts.c (common_handle_option): Change function argument to
57 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
58 OPT_Wvla_larger_than_.
59 * opts.h (enum cl_var_type): Add an enumerator.
60 * stor-layout.c (layout_decl): Print a more meaningful warning.
61 * toplev.c (output_stack_usage): Adjust.
63 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
65 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
66 call to inline_expand_builtin_string_cmp.
67 (expand_builtin_strcmp): Likewise.
68 (expand_builtin_strncmp): Likewise.
69 (inline_string_cmp): Delete the last parameter, change char_type_node
70 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
72 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
73 the inlining expansion on target where the type of the call has same or
74 narrower precision than unsigned char.
76 2018-07-20 David Malcolm <dmalcolm@redhat.com>
78 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
79 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
80 * common.opt (fsave-optimization-record): New option.
81 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
82 * doc/invoke.texi (-fsave-optimization-record): New option.
83 * dumpfile.c: Include "optinfo-emit-json.h".
84 (struct kv_pair): Move to coretypes.h.
85 (optgroup_options): Make non-static.
86 (dump_context::end_scope): Call
87 optimization_records_maybe_pop_dump_scope.
88 * dumpfile.h (optgroup_options): New decl.
91 * optinfo-emit-json.cc: New file.
92 * optinfo-emit-json.h: New file.
93 * optinfo.cc: Include "optinfo-emit-json.h".
94 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
95 (optinfo_enabled_p): Check optimization_records_enabled_p.
96 (optinfo_wants_inlining_info_p): Likewise.
97 * optinfo.h: Update comment.
98 * profile-count.c (profile_quality_as_string): New function.
99 * profile-count.h (profile_quality_as_string): New decl.
100 (profile_count::quality): New accessor.
101 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
102 and optinfo_emit_json_cc_tests.
103 * selftest.h (selftest::json_cc_tests): New decl.
104 (selftest::optinfo_emit_json_cc_tests): New decl.
105 * toplev.c: Include "optinfo-emit-json.h".
106 (compile_file): Call optimization_records_finish.
107 (do_compile): Call optimization_records_start.
108 * tree-ssa-live.c: Include optinfo.h.
109 (remove_unused_scope_block_p): Retain inlining information if
110 optinfo_wants_inlining_info_p returns true.
112 2018-07-20 Richard Biener <rguenther@suse.de>
115 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
116 to cover -flto-partition=none.
118 2018-07-20 Martin Liska <mliska@suse.cz>
120 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
121 (get_decl_source_range): Remove unused function.
123 2018-07-20 Richard Biener <rguenther@suse.de>
125 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
126 (struct vn_phi_s): Likewise.
127 (struct vn_reference_s): Likewise.
128 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
129 for searching the slot of an entry known to be in the hash itself.
130 (vn_phi_hasher::equal): Likewise.
131 (vn_reference_hasher::equal): Likewise.
132 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
134 (optimistic_info, current_info): Remove, keeping only valid_info.
135 (vn_reference_lookup_1): Remove fallback lookup.
136 (vn_reference_lookup_2): Likewise.
137 (vn_nary_op_lookup_1): Likewise.
138 (vn_phi_lookup): Likewise.
139 (vn_nary_build_or_lookup_1): Make sure to not chain the built
141 (vn_reference_insert): Adjust, chain the inserted hash element
142 at last_inserted_ref.
143 (vn_reference_insert_pieces): Likewise.
144 (visit_reference_op_call): Likewise.
145 (vn_nary_op_insert_into): Chain the inserted hash element at
147 (vn_nary_op_insert_pieces): Adjust.
148 (vn_nary_op_insert): Likewise.
149 (vn_nary_op_insert_stmt): Likewise.
150 (vn_phi_insert): Adjust, chain the inserted hash element at
152 (process_scc): Remove clearing and copying the optimistic
153 table. Instead remove elements inserted during an optimistic
154 iteration from the single table we maintain.
155 (init_scc_vn): Adjust.
156 (free_scc_vn): Likewise.
157 (sccvn_dom_walker::record_cond): Likewise.
158 (sccvn_dom_walker::after_dom_children): Likewise.
160 2018-07-19 Martin Sebor <msebor@redhat.com>
162 PR tree-optimization/84047
163 PR tree-optimization/83776
164 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
165 (check_array_bounds): Call it.
167 2018-07-19 Martin Sebor <msebor@redhat.com>
169 * align.h (align_flags): Use member initialization.
171 2018-07-19 David Malcolm <dmalcolm@redhat.com>
173 * Makefile.in (OBJS): Add optinfo.o.
174 * coretypes.h (class symtab_node): New forward decl.
175 (struct cgraph_node): New forward decl.
176 (class varpool_node): New forward decl.
177 * dump-context.h: New file.
178 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
180 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
181 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
182 (set_alt_dump_file): Likewise.
183 (dump_context::~dump_context): New dtor.
184 (dump_gimple_stmt): Move implementation to...
185 (dump_context::dump_gimple_stmt): ...this new member function.
186 Add the stmt to any pending optinfo, creating one if need be.
187 (dump_gimple_stmt_loc): Move implementation to...
188 (dump_context::dump_gimple_stmt_loc): ...this new member function.
189 Start a new optinfo and add the stmt to it.
190 (dump_gimple_expr): Move implementation to...
191 (dump_context::dump_gimple_expr): ...this new member function.
192 Add the stmt to any pending optinfo, creating one if need be.
193 (dump_gimple_expr_loc): Move implementation to...
194 (dump_context::dump_gimple_expr_loc): ...this new member function.
195 Start a new optinfo and add the stmt to it.
196 (dump_generic_expr): Move implementation to...
197 (dump_context::dump_generic_expr): ...this new member function.
198 Add the tree to any pending optinfo, creating one if need be.
199 (dump_generic_expr_loc): Move implementation to...
200 (dump_context::dump_generic_expr_loc): ...this new member
201 function. Add the tree to any pending optinfo, creating one if
203 (dump_printf): Move implementation to...
204 (dump_context::dump_printf_va): ...this new member function. Add
205 the text to any pending optinfo, creating one if need be.
206 (dump_printf_loc): Move implementation to...
207 (dump_context::dump_printf_loc_va): ...this new member function.
208 Start a new optinfo and add the stmt to it.
209 (dump_dec): Move implementation to...
210 (dump_context::dump_dec): ...this new member function. Add the
211 value to any pending optinfo, creating one if need be.
212 (dump_context::dump_symtab_node): New member function.
213 (dump_context::get_scope_depth): New member function.
214 (dump_context::begin_scope): New member function.
215 (dump_context::end_scope): New member function.
216 (dump_context::ensure_pending_optinfo): New member function.
217 (dump_context::begin_next_optinfo): New member function.
218 (dump_context::end_any_optinfo): New member function.
219 (dump_context::s_current): New global.
220 (dump_context::s_default): New global.
221 (dump_scope_depth): Delete global.
222 (dumpfile_ensure_any_optinfo_are_flushed): New function.
223 (dump_symtab_node): New function.
224 (get_dump_scope_depth): Reimplement in terms of dump_context.
225 (dump_begin_scope): Likewise.
226 (dump_end_scope): Likewise.
227 (selftest::temp_dump_context::temp_dump_context): New ctor.
228 (selftest::temp_dump_context::~temp_dump_context): New dtor.
229 (selftest::verify_item): New function.
230 (ASSERT_IS_TEXT): New macro.
231 (ASSERT_IS_TREE): New macro.
232 (ASSERT_IS_GIMPLE): New macro.
233 (selftest::test_capture_of_dump_calls): New test.
234 (selftest::dumpfile_c_tests): Call it.
235 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
236 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
237 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
239 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
240 (dump_node, dump_bb): Move these unrelated decls.
241 (class dump_manager): Add leading comment.
242 * optinfo.cc: New file.
243 * optinfo.h: New file.
245 2018-07-19 Michael Collison <michael.collison@arm.com>
246 Richard Henderson <rth@redhat.com>
248 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
249 (subti): Handle op1 zero.
250 (subvti4, usub4ti4): New.
251 (*sub<GPI>3_compare1_imm): New.
252 (sub<GPI>3_carryinCV): New.
253 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
254 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
256 2018-07-19 Michael Collison <michael.collison@arm.com>
257 Richard Henderson <rth@redhat.com>
259 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
260 (addti3): Create simpler code if low part is already known to be 0.
261 (addvti4, uaddvti4): New.
262 (*add<GPI>3_compareC_cconly_imm): New.
263 (*add<GPI>3_compareC_cconly): New.
264 (*add<GPI>3_compareC_imm): New.
265 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
266 handle constants within this pattern..
267 (*add<GPI>3_compareV_cconly_imm): New.
268 (*add<GPI>3_compareV_cconly): New.
269 (*add<GPI>3_compareV_imm): New.
270 (add<GPI>3_compareV): New.
271 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
272 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
273 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
274 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
275 with 'comparison' operator.
276 (*add<GPI>3_compareV_cconly_imm): Ditto.
277 (*add<GPI>3_compareV_cconly): Ditto.
278 (*add<GPI>3_compareV_imm): Ditto.
279 (add<GPI>3_compareV): Ditto.
280 (add<mode>3_carryinC): Ditto.
281 (*add<mode>3_carryinC_zero): Ditto.
282 (*add<mode>3_carryinC): Ditto.
283 (add<mode>3_carryinV): Ditto.
284 (*add<mode>3_carryinV_zero): Ditto.
285 (*add<mode>3_carryinV): Ditto.
287 2018-07-19 Michael Collison <michael.collison@arm.com>
288 Richard Henderson <rth@redhat.com>
290 * config/aarch64/aarch64-modes.def (CC_V): New.
291 * config/aarch64/aarch64-protos.h
292 (aarch64_addti_scratch_regs): Declare
293 (aarch64_subvti_scratch_regs): Declare.
294 (aarch64_expand_subvti): Declare.
295 (aarch64_gen_unlikely_cbranch): Declare
296 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
297 for signed overflow using CC_Vmode.
298 (aarch64_get_condition_code_1): Handle CC_Vmode.
299 (aarch64_gen_unlikely_cbranch): New function.
300 (aarch64_addti_scratch_regs): New function.
301 (aarch64_subvti_scratch_regs): New function.
302 (aarch64_expand_subvti): New function.
304 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
306 * config/aarch64/aarch64-option-extensions.def: New entry for profile
308 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
309 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
312 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
315 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
316 address check not strict.
318 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
320 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
322 (store_pair_lanes<mode>): Likewise.
323 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
324 enum value 'ADDR_QUERY_LDP_STP_N'.
325 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
326 (aarch64_print_address_internal): Add declaration.
327 (aarch64_print_ldpstp_address): Remove.
328 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
329 (aarch64_print_operand): Change printing of 'y'.
330 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
331 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
332 'true' rather than '1'.
333 * gcc/config/aarch64/constraints.md (Uml): Likewise.
334 (Uml): Rename to Umn.
337 2018-07-19 Richard Biener <rguenther@suse.de>
339 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
341 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
342 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
343 (vn_reference_hasher): Likewise.
344 (struct vn_tables_s): Remove obstack and alloc-pool members.
345 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
346 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
347 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
348 (vn_reference_insert_pieces): Likewise.
349 (alloc_vn_nary_op_noinit): Adjust.
350 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
352 (shared_lookup_phiargs): Remove.
353 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
354 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
355 (visit_reference_op_call): Likewise.
356 (copy_nary, copy_phi, copy_reference): Remove.
357 (process_scc): Rewind the obstack when iterating. Do not
358 copy the elements to valid_info but just move them from one
359 hashtable to the other.
360 (allocate_vn_table): Adjust.
361 (free_vn_table): Likewise.
362 (init_scc_vn): Likewise.
363 (free_scc_vn): Likewise.
365 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
368 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
369 indirect_return as function type attribute.
370 (ix86_attribute_table): Change indirect_return to function
372 * doc/extend.texi: Update indirect_return attribute.
374 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
376 * wide-int.h (widest2_int): New.
377 * gimple-fold.c (arith_overflowed_p): Use it.
378 * tree.h (widest2_int_cst): New.
379 * tree-vrp.c (wide_int_binop_overflow): Rename from
381 Rewrite to work on trees.
382 (extract_range_from_multiplicative_op_1): Abstract code to...
383 (wide_int_range_min_max): ...here.
384 (wide_int_range_cross_product): ...and here.
385 (extract_range_from_binary_expr_1): Abstract overflow code to...
386 (wide_int_range_mult_wrapping): ...here.
387 * tree-vrp.h (wide_int_range_cross_product): New.
388 (wide_int_range_mult_wrapping): New.
390 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
391 Julia Koval <julia.koval@intel.com>
393 * config/i386/x86-tune-costs.h (skylake_memcpy,
394 skylake_memset): Replace rep_prefix with unrolling for size 512.
396 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
399 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
402 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
404 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
405 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
406 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
407 previously known as "PowerPC AltiVec Built-in Functions". Move
408 some material to new subsubsections "PowerPC AltiVec Built-in
409 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
411 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
412 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
413 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
414 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
416 2018-07-18 Richard Biener <rguenther@suse.de>
418 PR tree-optimization/86557
419 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
422 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
424 * config/s390/s390.c (s390_function_profiler): Generate CFI.
426 2018-07-17 Jeff Law <law@redhat.com>
428 * config/arm/arm.c (get_label_padding): Update for recent
429 changes to label_to_alignment.
431 PR tree-optimization/86010
432 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
434 * config/mips/mips.c (vr4130_align_insns): Update for recent
435 changes to label_to_alignment.
437 * config/frv/frv.c (frv_label_align): Update for recent changes
438 to label_to_alignment.
440 * config/nios2/nios2.c (nios2_label_align): Update for recent
441 changes which dropped ALIGN_LABELS_LOG.
443 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
445 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
446 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
448 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
450 * config/arc/arc.c (arc_label_align): Use align_labels instead of
451 deprecated align_labels_log.
453 2018-07-17 Richard Biener <rguenther@suse.de>
456 * dwarf2out.c (init_sections_and_labels): Always generate
457 a debug_line_str_section for early LTO debug.
458 (dwarf2out_finish): Reset debug_line_str_hash output early.
459 Bump counter for extra dwarf5 .debug_loc labels to not conflict
461 (dwarf2out_early_finish): Output debug_line_str.
463 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
465 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
466 index register on z196 or later.
468 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
470 * config/s390/s390.c (s390_default_align): Set default function
472 (s390_override_options_after_change): Call s390_default align.
473 (s390_option_override_internal): Call s390_default_align.
474 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
476 2018-07-17 Jakub Jelinek <jakub@redhat.com>
479 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
480 to _looptemp_ clauses, other than the first two.
482 2018-07-17 Martin Liska <mliska@suse.cz>
484 * opts.c: Do not enable OPT_falign_* for -Os.
486 2018-07-17 Martin Liska <mliska@suse.cz>
488 * align.h (MAX_CODE_ALIGN): New.
489 (MAX_CODE_ALIGN_VALUE): New.
490 * common/config/i386/i386-common.c (ix86_handle_option):
491 (MAX_CODE_ALIGN): Moved to align.h.
492 * final.c (MAX_CODE_ALIGN): Likewise.
493 * opts.c (parse_and_check_align_values):
494 (MAX_CODE_ALIGN): Likewise.
495 (MAX_CODE_ALIGN_VALUE): Likewise.
497 2018-07-17 Martin Liska <mliska@suse.cz>
499 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
500 in order to fulfil coding style.
501 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
502 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
503 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
504 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
505 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
506 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
507 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
508 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
509 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
511 2018-07-17 Martin Liska <mliska@suse.cz>
514 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
515 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
516 align_flags of label_to_alignment.
517 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
519 * config/m68k/m68k.c: Do not use removed align_labels_value and
521 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
522 (LOOP_ALIGN): Likewise.
523 (LABEL_ALIGN): Likewise.
524 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
525 Remove not used macro.
526 (rs6000_loop_align): Change return type to align_flags.
527 (rs6000_loop_align_max_skip): Remove.
528 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
529 Change return type to align_flags.
530 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
531 Remove not used macro.
532 (rs6000_loop_align): Change return type to align_flags.
533 (rs6000_loop_align_max_skip): Remove.
534 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
535 * config/rx/rx-protos.h (rx_align_for_label): Make it
537 * config/rx/rx.c (rx_align_for_label): Change return type
539 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
541 into align_flags class.
542 (LABEL_ALIGN): Likewise.
543 (LOOP_ALIGN): Likewise.
544 * config/s390/s390.c (s390_label_align): Use align_flags
546 (s390_asm_output_function_label): Likewise.
547 * config/sh/sh.c (sh_override_options_after_change):
548 Use align_flags class directly without macros.
549 (find_barrier): Likewise.
550 (barrier_align): Likewise.
551 (sh_loop_align): Likewise.
552 * config/spu/spu.c (spu_option_override):
553 Use align_flags_tuple::get_value instead of removed macros.
554 (spu_sched_init): Likewise.
555 * config/spu/spu.h (GTY): Likewise.
556 * config/visium/visium.c (visium_option_override):
557 Set "8" as default secondary alignment.
558 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
559 in order to guarantee secondary alignment of 8.
560 * coretypes.h: Include align.h header file.
561 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
562 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
563 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
564 * doc/tm.texi.in: Likewise.
565 * final.c (struct label_alignment): Remove not used structure.
566 (LABEL_ALIGN): Change type to align_flags.
567 (LOOP_ALIGN): Likewise.
568 (JUMP_ALIGN): Likewise.
569 (default_loop_align_max_skip): Remove.
570 (default_label_align_max_skip): Likewise.
571 (default_jump_align_max_skip): Likewise.
572 (default_label_align_after_barrier_max_skip):
573 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
574 (LABEL_TO_MAX_SKIP): Remove.
575 (label_to_alignment): Return align_flags type instead of integer.
576 (label_to_max_skip): Remove.
577 (align_fuzz): Use align_flags type.
578 (compute_alignments): Use align_flags type and use align_flags::max
579 to combine multiple alignments.
580 (grow_label_align): Grow vec instead of C array.
581 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
582 (shorten_branches): Use align_flags type and use align_flags::max
583 to combine multiple alignments.
584 (final_scan_insn_1): Remove usage of secondary alignment that comes
585 from label alignment, but instead use proper secondary alignment
586 which is computed in grow_label_align.
587 * flags.h (struct align_flags_tuple): Move to align.h.
588 (struct align_flags): Likewise.
589 (state_align_loops): Rename to align_loops.
590 (state_align_jumps): Rename to align_jumps.
591 (state_align_labels): Rename to align_labels.
592 (state_align_functions): Rename to align_functions.
593 (align_loops_log): Remove.
594 (align_jumps_log): Remove.
595 (align_labels_log): Remove.
596 (align_functions_log): Remove.
597 (align_loops_max_skip): Remove.
598 (align_jumps_max_skip): Remove.
599 (align_labels_max_skip): Remove.
600 (align_functions_max_skip): Remove.
601 (align_loops_value): Remove.
602 (align_jumps_value): Remove.
603 (align_labels_value): Remove.
604 (align_functions_value): Remove.
605 * output.h (label_to_alignment): Change return type to align_flags.
606 (label_to_max_skip): Remove.
607 * target.def: Remove loop_align_max_skip, label_align_max_skip,
608 jump_align_max_skip macros.
609 * targhooks.h (default_loop_align_max_skip): Remove.
610 (default_label_align_max_skip): Likewise.
611 (default_jump_align_max_skip): Likewise.
612 (default_label_align_after_barrier_max_skip): Remove.
613 * toplev.c (read_log_maxskip): Use ::normalize function.
614 (parse_N_M): Remove not used argument and also call ::normalize.
615 (parse_alignment_opts): Do not pass unused arguments.
616 * varasm.c (assemble_start_function): Use directly align_functions
617 instead of removed macros.
618 * system.h: Do not poison removed macros.
620 2018-07-17 Jakub Jelinek <jakub@redhat.com>
623 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
624 and cond temporaries don't have reference type if iterator has
625 pointer type. For init use &for_pre_body instead of pre_p if
626 for_pre_body is non-empty.
628 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
630 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
631 double-double modes to SFmode directly directly.
632 (trunc<mode>sf2_fprs): Delete.
634 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
636 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
637 for conversions between IFmode and the decimal floating point modes.
638 (init_float128_ieee): Use the correct names for conversions between
639 KFmode and the decimal floating point modes.
641 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
643 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
644 for the conversions between TDmode and IFmode.
645 (init_float128_ieee): Use more correct names for the conversions
646 between TDmode and KFmode.
648 2018-07-16 Jakub Jelinek <jakub@redhat.com>
650 PR tree-optimization/86526
651 * builtins.c (expand_builtin_memcmp): Formatting fixes.
652 (inline_expand_builtin_string_cmp): Likewise.
653 (inline_string_cmp): Likewise. Use c_readstr instead of
654 builtin_memcpy_read_str. Add unit_mode temporary.
656 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
659 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
660 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
662 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
664 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
665 Alphabetize prototypes of built-in functions, separating out
666 built-in functions that are listed in this section but should be
669 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
672 * expmed.c (emit_store_flag): Do not emit setcc followed by a
673 conditional move when trapping comparison was split to a
674 non-trapping one (and vice versa).
676 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
678 * config/s390/s390.c (s390_function_profiler): Generate nops
679 instead of profiler call sequences.
680 * config/s390/s390.opt: Add the new option.
682 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
684 * config/s390/s390.c (s390_function_profiler): Generate
685 __mcount_loc section.
686 * config/s390/s390.opt: Add the new option.
688 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
690 * common.opt: Add the new warning.
691 * config/s390/s390.c (s390_function_profiler): Emit "brasl
692 %r0,__fentry__" when -mfentry is specified.
693 (s390_option_override_internal): Disallow -mfentry for 31-bit
695 * config/s390/s390.opt: Add the new option.
697 2018-07-16 Richard Biener <rguenther@suse.de>
700 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
701 for function-local FUNCTION_DECL and RESULT_DECL immediately.
703 2018-07-16 Martin Liska <mliska@suse.cz>
706 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
709 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
711 * config/arc/arcHS.md: Update ARCHS scheduling rules.
713 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
715 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
717 * config/arc/arc-cpus.def (hs4x): New cpu.
719 * config/arc/arc-tables.opt: Regenerate.
720 * config/arc/arc.c (arc_sched_issue_rate): New function.
721 (TARGET_SCHED_ISSUE_RATE): Define.
722 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
723 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
725 (attr tune): Add ARCHS4x tune values.
726 (attr tune_dspmpy): Define.
727 (*tst): Correct instruction type.
728 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
729 * config/arc/arcHS4x.md: New file.
730 * config/arc/fpu.md: Update instruction type attributes.
731 * config/arc/t-multilib: Regenerate.
733 2018-07-16 Tom de Vries <tdevries@suse.de>
736 * var-tracking.c (vt_initialize): Fix pre_dec handling.
738 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
740 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
743 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
745 PR tree-optimization/86514
746 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
747 conversion to a boolean type from a type with greater precision.
749 2018-07-16 Tom de Vries <tdevries@suse.de>
751 * var-tracking.c (vt_initialize): Print adjusted insn slim if
752 dump_flags request TDF_SLIM.
754 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
756 * fold-const.c (int_const_binop_1): Abstract...
757 (wide_int_binop): ...wide int code here.
758 (poly_int_binop): ...poly int code here.
759 Abstract the rest of int_const_binop_1 into int_const_binop.
760 * fold-const.h (wide_int_binop): New.
761 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
762 Remove useless PLUS/MINUS_EXPR case.
763 (zero_nonzero_bits_from_vr): Move wide int code...
764 (zero_nonzero_bits_from_bounds): ...here.
765 (extract_range_from_binary_expr_1): Move mask optimization code...
766 (range_easy_mask_min_max): ...here.
767 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
768 (range_easy_mask_min_max): New.
770 2018-07-15 Jeff Law <law@redhat.com>
773 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
776 2018-07-14 Jim Wilson <jimw@sifive.com>
778 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
780 2018-07-14 Paul Koning <ni1d@arrl.net>
782 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
784 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
786 * lto-streamer-out.c (copy_function_or_variable): Dump info about
789 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
790 Steve Munroe <munroesj52@gmail.com>
792 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
793 (_mm_andnot_si128): Likewise.
794 (_mm_or_si128): Likewise.
795 (_mm_xor_si128): Likewise.
797 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
800 * builtins.c (expand_builtin_memcmp): Inline the calls first
801 when result_eq is false.
802 (expand_builtin_strcmp): Inline the calls first.
803 (expand_builtin_strncmp): Likewise.
804 (inline_string_cmp): New routine. Expand a string compare
805 call by using a sequence of char comparison.
806 (inline_expand_builtin_string_cmp): New routine. Inline expansion
807 a call to str(n)cmp/memcmp.
808 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
810 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
812 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
814 * config/arm/driver-arm.c: Include arm-native.h.
815 (host_detect_local_cpu): Use auto-generated data tables.
816 (vendors, arm_cpu_table): Delete. Move part information to ...
817 * config/arm/arm-cpus.in: ... here.
818 * config/arm/parsecpu.awk (gen_native): New function.
819 (vendor, part): New CPU fields.
820 (END): Add support for building the native CPU detection tables.
821 * config/arm/t-arm (arm-native.h): Add build rule.
822 (driver-arm.o): Add dependency on arm-native.h.
824 2018-07-13 Richard Biener <rguenther@suse.de>
827 * match.pd (addr1 - addr2): Allow either of the operand to
830 2018-07-13 Tom de Vries <tdevries@suse.de>
832 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
835 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
837 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
838 arrays instead of numbered variables.
840 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
842 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
843 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
845 2018-07-13 Richard Biener <rguenther@suse.de>
848 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
849 instead of get_context_die.
851 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
852 Richard Biener <rguenther@suse.de>
855 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
856 that the loop latch destination where phi is defined.
858 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
860 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
861 (riscv_expand_epilogue): Add assertion to check interrupt mode.
862 (riscv_set_current_function): Extract getting interrupt type to new
864 (riscv_get_interrupt_type): New function.
865 (riscv_merge_decl_attributes): New function, checking interrupt type is
867 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
869 2018-07-12 Paul Koning <ni1d@arrl.net>
871 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
874 2018-07-12 Paul Koning <ni1d@arrl.net>
876 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
877 zero reference, add doloop_end instead.
878 * doc/md.texi (decrement_and_branch_until_zero): Remove.
879 (Looping patterns): Remove decrement_and_branch_until_zero. Add
880 detail for doloop_end.
882 2018-07-12 Martin Sebor <msebor@redhat.com>
885 * attribs.c (decl_attributes): Reject conflicting attributes before
886 calling attribute handlers.
888 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
890 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
892 (gcc::dump_manager::get_dump_file_name): likewise.
893 (dump_begin): Likewise.
894 * dumpfile.h (dump_begin): Update prototype.
895 (gcc::dump_manager::get_dump_file_name,
896 gcc::dump_manager::get_dump_file_name): Update prototype.
898 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
900 * internal-fn.h (vectorizable_internal_fn_p): New function.
901 * tree-vect-slp.c (compatible_calls_p): Likewise.
902 (vect_build_slp_tree_1): Remove nops argument. Handle calls
903 to internal functions.
904 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
906 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
908 * fold-const.h (inverse_conditions_p): Declare.
909 * fold-const.c (inverse_conditions_p): New function.
910 * match.pd: Use inverse_conditions_p. Add folds of view_converts
911 that test the inverse condition of a conditional internal function.
912 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
913 * internal-fn.c (internal_fn_mask_index): Handle conditional
915 (vectorized_internal_fn_supported_p): New function.
916 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
917 (any_pred_load_store): Replace with...
918 (need_to_predicate): ...this new variable.
919 (redundant_ssa_names): New variable.
920 (ifcvt_can_use_mask_load_store): Move initial checks to...
921 (ifcvt_can_predicate): ...this new function. Handle tree codes
922 for which a conditional internal function exists.
923 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
924 instead of ifcvt_can_use_mask_load_store. Update after variable
926 (predicate_load_or_store): New function, split out from
927 predicate_mem_writes.
928 (check_redundant_cond_expr): New function.
929 (value_available_p): Likewise.
930 (predicate_rhs_code): Likewise.
931 (predicate_mem_writes): Rename to...
932 (predicate_statements): ...this. Use predicate_load_or_store
933 and predicate_rhs_code.
934 (combine_blocks, tree_if_conversion): Update after above name changes.
935 (ifcvt_local_dce): Handle redundant_ssa_names.
936 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
937 general conditional functions.
938 * tree-vect-stmts.c (vectorizable_call): Likewise.
940 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
941 Alan Hayward <alan.hayward@arm.com>
942 David Sherwood <david.sherwood@arm.com>
944 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
945 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
946 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
947 plus and minus and convert them into IFN_COND_FMA-based sequences.
948 (convert_mult_to_fma): Handle conditional plus and minus.
950 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
952 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
953 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
954 (cond_fnms_optab): New optabs.
955 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
957 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
958 * internal-fn.h (get_conditional_internal_fn): Declare.
959 (get_unconditional_internal_fn): Likewise.
960 * internal-fn.c (cond_ternary_direct): New macro.
961 (expand_cond_ternary_optab_fn): Likewise.
962 (direct_cond_ternary_optab_supported_p): Likewise.
963 (FOR_EACH_COND_FN_PAIR): Likewise.
964 (get_conditional_internal_fn): New function.
965 (get_unconditional_internal_fn): Likewise.
966 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
967 (gimple_match_op::gimple_match_op): Add a new overload for 5
969 (gimple_match_op::set_op): Likewise.
970 (gimple_resimplify5): Declare.
971 * genmatch.c (decision_tree::gen): Generate simplifications for
973 * gimple-match-head.c (gimple_simplify): Define an overload for
974 5 operands. Handle calls with 5 arguments in the top-level overload.
975 (convert_conditional_op): Handle conversions from unconditional
976 internal functions to conditional ones.
977 (gimple_resimplify5): New function.
978 (build_call_internal): Pass a fifth operand.
979 (maybe_push_res_to_seq): Likewise.
980 (try_conditional_simplification): Try converting conditional
981 internal functions to unconditional internal functions.
982 Handle 3-operand unconditional forms.
983 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
984 Define ternary equivalents of the current rules for binary conditional
986 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
988 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
989 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
990 (optab): Handle them.
991 (SVE_COND_FP_TERNARY): New int iterator.
992 (sve_fmla_op, sve_fmad_op): New int attributes.
993 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
994 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
995 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
997 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
999 * target.def (preferred_else_value): New target hook.
1000 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1001 * doc/tm.texi: Regenerate.
1002 * targhooks.h (default_preferred_else_value): Declare.
1003 * targhooks.c (default_preferred_else_value): New function.
1004 * internal-fn.h (conditional_internal_fn_code): Declare.
1005 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1006 (get_conditional_internal_fn): Use it.
1007 (conditional_internal_fn_code): New function.
1008 * gimple-match.h (gimple_match_cond): New struct.
1009 (gimple_match_op): Add a cond member function.
1010 (gimple_match_op::gimple_match_op): Update all forms to take a
1012 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1013 for the suboperation, but don't specify a particular else_value.
1014 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1015 (visit_nary_op, visit_reference_op_load): Pass
1016 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1017 * gimple-match-head.c: Include tree-eh.h
1018 (convert_conditional_op): New function.
1019 (maybe_resimplify_conditional_op): Likewise.
1020 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1021 (gimple_resimplify2): Likewise.
1022 (gimple_resimplify3): Likewise.
1023 (gimple_resimplify4): Likewise.
1024 (maybe_push_res_to_seq): Return null for conditional operations.
1025 (try_conditional_simplification): New function.
1026 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1028 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1030 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1032 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1034 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1036 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1038 (hash_tree): Do not hash DECL_FCONTEXT
1039 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1040 Do not stream DECL_FCONTEXT.
1041 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1042 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1044 2018-07-12 Richard Biener <rguenther@suse.de>
1047 * dwarf2out.c (gen_block_die): Only output blocks when they have
1048 at least one !DECL_IGNORED_P variable.
1050 2018-07-12 Richard Biener <rguenther@suse.de>
1053 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1054 Remove -mieee-fp handling.
1056 2018-07-12 Richard Biener <rguenther@suse.de>
1058 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1059 left-over from last patch.
1061 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1063 PR tree-optimization/86492
1064 * gimple-ssa-store-merging.c
1065 (imm_store_chain_info::coalesce_immediate_stores): Call
1066 check_no_overlap even for the merge_overlapping case. Formatting fix.
1068 2018-07-12 Richard Biener <rguenther@suse.de>
1071 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1072 move possibly trapping operations into the conditional.
1074 2018-07-12 Richard Biener <rguenther@suse.de>
1076 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1077 (vn_lookup_simplify_result): Remove recursion limit applied
1079 (vn_nary_build_or_lookup_1): Adjust.
1080 (try_to_simplify): Likewise.
1081 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1083 (gimple_resimplify2): Likewise.
1084 (gimple_resimplify3): Likewise.
1085 (gimple_resimplify4): Likewise.
1087 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1089 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1090 Use __mmask64 type instead of __mmask8 for __M argument.
1091 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1092 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1094 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1095 __mmask16 for __M argument.
1096 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1097 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1098 to __mmask16 instead of __mmask8.
1099 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1100 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1101 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1102 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1103 instead of __mmask16 for __U argument.
1104 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1105 __mmask16 instead of __mmask8 for __U argument.
1106 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1108 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1110 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1112 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1114 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1116 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1118 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1120 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1122 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1124 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1126 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1127 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1128 return type as well as __M argument type and all casts from __mmask8
1130 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1131 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1132 return type as well as __M argument type and all casts from __mmask8
1134 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1135 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1136 return type as well as __M argument type and all casts from __mmask8
1138 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1139 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1140 return type as well as __M argument type and all casts from __mmask8
1142 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1143 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1146 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1148 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1149 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1152 2018-07-11 Paul Koning <ni1d@arrl.net>
1154 * doc/md.texi (define_subst): Document how multiple occurrences of
1155 the same argument in the replacement pattern are handled.
1157 2018-07-11 Paul Koning <ni1d@arrl.net>
1159 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1161 (Common Type Attributes): Add "mode" attribute.
1163 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1165 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1166 stream DECL_ORIGINAL_TYPE.
1167 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1168 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1169 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1170 Do not walk original type.
1171 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1173 (write_ts_decl_non_common_tree_pointers): Do not stream
1175 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1176 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1178 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1180 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1181 thread twice from the same starting edge.
1183 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1185 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1186 * gimple.c (gimple_call_nonnull_result_p): ...here...
1187 (gimple_call_nonnull_arg): ...and here.
1188 * gimple.h (gimple_call_nonnull_result_p): New.
1189 (gimple_call_nonnull_arg): New.
1191 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1193 * config/arm/arm-cpus.in: Move information from fpu field of each
1194 cpu definition to the isa field.
1195 * config/arm/parsecpu.awk (fpu): Delete match rule.
1196 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1198 2018-07-11 Richard Biener <rguenther@suse.de>
1201 * dwarf2out.c (init_sections_and_labels): Use
1202 output_asm_line_debug_info consistently.
1203 (dwarf2out_early_finish): Likewise.
1204 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1207 2018-07-11 Richard Biener <rguenther@suse.de>
1209 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1210 Rework father_bb setting in a way to avoid propagating constants
1211 multiple times on a loop body.
1213 2018-07-10 Mark Wielaard <mark@klomp.org>
1216 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1218 2018-07-10 Richard Biener <rguenther@suse.de>
1220 * hash-map.h (hash_map::iterator::operator*): Return
1221 references to key and value.
1223 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1226 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1227 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1228 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1229 (gimplify_omp_for): For composite loops, move outer
1230 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1231 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1232 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1233 TREE_LIST for both the original class iterator and the "last" helper
1234 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1235 loop, remember has_decl_expr from outer composite loops for the
1236 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1238 2018-07-09 Martin Sebor <msebor@redhat.com>
1242 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1243 accessing implicitly initialized array elements.
1244 * expr.c (string_constant): Handle string initializers of
1245 character arrays within aggregates.
1246 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1247 Store element offset. As a special case, handle zero size.
1248 (fold_nonarray_ctor_reference): Same.
1249 (fold_ctor_reference): Add argument. Store subobject offset.
1250 * gimple-fold.h (fold_ctor_reference): Add argument.
1252 2018-07-09 Paul Koning <ni1d@arrl.net>
1254 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1255 (pdp11_insn_cost): New function.
1256 (pdp11_md_asm_adjust): New function.
1257 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1258 (pdp11_rtx_costs): Update to match machine better.
1259 (output_addr_const_pdp11): Correct format mismatch warnings.
1260 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1261 * config/pdp11/pdp11.md: General change to add base_cost and/or
1262 length attributes for use by new pdp11_insn_cost function.
1263 (MIN_BRANCH): Correct definition.
1265 (doloop_end): Use standard pattern name for looping pattern.
1266 (doloop_end_nocc): New.
1267 (movsf): Add another constraint alternative.
1268 (zero_extendqihi2): Add constraint alternatives for not in place
1270 (zero_extendhisi2): Remove.
1271 (shift patterns): Add CC handling variants.
1275 (define_peephole2): New peephole to recognize mov that sets CC for
1278 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1280 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1281 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1284 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1287 * real.c (real_nextafter): Return true if result is denormal.
1289 2018-07-09 Martin Liska <mliska@suse.cz>
1291 * common.opt: Add back wrongly removed attribute.
1293 2018-07-09 Richard Biener <rguenther@suse.de>
1296 * dwarf2out.c (gen_block_die): For an early generated DIE
1297 always output high/low PC attributes.
1299 2018-07-09 Tom de Vries <tdevries@suse.de>
1301 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
1302 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
1303 onto VAR_DECL with abstract origin.
1305 2018-07-07 Jim Wilson <jimw@sifive.com>
1307 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
1309 2018-07-07 Tom de Vries <tdevries@suse.de>
1311 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
1313 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
1315 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
1317 (combine_bound): Use wide-int overflow calculation instead of
1319 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
1321 * fold-const.c (int_const_binop_2): Same.
1322 (extract_muldiv_1): Same.
1323 (fold_div_compare): Same.
1324 (fold_abs_const): Same.
1326 * poly-int.h (add): Same.
1330 * predict.c (predict_iv_comparison): Same.
1331 * profile-count.c (slow_safe_scale_64bit): Same.
1332 * simplify-rtx.c (simplify_const_binary_operation): Same.
1333 * tree-chrec.c (tree_fold_binomial): Same.
1334 * tree-data-ref.c (split_constant_offset_1): Same.
1335 * tree-if-conv.c (idx_within_array_bound): Same.
1336 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
1337 * tree-ssa-phiopt.c (minmax_replacement): Same.
1338 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
1339 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
1340 * vr-values.c (vr_values::adjust_range_with_scev): Same.
1341 * wide-int.cc (wi::add_large): Same.
1342 (wi::mul_internal): Same.
1343 (wi::sub_large): Same.
1344 (wi::divmod_internal): Same.
1345 * wide-int.h: Change overflow type to overflow_type for neg, add,
1346 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
1347 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
1348 mul_internal, divmod_internal.
1349 (overflow_type): New enum.
1350 (accumulate_overflow): New.
1352 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1354 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
1355 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
1357 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1359 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
1360 argument is checked for zero before entering loop, avoid checking again.
1362 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1364 * gimplify.h (generic_expr_could_trap_p): Set as global function.
1365 * gimplify.h (generic_expr_could_trap_p): Likwise.
1366 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
1368 2018-07-06 Jakub Jelinek <jakub@redhat.com>
1370 PR tree-optimization/86401
1371 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
1372 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
1373 (fold_bit_and_mask): ... here. New helper function for match.pd.
1374 * fold-const.h (fold_bit_and_mask): Declare.
1375 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
1377 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
1380 * target.def (translate_mode_attribute): New hook.
1381 * targhooks.h (default_translate_mode_attribute): Declare.
1382 * targhooks.c (default_translate_mode_attribute): New function.
1383 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
1384 * doc/tm.texi: Regenerate.
1385 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
1386 (rs6000_translate_mode_attribute): New function.
1388 2018-07-06 Paul Koning <ni1d@arrl.net>
1390 * doc/md.texi (define_split): Document DONE and FAIL.
1391 (define_peephole2): Ditto.
1393 2018-07-05 Jeff Law <law@redhat.com>
1395 PR tree-optimization/86010
1396 * tree-ssa-dse.c (compute_trims): More aggressively trim at
1397 both the head and tail of mem* and str* calls.
1399 2018-07-05 Jim Wilson <jimw@sifive.com>
1401 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
1403 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
1405 * config/aarch64/aarch64-simd.md: correct flags text for
1406 MIN_EXPR replacement.
1408 2018-07-05 James Clarke <jrtc27@jrtc27.com>
1410 * configure: Regenerated.
1412 2018-07-05 Carl Love <cel@us.ibm.com>
1414 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
1415 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
1416 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
1417 VSX_BUILTIN_DOUBLEL_V4SF.
1419 2018-07-05 Martin Sebor <msebor@redhat.com>
1422 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
1423 than its domain to compute its the upper bound of a char array.
1425 2018-07-05 Nathan Sidwell <nathan@acm.org>
1427 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1428 * doc/cpp.texi: Update comment.
1429 * doc/tm.texi: Rebuilt.
1430 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1431 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1432 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1433 headers affected by extern "C".
1434 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1435 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1436 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1437 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1438 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1439 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1440 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1441 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1442 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1443 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1444 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1445 config/sparc/sp64-elf.h, config/spu/spu.h,
1446 config/stormy16/stormy16.h, config/v850/v850.h,
1447 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1448 define NO_IMPLICIT_EXTERN_C.
1449 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1451 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1454 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1455 instead of GET_MODE_SIZE when comparing Units.
1457 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1460 * rtlanal.c (set_noop_p): Constrain on mode change,
1461 include hard-reg-set.h
1463 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1465 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1467 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1470 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1473 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1475 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1477 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1480 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1482 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1484 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1487 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1489 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1491 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1492 neg_*_op* variables.
1494 2018-07-04 Martin Liska <mliska@suse.cz>
1496 * tree-switch-conversion.c: Define
1497 max_ratio_for_speed and max_ratio_for_size constants.
1499 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1500 Martin Liska <mliska@suse.cz>
1505 * common.opt: Rename align options with 'str_' prefix.
1506 * common/config/i386/i386-common.c (set_malign_value): New
1508 (ix86_handle_option): Use it to set -falign-* options/
1509 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1510 type from int to string.
1511 * config/aarch64/aarch64.c: Update default values from int
1513 * config/alpha/alpha.c (alpha_override_options_after_change):
1515 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1516 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1517 max skip conditionally.
1518 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1519 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1520 max skip conditionally.
1521 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1522 max skip conditionally.
1523 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1524 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1525 max skip conditionally.
1526 * config/i386/i386.c (struct ptt): Change type from int to
1528 (ix86_default_align): Set default values.
1529 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1530 max skip conditionally.
1531 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1532 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1533 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1534 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1535 max skip conditionally.
1536 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1537 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1538 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1539 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1540 max skip conditionally.
1541 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1542 * config/ia64/ia64.c (ia64_option_override): Set default values
1543 for alignment options.
1544 * config/m68k/m68k.c: Handle new str_align_* options.
1545 * config/mips/mips.c (mips_set_compression_mode): Change
1547 (mips_option_override): Set default values for options.
1548 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1550 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1552 * config/rx/rx.c (rx_option_override): Likewise.
1553 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1554 (LABEL_ALIGN): Use align_labels_log.
1555 (LOOP_ALIGN): Use align_loops_align.
1556 * config/s390/s390.c (s390_asm_output_function_label): Use new
1558 * config/sh/sh.c (sh_override_options_after_change):
1559 Change type of constants.
1560 * config/spu/spu.c (spu_sched_init): Likewise.
1561 * config/sparc/sparc.c (sparc_option_override): Set default
1563 * config/visium/visium.c (visium_option_override): Likewise.
1564 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1565 emit p2align format with last argument if it's not needed.
1566 * doc/invoke.texi: Document extended format of -falign-*.
1567 * final.c: Use align_labels alignment.
1568 * flags.h (struct target_flag_state): Change type to use
1570 (struct align_flags_tuple): New.
1571 (struct align_flags): Likewise.
1572 (align_loops_log): Redefine macro to use new types.
1573 (align_loops_max_skip): Redefine macro to use new types.
1574 (align_jumps_log): Redefine macro to use new types.
1575 (align_jumps_max_skip): Redefine macro to use new types.
1576 (align_labels_log): Redefine macro to use new types.
1577 (align_labels_max_skip): Redefine macro to use new types.
1578 (align_functions_log): Redefine macro to use new types.
1579 (align_loops): Redefine macro to use new types.
1580 (align_jumps): Redefine macro to use new types.
1581 (align_labels): Redefine macro to use new types.
1582 (align_functions): Redefine macro to use new types.
1583 (align_functions_max_skip): Redefine macro to use new types.
1584 (align_loops_value): New macro.
1585 (align_jumps_value): New macro.
1586 (align_labels_value): New macro.
1587 (align_functions_value): New macro.
1588 * function.c (invoke_set_current_function_hook): Propagate
1589 alignment values from flags to global variables default in
1591 * ipa-icf.c (sem_function::equals_wpa): Use
1592 cl_optimization_option_eq instead of memcmp.
1593 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1595 (cl_optimization_stream_in): Likewise.
1596 * optc-save-gen.awk: Support strings in cl_optimization.
1597 * opth-gen.awk: Likewise.
1598 * opts.c (finish_options): Remove error checking of invalid
1600 (MAX_CODE_ALIGN): Remove.
1601 (MAX_CODE_ALIGN_VALUE): Likewise.
1602 (parse_and_check_align_values): New function.
1603 (check_alignment_argument): Likewise.
1604 (common_handle_option): Use check_alignment_argument.
1605 * opts.h (parse_and_check_align_values): Declare.
1606 * toplev.c (init_alignments): Remove.
1607 (read_log_maxskip): New.
1608 (parse_N_M): Likewise.
1609 (parse_alignment_opts): Likewise.
1610 (backend_init_target): Remove usage of init_alignments.
1611 * toplev.h (parse_alignment_opts): Declare.
1612 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1614 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1615 * tree.c (cl_option_hasher::equal): New.
1616 * varasm.c: Use new global macros.
1618 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1620 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1621 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1622 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1623 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1624 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1625 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1626 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1627 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1628 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1629 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1630 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1631 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1632 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1634 2018-07-04 Martin Liska <mliska@suse.cz>
1635 Jonathan Wakely <jwakely@redhat.com>
1637 * coverage.c: Use correct type.
1638 * doc/invoke.texi: Language correction.
1640 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1643 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1644 ENDBRANCH for non-tail call which may return via indirect branch.
1645 * doc/extend.texi: Document indirect_return attribute.
1647 2018-07-03 Martin Sebor <msebor@redhat.com>
1649 PR tree-optimization/86274
1650 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1652 (format_floating): Correct handling of infinities and NaNs.
1654 2018-07-03 Martin Sebor <msebor@redhat.com>
1656 * print-tree.c (print_real_cst): New function.
1657 (print_node_brief): Call it.
1658 (print_node): Ditto.
1660 2018-07-03 Jeff Law <law@redhat.com>
1662 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1663 into a single pattern.
1665 * config/h8300/h8300.md (ors code_iterator): New.
1666 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1667 a single pattern and single splitter.
1668 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1669 (iorqi3_1, xorqi3_1): Likewise.
1670 (iorqi3, xorqi3 expanders): Similarly.
1672 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1673 (movmd_internal) into a single pattern using the P mode iterator.
1674 (movmd splitters): Similarly.
1675 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1676 (movsd splitters): Similarly.
1678 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1679 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1680 right table based on the size of the operand.
1681 * config/h8300/h8300.md (length_table): Corresponding changes. All
1682 references to "addb", "addw" and "addl" changed to "add".
1683 (btst patterns): Merge two variants into a single pattern.
1684 (tstqi, tsthi): Likewise.
1685 (addhi3_incdec, addsi3_incdec): Likewise.
1686 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1687 (mulhi3, mulsi3): Likewise.
1688 (udivhi3, udivsi3): Likewise.
1689 (divhi3, divsi3): Likewise.
1690 (andorqi3, andorhi3, andorsi3): Likewise.
1692 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1695 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1696 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1698 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1700 PR tree-optimization/85694
1701 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1702 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1704 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1705 (<u>avg<mode>3_ceil): New patterns.
1707 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1709 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1710 scan-tree-dump directive.
1711 * gcc.dg/vect/slp-perm-2.c: Likewise.
1712 * gcc.dg/vect/slp-perm-3.c: Likewise.
1713 * gcc.dg/vect/slp-perm-5.c: Likewise.
1714 * gcc.dg/vect/slp-perm-6.c: Likewise.
1715 * gcc.dg/vect/slp-perm-7.c: Likewise.
1716 * gcc.dg/vect/slp-perm-8.c: Likewise.
1718 2018-07-03 Marek Polacek <polacek@redhat.com>
1721 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1723 2018-07-03 Richard Biener <rguenther@suse.de>
1726 * tree-ssa-structalias.c (find_func_clobbers): Properly
1727 handle indirect calls.
1729 2018-07-03 Jeff Law <law@redhat.com>
1731 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1732 (shifts): New code iterator.
1733 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1734 expander. Fix HImode handling on H8/SX.
1735 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1736 (subqi3, subhi3, subsi3 expanders): Likewise.
1737 (andqi3, andhi3, andsi3 expanders): Likewise.
1738 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1739 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1740 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1741 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1742 (zero_extendqihi2, zero_extendqisi2): Likewise.
1743 (extendqihi2, extendqisi2): Likewise.
1744 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1745 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1746 (rotlqi3_1, rotlhi3_1): Likewise.
1747 (logicalhi3_sn, logicalsi3_sn): Likewise.
1748 (logicalhi3, logicalsi3): Likewise.
1750 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1752 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1753 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1754 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1755 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1756 type to append_pattern_def_seq instead of creating a stmt_vec_info
1758 (build_mask_conversion): Likewise. Remove vinfo argument.
1759 (vect_add_conversion_to_patterm): Likewise, renaming to...
1760 (vect_add_conversion_to_pattern): ...this.
1761 (vect_recog_mask_conversion_pattern): Update call to
1762 build_mask_conversion. Pass the vector type to
1763 append_pattern_def_seq here too.
1764 (vect_recog_gather_scatter_pattern): Update call to
1765 vect_add_conversion_to_pattern.
1767 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1769 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1770 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1771 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1772 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1773 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1774 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1775 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1776 append_pattern_def_seq instead of new_pattern_def_seq.
1777 (vect_recog_divmod_pattern): Do both of the above.
1778 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1781 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1783 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1784 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1785 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1786 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1787 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1788 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1789 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1790 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1791 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1792 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1793 parameter with a single stmt_vec_info.
1794 (vect_recog_func_ptr): Likewise.
1795 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1796 (vect_try_gather_scatter_pattern): ...this.
1797 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1798 the stmt_vec_info of the statement to be matched. Don't clear
1799 STMT_VINFO_RELATED_STMT.
1800 (vect_pattern_recog): Update call accordingly.
1802 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1804 PR tree-optimization/85694
1805 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1806 (uavgM3_ceil): Document new optabs.
1807 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1808 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1810 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1811 (savg_ceil_optab): New optabs.
1812 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1813 (vect_vect_recog_func_ptrs): Add it.
1814 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1815 constant directly from the associated lhs.
1817 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1819 * tree-vect-patterns.c (vect_split_statement): New function.
1820 (vect_convert_input): Use it to try to split an existing cast.
1822 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1824 * poly-int.h (print_hex): New function.
1825 * dumpfile.h (dump_dec, dump_hex): Declare.
1826 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1827 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1828 min_input_precision, operation_precision and operation_sign.
1829 * tree-vect-patterns.c (vect_get_range_info): New function.
1830 (vect_same_loop_or_bb_p, vect_single_imm_use)
1831 (vect_operation_fits_smaller_type): Delete.
1832 (vect_look_through_possible_promotion): Add an optional
1833 single_use_p parameter.
1834 (vect_recog_over_widening_pattern): Rewrite to use new
1835 stmt_vec_info infomration. Handle one operation at a time.
1836 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1837 (vect_truncatable_operation_p, vect_set_operation_type)
1838 (vect_set_min_input_precision): New functions.
1839 (vect_determine_min_output_precision_1): Likewise.
1840 (vect_determine_min_output_precision): Likewise.
1841 (vect_determine_precisions_from_range): Likewise.
1842 (vect_determine_precisions_from_users): Likewise.
1843 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1844 (vect_vect_recog_func_ptrs): Put over_widening first.
1846 (vect_pattern_recog): Call vect_determine_precisions.
1848 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1850 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1851 statements that have been replaced by further pattern statements.
1852 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1854 2018-07-03 Richard Biener <rguenther@suse.de>
1856 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1857 always set *dt. Dump vectype in vectype overload.
1858 * dumpfile.h (dump_gimple_expr): New function.
1859 (dump_gimple_expr_loc): Likewise.
1860 * dumpfile.c (dump_gimple_expr): New function.
1861 (dump_gimple_expr_loc): Likewise.
1863 2018-07-02 Jeff Law <law@redhat.com>
1865 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1866 the H8/300, H8/300H and H8/S variants into a single pattern.
1867 (movhi_h8300, movqi_h8300hs): Similarly.
1868 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1869 (QHI mode iterator): New.
1871 * config/h8300/h8300.md: Remove trailing whitespace.
1873 2018-07-02 Jim Wilson <jimw@sifive.com>
1875 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1876 instead of emit_insn for interrupt returns.
1877 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1878 (riscv_sret, riscv_uret): Likewise.
1880 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1882 * pretty-print.c (selftest::test_pp_format): Move save and restore
1883 of quotes to class auto_fix_quotes, and add an instance.
1884 * selftest.c: Include "intl.h".
1885 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1886 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1887 * selftest.h (selftest::auto_fix_quotes): New class.
1889 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1891 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1892 (aarch64_sve_prepare_conditional_op): Remove.
1893 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1894 Allow aarch64_simd_reg_or_zero as select operand; remove
1895 the aarch64_sve_prepare_conditional_op call.
1896 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1897 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1898 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1899 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1900 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1901 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1902 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1903 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1904 and a splitters to match all of the *_any patterns.
1905 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1907 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1908 (SVE_COND_FP_BINARY_REV): Remove.
1909 (sve_int_op_rev, sve_fp_op_rev): New.
1910 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1911 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1912 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1913 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1914 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1915 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1916 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1917 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1918 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1920 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1921 Remove match_dup 1 from the inner unspec.
1922 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1924 * config/aarch64/aarch64.md (movprfx): New attr.
1925 (length): Default movprfx to 8.
1926 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1927 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1928 (*<su>mul<SVE_I>3_highpart): Likewise.
1929 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1930 (*v<ASHIFT><SVE_I>3): Likewise.
1931 (*<su><MAXMIN><SVE_I>3): Likewise.
1932 (*<su><MAXMIN><SVE_F>3): Likewise.
1933 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1934 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1935 (*div<SVE_F>4): Likewise.
1937 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1939 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1942 2018-07-02 Richard Biener <rguenther@suse.de>
1944 PR tree-optimization/86363
1945 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1946 memset argument refers to a non-variable address.
1948 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1950 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1951 {PLUS,MINUS}_EXPR code to...
1952 (adjust_symbolic_bound): ...here,
1953 (combine_bound): ...here,
1954 (set_value_range_with_overflow): ...and here.
1956 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1958 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1960 (extract_range_from_abs_expr): ...here.
1962 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1964 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1965 -fno-omit-frame-pointer when not optimizing.
1967 2018-07-02 Martin Liska <mliska@suse.cz>
1970 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1971 (propagate_nothrow): Likewise.
1973 2018-07-02 Martin Liska <mliska@suse.cz>
1976 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1978 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1980 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1981 function in r262149, changing "loc" param from source_location to
1982 const dump_location_t &.
1983 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1984 declaration, as above.
1986 2018-07-01 Paul Koning <ni1d@arrl.net>
1988 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1989 -munit-asm, -mgnu-asm, -mdec-asm.
1990 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1991 (pdp11_output_labelref): New.
1992 (pdp11_output_def): New.
1993 (pdp11_output_addr_vec_elt): New.
1994 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1995 %# and %@ format codes.
1996 (pdp11_option_override): New.
1997 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1998 (pdp11_output_ident): New.
1999 (pdp11_asm_named_section): New.
2000 (pdp11_asm_init_sections): New.
2001 (pdp11_file_start): New.
2002 (pdp11_file_end): New.
2003 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2004 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2005 %o, like %c but octal.
2006 (pdp11_option_override): New.
2007 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2009 (DATA_SECTION_ASM_OP): Ditto.
2010 (READONLY_DATA_SECTION_ASM_OP): New.
2011 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2012 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2013 (ASM_OUTPUT_LABELREF): Ditto.
2014 (ASM_OUTPUT_DEF): Ditto.
2015 (ASM_OUTPUT_EXTERNAL): New.
2016 (ASM_OUTPUT_SOURCE_FILENAME): New.
2017 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2018 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2019 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2020 %# and %@ format codes.
2021 * config/pdp11/pdp11.opt (mgnu-asm): New.
2022 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2023 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2024 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2026 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2028 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2029 dereferencing path[] beyond its length.
2031 (debug_all_paths): New.
2032 (rewire_first_differing_edge): New.
2033 (adjust_paths_after_duplication): New.
2034 (duplicate_thread_path): Call adjust_paths_after_duplication.
2036 (thread_through_all_blocks): Add new argument to
2037 duplicate_thread_path.
2039 2018-06-30 Jim Wilson <jimw@sifive.com>
2041 * config/riscv/predicates.md (p2m1_shift_operand): New.
2042 (high_mask_shift_operand): New.
2043 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2044 pattern using p2m1_shift_operand.
2045 (lshsi3_zero_extend_3+2): New combiner pattern using
2046 high_mask_shift_operand.
2048 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2050 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2052 (vect_recog_rotate_pattern): ...here.
2053 (vect_convert_input): Try to insert casts of invariants in the
2055 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2056 preheader to be empty.
2058 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2060 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2061 vector type. If given, install it in the new statement's
2063 (vect_element_precision): New function.
2064 (vect_unpromoted_value): New struct.
2065 (vect_unpromoted_value::vect_unpromoted_value): New function.
2066 (vect_unpromoted_value::set_op): Likewise.
2067 (vect_look_through_possible_promotion): Likewise.
2068 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2069 (vect_widened_op_tree, vect_convert_input): Likewise.
2070 (vect_convert_inputs, vect_convert_output): Likewise.
2071 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2072 to handle the optional cast of the multiplication result and
2073 vect_widened_op_tree to detect the widened multiplication itself.
2074 Do not require the input and output of promotion casts to have
2075 the same sign, but base the signedness of the operation on the
2076 input rather than the result. If the pattern includes two
2077 promotions, check that those promotions have the same sign.
2078 Do not restrict the MULT_EXPR handling to a double-width result;
2079 handle quadruple-width results and wider. Use vect_convert_inputs
2080 to convert the inputs to the common type.
2081 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2082 to handle the optional cast of the ABS result. Also allow a sign
2083 change or a sign extension between the ABS and MINUS.
2084 Use vect_widened_op_tree to detect the widened subtraction and use
2085 vect_convert_inputs to convert the inputs to the common type.
2086 (vect_handle_widen_op_by_const): Delete.
2087 (vect_recog_widen_op_pattern): New function.
2088 (vect_recog_widen_mult_pattern): Use it.
2089 (vect_recog_widen_shift_pattern): Likewise.
2090 (vect_recog_widen_sum_pattern): Use
2091 vect_look_through_possible_promotion to handle the promoted
2094 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2096 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2097 the containing gimple_seq *.
2098 * gimple-iterator.h (gsi_for_stmt): Declare it.
2099 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2100 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2101 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2102 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2103 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2105 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2107 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2108 statement being replaced is part of an existing pattern
2109 definition sequence, inserting the new pattern statements before
2111 (vect_pattern_recog_1): Don't return a bool. If the statement
2112 is already part of a pattern, instead apply pattern matching
2113 to the pattern definition statements. Don't clear the
2114 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2115 (vect_pattern_recog): Don't break after the first match;
2116 continue processing the pattern definition statements instead.
2117 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2119 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2121 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2122 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2123 (vect_recog_widen_sum_pattern): Use it.
2125 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2127 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2128 phi is not a pattern statement and has not been replaced by
2129 a pattern statement.
2130 * tree-vect-patterns.c (type_conversion_p): Don't check
2131 STMT_VINFO_IN_PATTERN_P.
2132 (vect_recog_vector_vector_shift_pattern): Likewise.
2133 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2134 the pattern statement rather than the original statement; check
2135 directly for a WIDEN_MULT_EXPR here.
2136 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2137 vect_is_simple_use to return the pattern statement rather
2138 than the original statement; use is_pattern_stmt_p to check
2139 for such a pattern statement.
2140 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2141 to return the pattern statement rather than the original statement;
2142 don't do the same transformation here.
2143 (vect_is_simple_use): If the defining statement has been replaced
2144 by a pattern statement, return the pattern statement instead.
2145 Remove the corresponding (local) transformation from the vectype
2148 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2150 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2151 end and default to null.
2152 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2153 (vectorizable_reduction): Update calls accordingly, dropping the
2154 gimple ** argument if the passed-back statement isn't needed.
2155 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2156 (vect_recog_rotate_pattern): Likewise.
2157 (vect_recog_mask_conversion_pattern): Likewise.
2158 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2159 (vect_mask_constant_operand_p): Likewise.
2160 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2161 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2162 (get_group_load_store_type, get_load_store_type): Likewise.
2163 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2164 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2165 (vectorizable_conversion, vectorizable_assignment): Likewise.
2166 (vectorizable_shift, vectorizable_operation): Likewise.
2167 (vectorizable_store, vect_is_simple_cond): Likewise.
2168 (vectorizable_condition, vectorizable_comparison): Likewise.
2169 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2170 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2171 and move it to the end. Cope with null def_stmt_outs.
2173 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2175 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2177 2018-06-29 Jeff Law <law@redhat.com>
2179 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2180 displacements for TARGET_V850E2V3 and newer.
2181 (TARGET_LRA_P): Remove. Defaults to LRA now.
2182 * config/v850/v850.md (sign23byte_load): Remove.
2183 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2184 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2186 2018-06-29 Martin Liska <mliska@suse.cz>
2189 * coverage.c (coverage_init): Mangle full path name.
2190 * doc/invoke.texi: Document the change.
2191 * gcov-io.c (mangle_path): New.
2192 * gcov-io.h (mangle_path): Likewise.
2193 * gcov.c (mangle_name): Use mangle_path for path mangling.
2195 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2197 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2198 if starting source register is not even.
2200 2018-06-29 Martin Liska <mliska@suse.cz>
2202 PR tree-optimization/86263
2203 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2204 Make edge redirection.
2206 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2208 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2209 (dump_scope_depth): New variable.
2210 (get_dump_scope_depth): New function.
2211 (dump_begin_scope): New function.
2212 (dump_end_scope): New function.
2213 * dumpfile.h (get_dump_scope_depth): New declaration.
2214 (dump_begin_scope): New declaration.
2215 (dump_end_scope): New declaration.
2216 (class auto_dump_scope): New class.
2217 (AUTO_DUMP_SCOPE): New macro.
2218 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2221 2018-06-29 Richard Biener <rguenther@suse.de>
2223 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2224 compute_all_dependences succeeds.
2225 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2226 exceed --param loop-max-datarefs-for-datadeps.
2228 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2230 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2232 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2235 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2236 alternative 0 in preferred_for_speed attribute.
2238 2018-06-28 Paul Koning <ni1d@arrl.net>
2240 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2241 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2242 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2243 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2244 for shift insn patterns.
2246 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2248 * cgraph.c (cgraph_node::get_body): Replace assignments to
2249 "dump_file" with calls to set_dump_file.
2250 * dumpfile.c (alt_dump_file): Make static, and group with...
2251 (alt_flags): ...this definition.
2252 (dumps_are_enabled): New variable.
2253 (refresh_dumps_are_enabled): New function.
2254 (set_dump_file): New function.
2255 (set_alt_dump_file): New function.
2256 (gcc::dump_manager::dump_start): Replace assignments to
2257 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2259 (gcc::dump_manager::dump_finish): Likewise.
2260 * dumpfile.h (alt_dump_file): Delete decl.
2261 (dumps_are_enabled): New variable decl.
2262 (set_dump_file): New function decl.
2263 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2265 * tree-nested.c (lower_nested_functions): Replace assignments to
2266 "dump_file" with calls to set_dump_file.
2268 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2270 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2271 goto_locus of each outgoing edge of each basic block.
2273 2018-06-28 Richard Biener <rguenther@suse.de>
2275 * dwarf2out.c (decl_scope_table): Remove.
2276 (push_decl_scope): Likewise.
2277 (pop_decl_scope): Likewise.
2278 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2279 (gen_struct_or_union_type_die): Likewise.
2280 (gen_tagged_type_die): Likewise.
2281 (dwarf2out_init): Do not initialize decl_scope_table.
2282 (dwarf2out_c_finalize): Do not free it.
2284 2018-06-28 Richard Biener <rguenther@suse.de>
2286 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2287 deciding whether to not re-use a DIE.
2289 2018-06-28 Richard Biener <rguenther@suse.de>
2291 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2292 DW_AT_abstract_origin attribute.
2294 2018-06-28 Martin Liska <mliska@suse.cz>
2296 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2297 Use newly introduced constants.
2298 * tree-switch-conversion.h (struct jump_table_cluster):
2299 Define max_ratio_for_size and max_ratio_for_speed.
2301 2018-06-28 Martin Liska <mliska@suse.cz>
2303 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2304 Add new checking assert to catch invalid state.
2305 (jump_table_cluster::can_be_handled): Handle single case
2307 (jump_table_cluster::is_beneficial): Bail out for such case.
2308 (bit_test_cluster::find_bit_tests):
2309 Add new checking assert to catch invalid state.
2310 (bit_test_cluster::can_be_handled): Handle single case
2312 (bit_test_cluster::is_beneficial): Bail out for such case.
2313 (switch_decision_tree::analyze_switch_statement):
2316 2018-06-28 Martin Liska <mliska@suse.cz>
2318 * common.opt: Introduce -completion option.
2319 * gcc.c (driver_handle_option): Handle it.
2320 (driver::main): Print completions if completion
2322 * opt-suggestions.c (option_proposer::get_completions):
2324 (option_proposer::suggest_completion): Likewise.
2325 (option_proposer::find_param_completions): Likewise.
2326 (verify_autocompletions): Likewise.
2327 (test_completion_valid_options): Likewise.
2328 (test_completion_valid_params): Likewise.
2329 (in_completion_p): Likewise.
2330 (empty_completion_p): Likewise.
2331 (test_completion_partial_match): Likewise.
2332 (test_completion_garbage): Likewise.
2333 (opt_proposer_c_tests): Likewise.
2334 * opt-suggestions.h: Declare new functions.
2335 * opts.c (common_handle_option): Handle OPT__completion_.
2336 * selftest-run-tests.c (selftest::run_tests): Add
2337 opt_proposer_c_tests.
2338 * selftest.c (assert_str_startswith): New.
2339 * selftest.h (assert_str_startswith): Likewise.
2340 (opt_proposer_c_tests): New.
2341 (ASSERT_STR_STARTSWITH): Likewise.
2343 2018-06-28 Martin Liska <mliska@suse.cz>
2345 * Makefile.in: Add opt-suggestions.o.
2346 * gcc-main.c: Include opt-suggestions.h.
2347 * gcc.c (driver::driver): Likewise.
2348 (driver::~driver): Remove m_option_suggestions.
2349 (driver::build_option_suggestions): Moved to option_proposer.
2350 (driver::suggest_option): Likewise.
2351 (driver::handle_unrecognized_options): Use option_proposer.
2352 * gcc.h (class driver): Add new memver m_option_proposer.
2353 * opt-suggestions.c: New file.
2354 * opt-suggestions.h: New file.
2356 2018-06-28 Martin Liska <mliska@suse.cz>
2358 * vec.h (class auto_string_vec): New (moved from auto_argvec).
2359 (auto_string_vec::~auto_string_vec): Likewise.
2361 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2363 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
2364 prevent_decl_creation_for_types fields up and add reset_location field.
2365 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
2366 statement if id->reset_location is true.
2367 (copy_edges_for_bb): Do not set goto_locus on the new edges if
2368 id->reset_location is true.
2369 (copy_phis_for_bb): Force input_location on the arguments if
2370 id->reset_location is true.
2371 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
2372 is set on the function to be inlined.
2374 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
2376 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
2378 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
2380 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
2381 registers for Pmode.
2382 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
2383 hard registers for the clobbered pseudo.
2385 2018-06-27 Paul Koning <ni1d@arrl.net>
2387 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2388 mutually exclusive options.
2389 * config/pdp11/constraints.md (h): New constraint.
2390 (O): Update definition to match shift code generation.
2391 (D): New constraint.
2392 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
2394 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
2396 (output_jump): Change arguments.
2397 (pdp11_fixed_cc_regs): New function.
2398 (pdp11_cc_mode): Ditto.
2399 (pdp11_expand_shift): Ditto.
2400 (pdp11_assemble_shift): Ditto.
2401 (pdp11_small_shift): Ditto.
2402 (pdp11_branch_cost): Remove.
2403 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
2405 (pdp11_register_move_cost): Update for CC registers.
2406 (pdp11_rtx_costs): Add case for LSHIFTRT.
2407 (pdp11_output_jump): Add CCNZ mode conditional branches.
2408 (notice_update_cc_on_set): Remove.
2409 (pdp11_cc_mode): New function.
2410 (simple_memory_operand): Correct pre/post decrement case.
2411 (no_side_effect_operand): New function.
2412 (pdp11_regno_reg_class): Add CC_REGS class.
2413 (pdp11_fixed_cc_regs): New function.
2414 (pdp11_small_shift): New function.
2415 (pdp11_expand_shift): New function to expand shift insns.
2416 (pdp11_assemble_shift): New function to output shifts.
2417 (pdp11_branch_cost): Remove.
2418 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
2419 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
2420 (WCHAR_TYPE): Ditto.
2421 (PTRDIFF_TYPE): Ditto.
2422 (ADJUST_INSN_LENGTH): New macro.
2423 (FIXED_REGISTERS): Add CC registers.
2424 (CALL_USED_REGISTERS): Ditto.
2426 (REG_CLASS_NAMES): Ditto.
2427 (REG_CLASS_CONTENTS): Ditto.
2428 (SELECT_CC_MODE): Use new function.
2429 (TARGET_FLAGS_REGNUM): New macro.
2430 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2431 (cc0_reg_rtx): Remove.
2432 (CC_STATUS_MDEP): Remove.
2433 (CC_STATUS_MDEFP_INIT): Remove.
2434 (CC_IN_FPU): Remove.
2435 (NOTICE_UPDATE_CC): Remove.
2436 (REGISTER_NAMES): Add CC registers.
2437 (BRANCH_COST): Change to constant 1.
2438 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2440 * config/pdp11/pdp11.opt (mbcopy): Remove.
2441 (mbcopy-builtin): Remove.
2442 (mbranch-cheap): Remove.
2443 (mbranch-expensive): Remove.
2444 * config/pdp11/predicates.md (expand_shift_operand): Update to
2445 match shift code generation.
2446 (ccnz_operator): New predicate.
2447 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2448 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2449 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2451 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2452 description of O constraint.
2454 2018-06-27 Jeff Law <law@redhat.com>
2455 Austin Law <austinklaw@gmail.com>
2457 * config/v850/v850.md (addsi3_set_flags): New pattern.
2458 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2459 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2460 (zero_extendhisi2_v850_set_flags): Likewise.
2461 (zero_extendqisi2_v850_set_flags): Likewise.
2462 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2463 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2464 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2466 * config/v850/v850-protos.h (notice_update_cc): Remove.
2467 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2468 (v850_print_operand): Handle 'D' and "d".
2469 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2470 Add handling of arithmetic/logical operations compared against zero.
2471 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2472 Do not look at v850_compare_op, instead get mode from last argument.
2473 (v850_gen_compare): Remove
2474 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2475 after reload for prologue insns.
2476 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2478 (construct_save_jarl): Likewise.
2479 (TARGET_FLAGS_REGNUM): Define.
2480 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2481 (NOTICE_UPDATE_CC): Remove.
2482 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2483 than cc0. Conditionalize on reload_completed.
2484 (cmpsi_insn, setfcc_insn): Likewise.
2485 (tst1 splitter): Turn into define_and_split which sets the flags
2487 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2488 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2489 (cstoresf4, cstoredf4): Clobber the flags.
2490 (cmpsi, cmpsf, cmpdf): Remove expanders.
2491 (setf_insn): Remove pattern.
2492 (addsi3): Turn into define_and_split which clobbers the flags after
2493 reload and a suitable pattern (addsi3_clobber_flags) for use after
2495 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2496 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2497 (ashrsi3, ashrsi3_v850e2): Likewise.
2498 (bins): Clobber the flags.
2499 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2500 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2501 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2502 (call_value_internal_short, call_value_internal_long): Likewise.
2503 (callt_save_interrupt, callt_return_interrupt): Likewise.
2504 (save_interrupt, return_interrupt): Likewise.
2505 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2506 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2507 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2508 (All FP comparisons): Only allow after reload has completed.
2510 (divh, divhu): Tweak output template.
2511 (branch_z_normal, branch_z_invert): Remove
2512 (branch_nz_normal, branch_nz_invert): Likewise.
2513 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2515 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2516 * config/v850/v850.c (notice_update_cc): Remove.
2517 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2518 (CC_NO_CARRY): Likewise.
2519 (NOTICE_UPDATE_CC): Define to nothing.
2520 * config/v850/v850.md: Remove block comment on cc0 handling
2521 Remove "cc" attribute from all patterns. Remove cc_status handling
2522 from all patterns. Minor formatting fixes.
2524 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2526 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2527 (cortex-a76.cortex-a55): Likewise.
2528 * config/aarch64/aarch64-tune.md: Regenerate.
2529 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2530 cortex-a76.cortex-a55.
2532 2018-06-27 Jeff Law <law@redhat.com>
2534 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2535 (MULTILIB_DIRNAMES): Similarly.
2537 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2539 * gimple.h (gimple_return_retbnd): Delete.
2540 (gimple_return_set_retbnd): Likewise.
2541 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2542 gimple_return_set_retbnd.
2543 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2544 gimple_return_retbnd and adjust.
2545 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2546 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2547 Explicitly return NULL in a couple more cases. Move assertion
2548 on debug statements and remove unreachable code.
2549 (reset_debug_binding): Do not test id->retbnd.
2550 (expand_call_inline): Do not set it.
2552 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2554 * configure.ac: Add --disable-gcov option.
2555 * configure: Regenerate.
2556 * Makefile.in: Honour @enable_gcov@.
2557 * doc/install.texi: Document --disable-gcov.
2559 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2561 * config/arm/arm-cpus.in (cortex-a76): New entry.
2562 (cortex-a76.cortex-a55): Likewise.
2563 * config/arm/arm-tables.opt: Regenerate.
2564 * config/arm/arm-tune.md: Likewise.
2565 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2566 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2567 cortex-a76.cortex-a55.
2569 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2572 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2574 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2576 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2578 (EPILOGUE_USES): Likewise.
2580 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2582 * tree-inline.c (remap_location): New function extracted from...
2583 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2584 (copy_phis_for_bb): ...here. Call remap_location.
2585 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2587 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2589 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2590 unaligned vsx for 16B memset.
2592 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2595 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2596 ieee128_float_type_node to long_double_type_node unless
2597 TARGET_LONG_DOUBLE_128 is set.
2599 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2601 * cfgloop.c (get_loop_location): Convert return type from
2602 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2603 by implicit construction from rtx_insn *, and using
2604 dump_user_location_t::from_function_decl for the fallback case.
2605 * cfgloop.h (get_loop_location): Convert return type from
2606 location_t to dump_user_location_t.
2607 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2608 dump_printf_loc to pass in a dump_location_t rather than a
2609 location_t, via the gimple stmt.
2610 * coverage.c (get_coverage_counts): Update calls to
2611 dump_printf_loc to pass in dump_location_t rather than a
2613 * doc/optinfo.texi (Dump types): Convert example of
2614 dump_printf_loc from taking "locus" to taking "insn". Update
2615 description of the "_loc" calls to cover dump_location_t.
2616 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
2618 (dump_user_location_t::dump_user_location_t): New constructors,
2619 from gimple *stmt and rtx_insn *.
2620 (dump_user_location_t::from_function_decl): New function.
2621 (dump_loc): Make static.
2622 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2623 const dump_location_t &.
2624 (dump_generic_expr_loc): Delete.
2625 (dump_printf_loc): Convert param "loc" from location_t to
2626 const dump_location_t &.
2627 (selftest::test_impl_location): New function.
2628 (selftest::dumpfile_c_tests): New function.
2629 * dumpfile.h: Include "profile-count.h".
2630 (class dump_user_location_t): New class.
2631 (struct dump_impl_location_t): New struct.
2632 (class dump_location_t): New class.
2633 (dump_printf_loc): Convert 2nd param from source_location to
2634 const dump_location_t &.
2635 (dump_generic_expr_loc): Delete.
2636 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2637 const dump_location_t &.
2638 * gimple-fold.c (fold_gimple_assign): Update call to
2639 dump_printf_loc to pass in a dump_location_t rather than a
2640 location_t, via the gimple stmt.
2641 (gimple_fold_call): Likewise.
2642 * gimple-loop-interchange.cc
2643 (loop_cand::analyze_iloop_reduction_var): Update for change to
2644 check_reduction_path.
2645 (tree_loop_interchange::interchange): Update for change to
2647 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2648 change in return-type of find_loop_location.
2649 (graphite_regenerate_ast_isl): Likewise.
2650 * graphite-optimize-isl.c (optimize_isl): Likewise.
2651 * graphite.c (graphite_transform_loops): Likewise.
2652 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2653 pass in a dump_location_t rather than a location_t, via the
2655 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2656 * ipa.c (walk_polymorphic_call_targets): Likewise.
2657 * loop-unroll.c (report_unroll): Convert "locus" param from
2658 location_t to dump_location_t.
2659 (decide_unrolling): Update for change to get_loop_location's
2661 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2662 location_t to dump_user_location_t.
2663 (grid_find_single_omp_among_assignments_1): Updates calls to
2664 dump_printf_loc to pass in a dump_location_t rather than a
2665 location_t, via the gimple stmt.
2666 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2667 location_t to dump_location_t. Updates calls to dump_printf_loc
2668 to pass in a dump_location_t rather than a location_t, via the
2670 (grid_inner_loop_gridifiable_p): Likewise.
2671 (grid_dist_follows_simple_pattern): Likewise.
2672 (grid_gfor_follows_tiling_pattern): Likewise.
2673 (grid_target_follows_gridifiable_pattern): Likewise.
2674 (grid_attempt_target_gridification): Convert initialization
2675 of local "grid" from memset to zero-initialization; FIXME: does
2676 this require C++11? Update call to dump_printf_loc to pass in a
2677 optinfo_location rather than a location_t, via the gimple stmt.
2678 * profile.c (read_profile_edge_counts): Updates call to
2679 dump_printf_loc to pass in a dump_location_t rather than a
2681 (compute_branch_probabilities): Likewise.
2682 * selftest-run-tests.c (selftest::run_tests): Call
2684 * selftest.h (dumpfile_c_tests): New decl.
2685 * tree-loop-distribution.c (pass_loop_distribution::execute):
2686 Update for change in return type of find_loop_location.
2687 * tree-parloops.c (parallelize_loops): Likewise.
2688 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2689 "locus" from location_t to dump_user_location_t.
2690 (canonicalize_loop_induction_variables): Likewise.
2691 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2692 for change in return type of find_loop_location.
2693 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2694 to dump_printf_loc to pass in a dump_location_t rather than a
2695 location_t, via the stmt.
2696 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2698 * tree-vect-loop-manip.c (find_loop_location): Convert return
2699 type from source_location to dump_user_location_t.
2700 (vect_do_peeling): Update for above change.
2701 (vect_loop_versioning): Update for change in type of
2703 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2704 from location_t to dump_user_location_t.
2705 (vect_estimate_min_profitable_iters): Update for change in type
2707 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2708 location_t to dump_location_t.
2709 (vect_slp_bb): Update for change in type of vect_location.
2710 * tree-vectorizer.c (vect_location): Convert from source_location
2711 to dump_user_location_t.
2712 (try_vectorize_loop_1): Update for change in vect_location's type.
2713 (vectorize_loops): Likewise.
2714 (increase_alignment): Likewise.
2715 * tree-vectorizer.h (vect_location): Convert from source_location
2716 to dump_user_location_t.
2717 (find_loop_location): Convert return type from source_location to
2718 dump_user_location_t.
2719 (check_reduction_path): Convert 1st param from location_t to
2720 dump_user_location_t.
2721 * value-prof.c (check_counter): Update call to dump_printf_loc to
2722 pass in a dump_user_location_t rather than a location_t; update
2723 call to error_at for change in type of "locus".
2724 (check_ic_target): Update call to dump_printf_loc to
2725 pass in a dump_user_location_t rather than a location_t, via the
2728 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2730 * config/s390/s390.h (enum processor_flags): Do not use
2731 default tune parameter when -march was specified.
2733 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2736 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2737 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2740 2018-06-26 Richard Biener <rguenther@suse.de>
2742 PR tree-optimization/86287
2744 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2745 (vect_analyze_loop): Initialize n_stmts.
2747 2018-06-26 Richard Biener <rguenther@suse.de>
2750 * fold-const.c (fold_convertible_p): Pointer extension
2753 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2756 * dwarf2out.c (loc_list_has_views): Adjust comments.
2757 (dw_loc_list): Split single cross-partition range with
2760 2018-06-25 Jeff Law <law@redhat.com>
2762 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2763 on -mbig-switch by default.
2765 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2767 (const_float_0_operand): Remove unused predicate.
2768 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2769 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2770 (recipsf2): New expander. Original pattern now called
2772 (recipdf2, recipdf2_insn): Similarly.
2773 (rsqrtsf2, rsqrtsf2_insn): Similarly
2774 (rsqrtdf2, rsqrtdf2_insn): Similarly
2776 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2778 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2779 Simplify logic for FreeBSD (twice).
2781 2018-06-25 Martin Sebor <msebor@redhat.com>
2783 PR tree-optimization/86204
2784 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2785 a strnlen result if it's less than the length of the string.
2787 2018-06-25 Martin Sebor <msebor@redhat.com>
2789 PR tree-optimization/85700
2790 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2791 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2792 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2794 2018-06-25 Martin Sebor <msebor@redhat.com>
2796 * doc/extend.texi (Zero-length arrays): Update and clarify.
2798 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2800 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2801 added IEEE/IBM long double multilib support on PowerPC little
2802 endian Linux systems.
2803 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2804 (MULTILIB_DEFAULTS): Likewise.
2805 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2807 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2808 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2809 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2811 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2814 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2815 (REORDER_45): Likewise.
2817 2018-06-25 Jeff Law <law@redhat.com>
2819 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2820 dividend to 32 bits. Adjust length.
2821 (udivmodhi4): Cleanup output template. Fix length.
2823 2018-06-25 Carl Love <cel@us.ibm.com>
2825 * config/rs6000/vsx.md: Change word selector to prefered location.
2827 2018-06-25 Richard Biener <rguenther@suse.de>
2829 PR tree-optimization/86304
2830 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2831 epilogue-if-converted loops as well.
2833 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2835 * lto-section-out.c (lto_begin_section): Do not print section
2836 name for noaddr and unnumbered dumps.
2838 2018-06-25 Richard Biener <rguenther@suse.de>
2840 * tree-vectorizer.h (struct vec_info_shared): New structure
2841 with parts split out from struct vec_info and loop_nest from
2842 struct _loop_vec_info.
2843 (struct vec_info): Adjust accordingly.
2844 (struct _loop_vec_info): Likewise.
2845 (LOOP_VINFO_LOOP_NEST): Adjust.
2846 (LOOP_VINFO_DATAREFS): Likewise.
2847 (LOOP_VINFO_DDRS): Likewise.
2848 (struct _bb_vec_info): Likewise.
2849 (BB_VINFO_DATAREFS): Likewise.
2850 (BB_VINFO_DDRS): Likewise.
2851 (struct _stmt_vec_info): Add dr_aux member.
2852 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2853 (DR_MISALIGNMENT_UNINITIALIZED): New.
2854 (set_dr_misalignment): Adjust.
2855 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2856 (vect_analyze_loop): Adjust prototype.
2857 (vect_analyze_loop_form): Likewise.
2858 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2859 Compute dependences lazily.
2860 (vect_record_base_alignments): Use shared datarefs/ddrs.
2861 (vect_verify_datarefs_alignment): Likewise.
2862 (vect_analyze_data_refs_alignment): Likewise.
2863 (vect_analyze_data_ref_accesses): Likewise.
2864 (vect_analyze_data_refs): Likewise.
2865 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2866 constructor parameter for shared part.
2867 (vect_analyze_loop_form): Pass in shared part and adjust.
2868 (vect_analyze_loop_2): Pass in storage for the number of
2869 stmts. Move loop nest finding to the caller. Compute
2871 (vect_analyze_loop): Pass in shared part.
2872 (vect_transform_loop): Verify shared datarefs are unchanged.
2873 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2874 constructor parameter for shared part.
2875 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2876 (vect_slp_bb): Verify shared datarefs are unchanged before
2878 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2880 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2881 DR_MISALIGNMENT_UNINITIALIZED.
2882 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2883 parameter for shared part.
2884 (vec_info::~vec_info): Adjust.
2885 (vec_info_shared::vec_info_shared): New.
2886 (vec_info_shared::~vec_info_shared): Likewise.
2887 (vec_info_shared::save_datarefs): Likewise.
2888 (vec_info_shared::check_datarefs): Likewise.
2889 (try_vectorize_loop_1): Construct shared part live for analyses
2890 of a single loop for multiple vector sizes.
2891 * tree-parloops.c (gather_scalar_reductions): Adjust.
2893 2018-06-25 Richard Biener <rguenther@suse.de>
2895 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2896 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2897 (vect_analyze_data_refs): Remove similar code from here and
2898 simplify accordingly.
2900 2018-06-25 Richard Biener <rguenther@suse.de>
2902 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2903 for reverse storage order accesses rather than asserting
2904 they cannot happen here.
2906 2018-06-25 Tom de Vries <tdevries@suse.de>
2909 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2910 Use data16 instead of .byte for insn prefix.
2912 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2915 * parser.c (make_char_string_pack): Pass this literal chars
2916 through cpp_interpret_string.
2917 (cp_parser_userdef_numeric_literal): Check the result of
2918 make_char_string_pack.
2920 2018-06-24 Maya Rashish <coypu@sdf.org>
2922 * ginclude/stddef.h: Simplify conditions around avoiding
2923 re-definition of __size_t.
2925 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2927 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2928 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2930 2018-06-22 Maya Rashish <coypu@sdf.org>
2932 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2933 made to OpenBSD and NetBSD through the years.
2935 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2937 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2938 behavior of vec_pack (vector double, vector double) to match
2939 behavior of vec_float2 (vector double, vector double).
2941 2018-06-22 Olivier Hainque <hainque@adacore.com>
2943 * gimplify.c (gimplify_function_tree): Prevent creation
2944 of a trampoline for the address of the current function
2945 passed to entry/exit instrumentation hooks.
2947 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2950 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2953 2018-06-22 Martin Liska <mliska@suse.cz>
2955 PR tree-optimization/86263
2956 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2957 Bail out if is_enabled is false.
2958 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2960 (jump_table_cluster::is_enabled): New function.
2962 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2964 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2965 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2966 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2967 (lto_input_ts_binfo_tree_pointers): Likewise.
2968 * tree-streamer-out.c (streamer_write_tree_bitfields,
2969 write_ts_binfo_tree_pointers): Likewise.
2970 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2972 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2974 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2976 2018-06-22 Martin Liska <mliska@suse.cz>
2978 * symbol-summary.h (get): Make it pure and inline move
2979 functionality from ::get function.
2980 (get): Remove and inline into ::get and ::get_create.
2981 (get_create): Move code from ::get function.
2983 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2986 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2987 -x assembler-with-cpp.
2989 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2991 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2992 _FILE_OFFSET_BITS=64 for C++.
2994 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2996 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2997 conversion insn that shows up when pr85657-3.c is compiled using
2998 IEEE 128-bit long double.
2999 (neg<mode>2_internal): Use the correct mode to check whether the
3000 mode is IBM extended.
3001 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3002 multiply and divide external functions from being created more
3005 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3007 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3009 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3010 the edge can be forwarded.
3011 (cfg_layout_merge_blocks): Likewise.
3013 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3015 * except.c (finish_eh_generation): Commit edge insertions only after
3016 the EH edges have been redirected from post-landing to landing pads.
3018 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3020 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3021 create_tmp_var_for to create the FRAME decl.
3022 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3024 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3026 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3027 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3028 debug statement when resetting its value.
3029 (expand_call_inline): Copy the locus of the call onto the assignment
3030 of the return value, if any. Use local variable in more cases.
3032 2018-06-21 Martin Liska <mliska@suse.cz>
3034 * ipa-pure-const.c (propagate_nothrow): Use
3035 funct_state_summaries->get.
3036 (dump_malloc_lattice): Likewise.
3037 (propagate_malloc): Likewise.
3039 2018-06-21 Richard Biener <rguenther@suse.de>
3041 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3042 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3043 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3045 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3046 BLOCK_ABSTRACT_ORIGIN unconditionally.
3048 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3050 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3052 * ipa-reference.c (ipa_reference_c_finalize): Delete
3053 ipa_ref_opt_sum_summaries and set it to NULL.
3055 2018-06-21 Tom de Vries <tdevries@suse.de>
3057 PR tree-optimization/85859
3058 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3059 test with comment from bb_no_side_effects_p.
3061 2018-06-21 Richard Biener <rguenther@suse.de>
3063 PR tree-optimization/86232
3064 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3065 max for constant niter.
3067 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3069 * config/aarch64/aarch64-simd.md
3070 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3072 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3074 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3075 Make opernads of the unspec commutative.
3077 2018-06-21 Richard Biener <rguenther@suse.de>
3079 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3080 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3081 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3082 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3084 (vect_record_base_alignments): Do not record base alignment
3085 for gathers or scatters.
3086 (vect_compute_data_ref_alignment): Drop return value that is always
3087 true. Bail out early for gathers or scatters.
3088 (vect_enhance_data_refs_alignment): Bail out early for gathers
3090 (vect_find_same_alignment_drs): Likewise.
3091 (vect_analyze_data_refs_alignment): Remove dead code.
3092 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3093 (vect_analyze_data_refs): For possible gathers or scatters do
3094 not create an alternate DR, just check their possible validity
3095 and mark them. Adjust DECL_NONALIASED handling to not rely
3097 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3098 update inits of gathers or scatters.
3099 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3100 Also copy gather/scatter flag to pattern vinfo.
3102 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3104 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3105 behavior of vec_packsu (vector unsigned long long, vector unsigned
3106 long long) to match behavior of vec_packs with same signature.
3108 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3109 Thomas Schwinge <thomas@codesourcery.com>
3110 Cesar Philippidis <cesar@codesourcery.com>
3112 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3113 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3114 (gimplify_adjust_omp_clauses): Likewise.
3115 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3116 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3117 (gimplify_omp_target_update): Update handling of acc update and
3119 * omp-low.c (install_var_field): Remove unused parameter
3120 base_pointers_restrict.
3121 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3122 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3124 (omp_target_base_pointers_restrict_p): Delete.
3125 (scan_omp_target): Update call to scan_sharing_clauses.
3126 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3128 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3129 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3130 (convert_local_omp_clauses): Likewise.
3131 * tree-pretty-print.c (dump_omp_clause): Likewise.
3132 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3134 (omp_clause_code_name): Likewise.
3136 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3139 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3142 PR tree-optimization/86231
3143 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3144 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3146 2018-06-20 Tom de Vries <tdevries@suse.de>
3148 PR tree-optimization/86097
3149 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3150 iv type if signedness of iv type is not the same as that of *nit.
3152 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3154 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3155 EDGE_EH edges, verify they are all EDGE_EH.
3157 2018-06-20 Maya Rashish <coypu@sdf.org>
3159 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3161 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3163 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3164 * config/aarch64/aarch64.c (xgene1_tunings): Add
3165 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3166 (aarch64_mode_valid_for_sched_fusion_p):
3167 Allow 16-byte modes.
3168 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3169 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3171 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3173 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3174 * config/aarch64/iterators.md (VQ2): New mode iterator.
3176 2018-06-20 Martin Liska <mliska@suse.cz>
3178 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3179 Change default ratio from 10 to 8.
3181 2018-06-20 Martin Liska <mliska@suse.cz>
3183 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3185 (bit_test_cluster::find_bit_tests): Likewise.
3186 (switch_decision_tree::analyze_switch_statement): Find clusters.
3187 * tree-switch-conversion.h (struct jump_table_cluster): Document
3190 2018-06-20 Martin Liska <mliska@suse.cz>
3192 * tree-switch-conversion.c (switch_conversion::collect):
3193 Record m_uniq property.
3194 (switch_conversion::expand): Bail out for special conditions.
3195 (group_cluster::~group_cluster): New.
3196 (group_cluster::group_cluster): Likewise.
3197 (group_cluster::dump): Likewise.
3198 (jump_table_cluster::emit): New.
3199 (switch_decision_tree::fix_phi_operands_for_edges): New.
3200 (struct case_node): Remove struct.
3201 (jump_table_cluster::can_be_handled): New.
3202 (case_values_threshold): Moved to header.
3203 (reset_out_edges_aux): Likewise.
3204 (jump_table_cluster::is_beneficial): New.
3205 (bit_test_cluster::can_be_handled): Likewise.
3206 (add_case_node): Remove.
3207 (bit_test_cluster::is_beneficial): New.
3208 (case_bit_test::cmp): New.
3209 (bit_test_cluster::emit): New.
3210 (expand_switch_as_decision_tree_p): Remove.
3211 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3212 (fix_phi_operands_for_edge): Likewise.
3213 (switch_decision_tree::analyze_switch_statement): New.
3214 (compute_cases_per_edge): Move ...
3215 (switch_decision_tree::compute_cases_per_edge): ... here.
3216 (try_switch_expansion): Likewise.
3217 (switch_decision_tree::try_switch_expansion): Likewise.
3218 (record_phi_operand_mapping): Likewise.
3219 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3220 (emit_case_decision_tree): Likewise.
3221 (switch_decision_tree::emit): Likewise.
3222 (balance_case_nodes): Likewise.
3223 (switch_decision_tree::balance_case_nodes): Likewise.
3224 (dump_case_nodes): Likewise.
3225 (switch_decision_tree::dump_case_nodes): Likewise.
3226 (emit_jump): Likewise.
3227 (switch_decision_tree::emit_jump): Likewise.
3228 (emit_cmp_and_jump_insns): Likewise.
3229 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3230 (emit_case_nodes): Likewise.
3231 (switch_decision_tree::emit_case_nodes): Likewise.
3232 (conditional_probability): Remove.
3233 * tree-switch-conversion.h (enum cluster_type): New.
3235 (struct cluster): Likewise.
3236 (cluster::cluster): Likewise.
3237 (struct simple_cluster): Likewise.
3238 (simple_cluster::simple_cluster): Likewise.
3239 (struct group_cluster): Likewise.
3240 (struct jump_table_cluster): Likewise.
3241 (struct bit_test_cluster): Likewise.
3242 (struct min_cluster_item): Likewise.
3243 (struct case_tree_node): Likewise.
3244 (case_tree_node::case_tree_node): Likewise.
3245 (jump_table_cluster::case_values_threshold): Likewise.
3246 (struct case_bit_test): Likewise.
3247 (struct switch_decision_tree): Likewise.
3248 (struct switch_conversion): Likewise.
3249 (switch_decision_tree::reset_out_edges_aux): Likewise.
3251 2018-06-20 Martin Liska <mliska@suse.cz>
3253 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3254 (hoist_edge_and_branch_if_true): Likewise.
3255 (expand_switch_using_bit_tests_p): Likewise.
3256 (struct case_bit_test): Likewise.
3257 (case_bit_test_cmp): Likewise.
3258 (emit_case_bit_tests): Likewise.
3259 (switch_conversion::switch_conversion): New class.
3260 (struct switch_conv_info): Remove old struct.
3261 (collect_switch_conv_info): More to ...
3262 (switch_conversion::collect): ... this.
3263 (check_range): Likewise.
3264 (switch_conversion::check_range): Likewise.
3265 (check_all_empty_except_final): Likewise.
3266 (switch_conversion::check_all_empty_except_final): Likewise.
3267 (check_final_bb): Likewise.
3268 (switch_conversion::check_final_bb): Likewise.
3269 (create_temp_arrays): Likewise.
3270 (switch_conversion::create_temp_arrays): Likewise.
3271 (free_temp_arrays): Likewise.
3272 (gather_default_values): Likewise.
3273 (switch_conversion::gather_default_values): Likewise.
3274 (build_constructors): Likewise.
3275 (switch_conversion::build_constructors): Likewise.
3276 (constructor_contains_same_values_p): Likewise.
3277 (switch_conversion::contains_same_values_p): Likewise.
3278 (array_value_type): Likewise.
3279 (switch_conversion::array_value_type): Likewise.
3280 (build_one_array): Likewise.
3281 (switch_conversion::build_one_array): Likewise.
3282 (build_arrays): Likewise.
3283 (switch_conversion::build_arrays): Likewise.
3284 (gen_def_assigns): Likewise.
3285 (switch_conversion::gen_def_assigns): Likewise.
3286 (prune_bbs): Likewise.
3287 (switch_conversion::prune_bbs): Likewise.
3288 (fix_phi_nodes): Likewise.
3289 (switch_conversion::fix_phi_nodes): Likewise.
3290 (gen_inbound_check): Likewise.
3291 (switch_conversion::gen_inbound_check): Likewise.
3292 (process_switch): Use the newly created class.
3293 (switch_conversion::expand): New.
3294 (switch_conversion::~switch_conversion): New.
3295 * tree-switch-conversion.h: New file.
3297 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3299 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
3300 tree-vect-patterns.c.
3301 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
3302 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
3303 (vect_recog_sad_pattern): Likewise.
3304 (vect_recog_widen_sum_pattern): Likewise.
3305 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
3306 (vect_recog_widen_shift_pattern): Remove the type_in argument.
3307 (vect_recog_rotate_pattern): Likewise.
3308 (vect_recog_mult_pattern): Likewise.
3309 (vect_recog_vector_vector_shift_pattern): Likewise.
3310 (vect_recog_divmod_pattern): Likewise.
3311 (vect_recog_mixed_size_cond_pattern): Likewise.
3312 (vect_recog_bool_pattern): Likewise.
3313 (vect_recog_mask_conversion_pattern): Likewise.
3314 (vect_try_gather_scatter_pattern): Likewise.
3315 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
3316 (vect_recog_over_widening_pattern): Likewise.
3317 (vect_recog_gather_scatter_pattern): Likewise.
3318 (vect_recog_func_ptr): Move from tree-vectorizer.h
3319 (vect_vect_recog_func_ptrs): Move further down the file.
3320 (vect_recog_func): Likewise. Remove the third argument.
3321 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
3322 (vect_pattern_recog_1): Expect the pattern function to do any
3323 necessary target tests. Also expect it to provide a vector type.
3324 Remove the type_in handling.
3326 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3328 * tree-vect-patterns.c (vect_pattern_detected): New function.
3329 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
3330 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
3331 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
3332 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
3333 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3334 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3335 (vect_recog_mask_conversion_pattern)
3336 (vect_try_gather_scatter_pattern): Likewise.
3338 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3340 * tree-vect-patterns.c (vect_get_internal_def): New function.
3341 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3342 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
3343 (search_type_for_mask_1): Use it.
3345 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3347 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3348 redundant WIDEN_SUM_EXPR handling.
3349 (vect_recog_sad_pattern): Likewise.
3351 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3353 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3354 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
3355 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
3356 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
3357 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
3359 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3361 * tree-vect-stmts.c (vectorizable_call): Make sure that we
3362 use the stmt_vec_info of the original bb statement for the
3363 new zero assignment, even if the call is part of a pattern.
3365 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3367 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
3368 that the sequence is attached to the original statement rather
3369 than the pattern statement.
3370 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
3371 PATTERN_DEF_SEQ from the original statement rather than
3372 the main pattern statement.
3373 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
3374 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
3375 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
3377 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3379 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
3380 definition statements before the early exit for statements that aren't
3382 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
3384 (vect_transform_loop): ...here. Process pattern definition
3385 statements without first checking whether the main pattern
3386 statement is live or relevant.
3388 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3390 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
3391 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
3393 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
3395 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
3396 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
3397 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
3398 (expand_block_compare): Change select_block_compare_mode call.
3399 (expand_strncmp_align_check): Use new functions, fix comment.
3400 (emit_final_str_compare_gpr): New function.
3401 (expand_strn_compare): Refactor and clean up code.
3402 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
3404 2018-06-19 Tony Reix <tony.reix@atos.com>
3405 Damien Bergamini <damien.bergamini@atos.com>
3406 David Edelsohn <dje.gcc@gmail.com>
3408 * collect2.c (static_obj): New variable.
3409 (static_libs): New variable.
3410 (is_in_list): Uncomment declaration.
3411 (main): Track AIX libraries linked statically.
3412 (is_in_list): Uncomment definition.
3413 (scan_prog_file): Don't add AIX shared libraries initializer
3414 to constructor list if linking statically.
3416 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3418 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3420 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3422 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
3424 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
3427 2018-06-19 Martin Liska <mliska@suse.cz>
3429 * config/i386/i386.c (ix86_can_inline_p): Do not use
3430 ipa_fn_summaries::get_create.
3431 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3433 (devirtualization_time_bonus): Likewise.
3434 (ipcp_propagate_stage): Likewise.
3435 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3436 (edge_set_predicate): Likewise.
3437 (evaluate_conditions_for_known_args): Likewise.
3438 (evaluate_properties_for_edge): Likewise.
3439 (ipa_call_summary::reset): Tranform to ...
3440 (ipa_call_summary::~ipa_call_summary): ... this.
3441 (ipa_fn_summary::reset): Transform to ...
3442 (ipa_fn_summary::~ipa_fn_summary): ... this.
3443 (ipa_fn_summary_t::remove): Rename to ...
3444 (ipa_fn_summary_t::remove_callees): ... this.
3445 (ipa_fn_summary_t::duplicate): Use placement new
3446 instead of memory copy.
3447 (ipa_call_summary_t::duplicate): Likewise.
3448 (ipa_call_summary_t::remove): Remove.
3449 (dump_ipa_call_summary): Change get_create to get.
3450 (ipa_dump_fn_summary): Dump only when summary exists.
3451 (analyze_function_body): Use symbol_summary::get instead
3453 (compute_fn_summary): Likewise.
3454 (estimate_edge_devirt_benefit): Likewise.
3455 (estimate_edge_size_and_time): Likewise.
3456 (inline_update_callee_summaries): Likewise.
3457 (remap_edge_change_prob): Likewise.
3458 (remap_edge_summaries): Likewise.
3459 (ipa_merge_fn_summary_after_inlining): Likewise.
3460 (write_ipa_call_summary): Likewise.
3461 (ipa_fn_summary_write): Likewise.
3462 (ipa_free_fn_summary): Likewise.
3463 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3464 (struct ipa_call_summary): Likewise.
3465 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3467 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3468 (estimate_size_after_inlining): Likewise.
3469 (estimate_growth): Likewise.
3470 (growth_likely_positive): Likewise.
3471 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3472 (inline_call): Likewise.
3473 * ipa-inline.c (caller_growth_limits): Likewise.
3474 (can_inline_edge_p): Likewise.
3475 (can_inline_edge_by_limits_p): Likewise.
3476 (compute_uninlined_call_time): Likewise.
3477 (compute_inlined_call_time): Likewise.
3478 (want_inline_small_function_p): Likewise.
3479 (edge_badness): Likewise.
3480 (update_caller_keys): Likewise.
3481 (update_callee_keys): Likewise.
3482 (inline_small_functions): Likewise.
3483 (inline_to_all_callers_1): Likewise.
3484 (dump_overall_stats): Likewise.
3485 (early_inline_small_functions): Likewise.
3486 (early_inliner): Likewise.
3487 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3488 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3489 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3490 * ipa-split.c (execute_split_functions): Likewise.
3491 * symbol-summary.h: Likewise.
3492 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3494 2018-06-19 Richard Biener <rguenther@suse.de>
3496 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3497 (vectorize_loops): ... here. Fix dbgcnt handling.
3498 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3500 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3503 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3504 ieee128 argument takes up only one (vector) register, not two (floating
3507 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3509 * gimplify.c (gimplify_init_constructor): Really never clear for an
3510 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3512 2018-06-19 Richard Biener <rguenther@suse.de>
3514 PR tree-optimization/86179
3515 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3516 after failed recognition.
3518 2018-06-18 Martin Sebor <msebor@redhat.com>
3521 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3522 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3523 Handle integer subtraction.
3524 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3525 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3527 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3529 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3530 param from rtx to rtx_insn *.
3531 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3533 (frv_ifcvt_modify_insn): Likwise.
3534 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3535 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3536 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3537 as_a <rtx_insn *> cast to local "unprotected_region" once
3538 it's been established that it's not NULL or pc_rtx.
3539 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3540 param "sethi" from rtx to rtx_insn *.
3541 (nds32_group_float_insns): Likewise for param "insn".
3542 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3544 (vax_output_int_subtract): Likewise.
3545 * config/vax/vax.c (vax_output_int_add): Likewise for param
3547 (vax_output_int_subtract): Likewise.
3548 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3549 (emit_pattern_after): Likewise for param "after".
3550 (emit_insn_after): Likewise.
3551 (emit_jump_insn_after): Likewise.
3552 (emit_call_insn_after): Likewise.
3553 (emit_debug_insn_after): Likewise.
3554 (emit_pattern_before): Likewise for param "before".
3555 (emit_insn_before): Likewise.
3556 (emit_jump_insn_before): Likewise.
3557 * final.c (get_insn_template): Likewise for param "insn", removing
3559 * output.h (get_insn_template): Likewise for 2nd param.
3560 * rtl.h (emit_insn_before): Likewise.
3561 (emit_jump_insn_before): Likewise.
3562 (emit_debug_insn_before_noloc): Likewise.
3563 (emit_insn_after): Likewise.
3564 (emit_jump_insn_after): Likewise.
3565 (emit_call_insn_after): Likewise.
3566 (emit_debug_insn_after): Likewise.
3567 (set_insn_deleted): Likewise for param.
3569 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3572 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3573 floating point modes, so that IFmode is numerically greater than
3574 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3575 to declare the ordering. This prevents IFmode from being
3576 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3577 machine. Include rs6000-modes.h to share the fractional values
3578 between genmodes* and the rest of the compiler.
3582 * config/rs6000/rs6000-modes.h: New file.
3583 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3584 meaning of rs6000_long_double_size so that 126..128 selects an
3585 appropriate 128-bit floating point type.
3586 (rs6000_option_override_internal): Likewise.
3587 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3588 (TARGET_LONG_DOUBLE_128): Change the meaning of
3589 rs6000_long_double_size so that 126..128 selects an appropriate
3590 128-bit floating point type.
3591 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3592 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3593 source and destination to match the standard usage.
3594 (truncifkf2): Likewise.
3595 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3596 ISA 2.07 to use an explicit clobber, instead of passing in a
3598 (copysign<mode>3_soft): Likewise.
3600 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3602 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3603 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3604 (vect_slp_analyze_instance_dependence): Likewise.
3605 (vect_enhance_data_refs_alignment): Likewise.
3606 (vect_analyze_data_refs_alignment): Likewise.
3607 (vect_slp_analyze_and_verify_instance_alignment
3608 (vect_analyze_data_ref_accesses): Likewise.
3609 (vect_prune_runtime_alias_test_list): Likewise.
3610 (vect_analyze_data_refs): Likewise.
3611 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3612 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3613 (vect_analyze_scalar_cycles_1): Likewise.
3614 (vect_get_loop_niters): Likewise.
3615 (vect_analyze_loop_form_1): Likewise.
3616 (vect_update_vf_for_slp): Likewise.
3617 (vect_analyze_loop_operations): Likewise.
3618 (vect_analyze_loop): Likewise.
3619 (vectorizable_induction): Likewise.
3620 (vect_transform_loop): Likewise.
3621 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3622 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3623 (vect_make_slp_decision): Likewise.
3624 (vect_detect_hybrid_slp): Likewise.
3625 (vect_slp_analyze_operations): Likewise.
3626 (vect_slp_bb): Likewise.
3627 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3628 (vectorizable_bswap): Likewise.
3629 (vectorizable_call): Likewise.
3630 (vectorizable_simd_clone_call): Likewise.
3631 (vectorizable_conversion): Likewise.
3632 (vectorizable_assignment): Likewise.
3633 (vectorizable_shift): Likewise.
3634 (vectorizable_operation): Likewise.
3635 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3637 2018-06-18 Martin Sebor <msebor@redhat.com>
3639 PR tree-optimization/81384
3640 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3641 * builtins.c (expand_builtin_strnlen): New function.
3642 (expand_builtin): Call it.
3643 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3644 * builtins.def (BUILT_IN_STRNLEN): New.
3645 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3646 Warn for bounds in excess of maximum object size.
3647 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3648 single-value ranges. Handle strnlen.
3649 (handle_builtin_strlen): Handle strnlen.
3650 (strlen_check_and_optimize_stmt): Same.
3651 * doc/extend.texi (Other Builtins): Document strnlen.
3653 2018-06-18 Maya Rashish <coypu@sdf.org>
3655 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3656 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3657 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3659 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3661 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3663 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3665 * tree.c (escaped_string::escape): Replace cast to char * by
3666 const_cast<char *> (unescaped).
3668 2018-06-18 Nick Clifton <nickc@redhat.com>
3671 * tree.c (escaped_string): New class. Converts an unescaped
3672 string into its escaped equivalent.
3673 (warn_deprecated_use): Use the new class to convert the
3674 deprecation message, if present.
3675 (test_escaped_strings): New self test.
3676 (test_c_tests): Add test_escaped_strings.
3677 * doc/extend.texi (deprecated): Add a note that the
3678 deprecation message is affected by the -fmessage-length
3679 option, and that control characters will be escaped.
3680 (#pragma GCC error): Document this pragma.
3681 (#pragma GCC warning): Likewise.
3682 * doc/invoke.texi (-fmessage-length): Document this option's
3683 effect on the #warning and #error preprocessor directives and
3684 the deprecated attribute.
3686 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3688 * tree.c (decl_value_expr_lookup): Revert latest change.
3689 (decl_value_expr_insert): Likewise.
3691 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3693 * gimplify.c (nonlocal_vlas): Delete.
3694 (nonlocal_vla_vars): Likewise.
3695 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3697 (gimplify_body): Do not create and destroy nonlocal_vlas.
3698 * tree-nested.c: Include diagnostic.h.
3699 (use_pointer_in_frame): Tweak.
3700 (lookup_field_for_decl): Add assertion and declare the transformation.
3701 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3702 internal error when the reference is in a wrong context. Do not
3703 create a debug decl by default.
3704 (note_nonlocal_block_vlas): Delete.
3705 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3706 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3707 create a debug decl by default.
3708 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3709 call is in a wrong context.
3710 (fixup_vla_decls): New function.
3711 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3712 debug variables were created.
3713 * tree.c (decl_value_expr_lookup): Add checking assertion.
3714 (decl_value_expr_insert): Likewise.
3716 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3719 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3720 * tree-scalar-evolution.c (interpret_expr): Likewise.
3721 (expression_expensive_p): Likewise.
3722 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3723 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3724 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3725 (ssa_defined_by_minus_one_stmt_p): New.
3727 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3730 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3731 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3732 * dojump.c (do_jump): Likewise.
3733 * expr.c (expand_expr_real_2): Check operand type's sign.
3734 * fold-const.c (const_unop): Handle ABSU_EXPR.
3735 (fold_abs_const): Likewise.
3736 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3737 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3738 (strip_sign_op_1): Likesise.
3739 * match.pd: Add new pattern to generate ABSU_EXPR.
3740 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3741 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3742 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3743 * tree-inline.c (estimate_operator_cost): Likewise.
3744 * tree-pretty-print.c (dump_generic_node): Likewise.
3745 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3746 * tree.def (ABSU_EXPR): New.
3748 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3751 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3752 documented as preserved for backward compatibility only.
3753 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3755 PR rtl-optimization/86108
3756 * bb-reorder.c (create_forwarder_block): Renamed to ...
3757 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3758 jump from new landing pad to the second part.
3759 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3762 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3765 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3766 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3767 Only call store_expr for halves if the mode is the same.
3770 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3771 Fix up comment formatting.
3773 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3775 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3776 (typed_splay_tree::closure,
3777 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3778 (typed_splay_tree::typed_splay_tree,
3779 typed_splay_tree::operator =): Declared private.
3780 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3781 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3782 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3783 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3784 typed_splay_tree::splay_tree_splay,
3785 typed_splay_tree::splay_tree_foreach_helper,
3786 typed_splay_tree::splay_tree_insert,
3787 typed_splay_tree::splay_tree_remove,
3788 typed_splay_tree::splay_tree_lookup,
3789 typed_splay_tree::splay_tree_predecessor,
3790 typed_splay_tree::splay_tree_successor,
3791 typed_splay_tree::splay_tree_min,
3792 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3793 (typed_splay_tree::root, typed_splay_tree::comp,
3794 typed_splay_tree::delete_key,
3795 typed_splay_tree::delete_value): New data members.
3796 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3797 typed_splay_tree::remove.
3799 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3801 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3802 -mginv and -mno-ginv to the assembler.
3803 * config/mips/mips.opt (-mcrc): New option.
3805 * doc/invoke.text (-mcrc): Document.
3808 2018-06-15 Nick Clifton <nickc@redhat.com>
3811 * tree.c (escaped_string): New class. Converts an unescaped
3812 string into its escaped equivalent.
3813 (warn_deprecated_use): Use the new class to convert the
3814 deprecation message, if present.
3815 (test_escaped_strings): New self test.
3816 (test_c_tests): Add test_escaped_strings.
3817 * doc/extend.texi (deprecated): Add a note that the
3818 deprecation message is affected by the -fmessage-length
3819 option, and that control characters will be escaped.
3820 (#pragma GCC error): Document this pragma.
3821 (#pragma GCC warning): Likewise.
3822 * doc/invoke.texi (-fmessage-length): Document this option's
3823 effect on the #warning and #error preprocessor directives and
3824 the deprecated attribute.
3826 2018-06-15 Richard Biener <rguenther@suse.de>
3828 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3829 here, also noting vector size used.
3830 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3831 size used in MSG_OPTIMIZED_LOCATIONS dump.
3832 (pass_slp_vectorize::execute): Adjust.
3834 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3837 * config/arc/arc.c (arc_return_address_register): Fix
3840 2018-06-15 Richard Biener <rguenther@suse.de>
3843 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3844 leave useless conversion stripping to force_gimple_operand_gsi.
3845 (gimplify_build2): Likewise.
3846 (gimplify_build1): Likewise.
3848 2018-06-15 Richard Biener <rguenther@suse.de>
3851 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3852 before adjusting their block.
3854 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3856 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3857 multilibs for *-*-rtems*.
3858 * config/riscv/t-rtems: New file.
3860 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3863 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3864 unsigned_type_for returns NULL.
3867 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3868 subregs of multi-word pseudos unless the float mode has word size.
3870 2018-06-14 Richard Biener <rguenther@suse.de>
3873 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3874 duplicate and harmful type_hash_canon.
3875 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3877 2018-06-14 Richard Biener <rguenther@suse.de>
3880 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3883 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3885 * config/rtems.h (STDINT_LONG32): Define.
3887 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3888 Prachi Godbole <prachi.godbole@imgtec.com>
3890 * config/mips/mips-cpus.def: Define P6600.
3891 * config/mips/mips-tables.opt: Regenerate.
3892 * config/mips/mips.c (mips_ucbranch_type): New enum.
3893 (mips_rtx_cost_data): Add support for P6600.
3894 (mips_issue_rate): Likewise.
3895 (mips_multipass_dfa_lookahead): Likewise.
3896 (mips_avoid_hazard): Likewise.
3897 (mips_reorg_process_insns): Likewise.
3898 (mips_classify_branch_p6600): New function.
3899 * config/mips/mips.h (TUNE_P6600): New define.
3900 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3901 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3902 * config/mips/mips.md: Include p6600.md.
3903 (processor): Add p6600.
3904 * config/mips/p6600.md: New file.
3905 * doc/invoke.texi: Add p6600 to supported architectures.
3907 2018-06-13 Martin Sebor <msebor@redhat.com>
3909 PR tree-optimization/86114
3910 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3912 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3914 2018-06-13 Richard Biener <rguenther@suse.de>
3916 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3917 Properly set vector type of the intermediate stmt.
3918 * tree-vect-stmts.c (vectorizable_operation): The destination
3919 var always has vectype_out type.
3921 2018-06-13 Jeff Law <law@redhat.com>
3923 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3924 integer 0 for argument to print_rtl_with_bb.
3925 (rl78_reorg): Likewise.
3927 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3929 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3930 from rtx to rtx_insn *.
3931 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3933 (add_sched_insns_for_speculation): Likewise for local "target",
3934 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3935 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3936 from rtx_insn ** to rtx_code_label **.
3937 (reorg_emit_nops): Likewise.
3938 (c6x_reorg): Likewise for local "call_labels".
3939 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3941 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3942 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3943 the loops over LABEL_REFS.
3944 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3946 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3947 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3948 (split_branches): Strengthen local "olabel" from rtx to
3949 rtx_insn *, adding a safe_as_a cast.
3950 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3952 (add_insn_after): Likewise for first two params.
3953 (add_insn_before): Likewise.
3954 (remove_insn): Likewise for param.
3955 (emit_pattern_before_noloc): Likewise for second and third params.
3956 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3957 (emit_call_insn_before_noloc): Likewise.
3958 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3960 (emit_barrier_before): Likewise.
3961 (emit_label_before): Strengthen "label" param from "rtx" to
3962 "rtx_code_label *". Strengthen "before" param from "rtx" to
3964 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3966 (emit_pattern_after_noloc): Likewise.
3967 (emit_insn_after_noloc): Likewise.
3968 (emit_jump_insn_after_noloc): Likewise.
3969 (emit_call_insn_after_noloc): Likewise.
3970 (emit_debug_insn_after_noloc): Likewise.
3971 (emit_barrier_after): Likewise.
3972 (emit_label_after): Likewise for both params.
3973 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3974 "loc" param from "int" to "location_t".
3975 (emit_insn_after_setloc): Likewise.
3976 (emit_jump_insn_after_setloc): Likewise.
3977 (emit_call_insn_after_setloc): Likewise.
3978 (emit_debug_insn_after_setloc): Likewise.
3979 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3980 "loc" param from "int" to "location_t".
3981 (emit_pattern_before): Convert NULL_RTX to NULL.
3982 (emit_insn_before_setloc): Convert "loc" param from "int" to
3984 (emit_jump_insn_before_setloc): Likewise.
3985 (emit_call_insn_before_setloc): Likewise.
3986 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3987 rtx_insn *. Convert "loc" param from "int" to "location_t".
3988 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3989 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3990 Convert 3rd param from "int" to "location_t".
3991 (emit_barrier_before, emit_barrier_after, next_real_insn):
3992 Strengthen param from rtx to rtx_insn *.
3993 (emit_label_before): Strengthen 1st param from "rtx" to
3994 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3996 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3997 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3998 Strengthen 2nd param from "rtx" to "rtx_insn *".
3999 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4000 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4001 Likewise. Convert 3rd param from "int" to "location_t".
4002 (emit_label_after): Strengthen 1st param from "rtx" to
4004 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4006 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4007 from "rtx" to "rtx_insn *".
4009 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4011 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4012 bodies streamed in with -Q.
4013 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4014 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4015 * gimple-streamer-out.c: Include gimple-pretty-print.h
4016 (output_bb): Dump stmts streamed.
4017 * lto-section-out.c: Include print-tree.h
4018 (lto_begin_section): Dump sections created.
4019 (lto_output_decl_index): Dump decl encoded.
4020 * lto-streamer-out.c: Include print-tree.h
4021 (create_output_block): Dump output block created.
4022 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4023 (output_function): Dump function output.
4024 (output_constructor): Dump constructor streamed.
4025 (write_global_stream): Output indexes encoded.
4026 (produce_asm_for_decls): Dump streams encoded.
4027 * lto-streamer.c (streamer_dump_file): New global var.
4028 * lto-streamer.h (streamer_dump_file): Declare.
4029 * passes.c (ipa_write_summaries): Initialize streamer dump.
4030 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4033 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4036 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4037 offsets for register save directives. Emit a second batch of save
4038 directives, if need be, when the function accesses prior frames.
4040 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4042 * config/arc/fpu.md (fmasf4): Force operand to register.
4043 (fnmasf4): Likewise.
4045 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4047 * config/arc/arc-protos.h (arc_pad_return): Remove.
4048 * config/arc/arc.c (machine_function): Remove force_short_suffix
4050 (arc_print_operand): Adjust printing of '&'.
4051 (arc_verify_short): Remove conditional printing of short suffix.
4052 (arc_final_prescan_insn): Remove reference to size_reason.
4053 (pad_return): New function.
4054 (arc_reorg): Call pad_return.
4055 (arc_pad_return): Remove.
4056 (arc_init_machine_status): Remove reference to force_short_suffix.
4057 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4058 (attr length): When attribute iscompact is true force to 2
4059 regardless; in the case of maybe check if we want to force the
4060 instruction to have 4 bytes length.
4061 (nopv): Change it to generate 4 byte long nop as well.
4062 (blockage): New pattern.
4063 (simple_return): Remove call to arc_pad_return.
4064 (p_return_i): Likewise.
4066 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4068 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4070 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4072 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4075 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4077 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4078 for ARC700 and ARCv2.
4080 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4083 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4084 operands[2] instead of operands[1].
4087 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4089 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4090 case, check whether the outer register overlaps an unallocatable
4091 register, not just whether it fits the required class.
4093 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4095 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4097 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4098 (memrefs_conflict_p): Likewise.
4099 (init_alias_analysis): Likewise.
4100 * cfgexpand.c (expand_debug_expr): Likewise.
4101 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4102 * cse.c (fold_rtx): Likewise.
4103 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4104 * expr.c (emit_block_move_hints): Likewise.
4105 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4106 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4107 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4108 (emit_group_store): Likewise.
4109 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4110 to read the PRE/POST_MODIFY increment.
4111 * calls.c (store_one_arg): Use strip_offset.
4112 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4114 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4116 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4117 (simplify_binary_operation_1): Extend CONST_INT handling to
4119 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4120 than a HOST_WIDE_INT.
4121 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4123 (adjust_mems, add_stores): Update accodingly.
4124 (vt_canonicalize_addr): Track polynomial offsets.
4125 (emit_note_insn_var_location): Likewise.
4126 (vt_add_function_parameter): Likewise.
4127 (vt_initialize): Likewise.
4129 2018-06-12 Jeff Law <law@redhat.com>
4131 * config.gcc (alpha*-*-freebsd*): Remove.
4132 * config/alpha/freebsd.h: Remove.
4134 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4137 * spellcheck-tree.c (levenshtein_distance): Rename to...
4138 (get_edit_distance): ...this, and update for underlying renaming.
4139 * spellcheck-tree.h (levenshtein_distance): Rename to...
4140 (get_edit_distance): ...this.
4141 * spellcheck.c (levenshtein_distance): Rename to...
4142 (get_edit_distance): ...this. Convert from Levenshtein distance
4143 to Damerau-Levenshtein distance by supporting transpositions of
4144 adjacent characters. Rename "v1" to "v_next" and "v0" to
4146 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4147 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4148 update for underlying renaming.
4149 (selftest::levenshtein_distance_unit_test): Rename to...
4150 (selftest::test_get_edit_distance_unit): ...this, and update for
4151 underlying renaming.
4152 (selftest::test_find_closest_string): Add example from PR 69968
4153 where transposition helps
4154 (selftest::test_metric_conditions): Update for renaming.
4155 (selftest::test_metric_conditions): Likewise.
4156 (selftest::spellcheck_c_tests): Likewise.
4157 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4158 (get_edit_distance): ...this.
4159 (best_match::consider): Update for renaming.
4161 2018-06-12 Martin Sebor <msebor@redhat.com>
4163 PR tree-optimization/85259
4164 * builtins.c (compute_objsize): Handle constant offsets.
4165 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4166 true iff a warning has been issued.
4167 * gimple.h (gimple_nonartificial_location): New function.
4168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4169 gimple_nonartificial_location and handle -Wno-system-headers.
4170 (handle_builtin_stxncpy): Same.
4172 2018-06-12 Martin Sebor <msebor@redhat.com>
4175 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4177 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4179 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4180 BUILTIN_VEC_XST entries for pointer to double and long long.
4182 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4185 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4187 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4190 2018-06-12 Martin Liska <mliska@suse.cz>
4192 * doc/options.texi: Document IntegerRange.
4194 2018-06-12 Martin Liska <mliska@suse.cz>
4196 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4197 * opt-functions.awk: Handle Deprecated flag.
4198 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4200 (read_cmdline_option): Report warning for a deprecated option.
4201 * opts.h (struct cl_option): Add new field cl_deprecated.
4202 (CL_ERR_DEPRECATED): New.
4204 2018-06-12 Martin Liska <mliska@suse.cz>
4206 * doc/options.texi: Document Deprecated option flag.
4208 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4210 * config/arc/arc-arch.h (arc_extras): New enum.
4211 (arc_cpu_t):Add field extra.
4212 (arc_cpu_types): Consider the extras.
4213 * config/arc/arc-cpus.def: Add extras info.
4214 * config/arc/arc-opts.h (processor_type): Consider extra field.
4215 * config/arc/arc.c (arc_override_options): Handle extra field.
4217 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4219 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4220 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4222 * config/arc/arc.c (arc_init): Update pic warning.
4223 (irq_range): Update irq range parsing warnings.
4224 (arc_override_options): Update various warning messages.
4225 (arc_handle_aux_attribute): Likewise.
4227 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4229 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4231 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4233 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4234 compiler messages directives.
4236 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4238 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4239 * config/mips/mips-tables.opt: Regenerate.
4240 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4242 * doc/invoke.texi: Document -march=i6500.
4244 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4246 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4247 (i6400_gpmul): Add cpu_unit.
4248 (i6400_gpdiv): Likewise.
4249 (i6400_msa_add_d): Update reservations.
4250 (i6400_msa_int_add) Likewise.
4251 (i6400_msa_short_logic3) Likewise.
4252 (i6400_msa_short_logic2) Likewise.
4253 (i6400_msa_short_logic) Likewise.
4254 (i6400_msa_move) Likewise.
4255 (i6400_msa_cmp) Likewise.
4256 (i6400_msa_short_float2) Likewise.
4257 (i6400_msa_div_d) Likewise.
4258 (i6400_msa_long_logic1) Likewise.
4259 (i6400_msa_long_logic2) Likewise.
4260 (i6400_msa_mult) Likewise.
4261 (i6400_msa_long_float2) Likewise.
4262 (i6400_msa_long_float4) Likewise.
4263 (i6400_msa_long_float5) Likewise.
4264 (i6400_msa_long_float8) Likewise.
4265 (i6400_fpu_fadd): Include frint type.
4266 (i6400_fpu_store): New define_insn_reservation.
4267 (i6400_fpu_load): Likewise.
4268 (i6400_fpu_move): Likewise.
4269 (i6400_fpu_fcmp): Likewise.
4270 (i6400_fpu_fmadd): Likewise.
4271 (i6400_int_mult): Include imul3nc type and update reservation.
4272 (i6400_int_div): Include idiv3 type and update reservation.
4273 (i6400_int_load): Update to check type not move_type.
4274 (i6400_int_store): Likewise.
4275 (i6400_int_prefetch): Set zero latency.
4277 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4279 * gcc.c: Document new %@{...} sequence.
4280 (LINK_COMMAND_SPEC): Use it for the -L switches.
4281 (cpp_unique_options): Use it for the -I switches.
4282 (at_file_argbuf): New global variable.
4283 (in_at_file): Likewise.
4284 (alloc_args): Create at_file_argbuf.
4285 (clear_args): Truncate at_file_argbuf.
4286 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4287 (open_at_file): New function.
4288 (close_at_file): Likewise.
4289 (create_at_file): Delete.
4290 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4293 (validate_switches_from_spec): Deal with %@{...} sequence.
4294 (validate_switches): Likewise.
4295 (driver::finalize): Call clear_args.
4297 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4299 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
4301 2018-06-11 Martin Sebor <msebor@redhat.com>
4303 * doc/invoke.texi (-Wall): List -Wc++17-compat.
4304 (Wno-class-memaccess): Add @opindex.
4305 (Wno-templates, Wno-multiple-inheritance): Same.
4306 (Wno-virtual-inheritance, Wno-namespaces): Same.
4307 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
4308 (Wno-format-overflow, Wno-format-truncation): Same.
4309 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
4310 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
4311 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
4312 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
4313 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
4314 (Wno-misspelled-isr): Same.
4316 2018-06-11 Martin Sebor <msebor@redhat.com>
4318 * PR tree-optimization/86083
4319 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
4321 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
4323 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
4325 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4328 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4329 on the correct operand.
4330 (*movdi_internal64): Ditto.
4332 2018-06-11 Martin Liska <mliska@suse.cz>
4334 PR tree-optimization/86089
4335 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
4337 2018-06-11 Julia Koval <julia.koval@intel.com>
4339 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
4340 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
4341 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
4343 2018-06-11 Olivier Hainque <hainque@adacore.com>
4345 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
4346 for Ada with strict dwarf2.
4348 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4351 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4354 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
4356 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
4358 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
4360 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
4363 2018-06-08 Martin Liska <mliska@suse.cz>
4365 * tree-cfg.h (debug_function): Fix argument type to match
4368 2018-06-08 Martin Liska <mliska@suse.cz>
4370 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
4371 Remove usage of MPX-related (and removed) fields.
4372 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
4374 2018-06-08 David Malcolm <dmalcolm@redhat.com>
4376 * cfg.c (debug): Use TDF_NONE rather than 0.
4377 * cfghooks.c (debug): Likewise.
4378 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
4379 (struct dump_option_value_info): Convert to...
4380 (struct kv_pair): ...this template type.
4381 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
4383 (optinfo_verbosity_options): Likewise.
4384 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
4386 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
4387 than int for "optgroup_flags" param.
4388 (dump_generic_expr_loc): Use dump_flags_t rather than int for
4390 (dump_dec): Likewise.
4391 (dump_finish): Use TDF_NONE rather than 0.
4392 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
4393 rather than int for "optgroup_flags" param. Use TDF_NONE rather
4394 than 0. Update for change to option_ptr.
4395 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
4396 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
4397 0. Update for changes to optinfo_verbosity_options and
4399 (opt_info_switch_p): Convert optgroup_flags from int to
4401 (dump_basic_block): Use dump_flags_t rather than int
4402 for "dump_kind" param.
4403 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
4404 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
4405 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
4406 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
4407 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
4408 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
4409 TDF_NONE): Convert from macros to...
4410 (enum dump_flag): ...this new enum.
4411 (dump_flags_t): Update to use enum.
4412 (operator|, operator&, operator~, operator|=, operator&=):
4413 Implement for dump_flags_t.
4414 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
4415 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
4416 Convert from macros to...
4417 (enum optgroup_flag): ...this new enum.
4418 (optgroup_flags_t): New typedef.
4419 (operator|, operator|=): Implement for optgroup_flags_t.
4420 (struct dump_file_info): Convert field "alt_flags" to
4421 dump_flags_t. Convert field "optgroup_flags" to
4423 (dump_basic_block): Use dump_flags_t rather than int for param.
4424 (dump_generic_expr_loc): Likewise.
4425 (dump_dec): Likewise.
4426 (dump_register): Convert param "optgroup_flags" to
4428 (opt_info_enable_passes): Likewise.
4429 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4431 * gimple-pretty-print.c (debug): Likewise.
4432 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4433 (merged_store_group::apply_stores): Likewise.
4434 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4435 * gimple.c (verify_gimple_pp): Likewise.
4436 * graphite-poly.c (print_pbb_body): Likewise.
4437 * passes.c (pass_manager::register_one_dump_file): Convert
4438 local "optgroup_flags" to optgroup_flags_t.
4439 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4441 (debug_body): Likewise.
4442 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4443 to optgroup_flags_t.
4444 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4446 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4447 (convert_mult_to_fma): Likewise.
4448 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4449 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4450 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4451 "dump_kind" to dump_flags_t.
4453 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4455 * config/rs6000/rs6000.c (min, max): Delete.
4457 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4459 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4460 -mabi=spe and -mabi=no-spe.
4462 2018-06-08 Martin Liska <mliska@suse.cz>
4464 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4465 where we expect an existing summary.
4467 2018-06-08 Martin Liska <mliska@suse.cz>
4469 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4470 * ipa-inline.h (estimate_edge_growth): Likewise.
4472 2018-06-08 Martin Liska <mliska@suse.cz>
4474 * cgraph.c (function_version_hasher::hash): Use
4475 cgraph_node::get_uid ().
4476 (function_version_hasher::equal):
4477 * cgraph.h (cgraph_node::get_uid): New method.
4478 * ipa-inline.c (update_caller_keys): Use
4479 cgraph_node::get_uid ().
4480 (update_callee_keys): Likewise.
4481 * ipa-utils.c (searchc): Likewise.
4482 (ipa_reduced_postorder): Likewise.
4483 * lto-cgraph.c (input_node): Likewise.
4484 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4485 * symbol-summary.h (symtab_insertion): Likewise.
4486 (symtab_removal): Likewise.
4487 (symtab_duplication): Likewise.
4488 * tree-pretty-print.c (dump_function_header): Likewise.
4489 * tree-sra.c (convert_callers_for_node): Likewise.
4491 2018-06-08 Martin Liska <mliska@suse.cz>
4493 * cgraph.c (symbol_table::create_edge): Always assign a new
4495 (symbol_table::free_edge): Do not recycle numbers.
4496 * cgraph.h (cgraph_edge::get): New method.
4497 * symbol-summary.h (symtab_removal): Use it.
4498 (symtab_duplication): Likewise.
4499 (call_summary::hashable_uid): Remove.
4501 2018-06-08 Martin Liska <mliska@suse.cz>
4503 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4504 (initialize_growth_caches): Remove.
4505 (free_growth_caches): Likewise.
4506 (do_estimate_edge_time): Use edge_growth_cache.
4507 (do_estimate_edge_size): Likewise.
4508 (do_estimate_edge_hints): Likewise.
4509 * ipa-inline.c (reset_edge_caches): Likewise.
4510 (recursive_inlining): Likewise.
4511 (inline_small_functions): Likewise.
4512 * ipa-inline.h (initialize_growth_caches): Remove.
4513 (estimate_edge_size): Likewise.
4514 (estimate_edge_time): Likewise.
4515 (estimate_edge_hints): Likewise.
4516 (reset_edge_growth_cache): Likewise.
4517 * symbol-summary.h (call_summary::remove): New method.
4519 2018-06-08 Martin Liska <mliska@suse.cz>
4521 * ipa-cp.c (class edge_clone_summary): New summary.
4522 (grow_edge_clone_vectors): Remove.
4523 (ipcp_edge_duplication_hook): Remove.
4524 (class edge_clone_summary_t): New call_summary class.
4525 (ipcp_edge_removal_hook): Remove.
4526 (edge_clone_summary_t::duplicate): New function.
4527 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4528 (create_specialized_node): Likewise.
4529 (ipcp_driver): Initialize edge_clone_summaries and do not
4532 2018-06-08 Martin Liska <mliska@suse.cz>
4534 * symbol-summary.h (get): New function.
4535 (call_summary::m_initialize_when_cloning): New class member.
4537 2018-06-08 Martin Liska <mliska@suse.cz>
4539 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4540 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4541 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4542 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4543 (remove_cgraph_node_from_order): Use the removed_nodes set.
4544 (do_per_function_toporder): Likwise.
4545 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4546 instead of summary_uid.
4547 (symtab_removal): Likewise.
4548 (symtab_duplication): Likewise.
4550 2018-06-08 Martin Liska <mliska@suse.cz>
4552 * ipa-cp.c (ipcp_store_bits_results): Use
4553 ipcp_transformation_sum.
4554 (ipcp_store_vr_results): Likewise.
4555 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4557 (ipcp_transformation_initialize): ... this.
4558 (ipa_set_node_agg_value_chain):
4559 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4560 (write_ipcp_transformation_info): Likewise.
4561 (read_ipcp_transformation_info): Likewise.
4562 (ipcp_update_bits): Likewise.
4563 (ipcp_update_vr): Likewise.
4564 (ipcp_transform_function): Likewise.
4565 * ipa-prop.h: Rename ipcp_transformation_summary to
4566 ipcp_transformation.
4567 (class ipcp_transformation_t): New function summary.
4568 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4569 (ipa_get_agg_replacements_for_node): Likewise.
4571 2018-06-08 Martin Liska <mliska@suse.cz>
4573 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4575 (class funct_state_summary_t): New function_summary class.
4576 (has_function_state): Remove.
4577 (get_function_state): Likewise.
4578 (set_function_state): Likewise.
4579 (add_new_function): Likewise.
4580 (funct_state_summary_t::insert): New function.
4581 (duplicate_node_data): Remove.
4582 (remove_node_data): Remove.
4583 (funct_state_summary_t::duplicate): New function.
4584 (register_hooks): Create new funct_state_summaries.
4585 (pure_const_generate_summary): Use it.
4586 (pure_const_write_summary): Likewise.
4587 (pure_const_read_summary): Likewise.
4588 (propagate_pure_const): Likewise.
4589 (propagate_nothrow): Likewise.
4590 (dump_malloc_lattice): Likewise.
4591 (propagate_malloc): Likewise.
4592 (execute): Do not register hooks, just remove summary
4594 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4597 2018-06-08 Martin Liska <mliska@suse.cz>
4599 * ipa-reference.c (remove_node_data): Remove.
4600 (duplicate_node_data): Likewise.
4601 (class ipa_ref_var_info_summary_t): New class.
4602 (class ipa_ref_opt_summary_t): Likewise.
4603 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4604 (get_reference_optimization_summary): Use
4605 ipa_ref_opt_sum_summaries.
4606 (set_reference_vars_info): Remove.
4607 (set_reference_optimization_summary): Likewise.
4608 (ipa_init): Create summaries.
4609 (init_function_info): Use function summary.
4610 (ipa_ref_opt_summary_t::duplicate): New function.
4611 (ipa_ref_opt_summary_t::remove): New function.
4612 (get_read_write_all_from_node): Fix GNU coding style.
4613 (propagate): Use function summary.
4614 (write_node_summary_p): Fix GNU coding style.
4615 (stream_out_bitmap): Likewise.
4616 (ipa_reference_read_optimization_summary): Use function summary.
4617 (ipa_reference_c_finalize): Do not release hooks.
4619 2018-06-08 Martin Liska <mliska@suse.cz>
4621 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4622 (analyze_function_body): Extract multiple calls of get_create.
4623 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4624 * ipa-inline.c (recursive_inlining): Use ::get method.
4625 * ipa-inline.h (estimate_edge_growth): Likewise.
4627 2018-06-08 Martin Liska <mliska@suse.cz>
4629 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4631 (hsa_function_summary::hsa_function_summary): Use the new enum
4633 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4634 * hsa-gen.c (hsa_get_host_function): Likewise.
4635 (get_brig_function_name): Likewise.
4636 * ipa-hsa.c (process_hsa_functions): Likewise.
4637 (ipa_hsa_write_summary): Likewise.
4638 * symbol-summary.h (symtab_duplication): Use ::get function/
4639 (get): New function.
4641 2018-06-08 Martin Liska <mliska@suse.cz>
4643 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4645 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4646 (hsa_register_kernel): Likewise.
4647 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4648 * hsa-gen.c (hsa_get_host_function): Likewise.
4649 (get_brig_function_name): Likewise.
4650 (generate_hsa): Likewise.
4651 (pass_gen_hsail::execute): Likewise.
4652 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4653 (devirtualization_time_bonus): Likewise.
4654 (ipcp_propagate_stage): Likewise.
4655 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4656 (edge_set_predicate): Likewise.
4657 (evaluate_conditions_for_known_args): Likewise.
4658 (evaluate_properties_for_edge): Likewise.
4659 (ipa_fn_summary::reset): Likewise.
4660 (ipa_fn_summary_t::duplicate): Likewise.
4661 (dump_ipa_call_summary): Likewise.
4662 (ipa_dump_fn_summary): Likewise.
4663 (analyze_function_body): Likewise.
4664 (compute_fn_summary): Likewise.
4665 (estimate_edge_devirt_benefit): Likewise.
4666 (estimate_edge_size_and_time): Likewise.
4667 (estimate_calls_size_and_time): Likewise.
4668 (estimate_node_size_and_time): Likewise.
4669 (inline_update_callee_summaries): Likewise.
4670 (remap_edge_change_prob): Likewise.
4671 (remap_edge_summaries): Likewise.
4672 (ipa_merge_fn_summary_after_inlining): Likewise.
4673 (ipa_update_overall_fn_summary): Likewise.
4674 (read_ipa_call_summary): Likewise.
4675 (inline_read_section): Likewise.
4676 (write_ipa_call_summary): Likewise.
4677 (ipa_fn_summary_write): Likewise.
4678 (ipa_free_fn_summary): Likewise.
4679 * ipa-hsa.c (process_hsa_functions): Likewise.
4680 (ipa_hsa_write_summary): Likewise.
4681 (ipa_hsa_read_section): Likewise.
4682 * ipa-icf.c (sem_function::merge): Likewise.
4683 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4684 (do_estimate_edge_time): Likewise.
4685 (estimate_size_after_inlining): Likewise.
4686 (estimate_growth): Likewise.
4687 (growth_likely_positive): Likewise.
4688 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4689 (inline_call): Likewise.
4690 * ipa-inline.c (caller_growth_limits): Likewise.
4691 (can_inline_edge_p): Likewise.
4692 (can_inline_edge_by_limits_p): Likewise.
4693 (compute_uninlined_call_time): Likewise.
4694 (compute_inlined_call_time): Likewise.
4695 (want_inline_small_function_p): Likewise.
4696 (edge_badness): Likewise.
4697 (update_caller_keys): Likewise.
4698 (update_callee_keys): Likewise.
4699 (recursive_inlining): Likewise.
4700 (inline_small_functions): Likewise.
4701 (inline_to_all_callers_1): Likewise.
4702 (dump_overall_stats): Likewise.
4703 (early_inline_small_functions): Likewise.
4704 (early_inliner): Likewise.
4705 * ipa-inline.h (estimate_edge_growth): Likewise.
4706 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4707 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4708 * ipa-prop.h (IPA_NODE_REF): Likewise.
4709 (IPA_EDGE_REF): Likewise.
4710 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4711 (propagate_malloc): Likewise.
4712 * ipa-split.c (execute_split_functions): Likewise.
4713 * symbol-summary.h: Rename get to get_create.
4715 (get_create): Likewise.
4716 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4718 2018-06-08 Martin Liska <mliska@suse.cz>
4720 * symbol-summary.h (release): Move definition out of class
4722 (symtab_removal): Likewise.
4723 (symtab_duplication): Likewise.
4725 2018-06-08 Martin Liska <mliska@suse.cz>
4727 * symbol-summary.h (function_summary): Move constructor
4728 implementation out of class declaration.
4729 (release): Likewise.
4730 (symtab_insertion): Likewise.
4731 (symtab_removal): Likewise.
4732 (symtab_duplication): Likewise.
4735 2018-06-08 Martin Liska <mliska@suse.cz>
4737 * Makefile.in: Remove support for MPX (macros, related functions,
4738 fields in cgraph_node, ...).
4739 * builtin-types.def (BT_BND): Likewise.
4740 (BT_FN_BND_CONST_PTR): Likewise.
4741 (BT_FN_CONST_PTR_BND): Likewise.
4742 (BT_FN_VOID_PTR_BND): Likewise.
4743 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4744 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4745 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4746 (expand_builtin_mempcpy_with_bounds): Likewise.
4747 (expand_builtin_memset_with_bounds): Likewise.
4748 (expand_builtin_memset_args): Likewise.
4749 (std_expand_builtin_va_start): Likewise.
4750 (expand_builtin): Likewise.
4751 (expand_builtin_with_bounds): Likewise.
4752 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4753 (DEF_LIB_BUILTIN_CHKP): Likewise.
4754 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4755 (DEF_CHKP_BUILTIN): Likewise.
4756 (BUILT_IN_MEMCPY): Likewise.
4757 (BUILT_IN_MEMMOVE): Likewise.
4758 (BUILT_IN_MEMPCPY): Likewise.
4759 (BUILT_IN_MEMSET): Likewise.
4760 (BUILT_IN_STPCPY): Likewise.
4761 (BUILT_IN_STRCAT): Likewise.
4762 (BUILT_IN_STRCHR): Likewise.
4763 (BUILT_IN_STRCPY): Likewise.
4764 (BUILT_IN_STRLEN): Likewise.
4765 (BUILT_IN_MEMCPY_CHK): Likewise.
4766 (BUILT_IN_MEMMOVE_CHK): Likewise.
4767 (BUILT_IN_MEMPCPY_CHK): Likewise.
4768 (BUILT_IN_MEMSET_CHK): Likewise.
4769 (BUILT_IN_STPCPY_CHK): Likewise.
4770 (BUILT_IN_STRCAT_CHK): Likewise.
4771 (BUILT_IN_STRCPY_CHK): Likewise.
4772 * calls.c (store_bounds): Likewise.
4773 (emit_call_1): Likewise.
4774 (special_function_p): Likewise.
4775 (maybe_warn_nonstring_arg): Likewise.
4776 (initialize_argument_information): Likewise.
4777 (finalize_must_preallocate): Likewise.
4778 (compute_argument_addresses): Likewise.
4779 (expand_call): Likewise.
4780 * cfgexpand.c (expand_call_stmt): Likewise.
4781 (expand_return): Likewise.
4782 (expand_gimple_stmt_1): Likewise.
4783 (pass_expand::execute): Likewise.
4784 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4785 (cgraph_node::remove): Likewise.
4786 (cgraph_node::dump): Likewise.
4787 (cgraph_node::verify_node): Likewise.
4788 * cgraph.h (chkp_function_instrumented_p): Likewise.
4789 (symtab_node::get_alias_target): Likewise.
4790 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4791 (cgraph_local_p): Likewise.
4792 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4793 (cgraph_edge::rebuild_references): Likewise.
4794 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4795 (walk_polymorphic_call_targets): Likewise.
4796 (cgraph_node::expand_thunk): Likewise.
4797 (symbol_table::output_weakrefs): Likewise.
4798 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4799 (ix86_handle_option): Likewise.
4800 * config/i386/constraints.md: Likewise.
4801 * config/i386/i386-builtin-types.def (BND): Likewise.
4805 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4806 (BDESC_FIRST): Likewise.
4808 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4809 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4810 * config/i386/i386.c (enum reg_class): Likewise.
4811 (ix86_target_string): Likewise.
4812 (ix86_option_override_internal): Likewise.
4813 (ix86_conditional_register_usage): Likewise.
4814 (ix86_valid_target_attribute_inner_p): Likewise.
4815 (ix86_set_indirect_branch_type): Likewise.
4816 (ix86_set_current_function): Likewise.
4817 (ix86_function_arg_regno_p): Likewise.
4818 (init_cumulative_args): Likewise.
4819 (ix86_function_arg_advance): Likewise.
4820 (ix86_function_arg): Likewise.
4821 (ix86_pass_by_reference): Likewise.
4822 (ix86_function_value_regno_p): Likewise.
4823 (ix86_function_value_1): Likewise.
4824 (ix86_function_value_bounds): Likewise.
4825 (ix86_return_in_memory): Likewise.
4826 (ix86_setup_incoming_vararg_bounds): Likewise.
4827 (ix86_va_start): Likewise.
4828 (indirect_thunk_need_prefix): Likewise.
4829 (print_reg): Likewise.
4830 (ix86_print_operand): Likewise.
4831 (ix86_expand_call): Likewise.
4832 (ix86_output_function_return): Likewise.
4833 (reg_encoded_number): Likewise.
4834 (BDESC_VERIFYS): Likewise.
4835 (ix86_init_mpx_builtins): Likewise.
4836 (ix86_init_builtins): Likewise.
4837 (ix86_emit_cmove): Likewise.
4838 (ix86_emit_move_max): Likewise.
4839 (ix86_expand_builtin): Likewise.
4840 (ix86_builtin_mpx_function): Likewise.
4841 (ix86_get_arg_address_for_bt): Likewise.
4842 (ix86_load_bounds): Likewise.
4843 (ix86_store_bounds): Likewise.
4844 (ix86_load_returned_bounds): Likewise.
4845 (ix86_store_returned_bounds): Likewise.
4846 (ix86_class_likely_spilled_p): Likewise.
4847 (ix86_hard_regno_mode_ok): Likewise.
4848 (x86_order_regs_for_local_alloc): Likewise.
4849 (ix86_mitigate_rop): Likewise.
4850 (ix86_bnd_prefixed_insn_p): Likewise.
4851 (ix86_mpx_bound_mode): Likewise.
4852 (ix86_make_bounds_constant): Likewise.
4853 (ix86_initialize_bounds): Likewise.
4854 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4855 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4856 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4857 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4858 (TARGET_CHKP_BOUND_MODE): Likewise.
4859 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4860 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4861 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4862 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4863 * config/i386/i386.h (TARGET_MPX): Likewise.
4864 (TARGET_MPX_P): Likewise.
4865 (VALID_BND_REG_MODE): Likewise.
4866 (FIRST_BND_REG): Likewise.
4867 (LAST_BND_REG): Likewise.
4868 (enum reg_class): Likewise.
4869 (BND_REG_P): Likewise.
4870 (BND_REGNO_P): Likewise.
4871 (BNDmode): Likewise.
4872 (ADJUST_INSN_LENGTH): Likewise.
4873 * config/i386/i386.md: Likewise.
4874 * config/i386/i386.opt: Likewise.
4875 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4876 (defined): Likewise.
4877 (LINK_MPX): Likewise.
4878 (MPX_SPEC): Likewise.
4879 (LIBMPX_SPEC): Likewise.
4880 (LIBMPXWRAPPERS_SPEC): Likewise.
4881 (CHKP_SPEC): Likewise.
4882 * config/i386/predicates.md: Likewise.
4883 * dbxout.c (dbxout_type): Likewise.
4884 * doc/extend.texi: Likewise.
4885 * doc/invoke.texi: Likewise.
4886 * doc/md.texi: Likewise.
4887 * doc/tm.texi: Likewise.
4888 * doc/tm.texi.in: Likewise.
4889 * dwarf2out.c (is_base_type): Likewise.
4890 (gen_formal_types_die): Likewise.
4891 (gen_subprogram_die): Likewise.
4892 (gen_type_die_with_usage): Likewise.
4893 (gen_decl_die): Likewise.
4894 (dwarf2out_late_global_decl): Likewise.
4895 * expr.c (expand_assignment): Likewise.
4896 (emit_storent_insn): Likewise.
4897 (store_expr_with_bounds): Likewise.
4898 (store_expr): Likewise.
4899 (expand_expr_real_1): Likewise.
4900 * expr.h (store_expr_with_bounds): Likewise.
4901 * function.c (use_register_for_decl): Likewise.
4902 (struct bounds_parm_data): Likewise.
4903 (assign_parms_augmented_arg_list): Likewise.
4904 (assign_parm_find_entry_rtl): Likewise.
4905 (assign_parm_is_stack_parm): Likewise.
4906 (assign_parm_load_bounds): Likewise.
4907 (assign_bounds): Likewise.
4908 (assign_parms): Likewise.
4909 (expand_function_start): Likewise.
4910 * gcc.c (CHKP_SPEC): Likewise.
4911 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4912 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4913 (wrestrict_dom_walker::check_call): Likewise.
4914 * gimple.c (gimple_build_call_from_tree): Likewise.
4915 * gimple.h (enum gf_mask): Likewise.
4916 (gimple_call_with_bounds_p): Likewise.
4917 (gimple_call_set_with_bounds): Likewise.
4918 * gimplify.c (gimplify_init_constructor): Likewise.
4919 * ipa-cp.c (initialize_node_lattices): Likewise.
4920 (propagate_constants_across_call): Likewise.
4921 (find_more_scalar_values_for_callers_subset): Likewise.
4922 * ipa-hsa.c (process_hsa_functions): Likewise.
4923 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4924 * ipa-icf.c (sem_function::merge): Likewise.
4925 * ipa-inline.c (early_inliner): Likewise.
4926 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4927 (warn_function_cold): Likewise.
4928 (propagate_pure_const): Likewise.
4929 * ipa-ref.h (enum GTY): Likewise.
4930 * ipa-split.c (find_retbnd): Likewise.
4931 (consider_split): Likewise.
4932 (split_function): Likewise.
4933 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4934 * ipa.c (walk_polymorphic_call_targets): Likewise.
4935 (symbol_table::remove_unreachable_nodes): Likewise.
4936 (process_references): Likewise.
4937 (cgraph_build_static_cdtor_1): Likewise.
4938 * lto-cgraph.c (lto_output_node): Likewise.
4939 (output_refs): Likewise.
4940 (compute_ltrans_boundary): Likewise.
4941 (input_overwrite_node): Likewise.
4942 (input_node): Likewise.
4943 (input_cgraph_1): Likewise.
4944 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4945 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4946 (class pass_chkp_instrumentation_passes): Likewise.
4947 (make_pass_chkp_instrumentation_passes): Likewise.
4948 * passes.def: Likewise.
4949 * rtl.h (struct GTY): Likewise.
4950 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4951 * stor-layout.c (layout_type): Likewise.
4952 * symtab.c: Likewise.
4953 * target.def: Likewise.
4954 * targhooks.c (default_chkp_bound_type): Likewise.
4955 (default_chkp_bound_mode): Likewise.
4956 (default_builtin_chkp_function): Likewise.
4957 (default_chkp_function_value_bounds): Likewise.
4958 (default_chkp_make_bounds_constant): Likewise.
4959 (default_chkp_initialize_bounds): Likewise.
4960 * targhooks.h (default_chkp_bound_type): Likewise.
4961 (default_chkp_bound_mode): Likewise.
4962 (default_builtin_chkp_function): Likewise.
4963 (default_chkp_function_value_bounds): Likewise.
4964 (default_chkp_make_bounds_constant): Likewise.
4965 (default_chkp_initialize_bounds): Likewise.
4966 * toplev.c (compile_file): Likewise.
4967 (process_options): Likewise.
4968 * tree-core.h (DEF_BUILTIN): Likewise.
4969 (DEF_BUILTIN_CHKP): Likewise.
4970 * tree-inline.c (declare_return_variable): Likewise.
4971 (remap_gimple_stmt): Likewise.
4972 (copy_bb): Likewise.
4973 (initialize_inlined_parameters): Likewise.
4974 (expand_call_inline): Likewise.
4975 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4976 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4977 (make_pass_ipa_chkp_produce_thunks): Likewise.
4978 (make_pass_chkp): Likewise.
4979 (make_pass_chkp_opt): Likewise.
4980 (make_pass_chkp_instrumentation_passes): Likewise.
4981 * tree-pretty-print.c (dump_generic_node): Likewise.
4982 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4983 * tree-ssa-dce.c (propagate_necessity): Likewise.
4984 (eliminate_unnecessary_stmts): Likewise.
4985 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4986 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4987 * tree-ssa-sccvn.h: Likewise.
4988 * tree-ssa-strlen.c (get_string_length): Likewise.
4989 (valid_builtin_call): Likewise.
4990 (adjust_last_stmt): Likewise.
4991 (handle_builtin_strchr): Likewise.
4992 (handle_builtin_strcpy): Likewise.
4993 (handle_builtin_stxncpy): Likewise.
4994 (handle_builtin_memcpy): Likewise.
4995 (handle_builtin_strcat): Likewise.
4996 (strlen_check_and_optimize_stmt): Likewise.
4997 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4998 * tree-streamer-in.c: Likewise.
4999 * tree-streamer.c (record_common_node): Likewise.
5000 * tree.c (tree_code_size): Likewise.
5001 (wide_int_to_tree_1): Likewise.
5002 (type_contains_placeholder_1): Likewise.
5003 (build_common_tree_nodes): Likewise.
5004 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5005 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5006 (POINTER_BOUNDS_P): Likewise.
5007 (BOUNDED_TYPE_P): Likewise.
5008 (BOUNDED_P): Likewise.
5009 (CALL_WITH_BOUNDS_P): Likewise.
5010 (pointer_bounds_type_node): Likewise.
5011 * value-prof.c (gimple_ic): Likewise.
5012 * var-tracking.c (vt_add_function_parameters): Likewise.
5013 * varasm.c (make_decl_rtl): Likewise.
5014 (assemble_start_function): Likewise.
5015 (output_constant): Likewise.
5016 (maybe_assemble_visibility): Likewise.
5017 * varpool.c (ctor_for_folding): Likewise.
5018 * chkp-builtins.def: Remove.
5019 * ipa-chkp.c: Remove.
5020 * ipa-chkp.h: Remove.
5021 * rtl-chkp.c: Remove.
5022 * rtl-chkp.h: Remove.
5023 * tree-chkp-opt.c: Remove.
5024 * tree-chkp.c: Remove.
5025 * tree-chkp.h: Remove.
5027 2018-06-07 Carl Love <cel@us.ibm.com>
5029 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
5030 vextract_fp_from_shortl): Add BE support.
5032 2018-06-07 Paul Koning <ni1d@arrl.net>
5034 * compare-elim.c (try_merge_compare): Don't merge compare if
5035 address contains a side effect.
5036 (try_eliminate_compare): Likewise.
5038 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5040 * config.gcc: Support "tremont".
5041 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5042 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5044 * config/i386/i386.c (m_TREMONT): Define.
5045 (processor_target_table): Add "tremont".
5046 (PTA_TREMONT): Define.
5047 (ix86_lea_outperforms): Add TARGET_TREMONT.
5048 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5049 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5050 and M_INTEL_GOLDMONT_PLUS.
5051 (fold_builtin_cpu): Add "tremont".
5052 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5053 (ix86_option_override_internal): Add "tremont".
5054 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5055 (processor_type): Add PROCESSOR_TREMONT.
5056 * config/i386/x86-tune.def: Add m_TREMONT.
5057 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5059 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5061 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5062 symbol defined for msp430i* devices to be lower case.
5064 gcc/testsuite/gcc.target/msp430/
5065 * msp430i-device-symbol.c: New test.
5066 * msp430f-device-symbol.c: New test.
5067 * msp430.h: New test header file.
5069 2018-06-07 Richard Biener <rguenther@suse.de>
5071 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5072 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5073 Properly wrap signed arithmetic if overflow wraps.
5075 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5077 PR tree-optimization/69615
5078 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5079 of a cast from a same precision integral SSA_NAME in a bb dominated
5080 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5081 cast to utype if rhs2 has already a compatible type.
5083 2018-06-07 Richard Biener <rguenther@suse.de>
5085 PR tree-optimization/85935
5086 * graphite-scop-detection.c (find_params_in_bb): Analyze
5087 condition operands with respect to the correct loop. Assert
5088 the analysis doesn't fail.
5090 2018-06-04 Carl Love <cel@us.ibm.com>
5092 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5093 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5094 as it is slightly cheaper.
5095 (first_match_or_eos_index_<mode>):
5096 Calculate index using natural element order.
5097 (first_match_index_<mode>):
5098 Calculate index using natural element order.
5099 (first_match_or_eos_index_<mode>):
5100 Calculate index using natural order.
5101 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5102 for BE and LE modes.
5103 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5104 P9V_BUILTIN_VCLZLSBB_V16QI.
5105 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5108 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5110 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5111 indentation and line wrap for many prototypes. Add missing
5112 @smallexample directives around block of prototypes for vec_xl and
5115 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5117 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5118 track if we pass or return IEEE 128-bit floating point.
5119 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5120 C++ mangling that is compatible with GCC 8.1.
5121 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5122 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5123 floating point types.
5124 (rs6000_function_arg_advance_1): Likewise.
5125 (rs6000_mangle_type): Optionally generate mangled names that match
5126 what GCC 8.1 generated for IEEE 128-bit floating point types.
5127 (rs6000_globalize_decl_name): If we have an external function that
5128 passes or returns IEEE 128-bit types, generate a weak reference
5129 from the mangled name used in GCC 8.1 to the current mangled
5131 (rs6000_init_builtins): Make __ibm128 use the long double type if
5132 long double is IBM extended double. Make __float128 use the long
5133 double type if long double is IEEE 128-bit.
5136 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5137 macro for __ibm128 built-in functions.
5138 (PACK_IF): Add __ibm128 pack/unpack functions.
5139 (UNPACK_IF): Likewise.
5140 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5141 enable long double built-in functions if long double is IEEE
5142 128-bit floating point.
5143 (rs6000_invalid_builtin): Update long double built-in function
5145 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5146 functions, adjust the built-in function to use the long double
5147 built-in function if __ibm128 and long double are the same type.
5148 * doc/extend.texi (PowerPC builtins): Update documention for
5149 __builtin_{,un}pack_longdouble. Add documentation for
5150 __builtin_{,un}pack_ibm128.
5152 2018-06-06 Jim Wilson <jimw@sifive.com>
5154 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5155 (struct machine_function): New field interrupt_mode.
5156 (riscv_handle_type_attribute): New function. Add forward declaration.
5157 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5158 (riscv_expand_epilogue): Check interrupt_mode field.
5159 (riscv_set_current_function): Check interrupt attribute args and
5160 set interrupt_mode field.
5161 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5162 (riscv_sret, riscv_uret): New.
5163 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5164 new arguments to interrupt attribute.
5166 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5169 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5170 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5172 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5174 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5175 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5176 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5177 several redundant entries.
5179 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5181 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5182 type from "rtx" to "rtx_insn *".
5183 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5184 for local "call_insn", removing cast.
5185 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5187 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5189 PR tree-optimization/86066
5190 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5191 for BIT_INSERT_EXPR stores.
5193 2018-06-06 Richard Biener <rguenther@suse.de>
5195 PR tree-optimization/86062
5196 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5197 component refs ontop
5198 of to be offsetted base.
5200 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5202 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5203 to be static and remove check on interrupt attribute name.
5205 gcc/testsuite/gcc.target/msp430/
5206 * function-attributes-4.c: New test.
5207 * static-interrupts.c: New test.
5209 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5211 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5212 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5214 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5217 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5219 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5220 Remove second argument, change how error is called.
5221 (aarch64_layout_arg): Remove second argument from
5222 aarch64_err_no_fpadvsimd call.
5223 (aarch64_init_cumulative_args): Ditto.
5224 (aarch64_gimplify_va_arg_expr): Ditto.
5225 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5227 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5229 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5230 (*simple_return_indirect_internal<mode>): Rename from
5231 simple_return_indirect_internal. Use W mode iterator.
5232 (rstorssp): New expander.
5233 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5234 (clrssbsy): New expander.
5235 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5237 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5239 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5241 (cmse_check_pointed_object): Likewise.
5243 2018-06-05 Martin Liska <mliska@suse.cz>
5245 PR gcov-profile/47618
5246 * doc/invoke.texi: Document how -fprofile-dir format
5249 2018-06-05 Richard Biener <rguenther@suse.de>
5251 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5252 removal pretend DOM info isn't available so we do not update
5253 it and only remove edges, not dominated blocks. Actually free
5254 DOM info in case we removed something. Remove unreachable blocks.
5255 (mfb_keep_latches): Work with either DOM info or marked backedges.
5256 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5257 first. Mark backedges if DOM info isn't available.
5258 (Re-)compute DOM info after cleanup_control_flow_pre.
5260 2018-06-05 Richard Biener <rguenther@suse.de>
5262 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5263 (locus_discrim_hasher::hash): Adjust.
5264 (locus_discrim_hasher::equal): Likewise.
5265 (next_discriminator_for_locus): Work on line directly.
5266 (same_line_p): Pass in expanded locus1 as well.
5267 (assign_discriminators): Avoid redundant location expansions.
5269 2018-06-05 Richard Biener <rguenther@suse.de>
5271 PR tree-optimization/86046
5272 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5273 if required after clearing TREE_ADDRESSABLE.
5275 2018-06-05 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/86047
5278 * tree-ssa-loop.c (for_each_index): Glob handling of all
5279 decls and constants and really handle all of them.
5281 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5284 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5285 qualifier_void_pointer and qualifier_const_void_pointer.
5286 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5287 (arm_init_builtins): Handle the above.
5288 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5289 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5290 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5293 2018-06-05 Martin Liska <mliska@suse.cz>
5295 * auto-profile.c (read_autofdo_file): Do not use
5296 gcov_ctr_summary struct.
5297 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5298 * coverage.c (struct counts_entry): Likewise.
5299 (read_counts_file): Read just single summary entry.
5300 (get_coverage_counts): Use gcov_summary struct.
5301 * coverage.h (get_coverage_counts): Likewise.
5302 * gcov-dump.c (dump_working_sets): Likewise.
5303 (tag_summary): Dump just single summary.
5304 * gcov-io.c (gcov_write_summary): Write just histogram
5306 (gcov_read_summary): Read just single summary.
5307 (compute_working_sets): Use gcov_summary struct.
5308 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
5309 of GCOV_COUNTERS_SUMMABLE.
5310 (GCOV_COUNTERS_SUMMABLE): Remove.
5311 (GCOV_FIRST_VALUE_COUNTER): Replace with
5312 GCOV_COUNTER_V_INTERVAL.
5313 (struct gcov_ctr_summary): Remove.
5314 (struct gcov_summary): Directly use fields of former
5316 (compute_working_sets): Use gcov_summary struct.
5317 * gcov.c (read_count_file): Do not use ctrs fields.
5318 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
5320 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
5322 * profile.c: Likewise.
5323 * profile.h: Likewise.
5325 2018-06-05 Martin Liska <mliska@suse.cz>
5327 PR gcov-profile/84846
5328 * gcov.c (output_lines): Print working directory only
5329 in intermediate format.
5331 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
5333 * config/s390/s390-builtin-types.def: Add void function type.
5334 * config/s390/s390-builtins.def: Use the function type for the
5337 2018-06-04 Jim Wilson <jimw@sifive.com>
5339 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5341 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
5342 and maybe_eh_return. Change regno to unsigned int. Use new args to
5343 handle EH_RETURN_DATA_REGNO registers properly.
5344 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
5345 (riscv_expand_epilogue): Update comment. Change argument name and
5346 type. Update code to use new name and type. Pass new args to
5347 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
5349 * config/riscv/riscv.md (NORMAL_RETURN): New.
5350 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
5351 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
5352 (eh_return): Call gen_eh_return_internal and emit barrier.
5353 (eh_return_internal): Call riscv_expand_epilogue.
5355 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
5357 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
5358 bit_insertion field and declare can_be_merged_into method.
5359 (merged_store_group::can_be_merged_into): New method.
5360 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
5361 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
5362 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
5364 2018-06-04 Richard Biener <rguenther@suse.de>
5366 PR tree-optimization/85955
5367 * builtins.c (fold_builtin_sincos): Convert pointers to
5368 destination to appropriate type before dereferencing.
5370 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
5372 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
5374 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
5376 * expr.c (expand_expr_real_1): Force the operand into memory if
5377 its TYPE_MODE is BLKmode and if there is no integer mode for
5378 the number of bits being extracted.
5380 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5384 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5385 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
5387 2018-06-04 Richard Biener <rguenther@suse.de>
5389 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
5390 (cleanup_tree_cfg_noloop): ... single caller. Do
5391 start_recording_case_labels later.
5393 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
5395 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
5396 to _IMMINTRIN_H_INCLUDED.
5397 * config/i386/pconfigintrin.h: Ditto.
5398 * config/i386/waitpkgintrin.h: Ditto.
5399 * config/i386/immintrin.h: Add includes for sgxintrin.h,
5400 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
5401 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
5402 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
5403 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
5404 waitpkgintrin.h and cldemoteintrin.h.
5406 2018-06-04 Richard Biener <rguenther@suse.de>
5408 PR tree-optimization/86038
5409 * tracer.c (find_best_successor): Check probability for
5410 being initialized, bail out if not.
5412 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
5415 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
5416 of bits to ignore when comparing architectures.
5418 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5420 PR tree-optimization/69615
5421 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
5422 maximum or minimum of the type, try to merge it also as if
5423 range1 is + [-, x - 1] or + [x + 1, -].
5426 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5428 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5430 PR tree-optimization/86034
5431 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5432 the unsigned bitfield type in a bit insertion sequence if it does not
5433 have a larger precision than the bitfield size.
5434 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5436 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5438 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5440 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5442 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5443 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5444 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5445 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5447 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5449 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5450 Disable -fdelete-null-pointer-checks for ELF toolchain.
5452 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5453 Kito Cheng <kito.cheng@gmail.com>
5455 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5456 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5457 (nds32*-*-*): Add glibc and uclibc conditions.
5458 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5459 (TARGET_EXCEPT_UNWIND_INFO): Define.
5460 * config/nds32/elf.h: New file.
5461 * config/nds32/linux.h: New file.
5462 * config/nds32/nds32-elf.opt: New file.
5463 * config/nds32/nds32-linux.opt: New file.
5464 * config/nds32/nds32-fp-as-gp.c
5465 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5466 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5468 (nds32_asm_file_end): Ditto.
5469 (nds32_print_operand): Ditto.
5470 (nds32_insert_attributes): Ditto.
5471 (nds32_init_libfuncs): New function.
5472 (TARGET_HAVE_TLS): Define.
5473 (TARGET_INIT_LIBFUNCS): Define.
5474 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5476 (TARGET_ELF): Apply different mcmodel setting.
5477 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5478 been migrated into elf.h and linux.h files.
5479 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5480 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5481 (mcmodel): The content has been migrated into nds32-elf.opt and
5482 nds32-linux.opt files.
5483 * config/nds32/t-elf: New file.
5484 * config/nds32/t-linux: New file.
5486 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5487 Shiva Chen <shiva0217@gmail.com>
5489 * config/nds32/constants.md (unspec_volatile_element): Add
5490 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5491 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5493 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5494 (make_pass_nds32_fp_as_gp): Declare.
5495 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5497 (nds32_asm_function_end_prologue): Remove unused asm output.
5498 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5499 (nds32_asm_file_start): Output necessary fp_as_gp information.
5500 (nds32_option_override): Adjust register usage.
5501 (nds32_expand_prologue): Consider fp_as_gp situation.
5502 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5503 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5506 (simple_return): Ditto.
5507 (omit_fp_begin): Output special directive for fp_as_gp.
5508 (omit_fp_end): Output special directive for fp_as_gp.
5509 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5510 mforbid-fp-as-gp): New options.
5512 2018-06-01 Mark Wielaard <mark@klomp.org>
5514 * dwarf2out.c (dwarf2out_finish): Remove generation of
5515 DW_AT_loclists_base.
5517 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5519 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5520 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5521 (struct merged_store_group): Add bit_insertion field.
5522 (dump_char_array): Use standard hexadecimal format.
5523 (merged_store_group::merged_store_group): Set bit_insertion to false.
5524 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5525 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5526 also print the mask in the dump file.
5527 (pass_store_merging::gate): Minor tweak.
5528 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5529 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5530 stores with INTEGER_CST stores.
5531 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5532 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5533 and use it throughout. Generate bit insertion sequences if need be.
5534 (pass_store_merging::process_store): Remove redundant condition.
5535 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5537 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5539 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5540 the 128-bit floating point types. Fix function comment.
5542 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5544 * config/aarch64/aarch64-simd.md
5545 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5547 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5550 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5552 PR tree-optimization/85989
5553 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5555 (backprop::intersect_uses): Check it when deciding whether this
5556 is a backedge reference.
5557 (backprop::process_block): Add each phi to m_visited_phis
5558 after visiting it, then clear it at the end.
5560 2018-06-01 Richard Biener <rguenther@suse.de>
5562 * tree-vectorizer.h (vect_dr_stmt): New function.
5563 (vect_get_load_cost): Adjust.
5564 (vect_get_store_cost): Likewise.
5565 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5566 Use vect_dr_stmt instead of DR_SMTT.
5567 (vect_record_base_alignments): Likewise.
5568 (vect_calculate_target_alignment): Likewise.
5569 (vect_compute_data_ref_alignment): Likewise and make static.
5570 (vect_update_misalignment_for_peel): Likewise.
5571 (vect_verify_datarefs_alignment): Likewise.
5572 (vector_alignment_reachable_p): Likewise.
5573 (vect_get_data_access_cost): Likewise. Pass down
5574 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5575 (vect_get_peeling_costs_all_drs): Likewise.
5576 (vect_peeling_hash_get_lowest_cost): Likewise.
5577 (vect_enhance_data_refs_alignment): Likewise.
5578 (vect_find_same_alignment_drs): Likewise.
5579 (vect_analyze_data_refs_alignment): Likewise.
5580 (vect_analyze_group_access_1): Likewise.
5581 (vect_analyze_group_access): Likewise.
5582 (vect_analyze_data_ref_access): Likewise.
5583 (vect_analyze_data_ref_accesses): Likewise.
5584 (vect_vfa_segment_size): Likewise.
5585 (vect_small_gap_p): Likewise.
5586 (vectorizable_with_step_bound_p): Likewise.
5587 (vect_prune_runtime_alias_test_list): Likewise.
5588 (vect_analyze_data_refs): Likewise.
5589 (vect_supportable_dr_alignment): Likewise.
5590 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5591 (vect_gen_prolog_loop_niters): Likewise.
5592 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5593 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5595 (vect_recog_mask_conversion_pattern): Likewise.
5596 (vect_try_gather_scatter_pattern): Likewise.
5597 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5598 to vect_get_store_cost.
5599 (vect_get_store_cost): Get stmt_info instead of DR.
5600 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5601 (vect_get_load_cost): Get stmt_info instead of DR.
5603 2018-06-01 Richard Biener <rguenther@suse.de>
5606 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5607 vars wrapped in MEM_REFs.
5609 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5611 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5612 Fix subreg tests so that we only return a choice between
5613 GENERAL_REGS and FP_REGS if the original classes included both.
5615 2018-06-01 Richard Biener <rguenther@suse.de>
5618 * tree-ssa-structalias.c (get_function_part_constraint):
5619 Handle NULL fi->decl.
5620 (find_func_aliases_for_call): Properly handle indirect
5621 fi from direct call.
5622 (find_func_clobbers): Likewise.
5623 (ipa_pta_execute): Likewise.
5624 (create_variable_info_for): For functions that are ifunc_resolver
5625 resolve to a varinfo that contains the result of the resolver call.
5626 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5629 2018-05-31 Michael Collison <michael.collison@arm.com>
5631 * config/aarch64/aarch64.md:
5632 (*fix_to_zero_extenddfdi2): New pattern.
5633 * gcc.target/aarch64/fix_extend1.c: New testcase.
5635 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5639 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5640 and BUILT_IN_STRNCMP_EQ.
5641 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5642 BUILT_IN_STRNCMP_EQ.
5643 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5644 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5645 (gimple_fold_builtin): Likewise.
5646 * tree-ssa-strlen.c (compute_string_length): New function.
5647 (determine_min_obsize): New function.
5648 (handle_builtin_string_cmp): New function to handle calls to
5649 string compare functions.
5650 (strlen_optimize_stmt): Add handling to builtin string compare
5652 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5653 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5654 * tree.c (build_common_builtin_nodes): Add new defines of
5655 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5657 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5660 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5661 functions with naked attribute.
5663 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5665 * config/i386/sse.md (avx_vec_concat<mode>):
5666 Substitute concat_tg_mode mode attribute with xtg_mode.
5667 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5668 (concat_tg_mode): Remove mode attribute.
5670 2018-05-31 Martin Sebor <msebor@redhat.com>
5673 * calls.c (alloc_max_size): Correct a logic error/typo.
5674 Treat excessive arguments as infinite. Warn for invalid arguments.
5675 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5677 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5680 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5681 and movx for Haswell.
5683 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5684 Cesar Philippidis <cesar@codesourcery.com>
5687 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5688 when emitting error on private/firstprivate reductions.
5689 * omp-low.c (lower_omp_target): Avoid reference-type processing
5690 on pointers for firstprivate clause.
5692 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5694 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5697 * config/aarch64/aarch64-simd.md
5698 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5699 (aarch64_ld1_x3_<mode>): Likewise
5700 (aarch64_st1x2<VALLDIF:mode>): Likewise
5701 (aarch64_st1_x2_<mode>): Likewise
5702 (aarch64_st1x3<VALLDIF:mode>): Likewise
5703 (aarch64_st1_x3_<mode>): Likewise
5704 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5705 (vld1_s8_x3): Likewise.
5706 (vld1_u16_x3): Likewise.
5707 (vld1_s16_x3): Likewise.
5708 (vld1_u32_x3): Likewise.
5709 (vld1_s32_x3): Likewise.
5710 (vld1_u64_x3): Likewise.
5711 (vld1_s64_x3): Likewise.
5712 (vld1_f16_x3): Likewise.
5713 (vld1_f32_x3): Likewise.
5714 (vld1_f64_x3): Likewise.
5715 (vld1_p8_x3): Likewise.
5716 (vld1_p16_x3): Likewise.
5717 (vld1_p64_x3): Likewise.
5718 (vld1q_u8_x3): Likewise.
5719 (vld1q_s8_x3): Likewise.
5720 (vld1q_u16_x3): Likewise.
5721 (vld1q_s16_x3): Likewise.
5722 (vld1q_u32_x3): Likewise.
5723 (vld1q_s32_x3): Likewise.
5724 (vld1q_u64_x3): Likewise.
5725 (vld1q_s64_x3): Likewise.
5726 (vld1q_f16_x3): Likewise.
5727 (vld1q_f32_x3): Likewise.
5728 (vld1q_f64_x3): Likewise.
5729 (vld1q_p8_x3): Likewise.
5730 (vld1q_p16_x3): Likewise.
5731 (vld1q_p64_x3): Likewise.
5732 (vst1_s64_x2): Likewise.
5733 (vst1_u64_x2): Likewise.
5734 (vst1_f64_x2): Likewise.
5735 (vst1_s8_x2): Likewise.
5736 (vst1_p8_x2): Likewise.
5737 (vst1_s16_x2): Likewise.
5738 (vst1_p16_x2): Likewise.
5739 (vst1_s32_x2): Likewise.
5740 (vst1_u8_x2): Likewise.
5741 (vst1_u16_x2): Likewise.
5742 (vst1_u32_x2): Likewise.
5743 (vst1_f16_x2): Likewise.
5744 (vst1_f32_x2): Likewise.
5745 (vst1_p64_x2): Likewise.
5746 (vst1q_s8_x2): Likewise.
5747 (vst1q_p8_x2): Likewise.
5748 (vst1q_s16_x2): Likewise.
5749 (vst1q_p16_x2): Likewise.
5750 (vst1q_s32_x2): Likewise.
5751 (vst1q_s64_x2): Likewise.
5752 (vst1q_u8_x2): Likewise.
5753 (vst1q_u16_x2): Likewise.
5754 (vst1q_u32_x2): Likewise.
5755 (vst1q_u64_x2): Likewise.
5756 (vst1q_f16_x2): Likewise.
5757 (vst1q_f32_x2): Likewise.
5758 (vst1q_f64_x2): Likewise.
5759 (vst1q_p64_x2): Likewise.
5760 (vst1_s64_x3): Likewise.
5761 (vst1_u64_x3): Likewise.
5762 (vst1_f64_x3): Likewise.
5763 (vst1_s8_x3): Likewise.
5764 (vst1_p8_x3): Likewise.
5765 (vst1_s16_x3): Likewise.
5766 (vst1_p16_x3): Likewise.
5767 (vst1_s32_x3): Likewise.
5768 (vst1_u8_x3): Likewise.
5769 (vst1_u16_x3): Likewise.
5770 (vst1_u32_x3): Likewise.
5771 (vst1_f16_x3): Likewise.
5772 (vst1_f32_x3): Likewise.
5773 (vst1_p64_x3): Likewise.
5774 (vst1q_s8_x3): Likewise.
5775 (vst1q_p8_x3): Likewise.
5776 (vst1q_s16_x3): Likewise.
5777 (vst1q_p16_x3): Likewise.
5778 (vst1q_s32_x3): Likewise.
5779 (vst1q_s64_x3): Likewise.
5780 (vst1q_u8_x3): Likewise.
5781 (vst1q_u16_x3): Likewise.
5782 (vst1q_u32_x3): Likewise.
5783 (vst1q_u64_x3): Likewise.
5784 (vst1q_f16_x3): Likewise.
5785 (vst1q_f32_x3): Likewise.
5786 (vst1q_f64_x3): Likewise.
5787 (vst1q_p64_x3): Likewise.
5789 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5791 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5792 user_label_prefix to name.
5794 * tree-core.h: Update comment about the format of NAME string
5795 passed to handler in attribute_spec.
5797 * config/msp430/msp430.md: Remove erroneous subreg expression from
5798 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5799 zero_extend{q,h}isi2.
5801 2018-05-30 Borislav Petkov <bp@suse.de>
5803 * doc/extend.texi: Document some architecture specific
5804 constraints and sort entries.
5806 2018-05-30 Martin Sebor <msebor@redhat.com>
5809 * builtins.c (expand_builtin_stpcpy_1): New function.
5810 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5811 only if the former succeeds.
5813 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5815 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5818 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5820 * doc/invoke.texi (-flinker-output): Document
5822 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5824 * passes.c (ipa_write_summaries): Only modify statements if body
5826 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5827 incrementally linking.
5828 (ipa_passes): Likewise.
5829 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5830 pass down resolution info.
5831 * common.opt (flag_incremental_link): Update info.
5832 * gcc.c (plugin specs): Turn flinker-output=* to
5833 -plugin-opt=-linker-output-known
5834 * toplev.c (compile_file): Also cut compilation when doing incremental
5836 * flag-types. (enum lto_partition_model): Add
5837 LTO_LINKER_OUTPUT_NOLTOREL.
5838 (invoke.texi): Add -flinker-output docs.
5839 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5840 link same way as WPA; do not stream in dead initializers.
5842 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5843 darf2out_register_external_decl): Support incremental link.
5845 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5847 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5849 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5851 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5852 it down to simple_object_copy_lto_debug_sections.
5853 (run_gcc): Determine incremental LTO link time and configure
5854 lto1 into non-wpa mode, disable renaming of debug sections.
5856 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5858 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5859 descriptions of various incorrectly documented functions.
5861 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5864 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5865 address check not strict.
5867 2018-05-30 Richard Biener <rguenther@suse.de>
5869 PR tree-optimization/85964
5870 * tracer.c (better_p): Drop initialized count check, we only
5871 call the function with initialized counts now.
5872 (find_best_successor): Do find a best edge if one
5873 has uninitialized count.
5874 (find_best_predecessor): Likewise. Do BB frequency check only
5875 if count is initialized.
5877 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5879 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5880 (aarch64_ldrstr_offset_compare): New.
5881 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5882 load/store orderings.
5883 (aarch64_gen_adjusted_ldpstp): Likewise.
5885 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5887 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5888 Check for subset of GENERAL_REGS and FP_REGS.
5889 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5892 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5894 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5895 and wi::to_poly_offset. Add the current offset and then check
5896 whether the sum fits, rather than using an unchecked addition of
5897 a checked term. Check for a shwi rather than a uhwi.
5898 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5899 (store_constructor): Use poly_int_tree_p.
5900 (expand_expr_real_1): Likewise.
5901 * function.c (assign_temp): Likewise.
5902 * fold-const.c (const_binop): Use poly_int_tree_p and
5904 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5906 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5907 to_poly_offset for MEM offsets.
5908 * ipa-icf.c (sem_variable::equals): Likewise.
5909 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5910 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5911 wi::to_poly_offset for BIT_FIELD_REF offsets.
5912 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5914 * var-tracking.c (emit_note_insn_var_location): Use
5915 tree_to_poly_uint64.
5917 2018-05-29 Jim Wilson <jimw@sifive.com>
5919 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5921 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5924 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5925 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5927 (sse4_1_round<mode>2): Use nonimmediate_operand
5928 for operand 1 predicate.
5930 2018-05-29 Martin Sebor <msebor@redhat.com>
5931 Richard Biener <rguenther@suse.de>
5934 * calls.c (get_size_range): Call determine_value_range instead
5935 of get_value_range..
5936 * tree-vrp.h (determine_value_range): Declared new function.
5937 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5939 2018-05-29 Richard Biener <rguenther@suse.de>
5941 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5942 sure to use non-pattern stmts for get_earlier_stmt arguments.
5943 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5944 called on pattern stmts.
5945 (get_later_stmt): Likewise.
5947 2018-05-29 Martin Liska <mliska@suse.cz>
5949 PR gcov-profile/85759
5950 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5953 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5955 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5957 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5961 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5962 VEC_PACK_FLOAT_EXPR): New tree codes.
5963 * tree-pretty-print.c (op_code_prio): Handle
5964 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5965 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5966 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5967 * tree-inline.c (estimate_operator_cost): Likewise.
5968 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5969 * fold-const.c (const_binop): Likewise.
5970 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5971 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5972 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5973 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5974 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5975 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5976 * expr.c (expand_expr_real_2): Likewise.
5977 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5978 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5979 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5981 * optabs.c (expand_widen_pattern_expr): For
5982 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5983 sign from result type rather than operand's type.
5984 (expand_binop_directly): For vec_packu_float_optab and
5985 vec_packs_float_optab allow result type to be different from operand's
5987 * optabs-tree.c (optab_for_tree_code): Handle
5988 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5989 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5990 * tree-vect-generic.c (expand_vector_operations_1): Handle
5991 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5992 VEC_PACK_FLOAT_EXPR.
5993 * tree-vect-stmts.c (supportable_widening_operation): Handle
5995 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5996 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5997 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5998 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5999 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6001 (vec_pack<floatprefix>_float_<mode>): New expander.
6002 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6004 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6005 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6006 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6007 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6008 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6010 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6011 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6012 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6013 VEC_PACK_FLOAT_EXPR): Document.
6015 2018-05-29 Richard Biener <rguenther@suse.de>
6017 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6019 (stmt_vec_info_vec): Make pointer.
6020 (init_stmt_vec_info_vec): Remove.
6021 (free_stmt_vec_info_vec): Likewise.
6022 (set_stmt_vec_info_vec): New function.
6023 (free_stmt_vec_infos): Likewise.
6024 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6025 (set_vinfo_for_stmt): Likewise.
6026 (get_earlier_stmt): Likewise.
6027 (get_later_stmt): Likewise.
6028 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6029 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6030 (vec_info::~vec_info): Free stmt_vec_infos.
6031 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6032 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6033 (pass_slp_vectorize::execute): Likewise.
6034 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6035 (free_stmt_vec_info_vec): Likewise.
6036 (set_stmt_vec_info_vec): New function.
6037 (free_stmt_vec_infos): Likewise.
6038 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6039 the global stmt_vec_info_vec.
6040 * tree-parloops.c (gather_scalar_reductions): Use
6041 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6044 2018-05-29 Richard Biener <rguenther@suse.de>
6046 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6048 2018-05-29 Martin Liska <mliska@suse.cz>
6049 David Malcolm <dmalcolm@redhat.com>
6051 * vec.c (test_reverse): New.
6052 (vec_c_tests): Add new test.
6053 * vec.h (vl_ptr>::reverse): New function.
6055 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6057 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6059 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6062 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6064 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6066 2018-05-28 Richard Biener <rguenther@suse.de>
6068 PR tree-optimization/85933
6069 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6070 look at stmts marked as vectorizable.
6072 2018-05-28 Richard Biener <rguenther@suse.de>
6074 PR tree-optimization/85934
6075 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6076 vector boolean check before scalar optimization.
6078 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6080 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6083 2018-05-28 Mark Wielaard <mark@klomp.org>
6085 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6086 if it is an expression containing a minus sign.
6088 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6090 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6092 2018-05-27 Paul Koning <ni1d@arrl.net>
6094 * config/pdp11/pdp11.md (truncsihi2): Remove.
6096 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6097 Chung-Ju Wu <jasonwucj@gmail.com>
6099 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6101 (unaligned_store_dw): Ditto.
6102 * config/nds32/nds32-memory-manipulation.c
6103 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6104 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6105 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6106 (emit_setmem_word_loop): Rename to ...
6107 (emit_setmem_doubleword_loop): ... this.
6108 (nds32_gen_dup_4_byte_to_word_value): New function.
6109 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6110 (nds32_expand_setmem_loop): Refine implementation.
6111 (nds32_expand_setmem_loop_v3m): Ditto.
6112 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6115 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6117 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6119 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6121 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6122 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6123 machine->attr_no_prologue_p.
6124 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6125 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6126 (nds32_expand_epilogue): Consider attr_naked_p.
6127 (nds32_expand_epilogue_v3pop): Likewise.
6128 (nds32_can_use_return_insn): Likewise.
6129 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6130 attr_no_prologue_p fields.
6131 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6133 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6136 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6138 * config/i386/sse.md
6139 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6141 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6143 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6145 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6147 (*<floatsuffix>floatv2div2sf2): Rename to ...
6148 (*float<floatunssuffix>v2div2sf2): ... this.
6149 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6150 (float<floatunssuffix>v2div2sf2_mask): ... this.
6151 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6152 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6153 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6155 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6157 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6159 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6161 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6163 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6165 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6166 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6167 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6168 gen_ufix_truncv8dfv8si2.
6169 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6170 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6171 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6172 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6173 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6174 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6175 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6176 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6178 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6182 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6184 2018-05-25 Jim Wilson <jimw@sifive.com>
6186 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6187 * config/riscv/riscv.c (struct machine_function): Add
6188 interrupt_handler_p and attribute_checked_p fields.
6189 (riscv_attribute_table): Add interrupt.
6190 (riscv_interrupt_type_p): New.
6191 (riscv_save_reg_p): Save extra regs for interrupt handler.
6192 (riscv_use_save_libcall): Return false for interrupt handler.
6193 (riscv_first_stack_step): Add forward declaration.
6194 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6195 for interrupt handler with large frame. Use it for saved reg list.
6196 (riscv_expand_prologue): Move flag_stack_usage_info support to
6197 eliminate duplication.
6198 (riscv_expand_epilogue): Generate mret for interrupt handler.
6199 (riscv_epilogue_uses): New.
6200 (riscv_can_use_return_insn): Return false for interrupt handler.
6201 (riscv_function_ok_for_sibcall): Likewise.
6202 (riscv_set_current_function): Add interrupt handler support.
6203 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6204 * config/riscv/riscv.md (UNSPECV_MRET): New.
6206 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6208 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6210 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6212 PR tree-optimization/85712
6213 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6214 this candidate has already been replaced in-situ by a copy.
6216 2018-05-25 Jason Merrill <jason@redhat.com>
6218 PR c++/80485 - inline function non-zero address.
6219 * symtab.c (nonzero_address): Check DECL_COMDAT.
6221 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6224 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6225 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6226 extension of SImode operation. Use const123_operand predicate.
6227 (*saddsi_1): Remove.
6228 (*saddl_se_1): Ditto.
6230 (*ssubl_se_1): Ditto.
6231 * config/alpha/predicates.md (const123_operand): New predicate.
6232 * config/alpha/constraints.md (P): Use IN_RANGE.
6234 2018-05-25 Richard Biener <rguenther@suse.de>
6236 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6238 (ref_maybe_used_by_stmt_p): Likewise.
6239 (stmt_may_clobber_ref_p): Likewise.
6240 (stmt_may_clobber_ref_p_1): Likewise.
6241 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6243 (ref_maybe_used_by_stmt_p): Likewise.
6244 (stmt_may_clobber_ref_p): Likewise.
6245 (stmt_may_clobber_ref_p_1): Likewise.
6246 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6247 the alias oracle to disambiguate DRs with stmts DR analysis
6249 (vect_analyze_data_refs): Do not give up on not analyzable
6250 DRs for BB vectorization. Remove code truncating the dataref
6253 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6256 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6257 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6258 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6260 2018-05-25 Richard Biener <rguenther@suse.de>
6262 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6263 function, combining stmt data ref gathering and fatal analysis
6265 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6266 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6267 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6268 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6269 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6271 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6273 PR tree-optimization/85720
6274 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6275 SCC if all partitions are builtins.
6276 (version_loop_by_alias_check): New parameter. Generate cancelable
6277 runtime alias check if all partitions are builtins.
6278 (distribute_loop): Update call to above function.
6280 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6282 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6283 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6284 (parm_default_def_partition_arg): Ditto.
6285 (set_parm_default_def_partition): Ditto.
6286 (get_parm_default_def_partitions): Ditto and make it static.
6287 (get_undefined_value_partitions): Ditto and make it static.
6288 (remove_ssa_form): Refactor call to init_var_map here.
6289 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6290 computation for loop region.
6291 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6292 (register_default_def): Delete.
6293 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6294 (parm_default_def_partition_arg): Ditto.
6295 (set_parm_default_def_partition): Ditto.
6296 (get_parm_default_def_partitions): Ditto and make it static.
6297 (get_undefined_value_partitions): Ditto and make it static.
6298 (coalesce_with_default, coalesce_with_default): Update comment.
6299 (create_coalesce_list_for_region): New func factored out from
6300 create_outofssa_var_map.
6301 (populate_coalesce_list_for_outofssa): New func factored out from
6302 create_outofssa_var_map and coalesce_ssa_name.
6303 (create_outofssa_var_map): Delete.
6304 (coalesce_ssa_name): Refactor to support live range computation.
6305 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6306 (get_parm_default_def_partitions): Delete.
6307 (get_undefined_value_partitions): Ditto.
6308 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
6309 computation for loop region.
6310 (new_tree_live_info, loe_visit_block): Ditto.
6311 (live_worklist, set_var_live_on_entry): Ditto.
6312 (calculate_live_on_exit, verify_live_on_entry): Ditto.
6313 * tree-ssa-live.h (struct _var_map): New fields.
6314 (init_var_map): Change decl.
6315 (region_contains_p): New.
6317 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6319 * tree-ssa-live.h (live_merge_and_clear): Delete.
6321 2018-05-25 Richard Biener <rguenther@suse.de>
6324 * tree-dump.c (dequeue_and_dump): Remove access to removed
6325 operand 2 of a SWITCH_EXPR.
6327 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6329 * doc/sourcebuild.texi (vect_double_cond_arith): Include
6330 multiplication and division.
6331 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6332 (cond_udiv@var{m}, cond_umod@var{m}): Document.
6333 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
6334 (cond_udiv_optab, cond_umod_optab): New optabs.
6335 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
6336 (IFN_COND_RDIV): New internal functions.
6337 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
6338 TRUNC_MOD_EXPR and RDIV_EXPR.
6339 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
6340 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
6342 (SVE_INT_BINARY): Include mult.
6343 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
6344 (optab, sve_int_op): Handle mult.
6345 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
6347 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
6348 for SVE_INT_BINARY_SD.
6350 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6352 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
6353 (optab, sve_int_op): Handle div and udiv.
6354 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
6355 for SVE_INT_BINARY_SD.
6356 (*<optab><mode>3): New insn for the same.
6358 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6360 * tree-vect-patterns.c: Include predict.h.
6361 (vect_recog_divmod_pattern): Restrict check for division support
6362 to when optimizing for size.
6364 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6366 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
6367 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
6368 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
6369 (gimple_match_op::set_op): Likewise.
6370 (gimple_resimplify4): Declare.
6371 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
6372 (expr::gen_transform): Likewise.
6373 (decision_tree::gen): Generate a simplification routine for 4 operands.
6374 * gimple-match-head.c (gimple_simplify): Add an overload for
6375 4 operands. In the top-level function, handle up to 4 call
6376 arguments and call gimple_resimplify4.
6377 (gimple_resimplify4): New function.
6378 (build_call_internal): Pass a fourth operand.
6379 (maybe_push_to_seq): Likewise.
6380 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
6381 Fold VEC_COND_EXPRs of an operation and a default value into
6382 an IFN_COND_* function if possible.
6383 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
6385 (SVE_COND_FP_BINARY): Include them.
6386 (optab, sve_fp_op): Handle them.
6387 (SVE_INT_BINARY_REV): New code iterator.
6388 (SVE_COND_FP_BINARY_REV): New int iterator.
6389 (commutative): New int attribute.
6390 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
6392 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
6394 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
6395 (*cond_<optab><mode>): New patterns for reversed operands.
6397 2018-05-25 Richard Biener <rguenther@suse.de>
6399 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
6400 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
6401 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
6402 (STMT_VINFO_GROUPED_ACCESS): Adjust.
6403 * tree-vect-data-refs.c (everywhere): Adjust users.
6404 * tree-vect-loop.c (everywhere): Likewise.
6405 * tree-vect-slp.c (everywhere): Likewise.
6406 * tree-vect-stmts.c (everywhere): Likewise.
6407 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6409 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6411 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
6413 (gcc_cv_as_section_exclude): ... this.
6414 Try Solaris as #exclude syntax.
6415 * configure: Regenerate.
6416 * config.in: Regenerate.
6417 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
6419 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
6420 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
6422 * varasm.c (default_elf_asm_named_section): Don't check if
6423 HAVE_GAS_SECTION_EXCLUDE is defined.
6425 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6427 * doc/md.texi: Update the documentation of the cond_* optabs
6428 to mention the new final operand. Fix GET_MODE_NUNITS call.
6429 Describe the scalar case too.
6430 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6431 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6433 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6434 (get_conditional_internal_fn): Update comment.
6435 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6436 accumulator value as a final argument to conditional functions.
6437 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6438 a define_expand and add an "else" operand. Assert for now that
6439 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6440 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6441 (*cond_<optab><mode>): New patterns.
6442 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6443 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6444 (UNSPEC_COND_EOR): Delete.
6445 (optab): Remove associated mappings.
6446 (SVE_INT_BINARY): New code iterator.
6447 (sve_int_op): Remove int attribute and add "minus" to the code
6449 (SVE_COND_INT_OP): Delete.
6450 (SVE_COND_FP_OP): Rename to...
6451 (SVE_COND_FP_BINARY): ...this.
6453 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6455 * optabs.c (can_reuse_operands_p): New function.
6456 (maybe_legitimize_operands): Try to reuse the results for
6459 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6461 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6462 Add {q} suffix to insn mnemonic.
6464 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6466 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6467 (msp430_warn_func_return): New.
6469 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6471 * fold-const.c (tree_nonzero_bits): New function.
6472 * fold-const.h (tree_nonzero_bits): Likewise.
6473 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6474 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6476 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6480 * varasm.c (assemble_alias): Check ifunc_resolver only on
6483 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6486 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6487 when memory input operand is handled.
6489 2018-05-24 Luis Machado <luis.machado@linaro.org>
6491 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6493 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6495 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6497 * match.pd: Delay FMA folds until after vectorization.
6499 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6502 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6503 address check not strict.
6505 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6507 * gimple-match.h (gimple_match_op): New class.
6508 (mprts_hook): Replace parameters with a gimple_match_op *.
6509 (maybe_build_generic_op): Likewise.
6510 (gimple_simplified_result_is_gimple_val): Replace parameters with
6511 a const gimple_match_op *.
6512 (gimple_simplify): Replace code_helper * and tree * parameters with
6513 a gimple_match_op * parameter.
6514 (gimple_resimplify1): Replace code_helper *, tree and tree *
6515 parameters with a gimple_match_op * parameter.
6516 (gimple_resimplify2): Likewise.
6517 (gimple_resimplify3): Likewise.
6518 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6519 parameters with a gimple_match_op * parameter.
6520 * gimple-match-head.c (gimple_simplify): Change prototypes of
6521 auto-generated functions to take a gimple_match_op * instead of
6522 separate code_helper * and tree * parameters. Make the same
6523 change in the top-level overload and update calls to the
6524 gimple_resimplify routines. Update calls to the auto-generated
6525 functions and to maybe_push_res_to_seq in the publicly-facing
6526 operation-specific gimple_simplify overloads.
6527 (gimple_match_op::MAX_NUM_OPS): Define.
6528 (gimple_resimplify1): Replace rcode and ops with a single res_op
6529 parameter. Update call to gimple_simplify.
6530 (gimple_resimplify2): Likewise.
6531 (gimple_resimplify3): Likewise.
6532 (mprts_hook): Replace parameters with a gimple_match_op *.
6533 (maybe_build_generic_op): Likewise.
6534 (build_call_internal): Replace type, nargs and ops with
6535 a gimple_match_op *.
6536 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6537 with a single gimple_match_op *. Update calls to mprts_hook,
6538 build_call_internal and gimple_simplified_result_is_gimple_val.
6539 Factor out code that is common to the tree_code and combined_fn cases.
6540 * genmatch.c (expr::gen_transform): Replace tem_code and
6541 tem_ops with a gimple_match_op called tem_op. Update calls
6542 to the gimple_resimplify functions and maybe_push_res_to_seq.
6543 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6544 res_ops. Update call to the gimple_resimplify functions.
6545 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6546 (decision_tree::gen): Make the functions take a gimple_match_op *
6547 called res_op instead of separate res_code and res_ops parameters.
6548 Update call accordingly.
6549 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6550 and ops with a single res_op parameter. Update calls to
6551 maybe_build_generic_op and maybe_push_res_to_seq.
6552 (fold_stmt_1): Update calls to gimple_simplify and
6553 replace_stmt_with_simplification.
6554 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6555 and gimple_simplified_result_is_gimple_val.
6556 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6558 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6559 with a gimple_match_op *.
6560 (vn_nary_build_or_lookup): Likewise. Update call to
6561 vn_nary_build_or_lookup_1.
6562 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6563 gimple_match_op *. Update calls to the gimple_resimplify routines
6564 and to gimple_simplified_result_is_gimple_val.
6565 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6566 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6567 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6568 (visit_nary_op): Likewise.
6569 (visit_reference_op_load): Likewise.
6571 2018-05-23 Luis Machado <luis.machado@linaro.org>
6573 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6574 modifier for printing the step amount.
6576 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6579 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6582 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6584 * doc/sourcebuild.texi (Endianness): New subsubsection.
6586 2018-05-23 Luis Machado <luis.machado@linaro.org>
6588 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6589 <prefetch_dynamic_strides>: New const bool field.
6590 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6591 prefetch_dynamic_strides.
6592 (exynosm1_prefetch_tune): Likewise.
6593 (thunderxt88_prefetch_tune): Likewise.
6594 (thunderx_prefetch_tune): Likewise.
6595 (thunderx2t99_prefetch_tune): Likewise.
6596 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6598 (aarch64_override_options_internal): Update to set
6599 PARAM_PREFETCH_DYNAMIC_STRIDES.
6600 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6601 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6602 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6603 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6604 prefetch-dynamic-strides setting.
6606 2018-05-23 Luis Machado <luis.machado@linaro.org>
6608 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6609 <minimum_stride>: New const int field.
6610 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6611 minimum_stride field defaulting to -1.
6612 (exynosm1_prefetch_tune): Likewise.
6613 (thunderxt88_prefetch_tune): Likewise.
6614 (thunderx_prefetch_tune): Likewise.
6615 (thunderx2t99_prefetch_tune): Likewise.
6616 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6617 <default_opt_level>: Set to 3.
6618 (aarch64_override_options_internal): Update to set
6619 PARAM_PREFETCH_MINIMUM_STRIDE.
6620 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6621 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6622 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6623 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6624 stride is constant and is below the minimum stride threshold.
6626 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6628 * config/arm/arm-cpus.in (mode26): Delete.
6629 (armv4): Delete mode26 reference.
6630 * config/arm/arm.c (arm_configure_build_target): Delete use of
6633 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6635 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6637 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6638 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6640 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6641 (floatunsdidf2): Ditto.
6643 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6645 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6646 (fixuns_trunc<mode>si2_avx512f): Ditto.
6647 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6648 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6649 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6651 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6653 PR rtl-optimization/79985
6654 * df-scan.c (df_insn_refs_collect): Remove special case for
6655 global registers and asm statements.
6657 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6659 * extend.texi (Global Register Variables): Rewrite the bullet list.
6660 Note that the register is available for allocation. Note that access
6661 via inline asm must use constraints. Add note about async-signal
6662 handlers. Remove paragraph about automagic register selection.
6664 2018-05-23 Richard Biener <rguenther@suse.de>
6666 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6667 of fixed offset from memset VN.
6669 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6671 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6673 (alloc_cand_and_find_basis): Initialize first_interp field.
6674 (slsr_process_mul): Modify first_interp field.
6675 (slsr_process_add): Likewise.
6676 (slsr_process_cast): Modify first_interp field for each new
6678 (slsr_process_copy): Likewise.
6679 (dump_candidate): Dump first_interp field.
6680 (replace_mult_candidate): Process all interpretations, not just
6682 (replace_rhs_if_not_dup): Likewise.
6683 (replace_one_candidate): Likewise.
6685 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6687 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6689 (aarch64_needs_frame_chain): New function.
6690 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6692 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6695 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6696 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6697 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6698 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6699 as true for strict-align.
6700 (aarch64_can_inline_p): Perform checks even when callee has no
6701 attributes to check for strict alignment.
6702 * doc/extend.texi (AArch64 Function Attributes): Document
6704 * doc/invoke.texi: (AArch64 Options): Likewise.
6706 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6708 PR tree-optimization/85853
6709 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6710 the handling of the root of the node to...
6711 (vect_slp_analyze_node_operations_1): ...this new function,
6712 and run the whole thing with the child nodes' def types
6713 set according to their SLP node's def type.
6715 2018-05-23 Richard Biener <rguenther@suse.de>
6718 * tree-data-ref.c (create_runtime_alias_checks): Defer
6719 and ignore overflow warnings.
6721 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6723 PR tree-optimization/85822
6724 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6727 2018-05-23 Richard Biener <rguenther@suse.de>
6729 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6730 memset constants via native_interpret_expr.
6732 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6735 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6737 (cgraph_node::create_alias): Likewise.
6738 (cgraph_node::get_availability): Check ifunc_resolver instead
6739 of looking up ifunc attribute.
6740 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6741 * varasm.c (do_assemble_alias): Likewise.
6742 (assemble_alias): Likewise.
6743 (default_binds_local_p_3): Likewise.
6744 * cgraph.h (cgraph_node): Add ifunc_resolver.
6745 (cgraph_node::only_called_directly_or_aliased_p): Return false
6747 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6749 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6750 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6751 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6752 instead of looking up ifunc attribute.
6754 2018-05-22 Luis Machado <luis.machado@linaro.org>
6756 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6758 2018-05-22 Martin Sebor <msebor@redhat.com>
6761 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6762 only when expasion succeeds.
6763 (expand_builtin_strcmp): Same.
6764 (expand_builtin_strncmp): Same.
6766 2018-05-22 Martin Sebor <msebor@redhat.com>
6768 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6770 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6771 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6773 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6774 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6776 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6777 Add check for legitimate_address.
6778 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6779 (aarch64_swap_ldrstr_operands): New.
6780 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6783 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6784 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6786 * config/aarch64/aarch64.md: New patterns to generate stp
6788 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6789 single words and double words.
6790 (load_pair_sw, load_pair_dw): Likewise.
6791 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6793 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6795 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6796 for different mode ldpstp and add peephole for merged zero stores.
6798 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6800 (aarch64_gen_store_pair): Rename calls to match new patterns.
6801 (aarch64_gen_load_pair): Rename calls to match new patterns.
6802 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6803 (load_pair<DREG:mode><DREG2:mode>): ... This.
6804 (store_pair<mode>): Rename to...
6805 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6806 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6808 (V_INT_EQUIV): Handle SImode.
6809 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6812 2018-05-22 Martin Sebor <msebor@redhat.com>
6815 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6816 or ajust the presumed bound on an operation to avoid unnecessary
6819 2018-05-22 Martin Sebor <msebor@redhat.com>
6821 PR tree-optimization/85826
6822 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6823 assuming that a DECL necesarily has a constant size.
6825 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6828 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6830 2018-05-22 Richard Biener <rguenther@suse.de>
6832 PR tree-optimization/85834
6833 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6834 non-constant and non-zero memset arguments.
6836 2018-05-22 Martin Liska <mliska@suse.cz>
6839 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6841 2018-05-22 Richard Biener <rguenther@suse.de>
6843 PR tree-optimization/85863
6844 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6845 comparisons when vectype is specified.
6846 (vectorizable_condition): Do not specify vectype for
6847 vect_is_simple_cond when SLP vectorizing.
6849 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6852 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6853 define __ibm128 as long double.
6854 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6855 as a distinct type when IEEE 128-bit support is enabled.
6856 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6857 128-bit types to use the correct functions.
6858 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6859 convert between 128-bit floating point types that have different
6860 modes but the same representation, instead of using gen_lowpart to
6862 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6864 (IFKF_reg): New attributes to give the register constraints for
6866 (extend<mode>tf2_internal): New insns to mark an explicit
6867 conversion between 128-bit floating point types that have a
6868 different mode but share the same representation.
6870 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6872 PR tree-optimization/85814
6873 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6874 a null return from get_strinfo when unsharing the next
6875 strinfo in the chain.
6877 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6880 * varasm.c (weak_finish): Clean up weak_decls.
6882 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6884 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6885 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6886 UNSPEC_UADALP values.
6887 * config/aarch64/iterators.md (ABAL): New int iterator.
6890 (sur): Add mappings for the above.
6891 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6893 (aarch64_<sur>abal<mode>_4): Likewise.
6894 (aarch64_<sur>adalp<mode>_3): Likewise.
6895 (<sur>sadv16qi): New define_expand.
6897 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6899 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6900 (*movdf_internal): Ditto.
6901 (*rcpsf2_sse): Ditto.
6902 (*rsqrtsf2_sse): Ditto.
6903 (*sqrt<mode>2_sse): Ditto.
6905 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6907 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6909 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6910 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6911 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6912 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6914 * config/aarch64/arm_neon.h: Likewise.
6915 * config/aarch64/iterators.md (VQ_I): New.
6917 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6919 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6921 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6924 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6926 * config/nds32/constraints.md (S): New constraint.
6927 * config/nds32/nds32.md (call_internal): Use constraint S.
6928 (call_value_internal): Likewise.
6929 (sibcall_internal): Likewise.
6930 (sibcall_value_internal): Likewise.
6932 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6933 Chung-Ju Wu <jasonwucj@gmail.com>
6935 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6938 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6939 Chung-Ju Wu <jasonwucj@gmail.com>
6941 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6942 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6943 (nds32_rtx_costs_impl): Simplify.
6944 (nds32_address_cost_impl): Simplify.
6945 (nds32_init_rtx_costs): New function.
6946 (nds32_rtx_costs_speed_prefer): Likewise.
6947 (nds32_rtx_costs_size_prefer): Likewise.
6948 (nds32_address_cost_speed_prefer): Likewise.
6949 (nds32_address_cost_speed_fwprop): Likewise.
6950 (nds32_address_cost_size_prefer): Likewise.
6951 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6952 * config/nds32/nds32.c (nds32_option_override): Use
6953 nds32_init_rtx_costs function.
6955 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6957 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6958 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6959 (TARGET_PIPELINE_N8): Likewise.
6960 (TARGET_PIPELINE_N10): Likewise.
6961 (TARGET_PIPELINE_N13): Likewise.
6962 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6964 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6966 * config/nds32/nds32-fpu.md: Update copyright year.
6968 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6970 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6972 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6974 * config/nds32/nds32.c
6975 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6976 * config/nds32/nds32.opt (minline-asm-r15): New option.
6978 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6980 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6982 * config/nds32/nds32.md (abssi2): New pattern.
6984 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6986 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6987 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6988 Merge insn pattern from sse_cvtsi2ss<round_name> and
6989 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6990 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6991 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6992 using SWI48 mode iterator.
6993 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6994 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6995 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6996 pattern from sse_cvttss2si<round_saeonly_name>
6997 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6998 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6999 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7000 using SWI48 mode iterator.
7001 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7002 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7003 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7004 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7005 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7006 using SWI48 mode iterator.
7007 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7008 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7009 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7010 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7011 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7012 SWI48 mode iterator.
7013 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7014 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7015 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7016 pattern from sse_cvttsd2si<round_saeonly_name>
7017 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7019 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7021 * config/nds32/nds32-md-auxiliary.c
7022 (nds32_valid_smw_lwm_base_p): Refine.
7023 (nds32_output_smw_single_word): Refine.
7024 (nds32_output_smw_double_word): New.
7025 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7027 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7029 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7030 (nds32_output_stack_pop): Refine.
7031 (nds32_expand_unaligned_load): Refine.
7032 (nds32_expand_unaligned_store): Refine.
7034 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7035 Chung-Ju Wu <jasonwucj@gmail.com>
7037 * config/nds32/constants.md: Add TP_REGNUM constant.
7038 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7039 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7041 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7042 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7043 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7044 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7045 and PIC code generation.
7046 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7048 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7050 * config/nds32/nds32.md: Support TLS and PIC.
7051 * config/nds32/nds32.c: Support TLS and PIC.
7052 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7053 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7056 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7058 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7059 mode with E_ prefix.
7061 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7062 Chung-Ju Wu <jasonwucj@gmail.com>
7064 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7065 * config/nds32/nds32-md-auxiliary.c
7066 (symbolic_reference_mentioned_p): New.
7067 (nds32_legitimize_ict_address): New.
7068 (nds32_expand_ict_move): New.
7069 (nds32_indirect_call_referenced_p): New.
7070 (nds32_symbol_binds_local_p): Delete.
7071 (nds32_long_call_p): Modify.
7072 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7073 * config/nds32/nds32-protos.h
7074 (symbolic_reference_mentioned_p): Declare.
7075 (nds32_legitimize_ict_address): Declare.
7076 (nds32_expand_ict_move): Declare.
7077 (nds32_indirect_call_referenced_p): Declare.
7078 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7079 (nds32_relax_group): Use nds32_ict_const_p as condition.
7080 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7081 (nds32_asm_file_start): Output ict_model directive in asm code.
7082 (nds32_legitimate_address_p): Consider indirect call.
7083 (nds32_print_operand): Consider indirect call.
7084 (nds32_print_operand_address): Consider indirect call.
7085 (nds32_insert_attributes): Handle "indirect_call" attribute.
7086 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7087 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7088 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7089 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7090 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7091 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7092 (TARGET_ICT_MODEL_SMALL): Define.
7093 (TARGET_ICT_MODEL_LARGE): Define.
7094 * config/nds32/nds32.md (movsi): Consider ict model.
7095 (call, call_value): Consider ict model.
7096 (sibcall, sibcall_value): Consider ict model.
7097 * config/nds32/nds32.opt (mict-model): New option.
7098 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7101 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7102 Monk Chiang <sh.chiang04@gmail.com>
7103 Jim Wilson <jimw@sifive.com>
7105 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7106 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7107 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7108 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7109 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7110 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7111 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7112 compute save_libcall_adjustment properly.
7113 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7114 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7115 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7116 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7117 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7118 (ABI_SPEC): Handle mabi=ilp32e.
7119 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7120 (RVE): Add RVE mask.
7121 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7122 <-march>: Add rv32e as an example.
7124 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7127 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7128 (intra_create_variable_infos): Handle C++ constructors.
7130 2018-05-18 Martin Liska <mliska@suse.cz>
7132 * passes.def: Remove a redundant pass.
7134 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7137 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7139 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7141 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7143 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7145 (armv2, armv2a, armv3, armv3m): Delete architectures.
7146 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7147 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7148 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7150 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7151 (*mulsidi3adddi): Likewise.
7152 (mulsidi3): Likewise.
7153 (*mulsidi3_nov6): Likewise.
7154 (umulsidi3): Likewise.
7155 (umulsidi3_nov6): Likewise.
7156 (umaddsidi4): Likewise.
7157 (*umulsidi3adddi): Likewise.
7158 (smulsi3_highpart): Likewise.
7159 (*smulsi3_highpart_nov6): Likewise.
7160 (umulsi3_highpart): Likewise.
7161 (*umulsi3_highpart_nov6): Likewise.
7162 * config/arm/arm.h (arm_arch3m): Delete.
7163 * config/arm/arm.c (arm_arch3m): Delete.
7164 (arm_option_override_internal): Update armv3-related comment.
7165 (arm_configure_build_target): Delete use of isa_bit_mode32.
7166 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7167 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7168 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7170 (mulusa3): Likewise.
7171 * config/arm/arm-protos.h (arm_arch3m): Delete.
7172 * config/arm/arm-tables.opt: Regenerate.
7173 * config/arm/arm-tune.md: Likewise.
7174 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7175 deleted architectures.
7177 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7179 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7180 (armv5t, armv5te): New features.
7181 (ARMv5, ARMv5e): Delete fgroups.
7182 (ARMv5t, ARMv5te): Adjust for above changes.
7184 (armv5, armv5e): Delete arches.
7185 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7187 (*call_reg_arm): Likewise.
7188 (*call_value_reg_armv5): Likewise.
7189 (*call_value_reg_arm): Likewise.
7190 (*call_symbol): Likewise.
7191 (*call_value_symbol): Likewise.
7192 (*sibcall_insn): Likewise.
7193 (*sibcall_value_insn): Likewise.
7195 (prefetch): Likewise.
7196 (define_split and define_peephole2 dependent on arm_arch5):
7198 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7200 (TARGET_ARM_QBIT): Likewise.
7201 (TARGET_DSP_MULTIPLY): Likewise.
7202 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7203 (arm_arch5, arm_arch5e): Delete.
7204 (arm_arch5t, arm_arch5te): Declare.
7205 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7206 (arm_arch5t): Declare.
7207 (arm_option_reconfigure_globals): Update for the above.
7208 (arm_options_perform_arch_sanity_checks): Update comment, replace
7209 use of arm_arch5 with arm_arch5t.
7210 (use_return_insn): Likewise.
7211 (arm_emit_call_insn): Likewise.
7212 (output_return_instruction): Likewise.
7213 (arm_final_prescan_insn): Likewise.
7214 (arm_coproc_builtin_available): Likewise.
7215 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7216 arm_arch5e with arm_arch5t and arm_arch5te.
7217 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7218 (arm_arch5t, arm_arch5te): Declare.
7219 * config/arm/arm-tables.opt: Regenerate.
7220 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7221 * config/arm/t-multilib: Likewise.
7222 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7223 instead of arm_arch5.
7224 (*call_reg_thumb1): Likewise.
7225 (*call_value_reg_thumb1_v5): Likewise.
7226 (*call_value_reg_thumb1): Likewise.
7227 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7229 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7231 2018-05-18 Martin Liska <mliska@suse.cz>
7233 PR gcov-profile/84846
7234 * doc/gcov.texi: Document -t option of gcov tool.
7236 2018-05-18 Martin Liska <mliska@suse.cz>
7238 PR gcov-profile/84846
7239 * gcov.c (print_usage): Add new -t option.
7240 (process_args): Handle the option.
7241 (generate_results): Use stdout as output when requested by
7244 2018-05-18 Martin Liska <mliska@suse.cz>
7246 PR gcov-profile/84846
7247 * coverage.c (coverage_init): Write PWD to .gcno file.
7248 * doc/gcov.texi: Document how working directory is printed.
7249 * gcov-dump.c (dump_gcov_file): Print PWD.
7250 * gcov.c (output_intermediate_file): Likewise.
7251 (read_graph_file): Read PWD string.
7252 (output_lines): Print PWD.
7254 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7257 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7258 for retval and return false if all args to phi are zero.
7260 2018-05-18 Richard Biener <rguenther@suse.de>
7262 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7264 (evrp_dom_walker::before_dom_children): Call it.
7266 2018-05-18 Richard Biener <rguenther@suse.de>
7268 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7269 results when processing array refs with variable index.
7271 2018-05-18 Toon Moene <toon@moene.org>
7273 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7274 directly after that of -floop-interchange. Indicate that both
7275 options are enabled by default when specifying -O3.
7277 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7279 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7280 iterator. Delete separate integer-mode vec_set<mode> expander.
7281 (aarch64_simd_vec_setv2di): Delete.
7282 (vec_setv2di): Delete.
7283 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7284 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7285 the "w, r" alternative.
7287 2018-05-18 Martin Liska <mliska@suse.cz>
7289 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7290 * tree-pass.h (make_pass_lower_switch_O0): New function.
7291 * tree-switch-conversion.c (node_has_low_bound): Remove.
7292 (node_has_high_bound): Likewise.
7293 (node_is_bounded): Likewise.
7294 (class pass_lower_switch): Make it a template type and create
7296 (pass_lower_switch::execute): Add template argument.
7297 (make_pass_lower_switch): New function.
7298 (make_pass_lower_switch_O0): New function.
7299 (do_jump_if_equal): Remove.
7300 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7301 all the hard work to tree optimization passes.
7303 2018-05-18 Martin Liska <mliska@suse.cz>
7305 * dbgcnt.c (limit_low): Renamed from limit.
7306 (limit_high): New variable.
7307 (dbg_cnt_is_enabled): Check for upper limit.
7308 (dbg_cnt): Adjust dumping.
7309 (dbg_cnt_set_limit_by_index): Add new argument for high
7311 (dbg_cnt_set_limit_by_name): Likewise.
7312 (dbg_cnt_process_single_pair): Parse new format.
7313 (dbg_cnt_process_opt): Use strtok.
7314 (dbg_cnt_list_all_counters): Remove 'value' and add
7316 * doc/invoke.texi: Document changes.
7318 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
7320 * doc/sourcebuild.texi (scalar_all_fma): Document.
7321 * tree.def (FMA_EXPR): Delete.
7322 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
7323 * internal-fn.c (ternary_direct): New macro.
7324 (expand_ternary_optab_fn): Likewise.
7325 (direct_ternary_optab_supported_p): Likewise.
7326 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
7327 * builtins.c (fold_builtin_fma): Delete.
7328 (fold_builtin_3): Don't call it.
7329 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
7330 * expr.c (expand_expr_real_2): Likewise.
7331 * fold-const.c (operand_equal_p): Likewise.
7332 (fold_ternary_loc): Likewise.
7333 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
7334 * gimple.c (DEFTREECODE): Likewise.
7335 * gimplify.c (gimplify_expr): Likewise.
7336 * optabs-tree.c (optab_for_tree_code): Likewise.
7337 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
7338 * tree-eh.c (operation_could_trap_p): Likewise.
7339 (stmt_could_throw_1_p): Likewise.
7340 * tree-inline.c (estimate_operator_cost): Likewise.
7341 * tree-pretty-print.c (dump_generic_node): Likewise.
7342 (op_code_prio): Likewise.
7343 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7344 * tree-ssa-operands.c (get_expr_operands): Likewise.
7345 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
7346 * fold-const-call.h (fold_fma): Delete.
7347 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
7348 CFN_FNMA and CFN_FNMS.
7350 * genmatch.c (combined_fn): New enum.
7351 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
7352 (commutative_op): New function.
7353 (commutate): Use it. Handle more than 2 operands.
7354 (dt_operand::gen_gimple_expr): Use commutative_op.
7355 (parser::parse_expr): Allow :c to be used with non-binary
7356 operators if the commutative operand is known.
7357 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
7358 CFN_FMS, CFN_FNMA and CFN_FNMS.
7359 (backprop::process_assign_use): Remove FMA_EXPR handling.
7360 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
7361 (gen_hsa_fma): New function.
7362 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
7363 IFN_FNMA and IFN_FNMS.
7364 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
7365 * gimple-fold.h (follow_all_ssa_edges): Declare.
7366 * gimple-fold.c (follow_all_ssa_edges): New function.
7367 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
7368 gimple_build interface and use follow_all_ssa_edges to fold the result.
7369 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
7370 instead of checking for optabs directly.
7371 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
7372 rather than FMA_EXPRs.
7373 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
7374 call to IFN_FMA instead of an FMA_EXPR.
7376 2018-05-17 Jim Wilson <jimw@sifive.com>
7378 * expr.c (do_tablejump): When converting index to Pmode, if we have a
7379 sign extended promoted subreg, and the range does not have the sign bit
7380 set, then do a sign extend.
7382 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
7383 test, check for sign extended subreg and/or constant operands, and
7384 do a sign extend in that case.
7386 2018-05-17 Steve Ellcey <sellcey@cavium.com>
7388 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
7389 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
7391 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7392 Change logics_shift_reg to logics_shift_imm.
7393 (thunderx2t99_fp_loadpair_basic): Delete.
7394 (thunderx2t99_fp_storepair_basic): Delete.
7395 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
7396 (thunderx2t99_asimd_polynomial): Delete.
7397 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
7398 and neon_fp_mul_d_scalar_q.
7399 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
7400 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
7401 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
7402 (thunderx2t99_asimd_lut): Add missing tbl types.
7403 (thunderx2t99_asimd_ext): Delete.
7404 (thunderx2t99_asimd_load1_1_mult): Delete.
7405 (thunderx2t99_asimd_load1_2_mult): Delete.
7406 (thunderx2t99_asimd_load1_ldp): New.
7407 (thunderx2t99_asimd_load1): New.
7408 (thunderx2t99_asimd_load2): Add missing *load2* types.
7409 (thunderx2t99_asimd_load3): New.
7410 (thunderx2t99_asimd_load4): New.
7411 (thunderx2t99_asimd_store1_1_mult): Delete.
7412 (thunderx2t99_asimd_store1_2_mult): Delete.
7413 (thunderx2t99_asimd_store2_mult): Delete.
7414 (thunderx2t99_asimd_store2_onelane): Delete.
7415 (thunderx2t99_asimd_store_stp): New.
7416 (thunderx2t99_asimd_store1): New.
7417 (thunderx2t99_asimd_store2): New.
7418 (thunderx2t99_asimd_store3): New.
7419 (thunderx2t99_asimd_store4): New.
7421 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
7423 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
7424 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
7426 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
7427 Segher Boessenkool <segher@kernel.crashing.org>
7430 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7433 2018-05-17 Richard Biener <rguenther@suse.de>
7435 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7436 for pruning loop and prune defs feeding only already visited PHIs.
7438 2018-05-17 Richard Biener <rguenther@suse.de>
7440 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7442 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7443 Richard Biener <rguenther@suse.de>
7445 PR tree-optimization/85793
7446 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7447 for VMAT_ELEMENTWISE.
7449 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7451 * internal-fn.h (lookup_internal_fn): Declare
7452 * internal-fn.c (lookup_internal_fn): New function.
7453 * gimple.c (gimple_build_call_from_tree): Handle calls to
7455 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7456 internal function names.
7457 * tree-pretty-print.c (dump_generic_node): Likewise.
7458 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7460 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7462 * gimple-fold.h (gimple_build): Make the function forms take
7463 combined_fn rather than built_in_function.
7464 (gimple_simplify): Likewise.
7465 * gimple-match-head.c (gimple_simplify): Likewise.
7466 * gimple-fold.c (gimple_build): Likewise.
7467 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7468 rather than gimple_build_call_internal.
7469 (get_initial_defs_for_reduction): Likewise.
7470 (vect_create_epilog_for_reduction): Likewise.
7471 (vectorizable_live_operation): Likewise.
7473 2018-05-17 Martin Liska <mliska@suse.cz>
7475 * gimple-ssa-sprintf.c (format_directive): Do not use
7476 space in between 'G_' and '('.
7478 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7481 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7482 even if the mask is not all ones.
7485 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7487 (ix86_gimple_fold_builtin): Likewise.
7490 * config/i386/i386.c: Include tree-vector-builder.h.
7491 (ix86_vector_shift_count): New function.
7492 (ix86_fold_builtin): Fold shift builtins by scalar count.
7493 (ix86_gimple_fold_builtin): Likewise.
7495 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7496 _mm512_setzero): New intrinsics.
7498 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7499 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7501 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7502 code generation for cases where splatting a value is not useful.
7503 * simplify-rtx.c (simplify_ternary_operation): Simplify
7504 vec_merge across a vec_duplicate and a paradoxical subreg forming
7505 a vector mode to a vec_concat.
7507 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7509 * config.gcc: Support "goldmont-plus".
7510 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7512 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7513 PROCESSOR_GOLDMONT_PLUS.
7514 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7515 (processor_target_table): Add "goldmont-plus".
7516 (PTA_GOLDMONT_PLUS): Define.
7517 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7518 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7519 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7520 (fold_builtin_cpu): Add "goldmont-plus".
7521 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7522 (ix86_option_override_internal): Add "goldmont-plus".
7523 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7524 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7525 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7526 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7528 2018-05-17 Richard Biener <rguenther@suse.de>
7530 PR tree-optimization/85757
7531 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7532 remove defs that only feed that PHI from further processing.
7534 2018-05-16 Jim Wilson <jimw@sifive.com>
7536 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7538 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7540 2018-05-16 Mark Wielaard <mark@klomp.org>
7542 * dwarf2out.c (count_index_strings): New function.
7543 (output_indirect_strings): Call count_index_strings and generate
7544 header for dwarf_version >= 5.
7546 2018-05-16 Mark Wielaard <mark@klomp.org>
7548 * dwarf2out.c (dwarf_FORM): New function.
7549 (set_indirect_string): Use dwarf_FORM.
7550 (reset_indirect_string): Likewise.
7551 (size_of_die): Likewise.
7552 (value_format): Likewise.
7553 (output_die): Likewise.
7554 (add_skeleton_AT_string): Likewise.
7555 (output_macinfo_op): Likewise.
7556 (index_string): Likewise.
7557 (output_index_string_offset): Likewise.
7558 (output_index_string): Likewise.
7559 (count_index_strings): Likewise.
7561 2018-05-16 Carl Love <cel@us.ibm.com>
7563 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7564 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7566 2018-05-16 Martin Jambor <mjambor@suse.cz>
7568 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7569 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7571 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7573 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7574 (fnma<mode>4): Likewise.
7575 (fms<mode>4): Likewise.
7576 (fnms<mode>4): Likewise.
7577 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7578 (aarch64_fnma<mode>4): Likewise.
7579 (aarch64_fms<mode>4): Likewise.
7580 (aarch64_fnms<mode>4): Likewise.
7581 (aarch64_fnmadd<mode>4): Likewise.
7583 2018-05-16 Jason Merrill <jason@redhat.com>
7585 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7587 2018-05-16 Richard Biener <rguenther@suse.de>
7589 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7590 (dump_stmt_cost): Declare.
7591 (add_stmt_cost): Dump cost we add.
7592 (add_stmt_costs): New function.
7593 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7595 (vect_analyze_stmt): Adjust prototype.
7596 (vectorizable_condition): Likewise.
7597 (vectorizable_live_operation): Likewise.
7598 (vectorizable_reduction): Likewise.
7599 (vectorizable_induction): Likewise.
7600 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7601 cost vector to pass to vectorizable_ and record afterwards.
7602 (vect_model_reduction_cost): Take cost vector argument and adjust.
7603 (vect_model_induction_cost): Likewise.
7604 (vectorizable_reduction): Likewise.
7605 (vectorizable_induction): Likewise.
7606 (vectorizable_live_operation): Likewise.
7607 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7608 SLP_TREE_NUMBER_OF_VEC_STMTS.
7609 (vect_analyze_slp_cost_1): Remove.
7610 (vect_analyze_slp_cost): Likewise.
7611 (vect_slp_analyze_node_operations): Take visited args and
7612 a target cost vector. Avoid processing already visited stmt sets.
7613 (vect_slp_analyze_operations): Use a local cost vector to gather
7614 costs and register those of non-discarded instances.
7615 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7616 (vect_schedule_slp_instance): Remove copying of
7617 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7619 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7620 adding cost. Record cost entry location.
7621 (vect_prologue_cost_for_slp_op): Function to compute cost of
7622 a constant or invariant generated for SLP vect in the prologue,
7623 split out from vect_analyze_slp_cost_1.
7624 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7625 (vect_model_promotion_demotion_cost): Likewise.
7626 (vect_model_store_cost): Likewise, make static.
7627 (vect_model_load_cost): Likewise.
7628 (vectorizable_bswap): Add cost vector arg and adjust.
7629 (vectorizable_call): Likewise.
7630 (vectorizable_simd_clone_call): Likewise.
7631 (vectorizable_conversion): Likewise.
7632 (vectorizable_assignment): Likewise.
7633 (vectorizable_shift): Likewise.
7634 (vectorizable_operation): Likewise.
7635 (vectorizable_store): Likewise.
7636 (vectorizable_load): Likewise.
7637 (vectorizable_condition): Likewise.
7638 (vectorizable_comparison): Likewise.
7639 (can_vectorize_live_stmts): Likewise.
7640 (vect_analyze_stmt): Likewise.
7641 (vect_transform_stmt): Adjust calls to vectorizable_*.
7642 * tree-vectorizer.c: Include gimple-pretty-print.h.
7643 (dump_stmt_cost): New function.
7645 2018-05-16 Richard Biener <rguenther@suse.de>
7647 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7648 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7649 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7650 (check_name): New callback.
7651 (dse_classify_store): Track cycles via a visited bitmap of PHI
7652 defs and simplify handling of in-loop and across loop dead stores
7653 and properly fail for loop-variant refs. Handle byte-tracking with
7654 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7657 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7659 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7660 (vect_get_mask_type_for_stmt): Likewise.
7661 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7663 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7664 to determine the statement's vector type and the vector type that
7665 should be used for calculating nunits. Deal with cases in which
7666 the type has to be deferred.
7667 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7668 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7669 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7670 (vect_determine_vf_for_stmt): New functions, split out from...
7671 (vect_determine_vectorization_factor): ...here.
7672 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7673 (vect_get_mask_type_for_stmt): New functions, split out from
7674 vect_determine_vectorization_factor.
7676 2018-05-16 Richard Biener <rguenther@suse.de>
7678 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7679 verify the [VEC_]COND_EXPR embedded comparison.
7681 2018-05-15 Martin Sebor <msebor@redhat.com>
7683 PR tree-optimization/85753
7684 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7685 RECORD_TYPE in addition to ARRAY_TYPE.
7687 2018-05-15 Martin Sebor <msebor@redhat.com>
7690 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7692 2018-05-15 Richard Biener <rguenther@suse.de>
7694 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7695 add by_clobber_p one. Change algorithm to collect all defs
7696 representing uses we need to walk and try reducing them to
7697 a single one before failing.
7698 (dse_dom_walker::dse_optimize_stmt): Adjust.
7700 2018-05-13 Mark Wielaard <mark@klomp.org>
7702 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7703 (size_of_loc_descr): Likewise.
7704 (output_loc_operands): Likewise.
7705 (output_loc_operands_raw): Likewise.
7706 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7707 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7708 (hash_loc_operands): Likewise.
7709 (compare_loc_operands): Likewise.
7711 2018-05-14 Mark Wielaard <mark@klomp.org>
7713 * dwarf2out.c (count_index_addrs): New function.
7714 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7716 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7718 PR tree-optimization/83648
7719 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7720 return value as malloc candidate.
7722 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7725 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7726 param as true in call to suggest_attribute.
7728 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7730 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7731 -mreadonly-in-sdata.
7733 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7735 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7737 (aarch64_crypto_aesd_fused): Likewise.
7739 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7741 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7742 (movsi_aarch64): Likewise.
7743 (load_pairsi): Likewise.
7744 (load_pairdi): Likewise.
7745 (store_pairsi): Likewise.
7746 (store_pairdi): Likewise.
7747 (load_pairsf): Likewise.
7748 (load_pairdf): Likewise.
7749 (store_pairsf): Likewise.
7750 (store_pairdf): Likewise.
7751 (zero_extend): Likewise.
7752 (trunc): Swap alternatives.
7753 (fcvt_target): Add '?' to prefer w over r.
7755 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7758 * config/i386/i386.md: Disallow non-commutative arithmetics in
7759 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7760 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7761 in the peephole2 before it.
7763 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7765 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7766 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7767 (ix86_handle_option): Handle -mcldemote.
7768 * config.gcc: New header.
7769 * config/i386/cldemoteintrin.h: New file.
7770 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7771 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7773 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7774 OPTION_MASK_ISA_CLDEMOTE.
7775 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7776 (ix86_valid_target_attribute_inner_p): Ditto.
7777 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7778 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7779 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7780 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7781 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7783 * config/i386/i386.opt: Add -mcldemote.
7784 * config/i386/x86intrin.h: New header.
7785 * doc/invoke.texi: Add -mcldemote.
7787 2018-05-14 Richard Biener <rguenther@suse.de>
7789 * doc/match-and-simplify.texi: Adjust :s documentation.
7791 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7793 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7794 intended memcpy size.
7795 (REORDER_45): Likewise.
7797 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7799 * sort.cc: New file.
7800 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7801 * vec.c (qsort_chk): Use gcc_qsort.
7802 * Makefile.in (OBJS-libcommon): Add sort.o.
7803 (build/sort.o): New target. Use it...
7804 (BUILD_RTL): ... here, and...
7805 (build/gencfn-macros): ... here, and...
7806 (build/genmatch): ... here.
7808 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7809 Chung-Ju Wu <jasonwucj@gmail.com>
7811 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7812 * config/nds32/nds32-graywolf.md: New file.
7813 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7814 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7816 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7817 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7818 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7819 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7820 * config/nds32/pipelines.md: Include n15 settings.
7822 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7823 Chung-Ju Wu <jasonwucj@gmail.com>
7825 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7826 * config/nds32/nds32-n13.md: New file.
7827 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7828 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7830 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7831 * config/nds32/nds32.md (pipeline_model): Add n13.
7832 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7833 * config/nds32/pipelines.md: Include n13 settings.
7835 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7836 Chung-Ju Wu <jasonwucj@gmail.com>
7838 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7839 * config/nds32/nds32-n10.md: New file.
7840 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7841 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7843 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7844 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7845 * config/nds32/nds32.md (pipeline_model): Add n10.
7846 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7847 * config/nds32/pipelines.md: Include n10 settings.
7849 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7850 Kito Cheng <kito.cheng@gmail.com>
7851 Chung-Ju Wu <jasonwucj@gmail.com>
7853 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7854 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7855 Add enum values for DSP extension instructions.
7856 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7858 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7859 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7861 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7862 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7863 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7864 * config/nds32/nds32-intrinsic.md: Likewise.
7865 * config/nds32/nds32_intrinsic.h: Likewise.
7866 * config/nds32/nds32-md-auxiliary.c: Likewise.
7867 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7868 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7869 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7870 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7871 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7872 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7873 TYPE_DMAC in switch statement.
7874 * config/nds32/nds32.c: New checking and implementation for DSP
7875 extension instructions.
7876 * config/nds32/nds32.h: Likewise.
7877 * config/nds32/nds32.md: Likewise.
7878 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7879 * config/nds32/predicates.md: Implement new predicates for DSP
7882 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7884 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7885 Reformat alternatives and attributes so it is easier to identify
7886 which constraints/attributes go with which instruction.
7887 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7888 (mov<mode>_softfloat32, FMOVE64): Likewise.
7889 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7890 (mov<mode>_softfloat64, FMOVE64): Likewise.
7892 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7894 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7896 (Basic PowerPC Built-in Functions): The new name of the
7897 subsection previously known as "PowerPC Built-in Functions".
7898 (Basic PowerPC Built-in Functions Available on all Configurations):
7900 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7901 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7902 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7903 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7905 2018-05-11 Martin Jambor <mjambor@suse.cz>
7908 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7911 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7914 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7916 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7918 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7919 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7920 (ix86_handle_option): Handle -mwaitpkg.
7921 * config.gcc: New header.
7922 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7923 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7924 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7926 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7927 OPTION_MASK_ISA_WAITPKG.
7928 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7929 (ix86_option_override_internal): Add PTA_WAITPKG.
7930 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7931 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7932 IX86_BUILTIN_TPAUSE.
7933 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7934 __builtin_ia32_umwait and __builtin_ia32_tpause.
7935 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7936 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7937 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7938 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7939 UNSPECV_TPAUSE): New.
7940 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7941 * config/i386/i386.opt: Add -mwaitpkg.
7942 * config/i386/waitpkgintrin.h: New file.
7943 * config/i386/x86intrin.h: New header.
7944 * doc/invoke.texi: Add -mwaitpkg.
7946 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7949 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7951 (cortex-m0): Use armv6s-m isa.
7952 (cortex-m0plus): Likewise.
7953 (cortex-m1): Likewise.
7954 (cortex-m0.small-multiply): Likewise.
7955 (cortex-m0plus.small-multiply): Likewise.
7956 (cortex-m1.small-multiply): Likewise.
7958 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7959 Jakub Jelinek <jakub@redhat.com>
7961 PR tree-optimization/85692
7962 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7963 source permute as well.
7965 2018-05-11 Martin Liska <mliska@suse.cz>
7968 * doc/extend.texi: Document LLVM style format for no_sanitize
7971 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7973 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7974 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7975 (mode_supports_vsx_dform_quad): Likewise.
7976 (mode_supports_vmx_dform): Move these functions to be next to the
7977 other mode_supports functions.
7978 (mode_supports_dq_form): Likewise.
7979 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7980 mode_supports_dq_form.
7981 (reg_offset_addressing_ok_p): Likewise.
7982 (offsettable_ok_by_alignment): Likewise.
7983 (rs6000_legitimate_offset_address_p): Likewise.
7984 (legitimate_lo_sum_address_p): Likewise.
7985 (rs6000_legitimize_address): Likewise.
7986 (rs6000_legitimize_reload_address): Likewise.
7987 (rs6000_secondary_reload_inner): Likewise.
7988 (rs6000_preferred_reload_class): Likewise.
7989 (rs6000_output_move_128bit): Likewise.
7991 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7993 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7994 Generate SImode target register for null target.
7995 <case IX86_BUILTIN_XGETBV>: Ditto.
7996 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7997 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7999 2018-05-10 Carl Love <cel@us.ibm.com>
8001 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8002 dcbtt and dcbtstt if operands[2] is 0.
8004 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8007 * config/i386/sse.md (usadv64qi): New expander.
8009 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8011 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8012 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8013 -maltivec=be support.
8014 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8015 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8016 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8017 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8018 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8019 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8020 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8021 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8022 altivec_vsumsws): Adjust.
8023 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8024 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8025 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8027 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8028 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8029 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8030 (altivec_lve<VI_char>x): Delete expand.
8031 (*altivec_lve<VI_char>x_internal): Rename to...
8032 (altivec_lve<VI_char>x): ... this.
8033 (altivec_lvxl_<mode>): Delete expand.
8034 (*altivec_lvxl_<mode>_internal): Rename to ...
8035 (altivec_lvxl_<mode>): ... this.
8036 (altivec_stvxl_<mode>): Delete expand.
8037 (*altivec_stvxl_<mode>_internal): Rename to ...
8038 (altivec_stvxl_<mode>): ... this.
8039 (altivec_stve<VI_char>x): Delete expand.
8040 (*altivec_stve<VI_char>x_internal): Rename to ...
8041 (altivec_stve<VI_char>x): ... this.
8042 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8043 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8044 reduc_plus_scal_<mode>): Adjust.
8045 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8047 (rs6000_cpu_cpp_builtins): Adjust.
8048 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8049 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8050 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8051 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8052 -maltivec=be support.
8053 (rs6000_split_vec_extract_var): Adjust.
8054 (rs6000_split_v4si_init): Adjust.
8055 (swap_selector_for_mode): Delete.
8056 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8057 altivec_expand_stvex_be): Delete.
8058 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8059 -maltivec=be support.
8060 (rs6000_gimple_fold_builtin): Ditto.
8061 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8063 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8064 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8065 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8066 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8067 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8068 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8069 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8070 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8071 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8072 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8073 anonymous split): Adjust.
8074 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8075 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8077 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8079 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8080 when --with-gxx-include-dir is also specified.
8081 * configure: Regenerate.
8083 2018-05-09 Jim Wilson <jimw@sifive.com>
8086 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8087 * config/riscv/t-withmultilib: New.
8088 * config/riscv/withmultilib.h: New.
8089 * doc/install.texi: Document RISC-V --with-multilib-list support.
8091 2018-05-09 Richard Biener <rguenther@suse.de>
8093 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8095 (vect_bb_vectorization_profitable_p): Adjust. Compute
8096 actual scalar cost using the cost vector and the add_stmt_cost
8099 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8101 PR rtl-optimization/85645
8102 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8103 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8105 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8107 PR rtl-optimization/85645
8108 * shrink-wrap.c (spread_components): Return a boolean saying if
8109 anything was changed.
8110 (try_shrink_wrapping_separate): Iterate spread_components until
8111 nothing changes anymore.
8113 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8115 PR rtl-optimization/85645
8116 * regrename.c (build_def_use): Also kill the chains that include the
8117 destination of a REG_CFA_REGISTER note.
8119 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8121 PR rtl-optimization/85645
8122 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8123 insn that has a REG_CFA_REGISTER note.
8125 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8127 * cfgexpand.c (expand_clobber): New function.
8128 (expand_gimple_stmt_1): Use it.
8129 * tree-vect-stmts.c (vect_clobber_variable): New function,
8131 (vectorizable_simd_clone_call): ...here.
8132 (vectorizable_store): Emit a clobber either side of an
8133 IFN_STORE_LANES sequence.
8134 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8136 2018-05-09 Tom de Vries <tom@codesourcery.com>
8139 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8140 (define_insn "trap_if_false"): Add exit after trap.
8142 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8144 PR rtl-optimization/85638
8145 * bb-reorder.c: Include common/common-target.h.
8146 (create_forwarder_block): New function extracted from...
8147 (fix_up_crossing_landing_pad): ...here. Rename into...
8148 (dw2_fix_up_crossing_landing_pad): ...this.
8149 (sjlj_fix_up_crossing_landing_pad): New function.
8150 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8151 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8152 from both partitions and exit the loop after one iteration.
8154 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8157 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8159 (Basic PowerPC Built-in Functions): The new name of the
8160 subsection previously known as "PowerPC Built-in Functions".
8161 (Basic PowerPC Built-in Functions Available on all Configurations):
8163 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8165 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8167 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8169 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8172 2018-05-08 Jim Wilson <jimw@sifive.com>
8174 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8175 (LD_EMUL_SUFFIX): New.
8176 (LINK_SPEC): Use it.
8178 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8180 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8182 (Basic PowerPC Built-in Functions): The new name of the
8183 subsection previously known as "PowerPC Built-in Functions".
8184 (Basic PowerPC Built-in Functions Available on all Configurations):
8186 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8188 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8190 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8192 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8195 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8198 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8199 after cmpelim optimization.
8201 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8203 * config.gcc: Support "goldmont".
8204 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8205 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8207 * config/i386/i386.c (m_GOLDMONT): Define.
8208 (processor_target_table): Add "goldmont".
8209 (PTA_GOLDMONT): Define.
8210 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8211 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8212 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8213 (fold_builtin_cpu): Add "goldmont".
8214 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8215 (ix86_option_override_internal): Add "goldmont".
8216 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8217 (processor_type): Add PROCESSOR_GOLDMONT.
8218 * config/i386/i386.md: Add CPU "glm".
8219 * config/i386/glm.md: New file.
8220 * config/i386/x86-tune.def: Add m_GOLDMONT.
8221 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8223 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8226 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8228 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8229 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8230 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8233 * config/i386/i386.c (ix86_fold_builtin): Handle
8234 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8237 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8238 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8240 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8243 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8244 (check_arch): Likewise.
8245 (check_fpu): Return the result rather than printing it.
8246 (end arch): Fix operator precedence.
8247 (end cpu): Likewise.
8248 (END): Print the result from check_fpu.
8250 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8251 Alan Hayward <alan.hayward@arm.com>
8252 David Sherwood <david.sherwood@arm.com>
8254 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8255 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8256 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8257 (*fcmuo<mode>_and): New patterns.
8259 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8261 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8262 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8263 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8264 (cmp_op, sve_imm_con): New code attributes.
8265 (SVE_COND_INT_CMP, imm_con): Delete.
8266 (cmp_op): Remove above unspecs from int attribute.
8267 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8269 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8270 comparison-specific unspecs.
8271 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8272 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8273 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8274 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8275 (*vec_fcm<cmp_op><mode>): Rename to...
8276 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8277 (*vec_fcmuo<mode>): Rename to...
8278 (*fcmuo<mode>): ...this and adjust likewise.
8279 (*pred_fcm<cmp_op><mode>): New pattern.
8280 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8281 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8283 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8285 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8286 (aarch64_emit_sve_predicated_cond): New function.
8287 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8288 (aarch64_emit_unspec_cond_or): Replace with...
8289 (aarch64_emit_sve_or_conds): ...this new function. Use
8290 aarch64_emit_sve_ptrue_op for the individual comparisons and
8291 aarch64_emit_binop to OR them together.
8292 (aarch64_emit_inverted_unspec_cond): Replace with...
8293 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8294 aarch64_emit_sve_ptrue_op for the comparison and
8295 aarch64_emit_unop to invert the result.
8296 (aarch64_expand_sve_vec_cmp_float): Update after the above
8297 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8299 2018-05-07 Nathan Sidwell <nathan@acm.org>
8301 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8302 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8303 (Backwards Compatibility): Likewise.
8305 2018-05-07 Luis Machado <luis.machado@linaro.org>
8309 2018-05-07 Luis Machado <luis.machado@linaro.org>
8311 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8312 <prefetch_dynamic_strides>: New const bool field.
8313 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8314 prefetch_dynamic_strides.
8315 (exynosm1_prefetch_tune): Likewise.
8316 (thunderxt88_prefetch_tune): Likewise.
8317 (thunderx_prefetch_tune): Likewise.
8318 (thunderx2t99_prefetch_tune): Likewise.
8319 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8321 (aarch64_override_options_internal): Update to set
8322 PARAM_PREFETCH_DYNAMIC_STRIDES.
8323 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8324 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8325 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8326 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8327 prefetch-dynamic-strides setting.
8329 2018-05-07 Luis Machado <luis.machado@linaro.org>
8331 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8332 <minimum_stride>: New const int field.
8333 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8334 minimum_stride field.
8335 (exynosm1_prefetch_tune): Likewise.
8336 (thunderxt88_prefetch_tune): Likewise.
8337 (thunderx_prefetch_tune): Likewise.
8338 (thunderx2t99_prefetch_tune): Likewise.
8339 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8340 (aarch64_override_options_internal): Update to set
8341 PARAM_PREFETCH_MINIMUM_STRIDE.
8342 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8343 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8344 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8345 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8346 stride is constant and is below the minimum stride threshold.
8348 2018-05-07 Luis Machado <luis.machado@linaro.org>
8350 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
8353 2018-05-07 Luis Machado <luis.machado@linaro.org>
8355 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8356 <prefetch_dynamic_strides>: New const bool field.
8357 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8358 prefetch_dynamic_strides.
8359 (exynosm1_prefetch_tune): Likewise.
8360 (thunderxt88_prefetch_tune): Likewise.
8361 (thunderx_prefetch_tune): Likewise.
8362 (thunderx2t99_prefetch_tune): Likewise.
8363 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8365 (aarch64_override_options_internal): Update to set
8366 PARAM_PREFETCH_DYNAMIC_STRIDES.
8367 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8368 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8369 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8370 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8371 prefetch-dynamic-strides setting.
8373 2018-05-07 Luis Machado <luis.machado@linaro.org>
8375 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8376 <minimum_stride>: New const int field.
8377 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8378 minimum_stride field.
8379 (exynosm1_prefetch_tune): Likewise.
8380 (thunderxt88_prefetch_tune): Likewise.
8381 (thunderx_prefetch_tune): Likewise.
8382 (thunderx2t99_prefetch_tune): Likewise.
8383 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8384 (aarch64_override_options_internal): Update to set
8385 PARAM_PREFETCH_MINIMUM_STRIDE.
8386 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8387 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8388 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8389 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8390 stride is constant and is below the minimum stride threshold.
8392 2018-05-06 Jakub Jelinek <jakub@redhat.com>
8395 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
8396 the type is addressable. Don't force op into register if it has
8399 2018-05-05 Roland McGrath <mcgrathr@google.com>
8402 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
8403 any section for which we don't know a specific type it should have,
8404 regardless of name. Previously this was done only for the exact
8405 names ".init_array", ".fini_array", and ".preinit_array".
8406 (default_elf_asm_named_section): Add comment about
8407 relationship with default_section_type_flags and SECTION_NOTYPE.
8408 (get_section): Don't consider it a type conflict if one side has
8409 SECTION_NOTYPE and the other doesn't, as long as neither has the
8410 SECTION_BSS et al used in the default_section_type_flags logic.
8412 2018-05-05 Tom de Vries <tom@codesourcery.com>
8415 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
8416 (workaround_barsyncs): New function.
8417 (nvptx_reorg): Use workaround_barsyncs.
8418 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8419 (define_expand "nvptx_membar_cta"): New define_expand.
8420 (define_insn "*nvptx_membar_cta"): New insn.
8422 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
8424 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
8425 To improve optimization opportunities.
8426 * builtin-types.def: The new needed builtin types for the above.
8428 2018-05-04 Richard Biener <rguenther@suse.de>
8430 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8431 * gimple-ssa-store-merging.c
8432 (imm_store_chain_info::output_merged_store): Remove redundant create,
8433 release split_store vector contents on failure.
8434 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8435 scalar stmt vector on cache hit.
8437 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8439 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8441 * config.gcc (powerpc-xilinx-eabi*): Remove.
8442 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8444 (fusion_addis_mem_combo_load): Ditto.
8445 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8447 (rs6000_cpu_cpp_builtins): Ditto.
8448 * config/rs6000/rs6000-linux.c
8449 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8450 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8451 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8453 (rs6000_setup_reg_addr_masks): Ditto.
8454 (rs6000_init_hard_regno_mode_ok): Ditto.
8455 (rs6000_option_override_internal): Ditto.
8456 (legitimate_lo_sum_address_p): Ditto.
8457 (rs6000_legitimize_address): Ditto.
8458 (rs6000_legitimize_reload_address): Ditto.
8459 (rs6000_legitimate_address_p): Ditto.
8460 (abi_v4_pass_in_fpr): Ditto.
8461 (setup_incoming_varargs): Ditto.
8462 (rs6000_gimplify_va_arg): Ditto.
8463 (rs6000_split_multireg_move): Ditto.
8464 (rs6000_savres_strategy): Ditto.
8465 (rs6000_emit_prologue_components): Ditto.
8466 (rs6000_emit_epilogue_components): Ditto.
8467 (rs6000_emit_prologue): Ditto.
8468 (rs6000_emit_epilogue): Ditto.
8469 (rs6000_elf_file_end): Ditto.
8470 (rs6000_function_value): Ditto.
8471 (rs6000_libcall_value): Ditto.
8472 * config/rs6000/rs6000.h: Ditto.
8473 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8474 (TARGET_MINMAX): ... this. New.
8475 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8476 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8477 (*movsi_internal1_single): Delete.
8478 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8479 mfpu=, mxilinx-fpu): Delete.
8480 * config/rs6000/singlefp.h: Delete.
8481 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8482 * config/rs6000/t-rs6000: Ditto.
8483 * config/rs6000/t-xilinx: Delete.
8484 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
8485 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
8486 (VStype_simple): Delete.
8487 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8488 * config/rs6000/xfpu.h: Delete.
8489 * config/rs6000/xfpu.md: Delete.
8490 * config/rs6000/xilinx.h: Delete.
8491 * config/rs6000/xilinx.opt: Delete.
8492 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8493 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8495 2018-05-04 Tom de Vries <tom@codesourcery.com>
8498 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8501 2018-05-04 Richard Biener <rguenther@suse.de>
8504 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8505 (expand_complex_div_wide): Likewise.
8506 (expand_complex_operations_1): Likewise.
8507 (expand_complex_libcall): Preserve EH info of the original stmt.
8508 (tree_lower_complex): Handle removed blocks.
8509 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8510 on complex multiplication and division libcall builtins.
8512 2018-05-04 Richard Biener <rguenther@suse.de>
8515 * fold-const.c (negate_expr_p): Restrict negation of operand
8516 zero of a division to when we know that can happen without
8518 (fold_negate_expr_1): Likewise.
8520 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8523 * real.h (real_nextafter): Declare.
8524 * real.c (real_nextafter): New function.
8525 * fold-const-call.c (fold_const_nextafter): New function.
8526 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8527 CASE_CFN_NEXTTOWARD.
8528 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8529 even when arg1_mode is different from arg0_mode.
8531 2018-05-03 Nathan Sidwell <nathan@acm.org>
8533 * doc/extend.texi (Deprecated Features): Remove
8535 (Backwards Compatibility): Likewise.
8536 * doc/invoke.texi (C++ Language Options): Likewise.
8537 (C++ Dialect Options): Likewise.
8539 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8542 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8543 _mm512_mask_mullox_epi64): New intrinsics.
8545 2018-05-03 Tom de Vries <tom@codesourcery.com>
8548 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8549 dump files): Add offload-tree.
8551 2018-05-03 Richard Biener <rguenther@suse.de>
8553 PR tree-optimization/85615
8554 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8555 to loops not nested in BBs loop father to avoid creating multi-entry
8558 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8560 PR tree-optimization/70291
8561 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8562 arguments. Change return type to tree. Emit libcall as a new
8563 statement rather than replacing existing one when inplace_p is true.
8564 (expand_complex_multiplication_components): New function.
8565 (expand_complex_multiplication): Expand floating-point complex
8566 multiplication using the above.
8567 (expand_complex_division): Rename inner_type parameter to type.
8568 Update expand_complex_libcall call-site.
8569 (expand_complex_operations_1): Update expand_complex_multiplication
8570 and expand_complex_division call-sites.
8572 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8575 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8576 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8577 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8578 the highest significant bit of the shift count mask is clear. In
8579 check whether and[sq]i3 is needed verify that all significant bits
8580 of the shift count other than the highest are set.
8582 2018-05-02 Tom de Vries <tom@codesourcery.com>
8585 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8586 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8587 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8588 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8589 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8590 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8591 __builtin_goacc_parlevel_size.
8593 2018-05-02 Richard Biener <rguenther@suse.de>
8595 PR tree-optimization/85597
8596 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8597 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8600 2018-05-02 Tom de Vries <tom@codesourcery.com>
8603 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8604 dump files): Add ltrans-tree.
8606 2018-05-02 Tom de Vries <tom@codesourcery.com>
8609 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8610 dump files): Add wpa-ipa.
8612 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8614 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8615 powerpc*-*-linux*paired* target.
8616 * config/rs6000/750cl.h: Delete.
8617 * config/rs6000/paired.h: Delete.
8618 * config/rs6000/paired.md: Delete.
8619 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8621 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8622 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8623 comment. Remove paired float support.
8624 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8625 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8627 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8628 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8630 * config/rs6000/rs6000.c: Remove paired float support.
8631 (paired_expand_vector_init, paired_expand_vector_move,
8632 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8633 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8634 paired_expand_builtin, paired_expand_predicate_builtin,
8635 paired_init_builtins): Delete.
8636 * config/rs6000/rs6000.h: Remove paired float support.
8637 * config/rs6000/rs6000.md: Remove paired float support.
8638 (move_from_CR_ov_bit): Delete.
8639 * config/rs6000/rs6000.opt (mpaired): Delete.
8640 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8641 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8643 2018-05-02 Richard Biener <rguenther@suse.de>
8646 * gimplify.c (gimplify_save_expr): When in SSA form allow
8647 SAVE_EXPRs to compute to SSA vars.
8649 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8652 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8653 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8654 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8655 clobber operands[2], instead use a new pseudo. Formatting fixes.
8657 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8659 PR tree-optimization/85586
8660 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8661 exit early for statements in the same group if the accesses are
8664 2018-05-02 Tom de Vries <tom@codesourcery.com>
8667 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8670 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8672 PR tree-optimization/85143
8673 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8675 2018-05-01 Tom de Vries <tom@codesourcery.com>
8678 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8679 not found" error message.
8681 2018-05-01 Tom de Vries <tom@codesourcery.com>
8684 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8685 * vec.c (test_ordered_remove_if): New function.
8686 (vec_c_tests): Call test_ordered_remove_if.
8687 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8688 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8689 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8690 VEC_ORDERED_REMOVE_IF.
8692 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8694 PR tree-optimization/82665
8695 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8696 pointer subtraction where arguments come from a memchr call.
8698 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8700 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8701 --push-state --as-needed and --pop-state instead of --as-needed and
8702 --no-as-needed if ld supports it.
8703 * configure: Regenerated.
8706 * doc/install.texi2html: Replace _002d with - and _002a with * in
8707 generated html files using sed.
8709 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8712 * gcc-rich-location.c (blank_line_before_p): New function.
8713 (use_new_line): New function.
8714 (gcc_rich_location::add_fixit_insert_formatted): New function.
8715 * gcc-rich-location.h
8716 (gcc_rich_location::add_fixit_insert_formatted): New function.
8718 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8720 * selftest.c (assert_streq): Rename "expected" and "actual" to
8721 "val1" and "val2". Extend NULL-handling to cover both inputs
8722 symmetrically, while still requiring both to be non-NULL for a pass.
8723 * selftest.h (assert_streq): Rename "expected" and "actual" to
8725 (ASSERT_EQ): Likewise.
8726 (ASSERT_EQ_AT): Likewise.
8727 (ASSERT_KNOWN_EQ): Likewise.
8728 (ASSERT_KNOWN_EQ_AT): Likewise.
8729 (ASSERT_NE): Likewise.
8730 (ASSERT_MAYBE_NE): Likewise.
8731 (ASSERT_MAYBE_NE_AT): Likewise.
8732 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8733 the assertion to pass.
8734 (ASSERT_STREQ_AT): Likewise.
8736 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8738 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8739 interaction with -pie.
8741 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8743 * selftest.h: Fix alphabetization of per-source-file selftest
8746 2018-04-30 Jason Merrill <jason@redhat.com>
8748 PR c++/61982 - dead stores to destroyed objects.
8749 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8752 2018-04-30 Jason Merrill <jason@redhat.com>
8754 * tree.c (build_clobber): New.
8755 * tree.h: Declare it.
8756 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8758 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8760 * diagnostic-show-locus.c (layout::layout): Update for
8761 location_get_source_line returning a char_span.
8762 (struct char_span): Move to input.h.
8763 (struct correction): Update for fields in char_span becoming
8765 (struct source_line): Update for location_get_source_line
8766 returning a char_span.
8767 (layout::print_line): Likewise.
8768 * edit-context.c (edited_file::print_content): Likewise.
8769 (edited_file::print_diff_hunk): Likewise.
8770 (edited_file::print_run_of_changed_lines): Likewise.
8771 (edited_file::get_num_lines): Likewise.
8772 (edited_line::edited_line): Likewise.
8773 * final.c (asm_show_source): Likewise.
8774 * input.c (location_get_source_line): Convert return type
8775 from const char * to char_span, losing the final "line_len"
8777 (dump_location_info): Update for the above.
8778 (get_substring_ranges_for_loc): Likewise. Use a char_span
8779 when handling the literal within the line.
8780 (test_reading_source_line): Update for location_get_source_line
8781 returning a char_span.
8782 * input.h (class char_span): Move here from
8783 diagnostic-show-locus.c, converting from a struct to a class.
8784 Make data members private.
8785 (char_span::operator bool): New.
8786 (char_span::length): New.
8787 (char_span::get_buffer): New.
8788 (char_span::operator[]): New.
8789 (char_span::subspan): Make const.
8790 (char_span::xstrdup): New.
8791 (location_get_source_line): Convert return type from const char *
8792 to char_span, losing the final "line_size" param.
8794 2018-04-30 Jan Hubicka <jh@suse.cz>
8796 * lto-wrapper.c (ltrans_priorities): New static var.
8797 (cmp_priority): New.
8798 (run_gcc): Read priorities and if doing parallel build order
8799 the Makefile by them.
8801 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8803 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8805 2018-04-30 Richard Biener <rguenther@suse.de>
8807 * tree-cfg.c (verify_address): Remove base argument, add
8808 flag whether to check TREE_ADDRESSABLE and do that.
8809 (verify_expr): Remove.
8810 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8811 (verify_gimple_assign_single): Likewise.
8812 (verify_gimple_switch): Likewise.
8813 (verify_expr_location_1): Dereference tp once. Add (disabled)
8814 piece from verify_expr.
8815 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8817 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8819 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8821 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8823 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8824 (small_data_pattern): Likewise.
8825 (arc_rewrite_small_data): Likewise.
8826 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8827 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8828 (get_symbol_alignment): New function.
8829 (legitimate_small_data_address_p): Likewise.
8830 (legitimate_scaled_address): Update, call
8831 legitimate_small_data_address_p.
8832 (output_sdata): New static variable.
8833 (arc_print_operand): Update how we handle small data operands.
8834 (arc_print_operand_address): Likewise.
8835 (arc_legitimate_address_p): Update, use
8836 legitimate_small_data_address_p.
8837 (arc_rewrite_small_data_p): Remove.
8838 (arc_rewrite_small_data_1): Likewise.
8839 (arc_rewrite_small_data): Likewise.
8840 (small_data_pattern): Likewise.
8841 (compact_sda_memory_operand): Update to use
8842 legitimate_small_data_address_p and get_symbol_alignment.
8843 (prepare_move_operands): Don't rewite sdata pattern.
8844 (prepare_extend_operands): Remove.
8845 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8847 (zero_extendqisi2): Likewise.
8848 (zero_extendhisi2): Likewise.
8849 (extendqihi2): Likewise.
8850 (extendqisi2): Likewise.
8851 (extendhisi2): Likewise.
8855 * config/arc/constraints.md (Usd): Change it to memory constraint.
8857 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8859 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8860 as source of std instructions.
8861 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8862 allow 6-bit constants as source for store instructions.
8863 (movdi_insn): Update instruction pattern to allow 6-bit constants
8864 as source for store instructions.
8866 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8868 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8870 2018-04-30 Nathan Sidwell <nathan@acm.org>
8871 Sandra Loosemore <sandra@codesourcery.com>
8873 * dumpfile.c (dump_open): Allow '-' for stdout.
8874 * doc/invoke.texi (Developer Options): Document dump filename
8875 determination early. Document stdin/stdout selection.
8877 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8879 Microblaze Target: PIC data text relative
8881 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8882 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8884 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8885 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8886 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8887 New addressing mode for data-text relative position indepenedent code.
8888 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8889 'ADDRESS_SYMBOLIC_TXT_REL'.
8890 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8891 (microblaze_legitimate_pic_operand): Exclude function calls from
8892 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8893 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8895 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8896 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8897 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8898 for 'address + offset'.
8899 (microblaze_expand_prologue): Add new function prologue call for
8901 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8902 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8903 table in case of TARGET_PIC_DATA_TEXT_REL.
8904 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8905 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8906 Add new macros 'UNSPEC_TEXT',
8907 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8908 + exclude function calls from 'UNSPEC_PLT' in case of data text
8910 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8911 new target hook for generating address diff vector tables in case of
8913 * doc/tm.texi : Regenerate.
8914 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8915 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8916 of addr diff vector generation.
8917 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8918 target hook definition.
8919 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8920 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8921 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8922 option for data text relative.
8924 2018-04-30 Richard Biener <rguenther@suse.de>
8926 * tree-chrec.h (evolution_function_is_constant_p): Remove
8928 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8930 2018-04-30 Richard Biener <rguenther@suse.de>
8933 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8935 2018-04-30 Richard Biener <rguenther@suse.de>
8937 PR tree-optimization/28364
8938 PR tree-optimization/85275
8939 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8940 copying first exit test.
8942 2018-04-28 Mark Wielaard <mark@klomp.org>
8944 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8946 (dwarf_AT): Handle DW_AT_addr_base.
8947 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8949 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8952 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8953 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8954 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8955 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8957 2018-04-28 Richard Biener <rguenther@suse.de>
8959 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8960 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8961 to reflect use. Only add interesting stmts.
8963 2018-04-27 Martin Jambor <mjambor@suse.cz>
8966 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8967 the jump function allows for passing through aggregate values.
8969 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8971 * input.h (in_system_header_at): Convert from macro to inline
8973 (from_macro_expansion_at): Likewise.
8974 (from_macro_definition_at): Likewise.
8976 2018-04-27 Jeff Law <law@redhat.com>
8978 * config.gcc: Mark tile* targets as deprecated/obsolete.
8980 2018-04-27 Richard Biener <rguenther@suse.de>
8982 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8985 2018-04-27 Richard Biener <rguenther@suse.de>
8987 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8989 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8991 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8992 with Yd constraint. Set "preferred_for_speed" attribute from
8993 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8995 (*movdi_internal): Ditto.
8996 (movti_interunit splitters): Remove
8997 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8998 (movdi_interunit splitters): Ditto.
8999 * config/i386/constraints.md (Ye): Remove.
9000 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9002 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9005 * config/aarch64/constraints.md (Usg): Limit to 31.
9008 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9010 PR tree-optimization/85529
9011 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9012 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9013 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9014 zero extension or masking of the MSB bit.
9015 (optimize_range_tests): Add FIRST_BB argument, pass it through
9016 to optimize_range_tests_var_bound.
9017 (maybe_optimize_range_tests, reassociate_bb): Adjust
9018 optimize_range_tests callers.
9020 2018-04-26 Richard Biener <rguenther@suse.de>
9021 Jakub Jelinek <jakub@redhat.com>
9023 * cgraph.h (symbol_table): Just declare debug method here.
9024 * symtab.c (symbol_table::debug): Define.
9026 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9028 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9030 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9032 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9033 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9034 (*movdi_internal): Substitute Yi and Yj constraint with x
9035 and Ym and Yn constraint with y constraint. Update "isa"
9036 attribute and set "preferred_for_speed" attribute from
9037 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9038 (*movsi_internal): Ditto.
9039 (*movdf_internal): Ditto.
9040 (*movsf_internal): Ditto.
9041 (*zero_extendsidi2): Ditto.
9042 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9043 (sse2_loadld): Ditto.
9044 (*vec_extract<ssevecmodelower>_0): Ditto.
9045 (*vec_extractv4si_0_zext_sse4): Ditto.
9046 (vec_concatv2di): Ditto.
9047 (*vec_dup<mode>): Ditto.
9048 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9049 * config/i386/constraints.md (Yi): Remove.
9054 2018-04-26 Nathan Sidwell <nathan@acm.org>
9056 * dumpfile.c (dump_open): New.
9057 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9058 (dump_finish): Detect stdio/stderr by value not name.
9060 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9062 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9064 2018-04-26 Tom de Vries <tom@codesourcery.com>
9067 * config/nvptx/nvptx.c (verify_neutering_jumps)
9068 (verify_neutering_labels): New function
9069 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9071 2018-04-26 Tom de Vries <tom@codesourcery.com>
9074 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9075 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9078 2018-04-26 Richard Biener <rguenther@suse.de>
9079 Tom de Vries <tom@codesourcery.com>
9082 * lto-streamer-out.c (output_function): Fixup loops if required to match
9083 discovery done in the reader.
9085 2018-04-26 Richard Biener <rguenther@suse.de>
9087 PR tree-optimization/85116
9088 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9089 have a loop exit from the single latch predecessor. Remove
9090 case of header with just condition.
9091 (ch_base::copy_headers): Exclude infinite loops from any
9093 (pass_ch::execute): Record exits.
9095 2018-04-26 Richard Biener <rguenther@suse.de>
9097 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9098 prologue cost vector and pass it to vect_get_load_cost.
9099 (vect_get_peeling_costs_all_drs): Likewise.
9100 (vect_peeling_hash_get_lowest_cost): Likewise.
9101 (vect_enhance_data_refs_alignment): Likewise.
9103 2018-04-26 Richard Biener <rguenther@suse.de>
9106 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9107 checking of integer<->pointer conversions.
9108 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9109 sign-/zero-extending pointer types.
9110 (expand_omp_for_static_chunk): Likewise.
9112 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9113 Jean Lee <xiaoyur347@gmail.com>
9115 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9116 (TARGET_ASAN_SHADOW_OFFSET): Define.
9117 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9118 true for -fsanitize=address.
9120 2018-04-25 Mark Wielaard <mark@klomp.org>
9122 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9125 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9127 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9128 than "alu", remove explicit "memory" and "imm_disp" attributes.
9129 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9132 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9133 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9134 gen_lowpart_no_emit.
9136 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9139 * config/i386/i386.c (ix86_expand_builtin): Change memory
9140 operand to XI, extend p0 to Pmode.
9141 * config/i386/i386.md: Change unspec volatile and operand
9142 1 mode to XI, change operand 0 mode to P.
9144 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9146 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9147 GET_MODE_MASK before any checking.
9148 (nds32_can_use_bset_p): Likewise.
9149 (nds32_can_use_btgl_p): Likewise.
9151 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9153 * config/nds32/nds32-doubleword.md: New define_split pattern for
9154 illegal register number.
9156 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9158 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9160 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9162 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9164 2018-04-25 Richard Biener <rguenther@suse.de>
9166 * lto-streamer.h (LTO_major_version): Bump to 8.
9168 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9170 * BASE-VER: Set to 9.0.0.
9172 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9174 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9175 in __abskf2 and __powikf2.
9177 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9180 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9181 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9182 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9183 Use the above on operand 2. Reindent.
9184 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9186 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9189 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9191 * config/i386/i386.opt (mcet): Removed.
9192 * doc/install.texi: Remove -mcet documentation.
9193 * doc/invoke.texi: Likewise.
9195 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9198 * doc/install.texi: Remove -mcet from bootstrap-cet.
9200 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9203 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9204 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9208 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9209 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9210 containing a CONST_VECTOR.
9212 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9214 * doc/install.texi: Update newlib dependency for nvptx.
9216 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9219 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9220 instead of INTVAL when shifting x left.
9222 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9224 PR tree-optimization/85478
9225 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9226 vect_grouped_store_supported for single element vectors.
9228 2018-04-24 Richard Biener <rguenther@suse.de>
9231 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9232 load cost increase to the case of non-constant step.
9234 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9237 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9238 destination if any_malformed_asm.
9240 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9243 * expr.c (store_field): In the bitfield case, if the value comes from
9244 a function call and is returned in registers by means of a PARALLEL,
9245 do not change the mode of the temporary unless BLKmode and VOIDmode.
9247 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9249 PR rtl-optimization/85423
9250 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9251 dependencies to debug insns when the previous insn is non-debug.
9253 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9255 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9256 enums into a single definition.
9257 (fls): Fix predicates and printing.
9260 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9262 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9263 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9264 and short u6 immediate.
9265 (check_if_valid_sleep_operand): Remove.
9266 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9268 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9270 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9271 flag_always_save_lp condition.
9272 * config/nds32/nds32.opt (malways-save-lp): New option.
9274 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9276 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9277 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9278 * config/nds32/nds32.h
9279 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9280 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9282 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9284 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9285 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9287 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9288 Chung-Ju Wu <jasonwucj@gmail.com>
9290 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9292 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9293 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9295 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9297 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9299 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9301 * config/nds32/nds32-protos.h (nds32_data_alignment,
9302 nds32_local_alignment): Declare.
9303 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9304 nds32_local_alignment): New functions.
9305 (TARGET_CONSTANT_ALIGNMENT): Define.
9306 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9308 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9310 * config/nds32/nds32.c
9311 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
9312 (TARGET_MODES_TIEABLE_P): Likewise.
9314 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9316 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
9319 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
9320 Chung-Ju Wu <jasonwucj@gmail.com>
9322 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9323 for unaligned access.
9324 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9326 * config/nds32/nds32-intrinsic.md: Likewise.
9327 * config/nds32/nds32_intrinsic.h: Likewise.
9328 * config/nds32/nds32.h (nds32_builtins): Likewise.
9329 * config/nds32/nds32.opt (munaligned-access): New option.
9330 * config/nds32/nds32.c (nds32_asm_file_start): Display
9331 flag_unaligned_access status.
9333 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9335 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
9336 -mno-relax is present.
9337 * config/riscv/linux.h (LINK_SPEC): Ditto.
9339 2018-04-20 Martin Sebor <msebor@redhat.com>
9342 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
9344 (gimple_fold_builtin_stxcpy_chk): Same.
9345 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
9347 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
9350 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
9351 __powikf2 when long double is IEEE 128-bit.
9353 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9355 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
9356 step to make sure stack always aligned.
9358 2018-04-20 Carl Love <cel@us.ibm.com>
9361 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
9362 size check for arg0.
9364 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
9365 Tom de Vries <tom@codesourcery.com>
9368 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
9369 Emit insns for calls too.
9370 (nvptx_find_par): Always look for worker-level predecessor insn.
9371 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
9373 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
9374 (nvptx_process_pars): Propagate frames for calls.
9376 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
9379 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
9381 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
9382 (ix86_handle_option): Don't handle OPT_mibt.
9383 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
9385 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
9387 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
9388 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
9389 (ix86_target_macros): Define __CET__ with flag_cf_protection
9390 for -fcf-protection.
9391 * config/i386/i386.c (isa2_opts): Remove -mibt.
9392 * config/i386/i386.h (TARGET_IBT): Removed.
9393 (TARGET_IBT_P): Likewise.
9394 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
9395 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
9396 * config/i386/i386.opt (mcet): Update help message.
9399 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
9400 -mcet as an alias for -mshstk.
9402 2018-04-20 Richard Biener <rguenther@suse.de>
9405 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
9406 complexity by forcing a single use of the multiply operand.
9408 2018-04-20 Martin Jambor <mjambor@suse.cz>
9411 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
9412 recursion dependency to only apply to non-clones.
9414 2018-04-20 Martin Jambor <mjambor@suse.cz>
9417 * ipa-cp.c (create_specialized_node): Check that clones of
9418 self-recursive edges exist during IPA-CP.
9420 2018-04-19 Toon Moene <toon@moene.org>
9422 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
9425 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9427 PR tree-optimization/85467
9428 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9429 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9430 VECTOR_CST element to type.
9432 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9435 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9436 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9437 (builtin_longjmp): Likewise.
9438 (save_stack_nonlocal): New pattern.
9439 (restore_stack_nonlocal): Likewise.
9441 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9444 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9445 Replace ASM_OUTPUT_LABEL with fprintf.
9447 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9450 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9451 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9452 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9453 define __IBT__ and __SHSTK__ for -fcf-protection.
9454 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9456 (ix86_trampoline_init): Likewise.
9457 (x86_output_mi_thunk): Likewise.
9458 (ix86_notrack_prefixed_insn_p): Likewise.
9459 (ix86_option_override_internal): Don't disallow -fcf-protection.
9460 * config/i386/i386.md (rdssp<mode>): Also enable for
9462 (incssp<mode>): Likewise.
9463 (nop_endbr): Likewise.
9464 * config/i386/i386.opt (mcet): Change help message to built-in
9468 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9469 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9470 enable CET built-in functions.
9472 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9474 * common/config/i386/i386-common.c
9475 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9476 OPTION_MASK_ISA_MOVDIRI_UNSET,
9477 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9478 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9479 * config.gcc (movdirintrin.h): New header.
9480 * config/i386/cpuid.h (bit_MOVDIRI,
9481 bit_MOVDIR64B): New bits.
9482 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9484 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9485 (VOID, PVOID, PCVOID)): New function types.
9486 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9487 __builtin_ia32_directstoreu_u64,
9488 __builtin_ia32_movdir64b): New builtins.
9489 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9490 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9492 (ix86_valid_target_attribute_inner_p): Ditto.
9493 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9494 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9495 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9496 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9497 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9498 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9499 (movdiri<mode>, movdir64b_<mode>): New.
9500 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9501 * config/i386/immintrin.h: Include movdirintrin.h.
9502 * config/i386/movdirintrin.h: New file.
9503 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9505 2018-04-19 Richard Biener <rguenther@suse.de>
9508 * cfg.c (clear_bb_flags): When loop state says we have
9509 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9511 2018-04-19 Richard Biener <rguenther@suse.de>
9513 PR tree-optimization/84737
9514 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9515 copying restrict info.
9516 (vect_setup_realignment): Use it.
9517 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9518 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9519 the first DR to all generated stores.
9520 (vectorizable_load): Likewise for loads.
9522 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9524 PR tree-optimization/85446
9525 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9526 the integral and pointer types to have the same precision.
9528 * doc/install.texi: Document --disable-cet being the default and
9531 2018-04-18 Martin Liska <mliska@suse.cz>
9533 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9536 2018-04-18 Martin Liska <mliska@suse.cz>
9539 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9542 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9543 arguments if they are comparable.
9545 2018-04-18 Martin Liska <mliska@suse.cz>
9548 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9551 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9554 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9557 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9558 ENDBR after calling __morestack.
9560 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9563 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9564 by using gcc_base_ver to generate a gcc_driver_version, and use
9565 it when generating GCC_DRIVER_NAME.
9566 * configure: Regenerate.
9568 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9571 * config.gcc: Obsolete powerpc*-*-*spe*.
9573 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9576 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9577 (stabstr_D): Change type of unum from unsigned int to
9578 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9581 2018-04-17 Jim Wilson <jimw@sifive.com>
9584 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9585 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9586 Set arg_pointer_offset after using pretend_args_size.
9588 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9590 PR rtl-optimization/85431
9591 * dse.c (record_store): Ignore zero width stores.
9594 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9595 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9596 __builtin_stack_restore rather than after it.
9597 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9598 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9599 argument instead of virtual_dynamic_stack_rtx.
9601 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9603 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9605 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9606 Add note to error message to explain internal mapping of overloaded
9607 built-in function name to non-overloaded built-in function name.
9608 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9611 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9614 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9615 where the inputs overlap with the output.
9617 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9619 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9620 (=v, v) alternative and explicit "memory" attribute.
9621 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9622 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9624 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9625 "sselog1" type instead of "sselog".
9626 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9627 "sselog". Remove explicit "memory" attribute.
9628 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9629 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9631 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9632 "sselog1" type instead of "sselog". Remove explicit "memory"
9634 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9635 use "sselog1" type instead of "sselog". Remove explicit "memory"
9637 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9638 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9640 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9641 "sselog1" type instead of "sselog". Remove explicit "memory"
9643 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9644 use "sselog1" type instead of "sselog". Remove explicit "memory"
9648 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9651 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9654 2018-04-17 Martin Jambor <mjambor@suse.cz>
9657 * ipa-cp.c (create_specialized_node): Call
9658 expand_all_artificial_thunks if necessary.
9660 2018-04-17 Martin Liska <mliska@suse.cz>
9663 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9664 in message, remote space in between '_G' and '('.
9666 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9669 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9670 avx512f_vmcmp<mode>3<round_saeonly_name>,
9671 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9672 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9673 avx512f_rndscale<mode><round_saeonly_name>,
9674 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9675 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9676 Use %<iptr>2 instead of %2 for -masm=intel.
9677 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9678 avx512f_vcvttss2usi<round_saeonly_name>,
9679 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9681 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9682 avx512f_vcvttsd2usi<round_saeonly_name>,
9683 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9684 Use %q1 instead of %1 for -masm=intel.
9685 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9686 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9687 of %3 for -masm=intel.
9688 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9690 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9692 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9694 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9695 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9697 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9699 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9700 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9701 %g1 and one with %0 and %1.
9702 (avx512er_vmrcp28<mode><round_saeonly_name>,
9703 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9705 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9706 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9707 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9708 of %0 and %{%4%} for -masm=intel.
9709 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9710 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9711 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9712 order of %0 and %{%5%}%{z%} for -masm=intel.
9714 2018-04-17 Jan Hubicka <jh@suse.cz>
9717 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9719 2018-04-17 Martin Liska <mliska@suse.cz>
9722 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9723 for target_clone error message. Make default implementation
9724 clone to be a local declaration.
9725 (separate_attrs): Add new argument and check for an empty
9727 (expand_target_clones): Handle it.
9728 (ipa_target_clone): Make redirection just for target_clones
9731 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9732 Tom de Vries <tom@codesourcery.com>
9735 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9736 tiled basic blocks without omp continue statements.
9738 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9741 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9742 vec_extract expression as having side effects to make sure it gets
9745 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9748 * config/i386/i386.c (get_builtin_code_for_version): Check
9751 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9754 * gcc/config.gcc: Support "skylake".
9755 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9757 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9758 (processor_target_table): Add "skylake".
9759 (ix86_option_override_internal): Add "skylake".
9760 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9761 PROCESSOR_CANNONLAKE.
9762 (get_builtin_code_for_version): Fix priority for
9763 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9764 PROCESSOR_SKYLAKE-AVX512.
9765 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9766 (processor_type): Add PROCESSOR_SKYLAKE.
9768 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9769 Jason Merrill <jason@redhat.com>
9772 * convert.c (convert_to_integer_1): Use direct recursion for
9773 enumeral types and types with a precision less than the number
9774 of bits in their mode.
9776 2018-04-16 Julia Koval <julia.koval@intel.com>
9779 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9780 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9782 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9785 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9786 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9787 and -mno-direct-move.
9789 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9792 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9793 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9794 Ensure negative shifts result in {0}.
9796 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9798 PR rtl-optimization/79916
9799 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9800 regs (if any) to define how to gnerate SD moves when LRA is in
9803 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9805 PR rtl-optimization/85393
9806 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9807 * except.c (expand_dw2_landing_pad_for_region): Make static.
9808 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9809 a label and unconditional jump to old_bb, rather than
9810 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9813 PR rtl-optimization/85376
9814 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9815 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9816 instead of a specific value.
9818 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9819 Bin Cheng <bin.cheng@arm.com>
9821 PR tree-optimization/82965
9822 PR tree-optimization/83991
9823 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9824 by_profile_only parameter.
9825 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9826 information if the loop was predicted to iterate too many times.
9827 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9829 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9832 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9835 2018-04-13 Martin Liska <mliska@suse.cz>
9836 Jakub Jelinek <jakub@redhat.com>
9839 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9840 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9841 * builtins.c (expand_builtin_memory_copy_args): Use
9842 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9843 handle dest_addr == pc_rtx.
9845 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9848 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9849 asked to not generate direct moves.
9850 (fix_trunc<mode>si2_stfiwx): Similar.
9851 (fix_trunc<mode>si2_internal): Similar.
9853 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9856 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9857 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9858 lookup if dest in some wider mode is known to be const0_rtx and
9859 if so, record permanent equivalence for it to be ZERO_EXTEND of
9860 the narrower mode destination.
9862 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9864 * lto-streamer-out.c (output_function): Revert 259346.
9865 * omp-expand.c (expand_oacc_for): Likewise.
9867 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9869 PR rtl-optimization/85354
9870 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9871 * sel-sched.c (sel_global_init): ... here.
9873 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9876 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9877 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9878 mode for PE-COFF targets.
9879 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9880 (i386_pe_asm_lto_end): Likewise.
9881 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9882 (TARGET_ASM_LTO_END): Likewise.
9883 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9884 (i386_pe_asm_lto_start): New function.
9885 (i386_pe_asm_lto_end): Likewise.
9887 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9888 Richard Biener <rguenther@suse.de>
9891 * lto-streamer-out.c (output_function): Fix CFG loop state before
9893 * omp-expand.c (expand_oacc_for): Handle calls to internal
9894 functions like regular functions.
9896 2018-04-12 Richard Biener <rguenther@suse.de>
9899 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9900 for the early LTO debug to properly generate references to it
9901 during DIE emission. Do not re-use that for the skeleton for
9903 (dwarf2out_early_finish): Likewise.
9905 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9908 * config/i386/sse.md
9909 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9910 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9911 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9912 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9913 and output is a reg, avoid creating invalid lowpart subreg, but
9914 instead split into a 512-bit move. Don't split if not AVX512VL,
9915 input is xmm16+ reg and output is a mem.
9916 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9917 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9918 xmm16+ reg and output is a mem.
9920 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9922 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9923 also for flag_dwarf2_cfi_asm.
9925 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9927 PR rtl-optimization/85342
9928 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9929 a bool scalar var inside of the loop instead. Don't try to update
9930 recog_data.operand after failed apply_change_group.
9932 2018-04-12 Tom de Vries <tom@codesourcery.com>
9935 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9936 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9937 array with flexible array member as array without given dimension.
9938 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9939 argument for undefined param to true.
9941 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9944 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9945 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9946 from PowerPC section.
9947 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9948 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9949 help text that is too long.
9950 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9951 help text that is too long.
9952 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9953 help text that is too long.
9955 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9957 * config/alpha/alpha.md (stack_probe_internal): Rename
9958 from "probe_stack". Update all callers.
9960 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9962 PR rtl-optimization/84566
9963 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9964 sched_macro_fuse_insns.
9966 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9969 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9970 (compute_block_dependences): ... from here.
9972 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9974 PR tree-optimization/85331
9975 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9976 from int to HOST_WIDE_INT.
9978 2018-04-11 Martin Jambor <mjambor@suse.cz>
9981 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9982 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9983 not the same as the source val.
9984 (cgraph_edge_brings_value_p): New parameter.
9985 (gather_edges_for_value): Pass destination value to
9986 cgraph_edge_brings_value_p.
9987 (perhaps_add_new_callers): Likewise.
9988 (get_info_about_necessary_edges): Likewise and exclude values brought
9989 only by self-recursive edges.
9990 (create_specialized_node): Redirect only clones of self-calling edges.
9991 (+self_recursive_pass_through_p): New function.
9992 (find_more_scalar_values_for_callers_subset): Use it.
9993 (find_aggregate_values_for_callers_subset): Likewise.
9994 (known_aggs_to_agg_replacement_list): Removed.
9995 (decide_whether_version_node): Re-calculate known constants for all
9996 remaining context clones.
9998 2018-04-11 Richard Biener <rguenther@suse.de>
10001 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10002 from early DWARF output.
10003 (dwarf2out_early_finish): Output line info unconditionally into
10004 early DWARF and add reference to it.
10006 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10009 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10010 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10011 other than V2DFmode using iptr mode attribute.
10012 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10014 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10016 PR rtl-optimization/84659
10017 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10019 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10022 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10024 (output_loc_list): Pass address of a dummy size variable even in the
10025 locview handling loop.
10026 (index_location_lists): Add comment on why skip_loc_list_entry can't
10029 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10032 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10035 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10038 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10039 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10040 and -mstring-compare-inline-limit.
10042 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10045 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
10046 for stack clash protection in a register whenever we need it to be in
10049 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10051 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10052 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10054 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10057 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10059 (mlong-double-): Ditto.
10060 * config/rs6000/sysv4.opt (msdata=): Ditto.
10061 (mtls-size=): Ditto.
10063 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10065 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10066 erroneous entries for
10067 "vector int vec_ldl (int, long int *)", and
10068 "vector unsigned int vec_ldl (int, unsigned long int *)".
10069 Add comments and entries for
10070 "vector bool char vec_ldl (int, bool char *)",
10071 "vector bool short vec_ldl (int, bool short *)",
10072 "vector bool int vec_ldl (int, bool int *)",
10073 "vector bool long long vec_ldl (int, bool long long *)",
10074 "vector pixel vec_ldl (int, pixel *)",
10075 "vector long long vec_ldl (int, long long *)",
10076 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10077 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10078 type tree bool_long_long_type_node and correct definition of
10079 bool_V2DI_type_node to make reference to this new type tree.
10080 (rs6000_mangle_type): Replace erroneous reference to
10081 bool_long_type_node with bool_long_long_type_node.
10082 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10083 comments to emphasize sign distinctions for char and int types and
10084 replace RS6000_BTI_bool_long constant with
10085 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10086 use of RS6000_BTI_pixel.
10087 (bool_long_type_node): Remove this macro definition.
10088 (bool_long_long_type_node): New macro definition
10090 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10092 PR rtl-optimization/85300
10093 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10094 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10095 simplify_unary_operation fails.
10097 2018-04-10 Martin Liska <mliska@suse.cz>
10099 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10100 cgraph_edge and ipa_ref.
10102 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10106 * config/i386/sse.md
10107 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10108 computation of the VEC_MERGE selector from mask.
10109 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10110 Fix decoding of the VEC_MERGE selector into mask.
10112 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10114 PR tree-optimization/85286
10115 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10117 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10119 * final.c (final_1): Set insn_last_address as well as
10120 insn_current_address.
10122 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10125 * explow.c (emit_stack_probe): Call validize_mem on memory location
10126 before passing it to gen_probe_stack. Create address operand and
10127 legitimize it for the probe_stack_address case.
10129 2018-04-09 Jan Hubicka <jh@suse.cz>
10132 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10133 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10134 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10136 (free_lang_data): Rebuild type inheritance graph.
10138 2018-04-09 Martin Sebor <msebor@redhat.com>
10140 * invoke.texi (-finline-small-functions): Mention other optimization
10142 (-findirect-inlining, -fpartial-inlining): Same.
10143 (-finline-functions-called-once): Same.
10144 (-freorder-blocks-and-partition): Same.
10146 2018-04-09 Jan Hubicka <jh@suse.cz>
10149 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10150 jumps; choose last target that matches the criteria (i.e.
10151 no partition changes for non-crossing jumps).
10152 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10153 support for redirecting crossing jumps to non-crossing.
10155 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10157 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10158 also for naked functions.
10160 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10162 * config/arc/arc.md (add_shift): New pattern.
10163 (add_shift2): Likewise.
10164 (sub_shift): Likewise.
10165 (sub_shift_cmp0_noout): Likewise.
10166 (compare_si_ashiftsi): Likewise.
10167 (xbfu_cmp0_noout): New combine pattern.
10168 (xbfu_cmp0"): Likewise.
10169 (movsi_set_cc_insn): Place the predicable variant first.
10170 (commutative_binary_cmp0_noout): Remove clobber.
10171 (commutative_binary_cmp0): New pattern.
10172 (noncommutative_binary_cmp0): Likewise.
10173 (noncommutative_binary_cmp0_noout): Likewise.
10174 (noncommutative_binary_comparison_result_used): Removed.
10175 (rsub_cmp0): New pattern.
10176 (rsub_cmp0_noout): Likewise.
10177 (extzvsi): Changed, keep only meaningful variants.
10178 (SQH, SEZ): New iterators.
10179 (SQH_postfix): New mode attribute.
10180 (SEZ_prefix): New code attribute.
10181 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10182 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10183 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10184 of numerical value.
10185 (noncommutative_operator): Check the availability of barrel
10188 2018-04-09 Richard Biener <rguenther@suse.de>
10190 PR tree-optimization/85284
10191 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10192 Only use the niter constraining form of simple_iv when the exit
10193 is always executed.
10195 2018-04-09 Tom de Vries <tom@codesourcery.com>
10198 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10199 (define_expand "*memory_barrier"): New define_expand.
10200 (define_insn "memory_barrier"): New insn.
10202 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10204 PR rtl-optimization/80463
10205 PR rtl-optimization/83972
10206 PR rtl-optimization/83480
10208 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10209 correct producer for the insn.
10210 (tidy_control_flow): Fixup seqnos in case of debug insns.
10212 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10214 PR rtl-optimization/83913
10216 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10217 different sched-times when merging exprs.
10219 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10221 PR rtl-optimization/83962
10223 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10224 tidy_fallthru_edge and tidy_control_flow.
10226 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10228 PR rtl-optimization/83530
10230 * sel-sched.c (force_next_insn): New global variable.
10231 (remove_insn_for_debug): When force_next_insn is true, also leave only
10232 next insn in the ready list.
10233 (sel_sched_region): When the region wasn't scheduled, make another pass
10234 over it with force_next_insn set to 1.
10236 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10238 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10240 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10241 for interrupt control.
10242 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10243 functions for interrupt control.
10244 * config/nds32/nds32-intrinsic.md: Likewise.
10245 * config/nds32/nds32_intrinsic.h: Likewise.
10246 * config/nds32/nds32.h (nds32_builtins): Likewise.
10248 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10250 * config/nds32/nds32.c (nds32_init_machine_status,
10251 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10252 strict_aligned_p field.
10253 (nds32_expand_to_rtl_hook): New function.
10254 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10255 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10257 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10258 Chung-Ju Wu <jasonwucj@gmail.com>
10260 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10261 * config/nds32/nds32-n7.md: New file.
10262 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10263 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10265 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10266 * config/nds32/nds32.md (pipeline_model): Add n7.
10267 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10268 * config/nds32/pipelines.md: Include n7 settings.
10270 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10271 Chung-Ju Wu <jasonwucj@gmail.com>
10273 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10274 * config/nds32/nds32-e8.md: New file.
10275 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10276 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10278 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10279 * config/nds32/nds32.md (pipeline_model): Add e8.
10280 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10281 * config/nds32/pipelines.md: Include e8 settings.
10283 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10284 Chung-Ju Wu <jasonwucj@gmail.com>
10286 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10287 * config/nds32/nds32-n8.md: New file.
10288 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10289 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10291 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10292 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10293 * config/nds32/nds32.md (pipeline_model): Add n8.
10294 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10295 * config/nds32/pipelines.md: Include n8 settings.
10297 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10298 Chung-Ju Wu <jasonwucj@gmail.com>
10300 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10301 * config/nds32/nds32-n9-2r1w.md: New file.
10302 * config/nds32/nds32-n9-3r2w.md: New file.
10303 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10304 nds32_register_ports): New or modify for cpu n9.
10305 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10307 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10308 * config/nds32/nds32-utils.c: New file.
10309 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
10310 TARGET_MUL_SLOW): Define.
10311 * config/nds32/nds32.md (pipeline_model): New attribute.
10312 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
10313 New options that support cpu n9.
10314 * config/nds32/pipelines.md: Include n9 settings.
10315 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
10317 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10319 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
10320 information if necessary.
10321 (output_cond_branch_compare_zero): Likewise.
10322 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
10323 (nds32_target_alignment): Refine for alignment.
10324 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
10325 (FUNCTION_BOUNDARY): Modify.
10326 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
10328 * config/nds32/nds32.opt (malways-align, malign-functions): New.
10330 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10332 * config/nds32/constants.md (unspec_volatile_element): Add values for
10333 TLB operation and data prefetch.
10334 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10335 functions for TLB operation and data prefetch.
10336 * config/nds32/nds32-intrinsic.md: Likewise.
10337 * config/nds32/nds32_intrinsic.h: Likewise.
10338 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10339 (nds32_print_operand): Likewise.
10340 * config/nds32/nds32.h (nds32_builtins): Likewise.
10342 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
10343 Andrew Pinski <pinsika@gcc.gnu.org>
10345 PR middle-end/82976
10346 * match.pd: Use constant_boolean_node of correct type instead of
10347 boolean_true_node or boolean_false_node for simplifying
10348 pointer comparisons to zero.
10350 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10352 PR tree-optimization/80021
10353 * tree.c (verify_type_variant): Make error call in verify_variant_match
10354 translatable and remove final full stop.
10356 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10358 * config/nds32/constants.md (unspec_volatile_element): Add
10359 UNSPEC_VOLATILE_EH_RETURN.
10360 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
10361 nds32_output_stack_pop): Support dwarf exception handling process.
10362 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
10363 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
10364 exception handling process.
10365 (nds32_compute_stack_frame): Likewise.
10366 (nds32_return_addr_rtx): Likewise.
10367 (nds32_initial_elimination_offset): Likewise.
10368 (nds32_expand_prologue): Likewise.
10369 (nds32_expand_epilogue): Likewise.
10370 (nds32_dynamic_chain_address): New function.
10371 * config/nds32/nds32.h (machine_function): Add fields for dwarf
10372 exception handling.
10373 (DYNAMIC_CHAIN_ADDRESS): Define.
10374 (EH_RETURN_DATA_REGNO): Define.
10375 (EH_RETURN_STACKADJ_RTX): Define.
10376 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
10377 patterns for dwarf exception handling.
10379 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10381 * config/nds32/nds32.h: Clean up obsolete macros.
10383 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10385 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10386 Add enum values for particular instructions.
10387 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10388 particular intrinsic functions.
10389 * config/nds32/nds32-intrinsic.md: Likewise.
10390 * config/nds32/nds32_intrinsic.h: Likewise.
10391 * config/nds32/nds32.h (nds32_builtins): Likewise.
10392 * config/nds32/nds32.md (type): Add pbsad and pbsada.
10393 (btst, ave): New patterns for particular instructions.
10395 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10397 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10398 Add enum values for atomic load/store and memory sync.
10399 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
10401 * config/nds32/nds32-intrinsic.md: Likewise.
10402 * config/nds32/nds32_intrinsic.h: Likewise.
10403 * config/nds32/nds32.h (nds32_builtins): Likewise.
10405 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10407 PR tree-optimization/85257
10408 * fold-const.c (native_encode_vector): If not all elts could fit
10409 and off is -1, return 0 rather than offset.
10410 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
10411 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
10412 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
10413 adjust buffer in native_interpret_expr call.
10415 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10417 * config/nds32/constants.md (unspec_volatile_element): Add cache
10418 control enum values.
10419 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
10420 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
10421 * config/nds32/nds32.c (nds32_cctl_names): New.
10422 (nds32_print_operand): Handle cache control register names.
10423 * config/nds32/nds32.h (nds32_builtins): New enum values.
10424 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
10426 * config/nds32/nds32.md (type): Add mmu.
10427 * config/nds32/pipelines.md (simple_insn): Add mmu.
10429 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10431 * config/nds32/nds32.md (type): Remove call.
10432 * config/nds32/pipelines.md (simple_insn): Likewise.
10434 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10436 * config/nds32/constants.md (unspec_volatile_element): Add
10437 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10438 UNSPEC_VOLATILE_FMFCFG.
10439 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10440 description for fmfcfg and fmfcsr.
10441 (bdesc_1arg): Add fmtcsr.
10442 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10443 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10444 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10445 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10446 unspec_fmfcfg): New patterns.
10447 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10448 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10449 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10450 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10451 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10452 __nds32__fmfcfg): Define.
10454 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10456 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10457 intrinsic register names.
10458 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10459 intrinsic register enum values and macros.
10461 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10463 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10464 for load/store addressing form.
10465 (nds32_print_operand_address): Likewise.
10467 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10470 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10471 based on LABEL_REF. Remove useless assertion.
10472 (pic_address_needs_scratch): Fix formatting.
10473 (sparc_legitimize_pic_address): Minor tweaks.
10474 (sparc_delegitimize_address): Adjust assertion accordingly.
10475 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10476 into symbolic_operand.
10477 (movsi_high_pic_label_ref): Likewise.
10478 (movsi_lo_sum_pic_label_ref): Likewise.
10479 (movdi_pic_label_ref): Likewise.
10480 (movdi_high_pic_label_ref): Likewise.
10481 (movdi_lo_sum_pic_label_ref): Likewise.
10483 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10485 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10486 custom LIB_SPEC setup.
10488 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10489 Kito Cheng <kito.cheng@gmail.com>
10491 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10492 * config/riscv/freebsd.h: New.
10494 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10496 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10497 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10500 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10501 Kito Cheng <kito.cheng@gmail.com>
10503 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10504 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10505 * config/nds32/nds32-protos.h (nds32_output_call,
10506 nds32_output_return): Declare.
10507 * config/nds32/nds32.md: Refine all the call and return patterns.
10509 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10512 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10513 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10515 PR rtl-optimization/84872
10516 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10517 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10518 EDGE_CROSSING edge.
10520 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10522 * expr.c (copy_blkmode_to_reg): Revert 254862.
10523 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10525 2018-04-06 Richard Biener <rguenther@suse.de>
10527 PR middle-end/85244
10528 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10529 after seeing a component reference with an adjacent field. Treat
10530 refs to arrays at struct end of external decls similar to
10531 refs to unconstrained commons.
10533 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10536 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10537 look through SAVE_EXPRs with non-side-effects argument. Adjust
10539 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10542 2018-04-06 Richard Biener <rguenther@suse.de>
10544 PR middle-end/85180
10545 * alias.c (find_base_term): New wrapper around find_base_term
10546 unwinding CSELIB_VAL_PTR changes.
10547 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10550 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10552 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10554 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10555 constant definitions.
10556 ("nop"): lr 0,0 -> nopr r0
10557 ("nop_lr0", "nop_lr1"): New insn definitions.
10559 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10561 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10562 NDS32_V3PUSH_AVAILABLE_P macro.
10564 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10565 Chung-Ju Wu <jasonwucj@gmail.com>
10567 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10568 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10569 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10570 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10571 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10572 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10573 * config/nds32/constraints.md: New constraints and checking for hard
10574 float configuration.
10575 * config/nds32/iterators.md: New mode iterator and attribute for hard
10576 float configuration.
10577 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10579 * config/nds32/nds32-fpu.md: New file.
10580 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10581 deal with hard float code generation.
10582 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10584 (abi_type, float_reg_number): New enum type.
10585 * config/nds32/nds32-predicates.c: New predicates for hard float.
10586 * config/nds32/nds32-protos.h: Declare functions for hard float.
10587 * config/nds32/nds32.c: Implementation for hard float configuration.
10588 * config/nds32/nds32.h: Definitions for hard float configuration.
10589 * config/nds32/nds32.md: Include hard float machine description and
10590 modify patterns for hard float configuration.
10591 * config/nds32/nds32.opt: New options for hard float configuration.
10592 * config/nds32/predicates.md: New predicates for hard float
10595 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10597 * common/config/nds32/nds32-common.c
10598 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10600 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10602 PR middle-end/85195
10603 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10604 CONSTRUCTOR_ELT (ctor, ...)->value.
10606 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10609 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10611 2018-04-05 Tom de Vries <tom@codesourcery.com>
10614 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10617 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10618 Kito Cheng <kito.cheng@gmail.com>
10620 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10621 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10622 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10623 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10625 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10626 Kito Cheng <kito.cheng@gmail.com>
10628 * config/nds32/constraints.md (Ufe): New memory constraint.
10629 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10630 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10631 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10633 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10634 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10636 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10638 * config/nds32/nds32.md: Use optimize_size in the condition for
10639 alu-shift instructions.
10641 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10643 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10645 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10647 * config/nds32/nds32.md (negsi2): Refine pattern.
10649 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10650 Chung-Ju Wu <jasonwucj@gmail.com>
10652 * config/nds32/iterators.md (shift_rotate): New code iterator.
10653 (shift): New code attribute.
10654 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10655 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10656 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10657 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10658 bit-wise operations.
10659 (andsi3, *andsi3): Ditto.
10660 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10661 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10662 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10663 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10664 nds32_ior_operand, nds32_xor_operand): New predicates.
10666 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10668 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10669 (addsi3, subsi3): ... this.
10671 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10673 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10675 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10677 * config/nds32/nds32.md: Adjust indention.
10679 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10681 * config/nds32/nds32.md (feature): New attribute.
10683 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10685 * config/nds32/nds32.md (subtype): New attribute.
10687 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10690 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10691 expansion to perform a bitwise AND of the argument followed by a
10692 boolean negation of the result.
10694 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10696 PR rtl-optimization/84878
10697 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10698 the basic block. Assert the use reference is not artificial and that
10699 it has an associated insn.
10701 2018-04-04 Michael Matz <matz@suse.de>
10703 * builtins.c (compute_objsize): Pass correct operand
10704 to array_at_struct_end_p.
10706 2018-04-04 Richard Biener <rguenther@suse.de>
10709 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10710 from contexts for DINFO_LEVEL_TERSE and below.
10712 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10714 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10715 resiter_operand condition.
10716 * config/nds32/nds32.md (*move<mode>): Ditto.
10718 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10719 Monk Chiang <sh.chiang04@gmail.com>
10721 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10723 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10725 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10727 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10728 Kito Cheng <kito.cheng@gmail.com>
10730 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10731 nds32_cond_code_str, output_cond_branch,
10732 output_cond_branch_compare_zero, nds32_expand_cbranch,
10733 nds32_expand_cstore, nds32_expand_movcc,
10734 nds32_output_cbranchsi4_equality_zero,
10735 nds32_output_cbranchsi4_equality_reg,
10736 nds32_output_cbranchsi4_equality_reg_or_const_int,
10737 nds32_output_cbranchsi4_greater_less_zero: New functions.
10738 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10739 nds32_expand_cstore, nds32_expand_movcc,
10740 nds32_output_cbranchsi4_equality_zero,
10741 nds32_output_cbranchsi4_equality_reg,
10742 nds32_output_cbranchsi4_equality_reg_or_const_int,
10743 nds32_output_cbranchsi4_greater_less_zero): Declare.
10744 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10745 nds32_rimm11s_operand): New predicates.
10746 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10747 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10750 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10752 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10753 * config/nds32/nds32.md: Ditto.
10754 * config/nds32/pipelines.md: Ditto.
10756 2018-04-04 Richard Biener <rguenther@suse.de>
10758 PR tree-optimization/85168
10759 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10760 propagating abnormals.
10762 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10764 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10766 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10767 Kito Cheng <kito.cheng@gmail.com>
10769 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10770 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10771 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10772 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10773 * config/nds32/nds32.md (sibcall_internal): New.
10774 (sibcall_register): Remove.
10775 (sibcall_immediate): Remove.
10776 (sibcall_value_internal): New.
10777 (sibcall_value_register): Remove.
10778 (sibcall_value_immediate): Remove.
10779 * config/nds32/predicates.md (nds32_general_register_operand): New.
10780 (nds32_call_address_operand): New.
10782 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10784 PR rtl-optimization/85167
10785 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10786 bb_defs if *split_p, instead preinitialize it to NULL.
10788 PR tree-optimization/85156
10789 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10790 evaluating the argument multiple times.
10792 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10794 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10796 (_mm_cvtpd_ps): Likewise.
10797 (_mm_cvttpd_epi32): Likewise.
10798 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10799 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10800 vector, pixel, and bool following altivec.h include.
10802 2018-04-03 Martin Sebor <msebor@redhat.com>
10804 * doc/extend.texi (Common Function Attributes): Clarify.
10805 (const attribute): Likewise.
10806 (pure attribute): Likewise.
10808 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10811 * config/i386/i386.c (ix86_expand_vector_set): Use
10812 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10814 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10816 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10817 instructions when changing rounding bits to preserve precision bits
10818 in the x87 control word.
10820 2018-04-03 Martin Liska <mliska@suse.cz>
10822 PR tree-optimization/82491
10823 * rtl.h (strip_offset_and_add): Replace += suboffset with
10824 poly_uint64 () + suboffset.
10826 2018-03-29 Martin Liska <mliska@suse.cz>
10827 Martin Jambor <mjambor@suse.cz>
10830 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10831 param_type is not an integral or pointer type.
10833 2018-04-03 Richard Biener <rguenther@suse.de>
10835 * sese.h (recompute_all_dominators): Remove.
10837 2018-04-02 Martin Sebor <msebor@redhat.com>
10839 * doc/invoke.texi (-Wrestrict): Fix typos.
10841 2018-04-02 Jim Wilson <jimw@sifive.com>
10843 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10844 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10845 (<optab>di3, <optab>si3_extend): Likewise.
10846 (<optab>si3_mask, <optab>si3_mask_1): New.
10847 (<optab>di3_mask, <optab>di3_mask_1): New.
10848 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10849 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10850 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10852 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10854 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10857 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10859 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10860 (nds32_canonicalize_comparison): New function.
10862 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10863 Kito Cheng <kito.cheng@gmail.com>
10864 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10866 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10867 * config/nds32/constants.md (unspec_volatile_element): Add
10868 UNSPEC_VOLATILE_RELAX_GROUP.
10869 * config/nds32/nds32-relax-opt.c: New file.
10870 * config/nds32/nds32-predicates.c
10871 (nds32_symbol_load_store_p): New function.
10872 * config/nds32/nds32-protos.h
10873 (nds32_symbol_load_store_p): Declare function.
10874 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10875 * config/nds32/nds32.c
10876 (nds32_register_pass): New function to register pass.
10877 (nds32_register_passes): New function to register passes.
10878 * config/nds32/nds32.md (relax_group): New pattern.
10879 * config/nds32/nds32.opt (mrelax-hint): New option.
10880 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10882 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10884 * config/nds32/t-nds32: Modify files dependency.
10886 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10888 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10889 (PROFILE_HOOK): Define its implementation.
10891 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10893 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10894 type and 32-bit size.
10896 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10898 PR middle-end/85090
10899 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10900 (V_128_256): New mode iterator.
10901 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10902 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10903 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10905 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10906 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10908 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10911 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10912 NaN inputs correctly.
10914 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10917 * config/rs6000/vsx.md (??r): New mode attribute.
10918 (*vsx_mov<mode>_64bit): Use it.
10919 (*vsx_mov<mode>_32bit): Likewise.
10921 2018-03-30 Martin Sebor <msebor@redhat.com>
10923 PR tree-optimization/84818
10924 * builtins.c (check_access): Use warning_n.
10926 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10929 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10931 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10934 2018-03-30 Julia Koval <julia.koval@intel.com>
10937 * x86-tune.def (movx, partial_reg_dependency): Enable for
10940 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10942 PR inline-asm/84985
10943 * lra-constraints.c (process_alt_operands): Move setting
10944 this_alternative_matches below.
10946 2018-03-29 Martin Liska <mliska@suse.cz>
10949 * doc/invoke.texi: Document how LTO works with debug info.
10950 Describe auto-load support of binutils. Mention 'x86-64'
10951 as valid option value of -march option.
10953 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10955 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10958 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10959 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10960 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10963 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10966 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10967 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10968 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10969 for RS6000_BTM_POWERPC64.
10970 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10971 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10972 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10975 (DIVDEU): Likewise.
10977 2018-03-28 Carl Love <cel@us.ibm.com>
10980 2017-09-27 Carl Love <cel@us.ibm.com>
10982 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10983 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10984 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10987 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10989 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10990 instead of __vector bool.
10991 (_mm_max_pu8): Likewise.
10992 (_mm_min_pi16): Likewise.
10994 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10997 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10998 (DIVWEUO): Likewise.
10999 (DIVDEO): Likewise.
11000 (DIVDEUO): Likewise.
11001 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11002 DIVWEUO and DIVDEUO.
11003 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11004 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11005 (div_extend): Likewise.
11006 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11008 (__builtin_divweuo): Likewise.
11009 (__builtin_divdeo): Likewise.
11010 (__builtin_divdeuo): Likewise.
11012 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11015 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11016 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11018 PR tree-optimization/82004
11019 * gimple-match-head.c (optimize_pow_to_exp): New function.
11020 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11021 Don't fold to exp if optimize_pow_to_exp is false.
11023 2018-03-28 Martin Liska <mliska@suse.cz>
11026 * calls.c (initialize_argument_information): Fix trailing space.
11027 * common.opt: Fix typo and provide better explanation for
11028 -fsanitize-coverage option.
11029 * config/i386/i386.opt: Fix typo.
11031 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11032 Martin Liska <mliska@suse.cz>
11035 * gimplify.c (asan_poison_variable): Don't do the check for
11036 gimplify_omp_ctxp here.
11037 (gimplify_decl_expr): Do it here.
11038 (gimplify_target_expr): Likewise.
11040 2018-03-28 Martin Liska <mliska@suse.cz>
11043 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11044 chkp_type_bounds_count if MPX is not enabled.
11046 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11048 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11050 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11053 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11054 function to create the function decl for complex long double
11055 multiply and divide for -mabi=ieeelongdouble.
11056 (init_float128_ieee): Call it.
11058 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11061 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11062 -fcf-protection=branch -mibt.
11063 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11065 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11068 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11069 arm_word_relocations.
11071 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11074 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11075 extern array declarations.
11077 2018-03-27 Richard Biener <rguenther@suse.de>
11079 PR middle-end/84067
11080 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11081 explicit single_use checks.
11083 2018-03-27 Richard Biener <rguenther@suse.de>
11085 PR tree-optimization/85082
11086 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11089 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11091 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11092 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11093 Turn on fasynchronous-unwind-tables and funwind-tables.
11095 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11098 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11099 (*bmi_blsr_<mode>_ccz): Ditto.
11101 2018-03-26 Tom de Vries <tom@codesourcery.com>
11103 PR tree-optimization/85063
11104 * omp-general.c (offloading_function_p): New function. Factor out
11106 * omp-offload.c (pass_omp_target_link::gate): ... here.
11107 * omp-general.h (offloading_function_p): Declare.
11108 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11109 with attribute omp declare target for offloading functions.
11111 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11113 PR tree-optimization/84005
11114 * tree-data-ref.h (get_base_for_alignment): Declare.
11115 * tree-data-ref.c (get_base_for_alignment_1): New function.
11116 (get_base_for_alignment): Likewise.
11117 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11118 get_base_for_alignment to find a suitable base object, instead
11119 of always using drb->base_address.
11121 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11123 PR inline-asm/85022
11124 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11125 known size by default.
11127 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11129 PR inline-asm/85030
11130 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11131 and non BLKmode operands.
11133 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11136 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11137 Clean up attributes.
11139 2018-03-23 Richard Biener <rguenther@suse.de>
11142 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11143 we are going to emit early debug for LTO.
11145 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11147 PR inline-asm/85034
11148 * function.c (match_asm_constraints_1): Don't optimize if input
11149 doesn't satisfy general_operand predicate for output's mode.
11151 PR inline-asm/85022
11152 * alias.c (write_dependence_p): Don't require for x_canonicalized
11153 non-VOIDmode if x has VOIDmode.
11156 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11157 just don't try to optimize it rather than assert it never happens.
11159 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11161 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11162 macro expansions for definition of ST_INTERNAL_<mode> and
11163 LD_INTERNAL_<mode> builtins.
11164 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11166 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11168 (altivec_expand_st_builtin): Likewise.
11169 (altivec_expand_builtin): Remove calls to deleted functions.
11170 (rs6000_address_for_altivec): Delete this function.
11171 * config/rs6000/vector.md: Remove expands for
11172 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11174 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11177 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11178 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11179 re-computing once computed.
11180 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11181 (arm_init_machine_status): Initialize
11182 machine->static_chain_stack_bytes.
11184 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11187 * doc/extend.texi: Add four new prototypes for vec_ld.
11188 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11189 definitions for more logical presentation.
11190 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11191 entries for V1TI variants of __builtin_altivec_ld builtin.
11192 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11193 handling of V1TI variant of LVX icode pattern.
11194 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11195 (rs6000_gimple_fold_builtin): Likewise.
11196 (altivec_init_builtins): Add code to define
11197 __builtin_altivec_lvx_v1ti function.
11199 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11201 PR inline-asm/84941
11202 * function.c (match_asm_constraints_1): Don't do the optimization
11203 if input isn't a REG, SUBREG, MEM or constant.
11205 2018-03-22 Tom de Vries <tom@codesourcery.com>
11207 PR tree-optimization/84956
11208 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11209 bb_has_abnormal_pred.
11211 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11214 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11215 DECL_INITIAL (decl) to decl at the end.
11216 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11217 adjust the comment.
11219 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11221 * doc/extend.texi (__builtin_tgmath): Document when complex
11222 integer types are treated as _Complex _Float64.
11224 2018-03-21 Tom de Vries <tom@codesourcery.com>
11226 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11228 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11230 PR tree-optimization/84960
11231 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11232 if it is ENTRY block, move them into single succ of ENTRY in that case.
11234 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11236 PR tree-optimization/84811
11237 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11238 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11239 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11240 (known_subrange_p): Update accordingly. Cast each value involved
11241 in the size comparison, rather than casting the result of the
11244 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11246 PR tree-optimization/84982
11247 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11248 by flipping the least significant bit rather than all bits from
11249 bitpos to bitpos + bitsize - 1.
11251 2018-03-21 Nathan Sidwell <nathan@acm.org>
11253 * doc/extend.texi (Deprecated Features): Remove mention of
11254 long-deleted deprecations.
11256 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11259 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11260 * configure: Regenerate.
11262 2018-03-21 Tom de Vries <tom@codesourcery.com>
11264 PR tree-optimization/83126
11265 * tree-parloops.c (num_phis): New function.
11266 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11268 2018-03-21 Nathan Sidwell <nathan@acm.org>
11270 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11271 mention anon-struct/union members and trailing attributes.
11273 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11275 PR tree-optimization/84969
11276 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11277 builtin memset partitions if they set different rhs values.
11279 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11281 PR rtl-optimization/84989
11282 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11283 VEC_DUPLICATE with scalar result mode.
11285 2018-03-21 Martin Liska <mliska@suse.cz>
11288 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11289 not intended return statement.
11291 2018-03-21 Martin Liska <mliska@suse.cz>
11294 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11295 (chkp_find_bound_slots_1): Limit number of iterations.
11297 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11300 * Minor grammar fixes for x86 options.
11302 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11305 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11306 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11308 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
11311 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
11312 (altivec_lvx_<mode>_1op): Likewise.
11313 (altivec_stvx_<mode>_2op): Likewise.
11314 (altivec_stvx_<mode>_1op): Likewise.
11315 (altivec_lvx_<VM2:mode>): New define_expand.
11316 (altivec_stvx_<VM2:mode>): Likewise.
11317 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
11318 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11319 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
11320 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11321 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
11322 (rs6000_gen_lvx): Likewise.
11323 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
11324 (altivec_expand_stv_builtin): Likewise.
11325 (altivec_expand_builtin): Likewise.
11326 * config/rs6000/vector.md: Likewise.
11328 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11331 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
11334 2018-03-20 Richard Biener <rguenther@suse.de>
11337 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
11338 sign-conversions as zero, fall back to standard scalar_stmt
11341 2018-03-20 Martin Liska <mliska@suse.cz>
11344 * predict.c (rebuild_frequencies): Handle case when we have
11345 PROFILE_ABSENT, but flag_guess_branch_prob is false.
11347 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11350 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
11351 flag_section_anchors.
11352 * varasm.c (use_blocks_for_decl_p): Remove hack for
11353 dw2_force_const_mem.
11356 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
11358 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
11359 be created, use lowpart_subreg of operands[0] rather than operands[0]
11361 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
11362 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
11363 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
11364 and n constraint instead of aarch64_shift_imm_di and Usd.
11365 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
11366 (*aarch64_<optab>_reg_minus<mode>3): ... this.
11368 2018-03-20 Sudakshina Das <sudi.das@arm.com>
11371 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
11372 to favor GPR over NEON registers.
11373 (<shift>di3_neon): Likewise.
11375 2018-03-20 Tom de Vries <tom@codesourcery.com>
11378 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
11379 (nvptx_process_pars): Emit bar.sync asap and alap.
11381 2018-03-20 Tom de Vries <tom@codesourcery.com>
11384 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
11385 seen_label if seen_label is already set.
11387 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11390 * config/i386/i386.c (fold_builtin_cpu): For features above 31
11391 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
11392 Use 1U instead of 1. Formatting fixes.
11395 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
11396 instead of TREE_TYPE (s1) for the return value.
11398 2018-03-19 Jakub Jelinek <jakub@redhat.com>
11400 PR tree-optimization/84946
11401 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
11402 bitsize + bitsize in poly_uint64 rather than poly_int64.
11405 * dwarf2asm.c: Include fold-const.c.
11406 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
11407 of decl rather than decl itself.
11409 PR rtl-optimization/84643
11410 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11412 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
11415 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
11416 calling assemble_variable.
11418 2018-03-19 Sudakshina Das <sudi.das@arm.com>
11421 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
11422 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
11424 2018-03-19 Jim Wilson <jimw@sifive.com>
11427 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11428 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11429 (riscv_first_stack_step): Likewise.
11430 (riscv_option_override): Use STACK_BOUNDARY instead of
11431 MIN_STACK_BOUNDARY.
11432 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11433 MIN_STACK_BOUNDARY.
11434 (BIGGEST_ALIGNMENT): Set to 128.
11435 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11436 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11439 2018-03-19 Richard Biener <rguenther@suse.de>
11441 PR tree-optimization/84933
11442 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11443 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11445 2018-03-19 Richard Biener <rguenther@suse.de>
11447 PR tree-optimization/84859
11448 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11449 (cond_if_else_store_replacement): Perform sinking operation on
11450 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11451 Generalize what a BB with a single eligible store is.
11453 2018-03-19 Richard Biener <rguenther@suse.de>
11455 PR tree-optimization/84929
11456 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11457 chrec_is_positive against non-chrec arg.
11459 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11462 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11464 2018-03-18 Martin Liska <mliska@suse.cz>
11466 PR rtl-optimization/84635
11467 * regrename.c (build_def_use): Use matches_mode only when
11470 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11472 PR tree-optimization/84913
11473 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11474 vectorize chains of COND_EXPRs.
11476 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11478 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11480 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11482 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11484 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11486 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11488 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11489 Kito Cheng <kito.cheng@gmail.com>
11491 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11492 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11493 (nds32_adjust_reg_alloc_order): New function.
11494 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11496 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11498 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11499 nds32_print_operand, nds32_print_operand_address): Use
11500 HOST_WIDE_INT_PRINT_DEC instead.
11502 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11504 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11506 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11509 * config/i386/i386.c (initial_ix86_tune_features,
11510 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11511 unsigned long long.
11512 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11513 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11514 rather than 1u << ix86_tune. Formatting fix.
11515 (ix86_option_override_internal): Change ix86_arch_mask from
11516 unsigned int to unsigned HOST_WIDE_INT, initialize to
11517 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11518 (ix86_function_specific_restore): Likewise.
11520 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11523 * postreload.c (reload_combine_recognize_pattern): Perform
11524 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11525 truncate_int_for_mode the result for the destination's mode.
11528 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11529 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11531 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11534 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11537 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11538 regno_allocno_class_array and sorted_pseudos.
11539 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11540 insns where regno is used.
11542 2018-03-16 Martin Liska <mliska@suse.cz>
11545 * multiple_target.c (create_dispatcher_calls): Redirect
11546 reference in the symbol table.
11548 2018-03-16 Martin Liska <mliska@suse.cz>
11551 * multiple_target.c (create_dispatcher_calls): Redirect also
11554 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11558 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11559 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11560 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11562 2018-03-16 Julia Koval <julia.koval@intel.com>
11564 * doc/invoke.texi (Skylake Server): Add CLWB.
11565 Cannonlake): Remove CLWB.
11567 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11569 PR tree-optimization/84841
11570 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11572 (FLOAT_ONE_CONST_TYPE): Define.
11573 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11574 (sort_by_operand_rank): Put entries with higher constant_type last
11575 rather than first to match comments.
11577 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11579 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11582 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11585 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11586 check and use address of target always.
11588 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11591 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11592 (indirect_thunk_bnd_needed): Likewise.
11593 (indirect_thunks_used): Likewise.
11594 (indirect_thunks_bnd_used): Likewise.
11595 (indirect_return_needed): New.
11596 (indirect_return_bnd_needed): Likewise.
11597 (output_indirect_thunk_function): Add a bool argument for
11599 (output_indirect_thunk_function): Don't generate alias for
11600 function return thunk.
11601 (ix86_code_end): Call output_indirect_thunk_function to generate
11602 function return thunks.
11603 (ix86_output_function_return): Set indirect_return_bnd_needed
11604 and indirect_return_needed instead of indirect_thunk_bnd_needed
11605 and indirect_thunk_needed.
11607 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11609 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11610 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11611 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11613 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11614 Paul Hua <paul.hua.gm@gmail.com>
11617 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11619 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11621 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11622 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11623 resp. SFmode cases.
11625 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11628 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11629 instead of GET_MODE_SIZE when comparing Units.
11631 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11634 * varasm.c (hash_section): Return an unchangeble hash value
11635 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11636 Return !aarch64_can_use_per_function_literal_pools_p ().
11638 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11641 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11642 rather than address of cmode as last argument to prepare_cmp_insn.
11644 2018-03-15 Julia Koval <julia.koval@intel.com>
11646 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11647 F_AVX512VNNI, F_AVX512BITALG): New.
11649 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11652 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11653 insn for floating-point loads and stores.
11655 2018-03-14 Carl Love <cel@us.ibm.com>
11657 * config/rs6000/rs6000-c.c: Add macro definitions for
11658 ALTIVEC_BUILTIN_VEC_PERMXOR.
11659 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11660 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11661 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11662 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11664 * config/doc/extend.texi: Add prototypes for vec_permxor.
11666 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11669 * diagnostic-show-locus.c (class layout_point): Convert m_line
11670 from int to linenum_type.
11671 (line_span::comparator): Use linenum "compare" function when
11672 comparing line numbers.
11673 (test_line_span): New function.
11674 (layout_range::contains_point): Convert param "row" from int to
11676 (layout_range::intersects_line_p): Likewise.
11677 (layout::will_show_line_p): Likewise.
11678 (layout::print_source_line): Likewise.
11679 (layout::should_print_annotation_line_p): Likewise.
11680 (layout::print_annotation_line): Likewise.
11681 (layout::print_leading_fixits): Likewise.
11682 (layout::annotation_line_showed_range_p): Likewise.
11683 (struct line_corrections): Likewise for field m_row.
11684 (line_corrections::line_corrections): Likewise for param "row".
11685 (layout::print_trailing_fixits): Likewise.
11686 (layout::get_state_at_point): Likewise.
11687 (layout::get_x_bound_for_row): Likewise.
11688 (layout::print_line): Likewise.
11689 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11690 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11691 * input.c (selftest::test_linenum_comparisons): New function.
11692 (selftest::input_c_tests): Call it.
11693 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11694 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11695 * selftest.h (ASSERT_GT): New macro.
11696 (ASSERT_GT_AT): New macro.
11697 (ASSERT_LT): New macro.
11698 (ASSERT_LT_AT): New macro.
11700 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11702 PR rtl-optimization/84780
11703 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11705 2018-03-14 Martin Liska <mliska@suse.cz>
11707 * tree.c (record_node_allocation_statistics): Use
11708 get_stats_node_kind.
11709 (get_stats_node_kind): New function extracted from
11710 record_node_allocation_statistics.
11711 (free_node): Use get_stats_node_kind.
11713 2018-03-14 Richard Biener <rguenther@suse.de>
11715 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11716 that the value-set of ANTIC_IN doesn't grow.
11719 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11721 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11722 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11724 2018-03-14 Julia Koval <julia.koval@intel.com>
11726 * config.gcc (icelake-client, icelake-server): New.
11728 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11729 (initial_ix86_arch_features): Ditto.
11730 (PTA_SKYLAKE): Add SGX.
11731 (PTA_ICELAKE): Remove.
11732 (PTA_ICELAKE_CLIENT): New.
11733 (PTA_ICELAKE_SERVER): New.
11734 (ix86_option_override_internal): Split up icelake on icelake client and
11736 (get_builtin_code_for_version): Ditto.
11737 (fold_builtin_cpu): Ditto.
11738 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11739 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11740 * config/i386/i386.h (processor_type): Ditto.
11741 * doc/invoke.texi: Ditto.
11743 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11746 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11747 INTEGER_CST offset, add it together with bitpos / 8 and
11748 sign extend based on POINTER_SIZE.
11752 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11755 * config/i386/constraints.md (Yc): New register constraint.
11756 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11757 Use Yc constraint for alternative 2 of operand 0. Remove
11758 preferred_for_speed attribute.
11760 2018-03-14 Richard Biener <rguenther@suse.de>
11762 PR tree-optimization/84830
11763 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11764 with the old one to avoid oscillations.
11766 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11769 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11771 (assign_by_spills): Return a flag of reload assignment failure.
11772 Do not process the reload assignment failures. Do not spill other
11773 reload pseudos if they has the same reg class. Update n if
11775 (lra_assign): Add a return arg. Set up from the result of
11776 assign_by_spills call.
11777 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11778 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11779 usage_insns if it is not NULL.
11780 (spill_hard_reg_in_range): New function.
11781 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11782 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11783 function prototypes.
11784 (lra_assign): Change prototype.
11785 * lra.c (lra): Add code to deal with fails by splitting hard reg
11788 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11790 * config/riscv/riscv.opt (mrelax): New option.
11791 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11792 "norelax" when riscv_mrelax is disabled.
11793 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11795 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11798 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11799 reassociation for int modes.
11801 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11803 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11804 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11806 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11807 * config/aarch64/aarch64-sve.md
11808 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11809 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11810 (*extend<mode><Vwide>2): Rename to...
11811 (aarch64_sve_extend<mode><Vwide>2): ...this.
11812 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11813 renaming the old pattern to...
11814 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11816 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11817 define_expand, renaming the old pattern to...
11818 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11819 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11820 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11821 account when deciding which SVE instruction the optab should use.
11822 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11824 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11826 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11827 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11828 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11829 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11830 (tlsdesc_small_<mode>): Turn a define_expand and use
11831 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11832 (tlsdesc_small_advsimd_<mode>): ...this.
11833 (tlsdesc_small_sve_<mode>): New pattern.
11835 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11837 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11838 (UNSPEC_UMUL_HIGHPART): New constants.
11839 (MUL_HIGHPART): New int iteraor.
11840 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11841 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11843 (*<su>mul<mode>3_highpart): New define_insn.
11845 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11848 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11851 2018-03-13 Martin Liska <mliska@suse.cz>
11854 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11856 (sem_item_optimizer::~sem_item_optimizer): Release it.
11857 (sem_item_optimizer::merge_classes): Register variable aliases.
11858 (sem_item_optimizer::fixup_pt_set): New function.
11859 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11860 * ipa-icf.h: Declare new variables and functions.
11862 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11864 PR middle-end/84834
11865 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11866 integer_pow2p@2 and test integer_pow2p in condition.
11867 (A < 0 ? C : 0): Similarly for @1.
11869 PR middle-end/84831
11870 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11871 characters starting at p contain '\0' character, don't look beyond
11875 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11876 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11879 * reg-stack.c (change_stack): Change update_end var from int to
11880 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11881 also call set_block_for_insn on the newly added insns and rescan.
11884 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11885 on the last operand.
11888 * tree.c (stabilize_reference_1): Return save_expr (e) for
11889 STATEMENT_LIST even if it doesn't have side-effects.
11891 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11893 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11895 2018-03-12 Renlin Li <renlin.li@arm.com>
11897 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11898 aarch64_output_scalar_simd_mov_immediate.
11900 2018-03-12 Martin Sebor <msebor@redhat.com>
11902 PR tree-optimization/83456
11903 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11904 for perfectly overlapping calls to memcpy.
11905 (gimple_fold_builtin_memory_chk): Same.
11906 (gimple_fold_builtin_strcpy): Handle no-warning.
11907 (gimple_fold_builtin_stxcpy_chk): Same.
11908 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11910 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11912 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11913 parameter. Use it for SFmode.
11914 (rs6000_function_arg_advance_1): Adjust.
11915 (rs6000_function_arg): Adjust.
11916 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11918 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11920 PR rtl-optimization/84169
11921 PR rtl-optimization/84780
11922 * combine.c (can_combine_p): Check for a 2-insn combination whether
11923 the destination register is used between the two insns, too.
11925 2018-03-12 Richard Biener <rguenther@suse.de>
11927 PR tree-optimization/84803
11928 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11929 for refs DR analysis didn't process.
11931 2018-03-12 Richard Biener <rguenther@suse.de>
11933 PR tree-optimization/84777
11934 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11935 force-vectorize loops ignore whether we are optimizing for size.
11937 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11939 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11940 (TARGET_MD_ASM_ADJUST): Define.
11942 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11943 Kito Cheng <kito.cheng@gmail.com>
11944 Chung-Ju Wu <jasonwucj@gmail.com>
11946 * config/nds32/nds32.c (nds32_compute_stack_frame,
11947 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11948 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11949 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11950 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11951 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11952 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11953 * config/nds32/nds32.md (prologue, epilogue): Use macro
11954 NDS32_V3PUSH_AVAILABLE_P to do checking.
11956 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11959 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11960 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11961 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11962 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11963 addition of most attributes on !orig_type_die or the attribute not
11964 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11966 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11967 Chung-Ju Wu <jasonwucj@gmail.com>
11969 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11970 __NDS32_VH__ macro.
11971 * config/nds32/nds32.opt (mvh): New option.
11973 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11974 Chung-Ju Wu <jasonwucj@gmail.com>
11976 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11978 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11979 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11982 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11983 Chung-Ju Wu <jasonwucj@gmail.com>
11985 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11987 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11988 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11990 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11991 Kito Cheng <kito.cheng@gmail.com>
11992 Chung-Ju Wu <jasonwucj@gmail.com>
11994 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11995 UNSPEC_FFMISM and UNSPEC_FLMISM.
11996 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11997 for ffb, ffmism and flmism.
11998 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11999 (unspec_ffmism): Ditto.
12000 (unspec_flmism): Ditto.
12001 (nds32_expand_builtin_impl): Check if string extension is available.
12002 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12003 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12005 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12008 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12011 * lra-assigns.c (assign_by_spills): Return a flag of reload
12012 assignment failure. Do not process the reload assignment
12013 failures. Do not spill other reload pseudos if they has the same
12015 (lra_assign): Add a return arg. Set up from the result of
12016 assign_by_spills call.
12017 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12018 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12019 usage_insns if it is not NULL.
12020 (spill_hard_reg_in_range): New function.
12021 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12022 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12023 function prototypes.
12024 (lra_assign): Change prototype.
12025 * lra.c (lra): Add code to deal with fails by splitting hard reg
12028 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12031 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12033 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12036 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12037 (dw_val_node): Add val_symbolic_view.
12038 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12039 (symview_upper_bound): New.
12040 (new_line_info_table): Initialize symviews_since_reset.
12041 (dwarf2out_source_line): Count symviews_since_reset and set
12042 symview_upper_bound.
12043 (dw_val_equal_p): Handle symview.
12044 (add_AT_symview): New.
12045 (print_dw_val): Handle symview.
12046 (attr_checksum, attr_checksum_ordered): Likewise.
12047 (same_dw_val_p, size_of_die): Likewise.
12048 (value_format, output_die): Likewise.
12049 (add_high_low_attributes): Use add_AT_symview for entry_view.
12050 (dwarf2out_finish): Reset symview_upper_bound, clear
12053 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12056 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12057 Add strict argument and use it.
12058 (rs6000_split_multireg_move): Update for new strict argument.
12059 (mem_operand_gpr): Disallow all non-offsettable addresses.
12060 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12062 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12065 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12066 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12067 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12070 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12071 decl, use remap_type if we want to use the type.
12073 2018-03-09 Martin Sebor <msebor@redhat.com>
12075 PR tree-optimization/84526
12076 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12078 (builtin_access::generic_overlap): Be prepared to handle non-array
12081 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12083 PR rtl-optimization/84682
12084 * lra-constraints.c (process_address_1): Check is_address flag
12085 for address constraints.
12086 (process_alt_operands): Likewise.
12087 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12088 preprocess_constraints.
12089 * recog.h (preprocess_constraints): Add oploc parameter.
12091 * recog.c (preprocess_constraints): Test address_operand for
12092 CT_ADDRESS constraints.
12094 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12097 * lra-assigns.c (assign_by_spills): Return a flag of reload
12098 assignment failure. Do not process the reload assignment
12099 failures. Do not spill other reload pseudos if they has the same
12101 (lra_assign): Add a return arg. Set up from the result of
12102 assign_by_spills call.
12103 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12104 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12105 usage_insns if it is not NULL.
12106 (spill_hard_reg_in_range): New function.
12107 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12108 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12109 function prototypes.
12110 (lra_assign): Change prototype.
12111 * lra.c (lra): Add code to deal with fails by splitting hard reg
12114 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12117 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12118 Accept complain bool parameter. Only emit errors if it is true.
12119 (arm_parse_cpu_option_name): Likewise.
12120 (arm_target_thumb_only): Adjust callers of the above.
12121 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12122 prototype to take a default true bool parameter.
12123 (arm_parse_arch_option_name): Likewise.
12125 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12126 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12130 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12131 * configure: Regenerate.
12132 * configure.ac ("linker --version-script option"): New.
12133 ("linker soname option"): New.
12135 2018-03-09 Richard Biener <rguenther@suse.de>
12137 PR tree-optimization/84775
12138 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12139 immediate uses of predicate stmts and mark them modified.
12142 PR tree-optimization/84178
12143 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12145 (version_loop_for_if_conversion): Delay update_ssa call.
12146 (tree_if_conversion): Delay update_ssa until after predicate
12149 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12152 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12153 when the function accesses prior frames.
12155 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12158 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12159 gen_llsym, otherwise call maybe_gen_llsym.
12161 PR inline-asm/84742
12162 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12163 has ',' character inside of it.
12165 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12168 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12169 as clobbering CC_REGNUM.
12171 2018-03-08 Richard Biener <rguenther@suse.de>
12173 PR middle-end/84552
12174 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12175 (follow_copies_to_constant): Do not follow SSA names registered
12178 2018-03-08 Richard Biener <rguenther@suse.de>
12180 PR tree-optimization/84178
12181 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12183 (version_loop_for_if_conversion): Delay update_ssa call.
12184 (tree_if_conversion): Delay update_ssa until after predicate
12187 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12189 PR tree-optimization/84178
12190 * tree-if-conv.c (release_bb_predicate): Remove the
12191 the assertion that the stmts have NULL use_ops.
12192 Discard the statements, asserting that they haven't
12193 yet been added to a BB.
12195 2018-03-08 Richard Biener <rguenther@suse.de>
12197 PR tree-optimization/84746
12198 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12199 (phi_translate): Pass in destination ANTIC_OUT set.
12200 (phi_translate_1): Likewise. For a simplified result lookup
12201 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12202 (phi_translate_set): Adjust.
12203 (do_pre_regular_insertion): Likewise.
12204 (do_pre_partial_partial_insertion): Likewise.
12206 2018-03-08 Martin Liska <mliska@suse.cz>
12208 PR gcov-profile/84735
12209 * doc/gcov.texi: Document usage of profile files.
12210 * gcov-io.h: Document changes in the format.
12212 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12216 * dwarf2out.c (struct dw_line_info_table): Update comments for
12218 (FORCE_RESET_NEXT_VIEW): New.
12219 (FORCE_RESETTING_VIEW_P): New.
12220 (RESETTING_VIEW_P): Check for -1 too.
12221 (ZERO_VIEW_P): Likewise.
12222 (new_line_info_table): Force-reset next view.
12223 (dwarf2out_begin_function): Likewise.
12224 (dwarf2out_source_line): Simplify zero_view_p initialization.
12225 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12226 view directly. Omit view when omitting .loc at line 0.
12228 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12230 PR tree-optimization/84740
12231 * tree-switch-conversion.c (process_switch): Call build_constructors
12232 only if info.phi_count is non-zero.
12234 PR tree-optimization/84739
12235 * tree-tailcall.c (find_tail_calls): Check call arguments against
12236 DECL_ARGUMENTS (current_function_decl) rather than
12237 DECL_ARGUMENTS (func) when checking for tail recursion.
12239 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12241 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12242 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12243 Volker Reichelt's entry and add entries for people that perform
12244 GCC fuzzy testing and report numerous bugs.
12246 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12249 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12250 readonly data in sdata, if that is disabled.
12251 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12252 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12253 -mreadonly-in-sdata option.
12255 2018-03-07 Martin Sebor <msebor@redhat.com>
12257 PR tree-optimization/84468
12258 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12259 basic block when looking for nul assignment.
12261 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12264 * except.h (output_function_exception_table): Adjust prototype.
12265 * except.c (output_function_exception_table): Remove FNNAME parameter
12266 and add SECTION parameter. Ouput one part of the table at a time.
12267 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12268 the first part of the exception table and emit unwind directives.
12269 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12270 (i386_pe_seh_cold_init): Likewise.
12271 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12272 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12273 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12274 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12275 * config/i386/winnt.c: Include except.h.
12276 (struct seh_frame_state): Add reg_offset, after_prologue and
12277 in_cold_section fields.
12278 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12279 (i386_pe_seh_cold_init): New function.
12280 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12281 to seh->in_cold_section.
12282 (seh_emit_push): Record the offset of the push.
12283 (seh_emit_save): Record the offet of the save.
12284 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12285 Test seh->after_prologue to disregard the epilogue.
12286 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12287 (i386_pe_end_cold_function): New function.
12289 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12292 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12293 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12296 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12298 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12299 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12301 PR middle-end/84723
12302 * multiple_target.c: Include tree-inline.h and intl.h.
12303 (expand_target_clones): Diagnose and fail if node->definition and
12304 !tree_versionable_function_p (node->decl).
12306 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12308 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12310 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
12311 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
12312 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
12314 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12317 * combine.c (try_combine): Use reg_or_subregno instead of handling
12318 just paradoxical SUBREGs and REGs.
12320 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
12322 * config/arc/arc.c (arc_finalize_pic): Remove function.
12323 (arc_must_save_register): We use single base PIC register, remove
12324 checks to save/restore the PIC register.
12325 (arc_expand_prologue): Likewise.
12326 * config/arc/arc-protos.h (arc_set_default_type_attributes):
12328 (arc_verify_short): Likewise.
12329 (arc_attr_type): Likewise.
12330 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
12331 (walk_stores): Likewise.
12332 (arc_address_cost): Make it static.
12333 (arc_verify_short): Likewise.
12334 (branch_dest): Likewise.
12335 (arc_attr_type): Likewise.
12336 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
12337 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
12338 (arc_final_prescan_insn): Remove inserting the nops due to
12339 hardware hazards. It is done in reorg step.
12340 (insn_length_variant_t): Remove.
12341 (insn_length_parameters_t): Likewise.
12342 (arc_insn_length_parameters): Likewise.
12343 (arc_get_insn_variants): Likewise.
12344 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
12346 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12348 PR inline-asm/84683
12349 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
12352 PR tree-optimization/84687
12353 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
12355 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12357 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12359 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
12360 Rename to ppc_speculation_barrier.
12361 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
12362 __builtin_ppc_speculation_barrier.
12364 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12367 * combine.c (combine_simplify_rtx): Don't try to simplify if
12368 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
12371 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
12373 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
12374 to 32 bytes when compiling for POWER9.
12376 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12379 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
12380 regparm >= 3 with no arg reg available also for calls with
12381 flag_force_indirect_call. Pass decl to ix86_function_regparm.
12384 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
12386 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
12388 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
12391 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
12393 2018-03-05 Richard Biener <rguenther@suse.de>
12395 PR tree-optimization/84486
12396 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
12397 When inserting a __builtin_assume_aligned call set the LHS
12398 SSA name alignment info accordingly.
12400 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
12402 PR tree-optimization/84114
12403 * config/aarch64/aarch64.c (aarch64_reassociation_width)
12404 Avoid reassociation of FLOAT_MODE addition.
12406 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
12408 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
12409 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
12410 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
12411 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
12412 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
12413 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
12414 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
12416 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
12417 __builtin_ia32_wbinvd): New builtins.
12418 (SPECIAL_ARGS2): New.
12419 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
12420 (SPECIAL_ARGS2): New.
12421 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
12422 (ix86_valid_target_attribute_inner_p): Ditto.
12423 (ix86_init_mmx_sse_builtins): Add special_args2.
12424 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
12425 TARGET_WBNOINVD_P): New.
12426 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
12427 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12428 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12429 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12430 * config/i386/pconfigintrin.h: New file.
12431 * config/i386/wbnoinvdintrin.h: Ditto.
12432 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12434 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12436 2018-03-05 Richard Biener <rguenther@suse.de>
12438 PR tree-optimization/84670
12439 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12441 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12442 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12443 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12444 doesn't grow if all successors (recursively) were visited at least
12447 2018-03-05 Richard Biener <rguenther@suse.de>
12449 PR tree-optimization/84650
12450 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12451 if executed in the loop pipeline.
12453 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12455 * doc/configfiles.texi (Configuration Files): Move info about
12456 conditionalizing $target-protos.h to...
12457 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12458 differs from $target-protos.h.
12460 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12461 Chung-Ju Wu <jasonwucj@gmail.com>
12463 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12464 * config/nds32/nds32-multiple.md (setmemsi): Define.
12465 * config/nds32/nds32-memory-manipulation.c
12466 (nds32_gen_dup_4_byte_to_word_value): New.
12467 (emit_setmem_word_loop): New.
12468 (emit_setmem_byte_loop): New.
12469 (nds32_expand_setmem_loop): New.
12470 (nds32_expand_setmem_loop_v3m): New.
12471 (nds32_expand_setmem_unroll): New.
12472 (nds32_expand_setmem): New.
12474 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12475 Chung-Ju Wu <jasonwucj@gmail.com>
12477 * config/nds32/nds32-memory-manipulation.c
12478 (nds32_emit_load_store): New.
12479 (nds32_emit_post_inc_load_store): New.
12480 (nds32_emit_mem_move): New.
12481 (nds32_emit_mem_move_block): New.
12482 (nds32_expand_movmemsi_loop_unknown_size): New.
12483 (nds32_expand_movmemsi_loop_known_size): New.
12484 (nds32_expand_movmemsi_loop): New.
12485 (nds32_expand_movmemsi_unroll): New.
12486 (nds32_expand_movmemqi): Rename ...
12487 (nds32_expand_movmemsi): ... to this.
12488 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12489 (movmemsi): ... to this.
12490 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12491 (nds32_expand_movmemsi): ... to this.
12493 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12494 Monk Chiang <sh.chiang04@gmail.com>
12495 Chung-Ju Wu <jasonwucj@gmail.com>
12497 * config/nds32/nds32-protos.h
12498 (nds32_expand_load_multiple): New arguments.
12499 (nds32_expand_store_multiple): Ditto.
12500 (nds32_valid_multiple_load_store): Rename ...
12501 (nds32_valid_multiple_load_store_p): ... to this.
12502 * config/nds32/nds32-memory-manipulation.c
12503 (nds32_expand_load_multiple): Refine implementation.
12504 (nds32_expand_store_multiple): Ditto.
12505 * config/nds32/nds32-multiple.md
12506 (load_multiple): Update nds32_expand_load_multiple interface.
12507 (store_multiple): Update nds32_expand_store_multiple interface.
12508 * config/nds32/nds32-predicates.c
12509 (nds32_valid_multiple_load_store): Rename ...
12510 (nds32_valid_multiple_load_store_p): ... to this and refine
12512 * config/nds32/predicates.md
12513 (nds32_load_multiple_and_update_address_operation): New predicate.
12514 (nds32_store_multiple_and_update_address_operation): New predicate.
12516 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12517 Chung-Ju Wu <jasonwucj@gmail.com>
12519 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12520 (combo): New attribute.
12521 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12523 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12525 * config/nds32/nds32.opt: Change -mcmodel= default value.
12527 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12528 Monk Chiang <sh.chiang04@gmail.com>
12529 Chung-Ju Wu <jasonwucj@gmail.com>
12531 * config/nds32/constants.md (unspec_element): New enum.
12532 * config/nds32/constraints.md (Umw): New constraint.
12533 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12534 * config/nds32/nds32-intrinsic.md: Likewise.
12535 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12536 (nds32_valid_smw_lwm_base_p): New.
12537 (nds32_output_smw_single_word): New.
12538 (nds32_output_lmw_single_word): New.
12539 (nds32_expand_unaligned_load): New.
12540 (nds32_expand_unaligned_store): New.
12541 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12542 (nds32_output_smw_single_word): Declare.
12543 (nds32_output_lmw_single_word): Declare.
12544 (nds32_expand_unaligned_load): Declare.
12545 (nds32_expand_unaligned_store): Declare.
12546 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12547 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12548 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12549 NDS32_BUILTIN_UASTORE_DW.
12550 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12553 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12554 Kito Cheng <kito.cheng@gmail.com>
12555 Chung-Ju Wu <jasonwucj@gmail.com>
12557 * config/nds32/nds32-intrinsic.c
12558 (nds32_expand_builtin_null_ftype_reg): Delete.
12559 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12560 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12561 (nds32_read_argument): New.
12562 (nds32_legitimize_target): Ditto.
12563 (nds32_legitimize_argument): Ditto.
12564 (nds32_check_constant_argument): Ditto.
12565 (nds32_expand_unop_builtin): Ditto.
12566 (nds32_expand_unopimm_builtin): Ditto.
12567 (nds32_expand_binop_builtin): Ditto.
12568 (nds32_builtin_decl_impl): Ditto.
12569 (builtin_description): Ditto.
12570 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12571 (nds32_init_builtins_impl): Ditto.
12572 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12573 (nds32_builtin_decl): New.
12574 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12575 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12577 2018-03-02 Jeff Law <law@redhat.com>
12579 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12580 (redundant_insn, fill_simple_delay_slots): Likewise.
12581 (fill_slots_from_thread): Likewise.
12582 * resource.c (mark_referenced_resources): Likewise.
12583 (mark_set_resources, find_dead_or_set_registers): Likewise.
12585 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12587 * substring-locations.h (format_warning_va): Formatting fix for
12588 ATTRIBUTE_GCC_DIAG.
12589 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12591 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12592 * substring-locations.c: Include intl.h.
12593 (format_warning_va): Turned into small wrapper around
12594 format_warning_n_va, renamed to ...
12595 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12596 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12598 (format_warning_at_substring_n): New function.
12599 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12600 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12601 format_warning_at_substring with just a shorter name instead of
12602 const function pointer.
12603 (fmtwarn_n): New function.
12604 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12605 appropriate, get rid of all the fmtstr temporaries, move conditionals
12606 with G_() wrapped string literals directly into fmtwarn arguments,
12607 cast dir.len to (int), formatting fixes.
12609 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12611 * doc/invoke.texi: Remove "Cilk Plus" references.
12613 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12614 Richard Biener <rguenther@suse.de>
12617 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12618 for error or warning attributes if CALL_FROM_THUNK_P is set.
12621 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12624 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12625 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12628 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12629 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12631 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12632 instead of -1U in last predictors element's probability member.
12634 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12637 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12638 arguments if they are comparable.
12640 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12642 PR tree-optimization/84634
12643 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12644 masks and masked_loop_p with a single loop_masks, making sure it's
12645 null for bb vectorization.
12647 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12649 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12650 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12651 loop->force_vectorize to check whether there is no alias.
12653 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12656 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12658 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12660 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12661 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12662 instead of a loop around prev_real_insn.
12663 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12666 PR inline-asm/84625
12667 * config/i386/i386.c (ix86_print_operand): Use conditional
12668 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12671 2018-03-02 Richard Biener <rguenther@suse.de>
12673 PR tree-optimization/84427
12674 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12675 (bitmap_set_subtract_values): Rewrite to handle multiple
12678 (prune_clobbered_mems): Likewise.
12679 (phi_translate): Take edge instead of pred/phiblock.
12680 (phi_translate_1): Likewise.
12681 (phi_translate_set): Likewise. Insert all translated
12682 exprs for a value into the set, keeping possibly multiple
12683 expressions per value.
12684 (compute_antic_aux): Adjust for phi_translate changes.
12685 When intersecting union the expressions and prune those
12686 not in the final value set, keeping possibly multiple
12687 expressions per value. Do not use value-insertion
12688 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12689 all expressions. Add verification that the value-sets
12690 only shrink during iteration.
12691 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12692 (do_pre_regular_insertion): Likewise.
12693 (do_pre_partial_partial_insertion): Likewise.
12695 2018-03-02 Richard Biener <rguenther@suse.de>
12698 * config/darwin.c (saved_debug_info_level): New static global.
12699 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12700 (darwin_asm_lto_end): Restore debug info generation settings.
12702 2018-03-01 Martin Liska <mliska@suse.cz>
12705 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12706 volatile arguments.
12708 2018-03-01 Richard Biener <rguenther@suse.de>
12711 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12712 type annotation with LTO when debug was disabled at compile-time.
12714 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12716 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12718 (mips_final_postscan_insn): Likewise.
12720 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12722 PR rtl-optimization/84528
12723 * alias.c (init_alias_target): Add commentary.
12724 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12725 a unique base value if the frame pointer is not eliminated
12726 to the stack pointer.
12728 2018-03-01 Tom de Vries <tom@codesourcery.com>
12730 PR rtl-optimization/83327
12731 * lra-int.h (hard_regs_spilled_into): Declare.
12732 * lra.c (hard_regs_spilled_into): Define.
12733 (init_reg_info): Init hard_regs_spilled_into.
12734 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12735 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12736 (process_bb_lives): Handle hard_regs_spilled_into.
12737 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12738 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12740 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12742 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12743 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12744 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12745 * config/rs6000/aix72.h: New file.
12747 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12749 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12750 instead of warning_at with conditional singular and plural messages
12754 * stor-layout.c (update_alignment_for_field): For
12755 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12756 && !DECL_PACKED (field), do the alignment update, just use
12757 only desired_align instead of MAX (type_align, desired_align)
12759 (place_field): Don't do known_align < desired_align handling
12760 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12761 is non-NULL, instead do it after rli->prev_field handling and
12762 only if not within a bitfield word. For DECL_PACKED (field)
12763 use type_align of BITS_PER_UNIT.
12765 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12767 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12768 superfluous parentheses and trailing spaces.
12770 2018-02-28 Richard Biener <rguenther@suse.de>
12772 PR tree-optimization/84584
12773 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12774 SCoPs with fake exit edge.
12776 2018-02-28 Martin Liska <mliska@suse.cz>
12779 * timevar.c (timer::print): Fix format to properly print 100%
12782 2018-02-28 Richard Biener <rguenther@suse.de>
12784 PR middle-end/84607
12785 * genmatch.c (capture_info::walk_match): Do not mark
12786 captured expressions without operands as expr_p given
12787 they act more like predicates and should be subject to
12788 "lost tail" side-effect preserving.
12790 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12792 PR rtl-optimization/81611
12793 * auto-inc-dec.c (attempt_change): Move dead note from
12794 mem_insn if it's the next use of regno
12795 (find_address): Take address use of reg holding
12796 non-incremented value. Add parm to limit search to the named
12798 (merge_in_block): Attempt to use a mem insn that is the next
12799 use of the original regno.
12801 2018-02-27 Martin Sebor <msebor@redhat.com>
12804 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12805 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12807 2018-02-27 Martin Sebor <msebor@redhat.com>
12809 PR translation/84207
12810 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12811 n argument to unsigned HOST_WIDE_INT.
12812 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12813 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12814 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12815 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12817 2018-02-27 Richard Biener <rguenther@suse.de>
12819 PR tree-optimization/84512
12820 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12821 Do not use the estimate returned from record_stmt_cost for
12822 the scalar iteration cost but sum properly using add_stmt_cost.
12824 2018-02-27 Richard Biener <rguenther@suse.de>
12826 PR tree-optimization/84466
12827 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12828 Adjust last change to less strictly validate use operands.
12830 2018-02-27 Martin Liska <mliska@suse.cz>
12832 PR gcov-profile/84548
12833 * gcov.c (process_file): Allow partial overlap and consider it
12834 also as group functions.
12835 (output_lines): Properly calculate range of lines for a group.
12837 2018-02-27 Martin Liska <mliska@suse.cz>
12839 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12840 'ggc' suffixes. Change first column width.
12841 (timer::print): Fix formatting of the column.
12843 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12845 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12846 preserve inline entry blocks for the sake of debug inline
12847 entry point markers alone.
12848 (remove_unused_locals): Suggest in comments a better place to
12849 force the preservation of inline entry blocks that are
12850 otherwise unused, but do not preserve them.
12852 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12854 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12856 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12859 * config/i386/constraints.md (Bs): Replace
12860 ix86_indirect_branch_register with
12861 TARGET_INDIRECT_BRANCH_REGISTER.
12863 * config/i386/i386.md (indirect_jump): Likewise.
12864 (tablejump): Likewise.
12865 (*sibcall_memory): Likewise.
12866 (*sibcall_value_memory): Likewise.
12867 Peepholes of indirect call and jump via memory: Likewise.
12868 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12869 (*sibcall_value_GOT_32): Likewise.
12870 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12871 (GOT_memory_operand): Likewise.
12872 (call_insn_operand): Likewise.
12873 (sibcall_insn_operand): Likewise.
12874 (GOT32_symbol_operand): Likewise.
12875 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12877 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12879 PR rtl-optimization/83496
12880 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12881 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12882 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12883 redundant insn, if any.
12884 (relax_delay_slots): Likewise.
12885 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12887 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12889 PR tree-optimization/83965
12890 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12891 that grouped statements are part of a reduction chain. Return
12892 true if the statement is not marked as a reduction itself but
12893 is part of a group.
12894 (vect_recog_dot_prod_pattern): Don't check whether the statement
12895 is part of a group here.
12896 (vect_recog_sad_pattern): Likewise.
12897 (vect_recog_widen_sum_pattern): Likewise.
12899 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12902 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12904 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12907 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12909 (ix86_output_indirect_function_return): New prototype.
12910 (ix86_split_simple_return_pop_internal): Likewise.
12911 * config/i386/i386.c (indirect_return_via_cx): New.
12912 (indirect_return_via_cx_bnd): Likewise.
12913 (indirect_thunk_name): Handle return va CX_REG.
12914 (output_indirect_thunk_function): Create alias for
12915 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12916 (ix86_output_indirect_jmp): Remove the bool argument.
12917 (ix86_output_indirect_function_return): New function.
12918 (ix86_split_simple_return_pop_internal): Likewise.
12919 * config/i386/i386.md (*indirect_jump): Don't pass false
12920 to ix86_output_indirect_jmp.
12921 (*tablejump_1): Likewise.
12922 (simple_return_pop_internal): Change it to define_insn_and_split.
12923 Call ix86_split_simple_return_pop_internal to split it for
12924 -mfunction-return=.
12925 (simple_return_indirect_internal): Call
12926 ix86_output_indirect_function_return instead of
12927 ix86_output_indirect_jmp.
12929 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12932 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12933 memset and value initialization afterwards.
12935 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12937 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12939 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12942 * common/config/aarch64/aarch64-common.c
12943 (aarch_option_optimization_table[]): Switch
12944 off fomit-frame-pointer
12946 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12947 Chung-Ju Wu <jasonwucj@gmail.com>
12949 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12951 (store_multiple): Ditto.
12953 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12955 * config.gcc: Add --with-cpu support for nds32 target.
12956 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12957 * config/nds32/nds32.opt: Add -mcpu= option.
12959 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12961 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12962 isel=yes): Warn for these deprecated options.
12964 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12966 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12967 ISA_2_5_MASKS_EMBEDDED.
12969 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12971 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12972 p->max as pointers rather than using iterative_hash_expr.
12974 2018-02-23 Carl Love <cel@us.ibm.com>
12976 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12977 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12979 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12980 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12981 P8V_BUILTIN_VEC_VUNSIGNED2.
12983 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12986 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12987 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12989 * lra-constraints.c (curr_insn_transform): Set up
12990 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12992 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12993 * lra-eliminations.c (spill_pseudos): Ditto.
12994 (process_insn_for_elimination): Ditto.
12995 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12996 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12997 LRA_NON_CLOBBERED_ALT.
12999 2018-02-22 Martin Sebor <msebor@redhat.com>
13001 PR tree-optimization/84480
13002 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13003 to maybe_diag_stxncpy_trunc. Call it.
13004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13005 from gimple_fold_builtin_strcpy. Print inlining stack.
13006 (handle_builtin_stxncpy): Print inlining stack.
13007 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13009 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13012 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13013 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13014 and -fcheck-pointer-bounds are used together.
13015 (indirect_thunk_prefix): New enum.
13016 (indirect_thunk_need_prefix): New function.
13017 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13018 "_nt" instead of "_bnd" for NOTRACK prefix.
13019 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13020 (output_indirect_thunk_function): Likewise.
13022 (ix86_code_end): Update output_indirect_thunk_function calls.
13023 (ix86_output_indirect_branch_via_reg): Replace
13024 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13025 (ix86_output_indirect_branch_via_push): Likewise.
13026 (ix86_output_function_return): Likewise.
13027 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13028 incompatible with -fcf-protection=branch and
13029 -fcheck-pointer-bounds.
13031 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13034 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13035 Change gcc_assert call to output_operand_lossage.
13037 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13039 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13041 2018-02-22 DJ Delorie <dj@redhat.com>
13042 Sebastian Perta <sebastian.perta@renesas.com>
13043 Oleg Endo <olegendo@gcc.gnu.org>
13045 * config/rx/rx.c (rx_rtx_costs): New function.
13046 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13048 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13050 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13052 2018-02-22 Martin Liska <mliska@suse.cz>
13055 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13056 Add "native" as a possible value.
13058 2018-02-22 Martin Liska <mliska@suse.cz>
13061 * config/i386/i386.c (ix86_option_override_internal):
13062 Add "native" as a possible value for -march and -mtune.
13064 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13067 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13068 to all type variants.
13070 PR tree-optimization/84503
13071 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13072 width as info->bitpos + info->bitsize - start.
13073 (merged_store_group::merge_overlapping): Simplify width computation.
13074 (check_no_overlap): New function.
13075 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13076 start + width and last_order of the group, fail if check_no_overlap
13078 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13079 to group if check_no_overlap fails.
13081 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13083 * config/rs6000/altivec.md: Delete contraint arguments to
13084 define_expand, define_split, and define_peephole2, and in
13085 define_insn_and_split if always unused.
13086 * config/rs6000/darwin.md: Ditto.
13087 * config/rs6000/dfp.md: Ditto.
13088 * config/rs6000/rs6000.md: Ditto.
13089 * config/rs6000/sync.md: Ditto.
13090 * config/rs6000/vector.md: Ditto.
13091 * config/rs6000/vsx.md: Ditto.
13093 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13095 * config/rs6000/altivec.md: Write output control strings as braced
13096 blocks instead of double-quoted strings.
13097 * config/rs6000/darwin.md: Ditto.
13098 * config/rs6000/rs6000.md: Ditto.
13099 * config/rs6000/vector.md: Ditto.
13100 * config/rs6000/vsx.md: Ditto.
13102 2018-02-21 Jason Merrill <jason@redhat.com>
13104 PR c++/84314 - ICE with templates and fastcall attribute.
13105 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13107 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13109 * ipa-cp.c (determine_versionability): Fix comment typos.
13111 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13114 * ipa-cp.c (determine_versionability): Do not version functions caling
13117 2018-02-21 Martin Liska <mliska@suse.cz>
13120 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13121 Add "native" as a possible value.
13122 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13123 the macro when native cpu detection is available.
13125 2018-02-21 Martin Liska <mliska@suse.cz>
13128 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13129 Add "native" as a possible value.
13130 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13131 when native cpu detection is available.
13133 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13134 Martin Sebor <msebor@redhat.com>
13136 PR tree-optimization/84478
13137 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13139 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13140 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13141 support which is conservatively correct, for 2 only stay conservative
13142 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13143 argument to the 2 argument get_range_strlen, adjust 6 arg
13144 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13146 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13147 (gimple_fold_builtin_strlen): Pass true as last argument to
13150 2018-02-20 Martin Sebor <msebor@redhat.com>
13152 PR middle-end/84095
13153 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13154 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13155 (builtin_memref::builtin_memref): Factor out parts into
13156 set_base_and_offset and call it.
13158 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13160 PR middle-end/84406
13161 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13162 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13163 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13164 search at the associated MODE_INT.
13166 2018-02-20 Jeff Law <law@redhat.com>
13168 PR middle-end/82123
13169 PR tree-optimization/81592
13170 PR middle-end/79257
13171 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13172 for range data rather than using global data.
13173 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13174 range data rather than using global data.
13175 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13176 pass it to children as needed.
13177 (struct directive::fmtresult): Similarly.
13178 (struct directive::set_width): Similarly.
13179 (struct directive::set_precision): Similarly.
13180 (format_integer, format_directive, parse_directive): Similarly.
13181 (format_none): Accept unnamed vr_values parameter.
13182 (format_percent, format_floating, format_character): Similarly.
13183 (format_string, format_plain): Similarly.
13184 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13185 the EVRP range analyzer for range data rather than using global data.
13186 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13187 gimple-ssa-evrp-analyze.h
13188 (class sprintf_dom_walker): Add after_dom_children member function.
13189 Add evrp_range_analyzer member.
13190 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13191 range analyzer as needed.
13192 (sprintf_dom_walker::after_dom_children): New member function.
13193 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13195 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13196 (evrp_range_analyzer::pop_to_marker): Likewise.
13198 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13200 PR tree-optimization/84419
13201 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13202 with the required type if its current type is compatible but
13205 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13207 PR middle-end/82004
13208 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13209 after vectorization.
13211 2018-02-20 Martin Liska <mliska@suse.cz>
13214 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13215 possible values if we don't have a hint.
13217 2018-02-20 Martin Liska <mliska@suse.cz>
13221 * final.c (shorten_branches): Build align_tab array with one
13223 * opts.c (finish_options): Add alignment option limit check.
13224 (MAX_CODE_ALIGN): Likewise.
13225 (MAX_CODE_ALIGN_VALUE): Likewise.
13226 * doc/invoke.texi: Document maximum allowed option value for
13227 all -falign-* options.
13229 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13232 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13233 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13234 * var-tracking.c (emit_note_insn_var_location): Remove all references
13235 to NOTE_INSN_CALL_ARG_LOCATION.
13236 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13237 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13238 Use copy_rtx_if_shared.
13239 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13240 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13241 (dwarf2out_var_location): Remove handling of
13242 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13244 * final.c (final_scan_insn): Remove all references to
13245 NOTE_INSN_CALL_ARG_LOCATION.
13246 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13247 before dumping final insns.
13248 * except.c (emit_note_eh_region_end): Remove all references to
13249 NOTE_INSN_CALL_ARG_LOCATION.
13250 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13251 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13252 * config/arc/arc.c (hwloop_optimize): Likewise.
13253 * config/arm/arm.c (create_fix_barrier): Likewise.
13254 * config/s390/s390.c (s390_chunkify_start): Likewise.
13255 * config/sh/sh.c (find_barrier): Likewise.
13256 * config/i386/i386.c (rest_of_insert_endbranch,
13257 ix86_seh_fixup_eh_fallthru): Likewise.
13258 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13259 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13260 * config/frv/frv.c (frv_function_prologue): Likewise.
13261 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13263 (note_outside_basic_block_p): Remove all references to
13264 NOTE_INSN_CALL_ARG_LOCATION.
13265 * gengtype.c (adjust_field_rtx_def): Likewise.
13266 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13268 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13269 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13272 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13275 PR tree-optimization/84452
13276 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13277 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13280 2018-02-19 Martin Liska <mliska@suse.cz>
13283 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13285 2018-02-19 Martin Liska <mliska@suse.cz>
13286 Richard Sandiford <richard.sandiford@linaro.org>
13288 PR tree-optimization/82491
13289 * gimple-fold.c (get_base_constructor): Make earlier bail out
13292 2018-02-19 Carl Love <cel@us.ibm.com>
13294 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13295 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13297 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13298 P8V_BUILTIN_VEC_NEG.
13300 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13302 * config/rl78/rl78.md (movdf): New define expand.
13304 2018-02-19 Martin Liska <mliska@suse.cz>
13307 * doc/invoke.texi: Fix typo.
13308 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
13310 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
13312 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
13313 handle rs6000_single_float and rs6000_double_float specially for
13316 2018-02-16 Jeff Law <law@redhat.com>
13318 * config/rx/rx.c (add_pop_cfi_notes): New function.;
13319 (pop_regs): Use it.
13321 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13324 * ipa-inline.c (inline_small_functions): Fix a typo.
13326 2018-02-16 Nathan Sidwell <nathan@acm.org>
13328 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
13330 2018-02-16 Carl Love <cel@us.ibm.com>
13332 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
13333 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
13334 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
13335 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
13336 expansion to P8V_BUILTIN_VEC_FLOAT2.
13338 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
13340 PR rtl-optimization/70023
13341 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
13342 src_regno into account.
13344 2018-02-16 Carl Love <cel@us.ibm.com>
13346 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
13347 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
13348 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
13349 * config/rs6000/rs6000.c: Remove case statements for
13350 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
13351 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
13352 and P9V_BUILTIN_VEC_VINSERT4B.
13353 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
13354 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
13355 * config/rs6000/vsx.md:
13356 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
13359 2018-02-16 Carl Love <cel@us.ibm.com>
13361 * config/rs6000/altivec.h: Add builtin names vec_extract4b
13363 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
13365 * config/rs6000/rs6000-c.c: Add the definitions for
13366 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
13367 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
13368 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
13369 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
13370 definition for insert4b and define insn *insert3b_internal.
13371 * doc/extend.texi: Add documentation for vec_extract4b.
13373 2018-02-16 Nathan Sidwell <nathan@acm.org>
13375 * doc/extend.texi (Backwards Compatibility): Mention friend
13376 injection. Note for-scope is deprecated.
13377 * doc/invoke.texi (-ffriend-injection): Deprecate.
13379 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
13381 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
13382 that moved to I2, also allow destinations that are a paradoxical
13383 subreg (instead of a normal reg).
13385 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
13388 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
13391 2018-02-16 Richard Biener <rguenther@suse.de>
13393 PR tree-optimization/84037
13394 PR tree-optimization/84016
13396 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13397 Adjust vec_construct for the fact we need additional higher latency
13398 128bit inserts for AVX256 and AVX512 vector builds.
13399 (ix86_add_stmt_cost): Scale vector construction cost for
13402 2018-02-16 Richard Biener <rguenther@suse.de>
13404 PR tree-optimization/84417
13405 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
13406 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
13407 (non_rewritable_lvalue_p): Likewise, use poly-ints.
13409 2018-02-16 Martin Liska <mliska@suse.cz>
13412 * internal-fn.def (ASAN_CHECK): Set proper flags.
13413 (ASAN_MARK): Likewise.
13415 2018-02-16 Julia Koval <julia.koval@intel.com>
13417 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
13418 from PTA_CANNONLAKE.
13420 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13423 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
13424 Use ++iter rather than iter++ for std::list iterators.
13425 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
13426 defer deleting them until all nodes in the forest are processed. Do
13427 free even leaf nodes. Change to_process into auto_vec.
13430 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13431 * vec.h (vec_default_construct): Use memset instead of placement new
13432 if BROKEN_VALUE_INITIALIZATION is defined.
13433 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13434 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13437 PR rtl-optimization/83723
13438 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13439 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13440 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13442 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13444 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13446 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13448 PR rtl-optimization/81443
13449 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13450 from inner REGs to paradoxical SUBREGs.
13452 2018-02-16 Richard Biener <rguenther@suse.de>
13454 PR tree-optimization/84399
13455 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13456 For operands we can analyze at their definition make sure we can
13457 analyze them at each use as well.
13459 2018-02-16 Richard Biener <rguenther@suse.de>
13461 PR tree-optimization/84190
13462 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13463 volatile accesses if the decl isn't volatile.
13465 2018-02-15 Jason Merrill <jason@redhat.com>
13467 PR c++/84314 - ICE with templates and fastcall attribute.
13468 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13469 TYPE_CANONICAL on an existing type.
13471 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13473 PR tree-optimization/84383
13474 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13475 dstoff nor call operand_equal_p if dstbase is NULL.
13477 PR tree-optimization/84334
13478 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13479 also a CONSTANT_CLASS_P, punt.
13481 2018-02-14 Jim Wilson <jimw@sifive.com>
13483 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13484 first SMALL_OPERAND check. New local min_second_step. Move assert
13485 to where locals are set. Add TARGET_RVC support.
13486 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13488 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13490 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13492 2018-02-14 Martin Sebor <msebor@redhat.com>
13494 PR tree-optimization/83698
13495 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13496 arrays constrain the offset range to their bounds.
13497 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13498 (builtin_access::overlap): Avoid setting the size of overlap if it's
13500 (maybe_diag_overlap): Also consider arrays when deciding what values
13501 of offsets to include in diagnostics.
13503 2018-02-14 Martin Sebor <msebor@redhat.com>
13506 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13507 that correspond to the kind of a declaration.
13509 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13512 * config/pa/pa.md: Load address of PIC label using the linkage table
13513 if the label is nonlocal.
13515 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13517 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13518 warning message if user requests -maltivec=be.
13519 * doc/invoke.texi: Document deprecation of -maltivec=be.
13521 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13524 * config/rs6000/rs6000-c.c: Update definitions for
13525 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13526 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13528 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13531 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13532 add _get_ssp intrinsics. Remove argument from
13533 __builtin_ia32_rdssp[d|q].
13534 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13535 * config/i386/i386-builtin.def: Remove argument from
13536 __builtin_ia32_rdssp[d|q].
13537 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13538 ix86_expand_special_args_builtin for _rdssp[d|q].
13539 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13540 Clear register before usage.
13541 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13542 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13544 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13546 PR tree-optimization/84357
13547 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13548 operand 1 of an ARRAY_REF too.
13550 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13553 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13554 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13556 (set_of_reg): New struct.
13557 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13558 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13559 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13561 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13562 Split into bitclr, bitset, bitinvert patterns if appropriate.
13563 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13564 use rx_fuse_in_memory_bitop.
13565 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13566 to named insn, correct maximum insn length.
13568 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13571 * machmode.def: Define a complex mode for PARTIAL_INT.
13572 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13574 * doc/rtl.texi: Document CSPImode.
13575 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13577 (msp430_hard_regno_nregs_with_padding): Likewise.
13579 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13582 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13584 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13586 PR rtl-optimization/84169
13587 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13588 we generated a parallel as new i3 and we split that to new i2 and i3
13589 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13590 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13591 those to i2, not i1. Partially rewrite this scan code.
13593 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13596 * stor-layout.c (place_field): For variable length fields, adjust
13597 offset_align afterwards not just based on the field's alignment,
13598 but also on the size.
13600 PR middle-end/84309
13601 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13602 of exps and logs in the use_exp2 case.
13604 2018-02-13 Jeff Law <law@redhat.com>
13606 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13607 entry for "vector".
13609 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13612 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13616 * common.opt (gas-loc-support, gas-locview-support): New.
13617 (ginline-points, ginternal-reset-location-views): New.
13618 * doc/invoke.texi: Document them. Use @itemx where intended.
13619 (gvariable-location-views): Adjust.
13620 * target.def (reset_location_view): New.
13621 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13622 (TARGET_RESET_LOCATION_VIEW): New.
13623 * doc/tm.texi: Rebuilt.
13624 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13625 (dwarf2out_default_as_locview_support): New.
13626 (output_asm_line_debug_info): Use option variables.
13627 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13628 (output_loc_list): Likewise.
13629 (add_high_low_attributes): Check option variables.
13630 Don't output entry view attribute in strict mode.
13631 (gen_inlined_subroutine_die): Check option variables.
13632 (dwarf2out_inline_entry): Likewise.
13633 (init_sections_and_labels): Likewise.
13634 (dwarf2out_early_finish): Likewise.
13635 (maybe_reset_location_view): New, from...
13636 (dwarf2out_var_location): ... here. Call it.
13637 * debug.h (dwarf2out_default_as_loc_support): Declare.
13638 (dwarf2out_default_as_locview_support): Declare.
13639 * hooks.c (hook_int_rtx_insn_0): New.
13640 * hooks.h (hook_int_rtx_insn_0): Declare.
13641 * toplev.c (process_options): Take -gas-loc-support and
13642 -gas-locview-support from dwarf2out. Enable
13643 -gvariable-location-views by default only with locview
13644 assembler support. Enable -ginternal-reset-location-views by
13645 default only if the target defines the corresponding hook.
13646 Enable -ginline-points by default if location views are
13647 enabled; force it disabled if statement frontiers are
13649 * tree-inline.c (expand_call_inline): Check option variables.
13650 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13652 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13654 PR tree-optimization/84321
13655 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13656 handling. Also check whether the anti-range contains any values
13657 that satisfy the mask; switch to a VR_RANGE if not.
13659 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13662 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13664 2018-02-13 Martin Jambor <mjambor@suse.cz>
13667 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13668 of call statements, also set location of a load to a temporary.
13670 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13672 * config/rl78/rl78.c (add_vector_labels): New function.
13673 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13674 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13675 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13676 which checks that no arguments are passed.
13677 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13678 * doc/extend.texi: Documentation for the new attribute.
13680 2018-02-13 Andreas Schwab <schwab@suse.de>
13682 * config/riscv/linux.h (CPP_SPEC): Define.
13684 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13687 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13688 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13689 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13690 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13691 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13692 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13693 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13694 temporarily for AES and PCLMUL builtins.
13696 PR tree-optimization/84339
13697 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13698 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13701 PR middle-end/84309
13702 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13703 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13704 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13706 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13708 * omp-simd-clone.h: New file.
13709 * omp-simd-clone.c: Include omp-simd-clone.h.
13710 (expand_simd_clones): No longer static.
13711 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13712 cgraph.h and omp-simd-clone.h.
13713 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13714 (vect_recog_widen_shift_pattern): Formatting fix.
13715 (vect_pattern_recog_1): Don't check optab for calls.
13718 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13719 operands[2] into a REG before using gen_lowpart on it.
13721 2018-02-12 Jeff Law <law@redhat.com>
13724 * config/sh/sh.c (find_barrier): Consider a sibling call
13727 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13728 successfully back substituting a reg.
13730 2018-02-12 Richard Biener <rguenther@suse.de>
13732 PR tree-optimization/84037
13733 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13734 parameter, move visited init to caller.
13735 (vect_slp_analyze_operations): Separate cost from validity
13736 check, initialize visited once for all instances.
13737 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13739 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13741 (vect_model_promotion_demotion_cost): Likewise.
13742 (vectorizable_bswap): Guard cost modeling with !slp_node
13743 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13745 (vectorizable_call): Likewise.
13746 (vectorizable_conversion): Likewise.
13747 (vectorizable_assignment): Likewise.
13748 (vectorizable_shift): Likewise.
13749 (vectorizable_operation): Likewise.
13750 (vectorizable_store): Likewise.
13751 (vectorizable_load): Likewise.
13752 (vectorizable_condition): Likewise.
13753 (vectorizable_comparison): Likewise.
13755 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13758 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13759 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13760 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13762 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13764 PR middle-end/83665
13765 * params.def (inline-min-speedup): Increase from 8 to 15.
13766 (max-inline-insns-auto): Decrease from 40 to 30.
13767 * ipa-split.c (consider_split): Add some buffer for function to
13768 be considered inlining candidate.
13769 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13772 2018-02-12 Richard Biener <rguenther@suse.de>
13774 PR tree-optimization/84037
13775 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13776 matched stmts if we cannot swap the non-matched ones.
13778 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13780 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13781 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13782 _mm_maskz_scalef_round_ss): New intrinsics.
13783 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13784 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13785 __builtin_ia32_scalefss_round): Remove.
13786 (__builtin_ia32_scalefsd_mask_round,
13787 __builtin_ia32_scalefss_mask_round): New intrinsics.
13788 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13789 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13790 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13791 "<round_constraint>")): Changed to ...
13792 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13793 "<round_scalar_constraint>")): ... this.
13794 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13795 %0, %1, %2<round_op3>}"): Changed to ...
13796 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13797 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13798 %2<round_scalar_mask_op3>}"): ... this.
13799 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13801 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13803 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13804 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13805 (_mm_maskz_sqrt_round_ss): New intrinsics.
13806 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13807 (__builtin_ia32_sqrtsd_mask_round)
13808 (__builtin_ia32_sqrtss_mask_round): New builtins.
13809 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13810 (__builtin_ia32_sqrtss_round): Remove.
13811 (__builtin_ia32_sqrtsd_mask_round)
13812 (__builtin_ia32_sqrtss_mask_round): New builtins.
13813 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13814 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13815 ((match_operand:VF_128 1 "vector_operand"
13816 "xBm,<round_constraint>")): Changed to ...
13817 ((match_operand:VF_128 1 "vector_operand"
13818 "xBm,<round_scalar_constraint>")): ... this.
13819 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13820 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13821 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13822 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13823 %<iptr>1<round_scalar_mask_op3>}): ... this.
13824 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13825 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13827 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13830 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13831 Cast vec_cmpeq result to correct type.
13832 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13833 Cast vec_cmpgt result to correct type.
13835 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13837 * final.c (final_scan_insn_1): Renamed from...
13838 (final_scan_insn): ... this. New wrapper, to recover
13839 seen from the outermost call in recursive ones.
13840 * config/sparc/sparc.c (output_return): Drop seen from call.
13841 (output_sibcall): Likewise.
13842 * config/visium/visium.c (output_branch): Likewise.
13844 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13846 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13849 2018-02-10 Alan Modra <amodra@gmail.com>
13852 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13853 Specify LR as an input.
13855 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13858 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13859 remove_member_access_dummy_vars): New functions.
13860 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13861 lower_omp_1, execute_lower_omp): Use them.
13863 PR rtl-optimization/84308
13864 * shrink-wrap.c (spread_components): Release todo vector.
13866 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13868 PR rtl-optimization/57193
13869 * ira-color.c (struct allocno_color_data): Add member
13870 conflict_allocno_hard_prefs.
13871 (update_conflict_allocno_hard_prefs): New.
13872 (bucket_allocno_compare_func): Add a preference based on
13873 conflict_allocno_hard_prefs.
13874 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13875 (color_allocnos): Remove a dead code. Initiate
13876 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13878 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13881 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13882 constraint from =wa to wa. Avoid a subreg on the output operand,
13883 instead use a pseudo and subreg it in a move.
13884 (p9_xxbrd_<mode>): Changed to ...
13885 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13886 (p9_xxbrd_v2df): New expander.
13887 (p9_xxbrw_<mode>): Changed to ...
13888 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13889 (p9_xxbrw_v4sf): New expander.
13891 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13893 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13895 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13898 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13899 multiply in 32-bit mode.
13900 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13901 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13904 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13906 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13907 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13908 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13909 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13911 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13914 * dwarf2out.c (is_trivial_indirect_ref): New function.
13915 (dwarf2out_late_global_decl): Do not generate a location
13916 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13917 and that are not defined in the current unit.
13919 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13921 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13922 instead of a libcall for UNORDERED.
13924 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13927 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13928 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13930 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13933 * config/s390/s390.c (s390_set_current_function): Invoke
13934 s390_indirect_branch_settings also if fndecl didn't change.
13936 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13938 * config/rs6000/rs6000.md (blockage): Set length to zero.
13940 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13942 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13944 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13947 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13948 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13952 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13953 PARALLEL incoming that failed vt_get_decl_and_offset check.
13955 PR middle-end/84237
13956 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13957 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13959 (get_variable_section): For decls in named .bss* sections pass true as
13960 second argument to bss_initializer_p.
13962 2018-02-09 Marek Polacek <polacek@redhat.com>
13963 Jakub Jelinek <jakub@redhat.com>
13966 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13967 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13968 Sync some changes from cxx_fold_indirect_ref.
13970 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13972 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13974 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13975 (BLOCK_INLINE_ENTRY_LABEL): New.
13976 (dwarf2out_var_location): Disregard inline entry markers.
13977 (inline_entry_data): New struct.
13978 (inline_entry_data_hasher): New hashtable type.
13979 (inline_entry_data_hasher::hash): New.
13980 (inline_entry_data_hasher::equal): New.
13981 (inline_entry_data_table): New variable.
13982 (add_high_low_attributes): Add DW_AT_entry_pc and
13983 DW_AT_GNU_entry_view attributes if a pending entry is found
13984 in inline_entry_data_table. Add old entry_pc attribute only
13985 if debug nonbinding markers are disabled.
13986 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13987 markers are enabled.
13988 (block_within_block_p, dwarf2out_inline_entry): New.
13989 (dwarf2out_finish): Check that no entries remained in
13990 inline_entry_data_table.
13991 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13992 (final_scan_insn, notice_source_line): Likewise.
13993 (rest_of_clean_state): Skip inline entry markers.
13994 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13996 * gimple.c (gimple_build_debug_inline_entry): New.
13997 * gimple.h (enum gimple_debug_subcode): Add
13998 GIMPLE_DEBUG_INLINE_ENTRY.
13999 (gimple_build_debug_inline_entry): Declare.
14000 (gimple_debug_inline_entry_p): New.
14001 (gimple_debug_nonbind_marker_p): Adjust.
14002 * insn-notes.def (INLINE_ENTRY): New.
14003 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14004 inline entry marker notes.
14005 (print_insn): Likewise.
14006 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14007 (INSN_DEBUG_MARKER_KIND): Likewise.
14008 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14009 * tree-inline.c (expand_call_inline): Build and insert
14010 debug_inline_entry stmt.
14011 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14012 inline entry blocks early, if nonbind markers are enabled.
14013 (dump_scope_block): Dump fragment info.
14014 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14015 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14016 (gimple_build_debug_inline_entry): New.
14017 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14018 Enable/disable inline entry points too.
14019 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14020 (DEBUG_INSN): Describe inline entry markers.
14022 * common.opt (gvariable-location-views): New.
14023 (gvariable-location-views=incompat5): New.
14024 * config.in: Rebuilt.
14025 * configure: Rebuilt.
14026 * configure.ac: Test assembler for view support.
14027 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14028 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14029 * dwarf2out.c (var_loc_view): New typedef.
14030 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14031 (dwarf2out_locviews_in_attribute): New.
14032 (dwarf2out_locviews_in_loclist): New.
14033 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14034 (enum dw_line_info_opcode): Add LI_adv_address.
14035 (struct dw_line_info_table): Add view.
14036 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14037 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14038 (zero_view_p): New variable.
14039 (ZERO_VIEW_P): New macro.
14040 (output_asm_line_debug_info): New.
14041 (struct var_loc_node): Add view.
14042 (add_AT_view_list, AT_loc_list): New.
14043 (add_var_loc_to_decl): Add view param. Test it against last.
14044 (new_loc_list): Add view params. Record them.
14045 (AT_loc_list_ptr): Handle loc and view lists.
14046 (view_list_to_loc_list_val_node): New.
14047 (print_dw_val): Handle dw_val_class_view_list.
14048 (size_of_die): Likewise.
14049 (value_format): Likewise.
14050 (loc_list_has_views): New.
14051 (gen_llsym): Set vl_symbol too.
14052 (maybe_gen_llsym, skip_loc_list_entry): New.
14053 (dwarf2out_maybe_output_loclist_view_pair): New.
14054 (output_loc_list): Output view list or entries too.
14055 (output_view_list_offset): New.
14056 (output_die): Handle dw_val_class_view_list.
14057 (output_dwarf_version): New.
14058 (output_compilation_unit_header): Use it.
14059 (output_skeleton_debug_sections): Likewise.
14060 (output_rnglists, output_line_info): Likewise.
14061 (output_pubnames, output_aranges): Update version comments.
14062 (output_one_line_info_table): Output view numbers in asm comments.
14063 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14064 Call maybe_gen_llsym.
14065 (loc_list_from_tree_1): Adjust.
14066 (add_AT_location_description): Create view list attribute if
14067 needed, check it's absent otherwise.
14068 (convert_cfa_to_fb_loc_list): Adjust.
14069 (maybe_emit_file): Call output_asm_line_debug_info for test.
14070 (dwarf2out_var_location): Reset views as needed. Precompute
14071 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14072 attribute. Set view.
14073 (new_line_info_table): Reset next view.
14074 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14075 (dwarf2out_source_line): Likewise. Output view resets and labels to
14076 the assembler, or select appropriate line info opcodes.
14077 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14078 (optimize_string_length): Catch it. Adjust.
14079 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14080 dw_val_class_view_list, and remove it if no longer needed.
14081 (hash_loc_list): Hash view numbers.
14082 (loc_list_hasher::equal): Compare them.
14083 (optimize_location_lists): Check whether a view list symbol is
14084 needed, and whether the locview attribute is present, and
14085 whether they match. Remove the locview attribute if no longer
14087 (index_location_lists): Call skip_loc_list_entry for test.
14088 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14089 Use output_dwarf_version.
14090 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14091 (struct dw_val_node): Add val_view_list.
14092 * final.c (SEEN_NEXT_VIEW): New.
14093 (set_next_view_needed): New.
14094 (clear_next_view_needed): New.
14095 (maybe_output_next_view): New.
14096 (final_start_function): Rename to...
14097 (final_start_function_1): ... this. Take pointer to FIRST,
14098 add SEEN parameter. Emit param bindings in the initial view.
14099 (final_start_function): Reintroduce SEEN-less interface.
14100 (final): Rename to...
14101 (final_1): ... this. Take SEEN parameter. Output final pending
14102 next view at the end.
14103 (final): Reintroduce seen-less interface.
14104 (final_scan_insn): Output pending next view before switching
14105 sections or ending a block. Mark the next view as needed when
14106 outputting variable locations. Notify debug backend of section
14107 changes, and of location view changes.
14108 (rest_of_handle_final): Adjust.
14109 * toplev.c (process_options): Autodetect value for debug variable
14110 location views option. Warn on incompat5 without -gdwarf-5.
14111 * doc/invoke.texi (gvariable-location-views): New.
14112 (gvariable-location-views=incompat5): New.
14113 (gno-variable-location-views): New.
14115 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14117 PR tree-optimization/84136
14118 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14119 that the result of find_edge is non-NULL.
14121 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14124 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14125 storing integer register in SImode. Fix cost of 256 and 512
14126 byte aligned SSE register store.
14128 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14130 * config/i386/i386.c (ix86_multiplication_cost): Fix
14131 multiplication cost for TARGET_AVX512DQ.
14133 2018-02-08 Marek Polacek <polacek@redhat.com>
14135 PR tree-optimization/84238
14136 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14139 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14141 PR tree-optimization/84265
14142 * tree-vect-stmts.c (vectorizable_store): Don't treat
14143 VMAT_CONTIGUOUS accesses as grouped.
14144 (vectorizable_load): Likewise.
14146 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14148 PR tree-optimization/81635
14149 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14150 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14151 (test_round_for_mask): New functions.
14152 (wide_int_cc_tests): Call test_round_for_mask.
14153 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14154 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14155 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14156 range returned by get_range_info.
14158 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14161 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14162 * symtab.c: Include builtins.h
14163 (symtab_node::output_to_lto_symbol_table_p): Move here
14164 from lto-streamer-out.c:output_symbol_p.
14165 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14166 (output_symbol_p): Move all logic to symtab.c
14167 (produce_symtab): Update.
14169 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14171 * config/s390/s390-opts.h (enum indirect_branch): Define.
14172 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14173 (s390_indirect_branch_via_thunk)
14174 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14175 (enum s390_indirect_branch_type): Define.
14176 * config/s390/s390.c (struct s390_frame_layout, struct
14177 machine_function): Remove.
14178 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14179 (indirect_branch_table_label_no, indirect_branch_table_name):
14181 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14182 (enum s390_indirect_branch_option): Define.
14183 (s390_return_addr_from_memory): New function.
14184 (s390_handle_string_attribute): New function.
14185 (s390_attribute_table): Add new attribute handler.
14186 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14187 (s390_indirect_branch_via_thunk): New function.
14188 (s390_indirect_branch_via_inline_thunk): New function.
14189 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14190 sibling call optimization for non z10 compiles.
14191 (s390_emit_call): Force indirect branch target to be a single
14192 register. Add r1 clobber for non-z10 compiles.
14193 (s390_emit_epilogue): Emit return jump via return_use expander.
14194 (s390_reorg): Handle JUMP_INSNs as execute targets.
14195 (s390_option_override_internal): Perform validity checks for the
14196 new command line options.
14197 (s390_indirect_branch_attrvalue): New function.
14198 (s390_indirect_branch_settings): New function.
14199 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14200 (s390_output_indirect_thunk_function): New function.
14201 (s390_code_end): Implement target hook.
14202 (s390_case_values_threshold): Implement target hook.
14203 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14205 * config/s390/s390.h (struct s390_frame_layout)
14206 (struct machine_function): Move here from s390.c.
14207 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14208 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14209 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14210 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14211 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14212 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14213 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14214 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14215 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14216 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14217 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14218 (mnemonic attribute): Add values which aren't recognized
14220 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14221 pattern for branch conversion. Fix mnemonic attribute.
14222 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14223 indirect branch via thunk if requested.
14224 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14225 ("*indirect_jump"): Disable for branch conversion using out of
14227 ("indirect_jump_via_thunk<mode>_z10")
14228 ("indirect_jump_via_thunk<mode>")
14229 ("indirect_jump_via_inlinethunk<mode>_z10")
14230 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14231 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14232 ("casesi_jump_via_inlinethunk<mode>_z10")
14233 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14234 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14235 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14236 ("*indirect2_jump"): Disable for branch conversion.
14237 ("casesi_jump"): Turn into expander and expand patterns for branch
14239 ("return_use"): New expander.
14240 ("*return"): Emit return via thunk and rename it to ...
14241 ("*return<mode>"): ... this one.
14242 * config/s390/s390.opt: Add new options and and enum for the
14245 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14247 * lra-constraints.c (match_reload): Unconditionally use
14248 gen_lowpart_SUBREG, rather than selecting between that
14249 and equivalent gen_rtx_SUBREG code.
14251 2018-02-08 Richard Biener <rguenther@suse.de>
14253 PR tree-optimization/84233
14254 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14255 changed flag instead of boguously re-using phi_inserted.
14257 2018-02-08 Martin Jambor <mjambor@suse.cz>
14259 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14260 static local variables.
14262 2018-02-08 Richard Biener <rguenther@suse.de>
14264 PR tree-optimization/84278
14265 * tree-vect-stmts.c (vectorizable_store): When looking for
14266 smaller vector types to perform grouped strided loads/stores
14267 make sure the mode is supported by the target.
14268 (vectorizable_load): Likewise.
14270 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14272 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14273 Increase LDP/STP opportunities by adding adjacent callee-saves.
14275 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14277 PR rtl-optimization/84068
14278 PR rtl-optimization/83459
14279 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14281 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14283 PR tree-optimization/84224
14284 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14285 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14286 non-zero arguments.
14288 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14291 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14292 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14293 count, remove one USE.
14295 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14297 * doc/install.texi (Configuration): Document the
14298 --with-long-double-format={ibm,ieee} PowerPC configuration
14302 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14303 Convert from define_expand to be define_insn_and_split. Rework
14304 float/double/_Float128 conversions to QI/HI/SImode to work with
14305 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14306 conversions to QI/HImode types did a store and then a load to
14307 truncate the value. For conversions to VSX registers, don't split
14308 the insn, instead emit the code directly. Use the code iterator
14309 any_fix to combine signed and unsigned conversions.
14310 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
14311 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
14312 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14313 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
14314 (fix_<mode>di2_hw): Likewise.
14315 (fixuns_<mode>di2_hw): Likewise.
14316 (fix_<mode>si2_hw): Likewise.
14317 (fixuns_<mode>si2_hw): Likewise.
14318 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
14319 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14320 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
14321 fix<uns>_trunc<SFDF:mode>si2_p8.
14322 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
14323 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
14324 (fix<uns>_<mode>_mem): Likewise.
14325 (fctiw<u>z_<mode>_mem): Likewise.
14326 (fix<uns>_<mode>_mem): Likewise.
14327 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
14328 the register allocator from doing a direct move to the GPRs to do
14329 a store, and instead use the ISA 3.0 store byte/half-word from
14330 vector register instruction. For IEEE 128-bit floating point,
14331 also optimize stores of 32-bit ints.
14332 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
14334 2018-02-07 Alan Hayward <alan.hayward@arm.com>
14336 * genextract.c (push_pathstr_operand): New function to support
14338 (walk_rtx): Call push_pathstr_operand.
14339 (print_path): Support [a-zA-Z].
14341 2018-02-07 Richard Biener <rguenther@suse.de>
14343 PR tree-optimization/84037
14344 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
14345 (cse_and_gimplify_to_preheader): Declare.
14346 (vect_get_place_in_interleaving_chain): Likewise.
14347 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14349 (_loop_vec_info::~_loop_vec_info): Delete it.
14350 (cse_and_gimplify_to_preheader): New function.
14351 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
14352 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
14353 (vectorizable_load): Likewise. For grouped stores always base
14354 the IV on the first element.
14355 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
14356 condition before gimplifying.
14358 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14360 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
14361 *DIV_EXPR and *MOD_EXPR.
14363 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
14366 * config/i386/i386.c (ix86_option_override_internal): Mask out
14367 the CF_SET bit when checking -fcf-protection.
14369 2018-02-07 Tom de Vries <tom@codesourcery.com>
14372 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
14375 2018-02-07 Richard Biener <rguenther@suse.de>
14377 PR tree-optimization/84204
14378 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
14381 PR tree-optimization/84205
14382 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
14383 special-case isl_ast_op_zdiv_r.
14385 PR tree-optimization/84223
14386 * graphite-scop-detection.c (gather_bbs::before_dom_children):
14387 Only add conditions from within the region.
14388 (gather_bbs::after_dom_children): Adjust.
14390 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
14393 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
14394 * config/avr/avr.md: Only post-reload split REG-REG moves if
14395 either register is GENERAL_REG_P.
14397 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14399 PR tree-optimization/84235
14400 * tree-ssa-scopedtables.c
14401 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
14402 if the subtraction is performed in floating point type where NaNs are
14403 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
14404 build 1. Formatting fix.
14406 2018-02-06 Jakub Jelinek <jakub@redhat.com>
14409 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
14410 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
14411 and skip it regardless of bb boundaries. Use CALL_P macro,
14412 don't test INSN_P (insn) together with CALL_P or JUMP_P check
14413 unnecessarily, formatting fix.
14415 2018-02-06 Michael Collison <michael.collison@arm.com>
14417 * config/arm/thumb2.md:
14418 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
14419 (*thumb_mov_notscc): Ditto.
14421 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
14424 * config/rs6000/rs6000.md (su code attribute): Use "u" for
14425 unsigned_fix, not "s".
14427 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14429 * configure.ac (gcc_fn_eh_frame_ro): New function.
14430 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14431 correct .eh_frame permissions.
14432 * configure: Regenerate.
14434 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14436 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14437 irrelevant options.
14439 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14441 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14442 Display warning message for -mno-speculate-indirect-jumps.
14444 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14446 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14448 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14450 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14452 PR tree-optimization/84225
14453 * tree-eh.c (find_trapping_overflow): Only call
14454 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14456 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14459 * config/i386/i386.c: Reimplement the check of possible options
14460 -mibt/-mshstk conbination. Change error messages.
14461 * doc/invoke.texi: Fix a typo: remove extra '='.
14463 2018-02-06 Marek Polacek <polacek@redhat.com>
14465 PR tree-optimization/84228
14466 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14468 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14471 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14472 emitted arch directives.
14473 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14474 __ARM_FEATURE_COPROC before changing architectures.
14476 2018-02-06 Richard Biener <rguenther@suse.de>
14478 * config/i386/i386.c (print_reg): Fix typo.
14479 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14481 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14483 * configure: Regenerate.
14485 2018-02-05 Martin Sebor <msebor@redhat.com>
14487 PR tree-optimization/83369
14488 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14491 2018-02-05 Martin Liska <mliska@suse.cz>
14493 * doc/invoke.texi: Cherry-pick upstream r323995.
14495 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14497 * ira.c (ira_init_register_move_cost): Adjust comment.
14499 2018-02-05 Martin Liska <mliska@suse.cz>
14501 PR gcov-profile/84137
14502 * doc/gcov.texi: Fix typo in documentation.
14504 2018-02-05 Martin Liska <mliska@suse.cz>
14506 PR gcov-profile/83879
14507 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14508 using dlopen functionality.
14510 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14512 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14513 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14514 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14515 _mm_maskz_range_round_ss): New intrinsics.
14516 (__builtin_ia32_rangesd128_round)
14517 (__builtin_ia32_rangess128_round): Remove.
14518 (__builtin_ia32_rangesd128_mask_round,
14519 __builtin_ia32_rangess128_mask_round): New builtins.
14520 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14521 __builtin_ia32_rangess128_round): Remove.
14522 (__builtin_ia32_rangesd128_mask_round,
14523 __builtin_ia32_rangess128_mask_round): New builtins.
14524 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14525 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14526 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14527 "<round_saeonly_constraint>")): Changed to ...
14528 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14529 "<round_saeonly_scalar_constraint>")): ... this.
14530 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14531 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14532 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14533 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14534 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14536 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14538 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14540 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14541 Remove all values except native, 8540 and 8548.
14543 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14545 * config/i386/i386.c (ix86_output_function_return): Pass
14546 INVALID_REGNUM, instead of -1, as invalid register number to
14547 indirect_thunk_name and output_indirect_thunk.
14549 2018-02-02 Julia Koval <julia.koval@intel.com>
14551 * config.gcc: Add -march=icelake.
14552 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14553 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14554 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14555 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14556 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14557 (processor_target_table): Add icelake.
14558 (ix86_option_override_internal): Handle new PTAs.
14559 (get_builtin_code_for_version): Handle icelake.
14560 (M_INTEL_COREI7_ICELAKE): New.
14561 (fold_builtin_cpu): Handle icelake.
14562 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14563 * doc/invoke.texi: Add -march=icelake.
14565 2018-02-02 Julia Koval <julia.koval@intel.com>
14567 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14568 to wide_int_bitmask.
14569 * wide-int-bitmask.h: New.
14571 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14574 * config/i386/i386.md: Replace Pmode with word_mode in
14575 builtin_setjmp_setup and builtin_longjmp to support x32.
14577 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14581 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14583 (rs6000_supported_cpu_names): New static variable.
14584 (linux_cpu_translation_table): Likewise.
14585 (elf_platform) <cpu>: Define new static variable and use it.
14586 Translate kernel AT_PLATFORM name to canonical name if needed.
14587 Error if platform name is unknown.
14589 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14592 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14594 2018-02-01 Jeff Law <law@redhat.com>
14597 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14598 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14599 the scratch if RELEASE_VIA_POP is false.
14600 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14601 If we have to save a temporary register, decrement SIZE appropriately.
14602 Pass new arguments to release_scratch_register_on_entry.
14603 (ix86_adjust_stack_and_probe): Likewise.
14604 (ix86_emit_probe_stack_range): Pass new arguments to
14605 release_scratch_register_on_entry.
14607 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14609 PR rtl-optimization/84157
14610 * combine.c (change_zero_ext): Use REG_P predicate in
14611 front of HARD_REGISTER_P predicate.
14613 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14615 * config/avr/avr.c (avr_option_override): Move disabling of
14616 -fdelete-null-pointer-checks to...
14617 * common/config/avr/avr-common.c (avr_option_optimization_table):
14620 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14622 PR tree-optimization/81635
14623 * tree-data-ref.c (split_constant_offset_1): For types that
14624 wrap on overflow, try to use range info to prove that wrapping
14627 2018-02-01 Renlin Li <renlin.li@arm.com>
14630 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14631 TAILCALL_ADDR_REGS.
14632 (aarch64_register_move_cost): Likewise.
14633 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14634 TAILCALL_ADDR_REGS.
14635 (REG_CLASS_NAMES): Likewise.
14636 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14637 TAILCALL_ADDR_REGS. Remove IP registers.
14638 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14640 2018-02-01 Richard Biener <rguenther@suse.de>
14642 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14643 for specifying RPO order and allow NULL for that.
14644 * domwalk.c (dom_walker::dom_walker): Likewise.
14645 (dom_walker::walk): Handle NULL RPO order.
14646 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14648 (rewrite_update_dom_walker): Likewise.
14649 (mark_def_dom_walker): Likewise.
14651 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14653 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14654 (aarch64_maybe_expand_sve_subreg_move): Declare.
14655 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14656 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14658 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14659 that are semantically a reverse operation.
14660 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14661 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14662 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14664 (aarch64_can_change_mode_class): For big-endian, forbid changes
14665 between two SVE modes if they have different element sizes.
14667 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14669 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14670 the TImode handling for big-endian targets.
14672 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14674 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14675 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14677 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14678 Remove BSWAP handing for big-endian targets and use the form of
14679 LD1RQ appropariate for the mode.
14681 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14683 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14684 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14685 duplicated element.
14687 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14690 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14691 check for operands that need to go through aarch64_sve_reload_be.
14693 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14695 PR tree-optimization/81661
14696 PR tree-optimization/84117
14697 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14698 * tree-eh.c: Include gimplify.h.
14699 (find_trapping_overflow, replace_trapping_overflow,
14700 rewrite_to_non_trapping_overflow): New functions.
14701 * tree-vect-loop.c: Include tree-eh.h.
14702 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14703 * tree-data-ref.c: Include tree-eh.h.
14704 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14706 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14708 PR rtl-optimization/84123
14709 * combine.c (change_zero_ext): Check if hard register satisfies
14710 can_change_dest_mode before calling gen_lowpart_SUBREG.
14712 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14715 * ira.c (ira_init_register_move_cost): Remove assert.
14717 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14719 PR rtl-optimization/84071
14720 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14721 * doc/tm.texi: Regenerate.
14723 2018-01-31 Richard Biener <rguenther@suse.de>
14725 PR tree-optimization/84132
14726 * tree-data-ref.c (analyze_miv_subscript): Properly
14727 check whether evolution_function_is_affine_multivariate_p
14728 before calling gcd_of_steps_may_divide_p.
14730 2018-01-31 Julia Koval <julia.koval@intel.com>
14733 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14734 * config/i386/i386.md (rdpid_rex64) New.
14735 (rdpid): Make 32bit only.
14737 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14740 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14741 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14743 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14746 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14748 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14750 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14752 PR rtl-optimization/84071
14753 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14754 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14756 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14758 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14759 (arc_attribute_table): Add 'aux' attribute.
14760 (arc_in_small_data_p): Consider aux like variables.
14761 (arc_is_aux_reg_p): New function.
14762 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14763 (arc_get_aux_arg): New function.
14764 (prepare_move_operands): Handle aux-register access.
14765 (arc_handle_aux_attribute): New function.
14766 * doc/extend.texi (ARC Variable attributes): Add subsection.
14768 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14770 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14771 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14772 (arc_attribute_table): Add 'uncached' attribute.
14773 (arc_print_operand): Print '.di' flag for uncached memory
14775 (arc_in_small_data_p): Do not consider for small data the uncached
14777 (arc_is_uncached_mem_p): New function.
14778 * config/arc/predicates.md (compact_store_memory_operand): Check
14779 for uncached memory accesses.
14780 (nonvol_nonimm_operand): Likewise.
14781 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14783 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14786 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14787 falign-loops=): Add Optimization flag.
14789 2018-01-30 Jeff Law <law@redhat.com>
14792 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14793 INT_REGISTERS_SAVED. Check it prior to calling
14794 get_scratch_register_on_entry.
14795 (ix86_adjust_stack_and_probe): Similarly.
14796 (ix86_emit_probe_stack_range): Similarly.
14797 (ix86_expand_prologue): Corresponding changes.
14799 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14802 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14803 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14805 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14808 * lra-constraints.c (curr_insn_transform): Process AND in the
14811 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14813 PR rtl-optimization/83986
14814 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14815 dependence against last_pending_memory_flush in addition to
14816 pending_jump_insns.
14818 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14820 PR tree-optimization/81611
14821 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14824 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14827 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14830 2018-01-30 Richard Biener <rguenther@suse.de>
14831 Jakub Jelinek <jakub@redhat.com>
14833 PR tree-optimization/84111
14834 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14835 inner loops added during recursion, as they don't have up-to-date
14838 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14841 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14842 (can_inline_edge_by_limits_p): ... here.
14843 (can_early_inline_edge_p, check_callers,
14844 update_caller_keys, update_callee_keys, recursive_inlining,
14845 add_new_edges_to_heap, speculation_useful_p,
14846 inline_small_functions,
14847 inline_small_functions, flatten_function,
14848 inline_to_all_callers_1): Update.
14850 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14852 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14855 2018-01-30 Richard Biener <rguenther@suse.de>
14857 PR tree-optimization/83008
14858 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14859 invariant and constant vector uses in stmts when they need
14860 more than one stmt.
14862 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14865 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14866 * configure: Regenerate.
14868 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14870 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14872 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14873 Use gen_rtx_REG rather than gen_lowpart.
14875 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14877 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14878 rather than 0 when creating partial subregs.
14880 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14882 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14885 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14888 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14889 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14890 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14891 flags. This restores the settings used before the 2017-07-24.
14892 Turning off pre increment/decrement/modify allows IVOPTS to
14893 optimize DF/SF loops where the index is an int.
14895 2018-01-29 Richard Biener <rguenther@suse.de>
14896 Kelvin Nilsen <kelvin@gcc.gnu.org>
14899 * tree-vect-stmts.c (vectorizable_call): Don't call
14900 targetm.vectorize_builtin_md_vectorized_function if callee is
14903 2018-01-22 Carl Love <cel@us.ibm.com>
14905 * doc/extend.tex: Fix typo in second arg in
14906 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14908 2018-01-29 Richard Biener <rguenther@suse.de>
14910 PR tree-optimization/84086
14911 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14912 (flush_ssaname_freelist): When SSA names were released reset
14913 the SCEV hash table.
14915 2018-01-29 Richard Biener <rguenther@suse.de>
14917 PR tree-optimization/84057
14918 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14919 removed paths when removing edges.
14921 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14923 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14924 -mfunction-return=@var{choice}.
14926 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14928 PR diagnostic/84034
14929 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14930 Handle CR like TAB.
14931 (layout::print_source_line): Likewise.
14932 (test_get_line_width_without_trailing_whitespace): Add test cases.
14934 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14936 PR middle-end/84040
14937 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14940 2018-01-26 Jim Wilson <jimw@sifive.com>
14942 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14944 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14947 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14949 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14950 and CMP + SUB-immediate -> SUBS.
14952 2018-01-26 Martin Sebor <msebor@redhat.com>
14954 PR tree-optimization/83896
14955 * tree-ssa-strlen.c (get_string_len): Rename...
14956 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14957 Avoid assuming length is constant.
14958 (handle_char_store): Use HOST_WIDE_INT for string length.
14960 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14963 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14964 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14966 2018-01-26 Richard Biener <rguenther@suse.de>
14968 PR rtl-optimization/84003
14969 * dse.c (record_store): Only record redundant stores when
14970 the earlier store aliases at least all accesses the later one does.
14972 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14974 PR rtl-optimization/83985
14975 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14976 REG_CFA_RESTORE insns.
14977 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14978 REG_CFA_RESTORE insns here.
14981 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14982 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14984 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14986 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14987 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14988 (arc_init): Likewise.
14989 (arc_override_options): Likewise.
14990 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14992 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14994 * config/arc/arc.h (TARGET_DBNZ): Define.
14995 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14996 properly set the tune attribute.
14997 (dbnz): Use TARGET_DBNZ guard.
14998 * config/arc/arc.opt (mtune): Add core3 option.
15000 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15002 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15003 recognize new pic like addresses.
15004 (arc_delegitimize_address): Clean up.
15006 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15008 * config/arc/arc-arches.def: Option mrf16 valid for all
15010 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15011 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15012 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15013 * config/arc/arc-tables.opt: Regenerate.
15014 * config/arc/arc.c (arc_conditional_register_usage): Handle
15015 reduced register file case.
15016 (arc_file_start): Set must have build attributes.
15017 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15018 mrf16 option value.
15019 * config/arc/arc.opt (mrf16): Add new option.
15020 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15021 * config/arc/genmultilib.awk: Handle new mrf16 option.
15022 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15023 * config/arc/t-multilib: Regenerate.
15024 * doc/invoke.texi (ARC Options): Document mrf16 option.
15026 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15028 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15029 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15030 (arc_attribute_table): Add 'secure_call' attribute.
15031 (arc_print_operand): Print secure call operand.
15032 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15034 (arc_is_secure_call_p): New function. * config/arc/arc.md
15035 (call_i): Add support for sjli instruction.
15036 (call_value_i): Likewise.
15037 * config/arc/constraints.md (Csc): New constraint.
15039 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15040 John Eric Martin <John.Martin@emmicro-us.com>
15042 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15043 * config/arc/arc.c (_arc_jli_section): New struct.
15044 (arc_jli_section): New type.
15045 (rc_jli_sections): New static variable.
15046 (arc_handle_jli_attribute): New function.
15047 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15048 (arc_file_end): New function.
15049 (TARGET_ASM_FILE_END): Define.
15050 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15051 (arc_add_jli_section): New function.
15052 (jli_call_scan): Likewise.
15053 (arc_reorg): Call jli_call_scan.
15054 (arc_output_addsi): Remove 'S' from printing asm operand.
15055 (arc_is_jli_call_p): New function.
15056 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15058 (movhi_insn): Likewise.
15059 (movsi_insn): Likewise.
15060 (movsi_set_cc_insn): Likewise.
15061 (loadqi_update): Likewise.
15062 (load_zeroextendqisi_update): Likewise.
15063 (load_signextendqisi_update): Likewise.
15064 (loadhi_update): Likewise.
15065 (load_zeroextendhisi_update): Likewise.
15066 (load_signextendhisi_update): Likewise.
15067 (loadsi_update): Likewise.
15068 (loadsf_update): Likewise.
15069 (movsicc_insn): Likewise.
15070 (bset_insn): Likewise.
15071 (bxor_insn): Likewise.
15072 (bclr_insn): Likewise.
15073 (bmsk_insn): Likewise.
15074 (bicsi3_insn): Likewise.
15075 (cmpsi_cc_c_insn): Likewise.
15076 (movsi_ne): Likewise.
15077 (movsi_cond_exec): Likewise.
15078 (clrsbsi2): Likewise.
15079 (norm_f): Likewise.
15089 (call_i): Remove 'S' asm letter, add jli instruction.
15090 (call_value_i): Likewise.
15091 * config/arc/arc.op (mjli-always): New option.
15092 * config/arc/constraints.md (Cji): New constraint.
15093 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15095 (subsf3_fpx): Likewise.
15096 (mulsf3_fpx): Likewise.
15097 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15099 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15100 function attrbutes.
15101 * doc/invoke.texi (ARC): Document mjli-always option.
15103 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15105 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15106 avoid addition with 0 and use incw and decw where possible.
15108 2018-01-26 Richard Biener <rguenther@suse.de>
15110 PR tree-optimization/81082
15111 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15112 association if it requires casting to unsigned.
15113 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15114 from fold_plusminus_mult_expr to catch important cases late when
15115 range info is available.
15117 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15119 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15120 * configure.ac (hidden_linkonce): New test.
15121 * configure: Regenerate.
15122 * config.in: Regenerate.
15124 2018-01-26 Julia Koval <julia.koval@intel.com>
15126 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15127 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15128 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15129 _mm_mask_bitshuffle_epi64_mask): Fix type.
15130 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15131 USI_FTYPE_V4DI_V4DI_USI): Remove.
15132 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15133 __builtin_ia32_vpshufbitqmb256_mask,
15134 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15135 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15136 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15138 2018-01-26 Alan Modra <amodra@gmail.com>
15141 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15142 UNSPEC_VBPERMQ. Sort other unspecs.
15144 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15146 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15148 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15150 PR middle-end/83055
15151 * predict.c (drop_profile): Do not push/pop cfun; update also
15153 (handle_missing_profiles): Fix logic looking for zero profiles.
15155 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15157 PR middle-end/83977
15158 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15159 on functions with #pragma omp declare simd or functions with simd
15161 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15162 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15163 Remove trailing \n from warning_at calls.
15165 2018-01-25 Tom de Vries <tom@codesourcery.com>
15168 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15169 for neutered workers.
15171 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15174 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15175 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15177 2018-01-24 Jeff Law <law@redhat.com>
15180 * i386.c (get_probe_interval): Move to earlier point.
15181 (ix86_compute_frame_layout): If -fstack-clash-protection and
15182 the frame is larger than the probe interval, then use pushes
15183 to save registers rather than reg->mem moves.
15184 (ix86_expand_prologue): Remove conditional for int_registers_saved
15187 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15190 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15191 min/max for never referenced object.
15193 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15195 PR middle-end/83977
15196 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15198 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15199 attributes from DECL_ATTRIBUTES (decl) without affecting
15200 DECL_ATTRIBUTES (current_function_decl).
15201 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15202 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15204 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15206 PR tree-optimization/83979
15207 * fold-const.c (fold_comparison): Use constant_boolean_node
15208 instead of boolean_{true,false}_node.
15210 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15212 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15215 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15217 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15218 Simplify the clause that sets the length attribute.
15219 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15220 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15221 clause that sets the length attribute.
15222 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15224 2018-01-24 Tom de Vries <tom@codesourcery.com>
15227 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15228 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15229 Add strict parameter.
15230 (prevent_branch_around_nothing): Insert dummy insn between branch to
15231 label and label with no ptx insn inbetween.
15232 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15234 2018-01-24 Tom de Vries <tom@codesourcery.com>
15237 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15238 for neutered threads in warp.
15239 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15241 2018-01-24 Richard Biener <rguenther@suse.de>
15243 PR tree-optimization/83176
15244 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15247 2018-01-24 Richard Biener <rguenther@suse.de>
15249 PR tree-optimization/82819
15250 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15251 code generating pluses that are no-ops in the target precision.
15253 2018-01-24 Richard Biener <rguenther@suse.de>
15255 PR middle-end/84000
15256 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15258 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15260 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15261 to merge probabilities.
15262 * predict.c (probably_never_executed): Also mark as cold functions
15263 with global 0 profile and guessed local profile.
15264 * profile-count.c (profile_probability::combine_with_count): New
15266 * profile-count.h (profile_probability::operator*,
15267 profile_probability::operator*=, profile_probability::operator/,
15268 profile_probability::operator/=): Reduce precision to adjusted
15269 and set value to guessed on contradictory divisions.
15270 (profile_probability::combine_with_freq): Remove.
15271 (profile_probability::combine_wiht_count): Declare.
15272 (profile_count::force_nonzero):: Set to adjusted.
15273 (profile_count::probability_in):: Set quality to adjusted.
15274 * tree-ssa-tail-merge.c (replace_block_by): Use
15275 combine_with_count.
15277 2018-01-23 Andrew Waterman <andrew@sifive.com>
15278 Jim Wilson <jimw@sifive.com>
15280 * config/riscv/riscv.c (riscv_stack_boundary): New.
15281 (riscv_option_override): Set riscv_stack_boundary. Handle
15282 riscv_preferred_stack_boundary_arg.
15283 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15284 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15285 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15286 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15287 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15288 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15290 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15293 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15294 of struct ix86_frame.
15295 (ix86_expand_epilogue): Likewise. Add a local variable for
15296 the reg_save_offset field in struct ix86_frame.
15298 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15300 PR tree-optimization/82604
15301 * tree-loop-distribution.c (enum partition_kind): New enum item
15302 PKIND_PARTIAL_MEMSET.
15303 (partition_builtin_p): Support above new enum item.
15304 (generate_code_for_partition): Ditto.
15305 (compute_access_range): Differentiate cases that equality can be
15306 proven at all loops, the innermost loops or no loops.
15307 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15308 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
15309 (finalize_partitions, distribute_loop): Don't fuse partition of
15310 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
15311 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
15312 parloop is enabled.
15314 2018-01-23 Martin Liska <mliska@suse.cz>
15316 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
15317 order to ignore the predictor.
15318 (PRED_POLYMORPHIC_CALL): Likewise.
15319 (PRED_RECURSIVE_CALL): Likewise.
15321 2018-01-23 Martin Liska <mliska@suse.cz>
15323 * tree-profile.c (tree_profiling): Print function header to
15324 aware reader which function we are working on.
15325 * value-prof.c (gimple_find_values_to_profile): Do not print
15326 not interesting value histograms.
15328 2018-01-23 Martin Liska <mliska@suse.cz>
15330 * profile-count.h (enum profile_quality): Add
15331 profile_uninitialized as the first value. Do not number values
15332 as they are zero based.
15333 (profile_count::verify): Update sanity check.
15334 (profile_probability::verify): Likewise.
15336 2018-01-23 Nathan Sidwell <nathan@acm.org>
15338 * doc/invoke.texi (ffor-scope): Deprecate.
15340 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15342 PR tree-optimization/83510
15343 * domwalk.c (set_all_edges_as_executable): New function.
15344 (dom_walker::dom_walker): Convert bool param
15345 "skip_unreachable_blocks" to enum reachability. Move setup of
15346 edge flags to set_all_edges_as_executable and only do it when
15347 reachability is REACHABLE_BLOCKS.
15348 * domwalk.h (enum dom_walker::reachability): New enum.
15349 (dom_walker::dom_walker): Convert bool param
15350 "skip_unreachable_blocks" to enum reachability.
15351 (set_all_edges_as_executable): New decl.
15352 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
15353 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
15355 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
15356 but converting true to REACHABLE_BLOCKS.
15357 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
15359 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
15360 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
15361 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
15363 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
15364 if check_all_array_refs will be called.
15366 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15368 * tree.c (selftest::test_location_wrappers): Add more test
15371 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15373 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
15374 (selftest::test_bit_in_range): Likewise.
15376 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15379 * doc/sourcebuild.texi (vect_float): Say that the selector
15380 only describes the situation when -funsafe-math-optimizations is on.
15381 (vect_float_strict): Document.
15383 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15385 PR tree-optimization/83965
15386 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15387 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
15388 instead of checking only for a reduction.
15389 (vect_recog_widen_sum_pattern): Likewise.
15391 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15393 * predict.c (probably_never_executed): Only use precise profile info.
15394 (compute_function_frequency): Skip after inlining hack since we now
15395 have quality checking.
15397 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15399 * profile-count.h (profile_probability::very_unlikely,
15400 profile_probability::unlikely, profile_probability::even): Set
15401 precision to guessed.
15403 2018-01-23 Richard Biener <rguenther@suse.de>
15405 PR tree-optimization/83963
15406 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
15407 Properly terminate dominator walk when crossing the exit edge not
15408 when visiting its source block.
15410 2018-01-23 Jakub Jelinek <jakub@redhat.com>
15413 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
15414 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
15416 2018-01-22 Jakub Jelinek <jakub@redhat.com>
15418 PR tree-optimization/83957
15419 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
15420 semicolon after for body surrounded by braces.
15422 PR tree-optimization/83081
15423 * profile-count.h (profile_probability::split): New method.
15424 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
15425 Use profile_probability::split.
15426 (do_compare_rtx_and_jump): Fix adjustment of probabilities
15427 when splitting a single conditional jump into 2.
15429 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15431 PR tree-optimization/69452
15432 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15435 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15437 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15438 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15439 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15441 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15443 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15445 * config/rl78/rl78.md (movdi): New define_expand.
15446 * config/rl78/rl78.c (rl78_split_movdi): New function.
15448 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15451 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15453 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15454 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15455 128-bit to produce an UNSPEC move to get the double word with the
15456 signbit and then a shift directly to do signbit.
15457 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15458 implementation with a new version that just does either a direct
15459 move or a regular move. Move memory interface to separate insns.
15460 Move insns so they are next to the expander.
15461 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15462 with signbit move. Split big and little endian case.
15463 (signbit<mode>2_dm_mem_le): Likewise.
15464 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15465 (signbit<mode>2_dm2): Likewise.
15467 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15469 * config/rl78/rl78.md (anddi3): New define_expand.
15471 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15473 * config/rl78/rl78.md (umindi3): New define_expand.
15475 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15477 * config/rl78/rl78.md (smindi3): New define_expand.
15479 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15481 * config/rl78/rl78.md (smaxdi3): New define_expand.
15483 2018-01-22 Carl Love <cel@us.ibm.com>
15485 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15486 LVX_V1TI): Add macro expansion.
15487 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15488 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15489 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15490 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15491 Change check to determine if the instruction is a byte reversing
15492 entry. Fix typo in comment.
15493 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15494 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15495 Add def_builtin calls for new builtins.
15496 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15497 Add define_insn expansion.
15499 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15501 * config/rl78/rl78.md (umaxdi3): New define_expand.
15503 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15505 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15506 for non-QImode registers.
15508 2018-01-22 Richard Biener <rguenther@suse.de>
15510 PR tree-optimization/83963
15511 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15512 including the loop exit block.
15513 (scop_detection::merge_sese): Likewise.
15514 (scop_detection::add_scop): Do it here instead.
15516 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15518 * doc/sourcebuild.texi (arm_softfloat): Document.
15520 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15523 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15524 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15525 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15527 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15528 David Edelsohn <dje.gcc@gmail.com>
15531 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15532 Change "crset eq" to "crset 2".
15533 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15534 (*call_indirect_aix<mode>_nospec): Likewise.
15535 (*call_value_indirect_aix<mode>_nospec): Likewise.
15536 (*call_indirect_elfv2<mode>_nospec): Likewise.
15537 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15538 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15539 change assembly output from . to $.
15540 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15541 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15542 (*tablejump<mode>_internal1_nospec): Likewise.
15544 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15547 * config/sh/sh_optimize_sett_clrt.cc:
15548 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15550 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15552 PR tree-optimization/83940
15553 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15554 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15555 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15556 use it to pass back the definition type.
15557 (vect_check_store_rhs): Likewise rhs_dt_out.
15558 (vect_build_gather_load_calls): Add a mask_dt argument and use
15559 it instead of a call to vect_is_simple_use.
15560 (vectorizable_store): Update calls to vect_check_load_store_mask
15561 and vect_check_store_rhs. Use the dt returned by the latter instead
15562 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15563 instead of calls to vect_is_simple_use. Pass the scalar rather
15564 than the vector operand to vect_is_simple_use when handling
15565 second and subsequent copies of an rhs value.
15566 (vectorizable_load): Update calls to vect_check_load_store_mask
15567 and vect_build_gather_load_calls. Use the cached mask_dt and
15568 gs_info.offset_dt instead of calls to vect_is_simple_use.
15570 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15572 PR middle-end/83945
15573 * tree-emutls.c: Include gimplify.h.
15574 (lower_emutls_2): New function.
15575 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15576 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15577 it before further processing.
15580 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15581 UINTVAL (trueop1) instead of INTVAL (op1).
15583 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15587 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15588 INCOMING_FRAME_SP_OFFSET if not defined.
15589 (scan_trace): Add ENTRY argument. If true and
15590 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15591 emit a note to adjust the CFA offset.
15592 (create_cfi_notes): Adjust scan_trace callers.
15593 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15594 INCOMING_FRAME_SP_OFFSET in the CIE.
15595 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15596 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15598 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15599 * doc/tm.texi: Regenerated.
15601 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15603 PR rtl-optimization/83147
15604 * lra-constraints.c (remove_inheritance_pseudos): Use
15605 lra_substitute_pseudo_within_insn.
15607 2018-01-19 Tom de Vries <tom@codesourcery.com>
15608 Cesar Philippidis <cesar@codesourcery.com>
15611 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15613 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15616 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15617 spaces for function labels.
15619 2018-01-19 Martin Liska <mliska@suse.cz>
15621 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15622 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15623 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15624 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15625 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15626 (PRED_CONST_RETURN): Change from 69 to 65.
15627 (PRED_NULL_RETURN): Change from 91 to 71.
15628 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15629 (PRED_LOOP_GUARD): Change from 66 to 73.
15631 2018-01-19 Martin Liska <mliska@suse.cz>
15633 * predict.c (predict_insn_def): Add new assert.
15634 (struct branch_predictor): Change type to signed integer.
15635 (test_prediction_value_range): Amend test to cover
15636 PROB_UNINITIALIZED.
15637 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15638 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15639 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15640 (PRED_LOOP_IV_COMPARE): Likewise.
15641 * predict.h (PROB_UNINITIALIZED): Define new constant.
15643 2018-01-19 Martin Liska <mliska@suse.cz>
15645 * predict.c (dump_prediction): Add new format for
15646 analyze_brprob.py script which is enabled with -details
15648 * profile-count.h (precise_p): New function.
15650 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15652 PR tree-optimization/83922
15653 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15654 there are no statements that need masking.
15655 (vect_active_double_reduction_p): New function.
15656 (vect_analyze_loop_operations): Use it when handling phis that
15657 are not in the loop header.
15659 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15661 PR tree-optimization/83914
15662 * tree-vect-loop.c (vectorizable_induction): Don't convert
15663 init_expr or apply the peeling adjustment for inductions
15664 that are nested within the vectorized loop.
15666 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15668 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15671 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15675 * function.h (gimplify_parameters): Add gimple_seq * argument.
15676 * function.c: Include gimple.h and options.h.
15677 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15678 for the added local temporaries if needed.
15679 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15680 if there are any parameter cleanups, wrap whole body into a
15681 try/finally with the cleanups.
15683 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15686 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15687 Use GET_MODE_CLASS for scalar floating point.
15689 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15693 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15694 Fix call of call_cgraph_insertion_hooks.
15696 2018-01-18 Martin Sebor <msebor@redhat.com>
15698 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15700 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15703 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15706 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15709 * common.opt: (-ffile-prefix-map): New option.
15710 * opts.c (common_handle_option): Defer it.
15711 * opts-global.c (handle_common_deferred_options): Handle it.
15712 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15713 * file-prefix-map.h: New file.
15714 (remap_debug_filename, add_debug_prefix_map): ...here.
15715 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15716 * final.c (debug_prefix_map, add_debug_prefix_map
15717 remap_debug_filename): Move to...
15718 * file-prefix-map.c: New file.
15719 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15720 generalize, get rid of alloca(), use strrchr() instead of strchr().
15721 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15722 Implement in terms of add_prefix_map().
15723 (remap_macro_filename, remap_debug_filename): Implement in term of
15725 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15726 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15727 * dbxout.c: Include file-prefix-map.h.
15728 * varasm.c: Likewise.
15729 * vmsdbgout.c: Likewise.
15730 * xcoffout.c: Likewise.
15731 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15732 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15733 * doc/invoke.texi (-ffile-prefix-map): Document.
15734 (-fdebug-prefix-map): Update description.
15736 2018-01-18 Martin Liska <mliska@suse.cz>
15738 * config/i386/i386.c (indirect_thunk_name): Document that also
15740 (output_indirect_thunk): Document why both instructions
15741 (pause and lfence) are generated.
15743 2018-01-18 Richard Biener <rguenther@suse.de>
15745 PR tree-optimization/83887
15746 * graphite-scop-detection.c
15747 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15748 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15749 (scop_detection::merge_sese): Re-implement with a flood-fill
15750 algorithm that properly finds a SESE region if it exists.
15752 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15755 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15756 pointer_diff optimizations use view_convert instead of convert.
15758 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15760 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15761 Generate different code for -mno-speculate-indirect-jumps.
15762 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15763 (*call_indirect_aix<mode>): Disable for
15764 -mno-speculate-indirect-jumps.
15765 (*call_indirect_aix<mode>_nospec): New define_insn.
15766 (*call_value_indirect_aix<mode>): Disable for
15767 -mno-speculate-indirect-jumps.
15768 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15769 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15770 -mno-speculate-indirect-jumps.
15771 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15773 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15775 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15776 long double type, set the flags for noting the default long double
15777 type, even if we don't pass or return a long double type.
15779 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15782 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15785 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15787 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15788 support for merge[hl].
15789 (fold_mergehl_helper): New helper function.
15790 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15791 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15792 (altivec_vmrglw_direct): Add xxmrglw insn.
15794 2018-01-17 Andrew Waterman <andrew@sifive.com>
15796 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15797 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15799 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15802 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15803 call the lto_location_cache before reading the
15804 DECL_SOURCE_LOCATION of the types.
15806 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15807 Richard Sandiford <richard.sandiford@linaro.org>
15809 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15810 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15811 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15812 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15813 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15815 * config/aarch64/constraints.md (aarch64_movti_operand):
15817 * config/aarch64/predicates.md (Uti): Add new constraint.
15819 2018-01-17 Carl Love <cel@us.ibm.com>
15821 * config/rs6000/vsx.md (define_expand xl_len_r,
15822 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15823 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15825 (define_expand, define_insn): Move the shift left from the
15826 define_insn to the define_expand for lxvl and stxvl instructions.
15827 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15828 and XL_LEN_R definitions to PURE.
15830 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15832 * config/i386/i386.c (indirect_thunk_name): Declare regno
15833 as unsigned int. Compare regno with INVALID_REGNUM.
15834 (output_indirect_thunk): Ditto.
15835 (output_indirect_thunk_function): Ditto.
15836 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15837 in the call to output_indirect_thunk_function.
15839 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15841 PR middle-end/83884
15842 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15843 rather than the size of inner_type to determine the stack slot size
15844 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15846 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15849 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15852 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15854 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15855 endian Linux systems to optionally enable multilibs for selecting
15856 the long double type if the user configured an explicit type.
15857 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15858 have no long double multilibs if not defined.
15859 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15860 warn if the user used -mabi={ieee,ibm}longdouble and we built
15861 multilibs for long double.
15862 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15863 appropriate multilib option.
15864 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15866 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15867 for building long double multilibs.
15868 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15870 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15872 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15875 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15877 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15880 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15883 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15886 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15888 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15889 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15891 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15893 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15894 different rtl trees depending on TARGET_64BIT.
15895 (rs6000_gen_lvx): Likewise.
15897 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15899 * config/visium/visium.md (nop): Tweak comment.
15900 (hazard_nop): Likewise.
15902 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15904 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15905 -mspeculate-indirect-jumps.
15906 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15907 for -mno-speculate-indirect-jumps.
15908 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15909 (*call_value_indirect_elfv2<mode>): Disable for
15910 -mno-speculate-indirect-jumps.
15911 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15912 (indirect_jump): Emit different RTL for
15913 -mno-speculate-indirect-jumps.
15914 (*indirect_jump<mode>): Disable for
15915 -mno-speculate-indirect-jumps.
15916 (*indirect_jump<mode>_nospec): New define_insn.
15917 (tablejump): Emit different RTL for
15918 -mno-speculate-indirect-jumps.
15919 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15920 (tablejumpsi_nospec): New define_expand.
15921 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15922 (tablejumpdi_nospec): New define_expand.
15923 (*tablejump<mode>_internal1): Disable for
15924 -mno-speculate-indirect-jumps.
15925 (*tablejump<mode>_internal1_nospec): New define_insn.
15926 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15929 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15931 * caller-save.c (insert_save): Drop unnecessary parameter. All
15934 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15935 Richard Biener <rguenth@suse.de>
15938 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15939 return early, inline manually is_gimple_sizepos. Make sure if we
15940 call gimplify_expr we don't end up with a gimple constant.
15941 * tree.c (variably_modified_type_p): Don't return true for
15942 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15943 * gimplify.h (is_gimple_sizepos): Remove.
15945 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15947 PR tree-optimization/83857
15948 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15949 vectorizable_live_operation for pure SLP statements.
15950 (vectorizable_live_operation): Handle PHIs.
15952 2018-01-16 Richard Biener <rguenther@suse.de>
15954 PR tree-optimization/83867
15955 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15956 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15958 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15961 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15962 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15963 If off is not INTEGER_CST, issue a may not be aligned warning
15964 rather than isn't aligned. Use isn%'t rather than isn't.
15965 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15967 <case MULT_EXPR>: Improve the case when bottom and one of the
15968 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15969 operand, in that case check if the other operand is multiple of
15970 bottom divided by the INTEGER_CST operand.
15972 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15975 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15976 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15977 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15978 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15979 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15980 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15981 (pa_function_arg_size): New function.
15983 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15985 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15986 in a separate statement.
15988 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15990 PR tree-optimization/83847
15991 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15992 group gathers and scatters.
15994 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15996 PR rtl-optimization/86620
15997 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15999 PR rtl-optimization/83213
16000 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16001 to last if both are JUMP_INSNs.
16003 PR tree-optimization/83843
16004 * gimple-ssa-store-merging.c
16005 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16006 store_immediate_info for bswap/nop orig_stores.
16008 2018-01-15 Andrew Waterman <andrew@sifive.com>
16010 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16012 <UDIV>: Increase cost if !TARGET_DIV.
16014 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16016 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16017 (define_attr "cr_logical_3op"): New.
16018 (cceq_ior_compare): Adjust.
16019 (cceq_ior_compare_complement): Adjust.
16020 (*cceq_rev_compare): Adjust.
16021 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16022 (is_cracked_insn): Adjust.
16023 (insn_must_be_first_in_group): Adjust.
16024 * config/rs6000/40x.md: Adjust.
16025 * config/rs6000/440.md: Adjust.
16026 * config/rs6000/476.md: Adjust.
16027 * config/rs6000/601.md: Adjust.
16028 * config/rs6000/603.md: Adjust.
16029 * config/rs6000/6xx.md: Adjust.
16030 * config/rs6000/7450.md: Adjust.
16031 * config/rs6000/7xx.md: Adjust.
16032 * config/rs6000/8540.md: Adjust.
16033 * config/rs6000/cell.md: Adjust.
16034 * config/rs6000/e300c2c3.md: Adjust.
16035 * config/rs6000/e500mc.md: Adjust.
16036 * config/rs6000/e500mc64.md: Adjust.
16037 * config/rs6000/e5500.md: Adjust.
16038 * config/rs6000/e6500.md: Adjust.
16039 * config/rs6000/mpc.md: Adjust.
16040 * config/rs6000/power4.md: Adjust.
16041 * config/rs6000/power5.md: Adjust.
16042 * config/rs6000/power6.md: Adjust.
16043 * config/rs6000/power7.md: Adjust.
16044 * config/rs6000/power8.md: Adjust.
16045 * config/rs6000/power9.md: Adjust.
16046 * config/rs6000/rs64.md: Adjust.
16047 * config/rs6000/titan.md: Adjust.
16049 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16051 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16052 ix86_indirect_branch_register logic.
16054 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16056 * config/i386/constraints.md (Bs): Update
16057 ix86_indirect_branch_register check. Don't check
16058 ix86_indirect_branch_register with GOT_memory_operand.
16060 * config/i386/predicates.md (GOT_memory_operand): Don't check
16061 ix86_indirect_branch_register here.
16062 (GOT32_symbol_operand): Likewise.
16064 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16066 * config/i386/predicates.md (constant_call_address_operand):
16067 Rewrite ix86_indirect_branch_register logic.
16068 (sibcall_insn_operand): Likewise.
16070 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16072 * config/i386/constraints.md (Bs): Replace
16073 ix86_indirect_branch_thunk_register with
16074 ix86_indirect_branch_register.
16076 * config/i386/i386.md (indirect_jump): Likewise.
16077 (tablejump): Likewise.
16078 (*sibcall_memory): Likewise.
16079 (*sibcall_value_memory): Likewise.
16080 Peepholes of indirect call and jump via memory: Likewise.
16081 * config/i386/i386.opt: Likewise.
16082 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16083 (GOT_memory_operand): Likewise.
16084 (call_insn_operand): Likewise.
16085 (sibcall_insn_operand): Likewise.
16086 (GOT32_symbol_operand): Likewise.
16088 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16090 PR middle-end/83837
16091 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16092 type rather than type addr's type points to.
16093 (expand_omp_atomic_mutex): Likewise.
16094 (expand_omp_atomic): Likewise.
16096 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16099 * config/i386/i386.c (output_indirect_thunk_function): Use
16100 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16101 for __x86_return_thunk.
16103 2018-01-15 Richard Biener <rguenther@suse.de>
16105 PR middle-end/83850
16106 * expmed.c (extract_bit_field_1): Fix typo.
16108 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16111 * config/arm/iterators.md (VF): New mode iterator.
16112 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16113 Remove integer-related logic from pattern.
16114 (neon_vabd<mode>_3): Likewise.
16116 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16118 PR middle-end/82694
16119 * common.opt (fstrict-overflow): No longer an alias.
16120 (fwrapv-pointer): New option.
16121 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16122 also for pointer types based on flag_wrapv_pointer.
16123 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16124 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16125 opts->x_flag_wrapv got set.
16126 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16127 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16128 POINTER_TYPE_OVERFLOW_UNDEFINED.
16129 * match.pd: Likewise in address comparison pattern.
16130 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16132 2018-01-15 Richard Biener <rguenther@suse.de>
16135 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16136 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16137 Reset type names to their identifier if their TYPE_DECL doesn't
16138 have linkage (and thus is used for ODR and devirt).
16139 (save_debug_info_for_decl): Remove.
16140 (save_debug_info_for_type): Likewise.
16141 (add_tree_to_fld_list): Adjust.
16142 * tree-pretty-print.c (dump_generic_node): Make dumping of
16143 type names more robust.
16145 2018-01-15 Richard Biener <rguenther@suse.de>
16147 * BASE-VER: Bump to 8.0.1.
16149 2018-01-14 Martin Sebor <msebor@redhat.com>
16152 * builtins.c (check_access): Avoid warning when the no-warning bit
16155 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16157 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16158 * ira-color (allocno_hard_regs_compare): Likewise.
16160 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16163 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16164 Use .pushsection/.popsection.
16166 2018-01-14 Martin Sebor <msebor@redhat.com>
16169 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16171 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16173 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16174 entry from extra_headers.
16175 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16176 extra_headers, make the list bitwise identical to the i?86-*-* one.
16178 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16180 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16181 -mcmodel=large with -mindirect-branch=thunk,
16182 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16183 -mfunction-return=thunk-extern.
16184 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16185 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16186 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16188 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16190 * config/i386/i386.c (print_reg): Print the name of the full
16191 integer register without '%'.
16192 (ix86_print_operand): Handle 'V'.
16193 * doc/extend.texi: Document 'V' modifier.
16195 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16197 * config/i386/constraints.md (Bs): Disallow memory operand for
16198 -mindirect-branch-register.
16200 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16201 (GOT_memory_operand): Likewise.
16202 (call_insn_operand): Likewise.
16203 (sibcall_insn_operand): Likewise.
16204 (GOT32_symbol_operand): Likewise.
16205 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16206 for -mindirect-branch-register.
16207 (tablejump): Likewise.
16208 (*sibcall_memory): Likewise.
16209 (*sibcall_value_memory): Likewise.
16210 Disallow peepholes of indirect call and jump via memory for
16211 -mindirect-branch-register.
16212 (*call_pop): Replace m with Bw.
16213 (*call_value_pop): Likewise.
16214 (*sibcall_pop_memory): Replace m with Bs.
16215 * config/i386/i386.opt (mindirect-branch-register): New option.
16216 * doc/invoke.texi: Document -mindirect-branch-register option.
16218 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16220 * config/i386/i386-protos.h (ix86_output_function_return): New.
16221 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16222 set function_return_type.
16223 (indirect_thunk_name): Add ret_p to indicate thunk for function
16225 (output_indirect_thunk_function): Pass false to
16226 indirect_thunk_name.
16227 (ix86_output_indirect_branch_via_reg): Likewise.
16228 (ix86_output_indirect_branch_via_push): Likewise.
16229 (output_indirect_thunk_function): Create alias for function
16230 return thunk if regno < 0.
16231 (ix86_output_function_return): New function.
16232 (ix86_handle_fndecl_attribute): Handle function_return.
16233 (ix86_attribute_table): Add function_return.
16234 * config/i386/i386.h (machine_function): Add
16235 function_return_type.
16236 * config/i386/i386.md (simple_return_internal): Use
16237 ix86_output_function_return.
16238 (simple_return_internal_long): Likewise.
16239 * config/i386/i386.opt (mfunction-return=): New option.
16240 (indirect_branch): Mention -mfunction-return=.
16241 * doc/extend.texi: Document function_return function attribute.
16242 * doc/invoke.texi: Document -mfunction-return= option.
16244 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16246 * config/i386/i386-opts.h (indirect_branch): New.
16247 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16248 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16249 with local indirect jump when converting indirect call and jump.
16250 (ix86_set_indirect_branch_type): New.
16251 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16252 (indirectlabelno): New.
16253 (indirect_thunk_needed): Likewise.
16254 (indirect_thunk_bnd_needed): Likewise.
16255 (indirect_thunks_used): Likewise.
16256 (indirect_thunks_bnd_used): Likewise.
16257 (INDIRECT_LABEL): Likewise.
16258 (indirect_thunk_name): Likewise.
16259 (output_indirect_thunk): Likewise.
16260 (output_indirect_thunk_function): Likewise.
16261 (ix86_output_indirect_branch_via_reg): Likewise.
16262 (ix86_output_indirect_branch_via_push): Likewise.
16263 (ix86_output_indirect_branch): Likewise.
16264 (ix86_output_indirect_jmp): Likewise.
16265 (ix86_code_end): Call output_indirect_thunk_function if needed.
16266 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16268 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16269 (ix86_attribute_table): Add indirect_branch.
16270 * config/i386/i386.h (machine_function): Add indirect_branch_type
16271 and has_local_indirect_jump.
16272 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16274 (tablejump): Likewise.
16275 (*indirect_jump): Use ix86_output_indirect_jmp.
16276 (*tablejump_1): Likewise.
16277 (simple_return_indirect_internal): Likewise.
16278 * config/i386/i386.opt (mindirect-branch=): New option.
16279 (indirect_branch): New.
16282 (thunk-inline): Likewise.
16283 (thunk-extern): Likewise.
16284 * doc/extend.texi: Document indirect_branch function attribute.
16285 * doc/invoke.texi: Document -mindirect-branch= option.
16287 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16290 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16292 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16294 * ipa-inline.c (want_inline_small_function_p): Return false if
16295 inlining has already failed with CIF_FINAL_ERROR.
16296 (update_caller_keys): Call want_inline_small_function_p before
16298 (update_callee_keys): Likewise.
16300 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16302 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16304 (rs6000_quadword_masked_address_p): Likewise.
16305 (quad_aligned_load_p): Likewise.
16306 (quad_aligned_store_p): Likewise.
16307 (const_load_sequence_p): Add comment to describe the outer-most loop.
16308 (mimic_memory_attributes_and_flags): New function.
16309 (rs6000_gen_stvx): Likewise.
16310 (replace_swapped_aligned_store): Likewise.
16311 (rs6000_gen_lvx): Likewise.
16312 (replace_swapped_aligned_load): Likewise.
16313 (replace_swapped_load_constant): Capitalize argument name in
16314 comment describing this function.
16315 (rs6000_analyze_swaps): Add a third pass to search for vector loads
16316 and stores that access quad-word aligned addresses and replace
16317 with stvx or lvx instructions when appropriate.
16318 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
16319 New function prototype.
16320 (rs6000_quadword_masked_address_p): Likewise.
16321 (rs6000_gen_lvx): Likewise.
16322 (rs6000_gen_stvx): Likewise.
16323 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
16324 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
16325 when memory address is aligned.
16326 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
16327 this split to select lvx instruction when memory address is aligned.
16328 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
16329 instruction when memory address is aligned.
16330 (*vsx_le_perm_load_v16qi): Likewise.
16331 (four unnamed splitters): Modify to select the stvx instruction
16332 when memory is aligned.
16334 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
16336 * predict.c (determine_unlikely_bbs): Handle correctly BBs
16337 which appears in the queue multiple times.
16339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16340 Alan Hayward <alan.hayward@arm.com>
16341 David Sherwood <david.sherwood@arm.com>
16343 * tree-vectorizer.h (vec_lower_bound): New structure.
16344 (_loop_vec_info): Add check_nonzero and lower_bounds.
16345 (LOOP_VINFO_CHECK_NONZERO): New macro.
16346 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
16347 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
16348 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
16349 fields. Make seg_len the distance travelled, not including the
16351 (dr_direction_indicator): Declare.
16352 (dr_zero_step_indicator): Likewise.
16353 (dr_known_forward_stride_p): Likewise.
16354 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
16356 (runtime_alias_check_p): Allow runtime alias checks with
16358 (operator ==): Compare access_size and align.
16359 (prune_runtime_alias_test_list): Rework for new distinction between
16360 the access_size and seg_len.
16361 (create_intersect_range_checks_index): Likewise. Cope with polynomial
16363 (get_segment_min_max): New function.
16364 (create_intersect_range_checks): Use it.
16365 (dr_step_indicator): New function.
16366 (dr_direction_indicator): Likewise.
16367 (dr_zero_step_indicator): Likewise.
16368 (dr_known_forward_stride_p): Likewise.
16369 * tree-loop-distribution.c (data_ref_segment_size): Return
16370 DR_STEP * (niters - 1).
16371 (compute_alias_check_pairs): Update call to the dr_with_seg_len
16373 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
16374 (vect_preserves_scalar_order_p): New function, split out from...
16375 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
16376 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
16377 (vect_vfa_access_size): New function.
16378 (vect_vfa_align): Likewise.
16379 (vect_compile_time_alias): Take access_size_a and access_b arguments.
16380 (dump_lower_bound): New function.
16381 (vect_check_lower_bound): Likewise.
16382 (vect_small_gap_p): Likewise.
16383 (vectorizable_with_step_bound_p): Likewise.
16384 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
16385 depencies if the vectorization factor is 1. Convert the checks
16386 for nonzero steps into checks on the bounds of DR_STEP. Try using
16387 a bunds check for variable steps if the minimum required step is
16388 relatively small. Update calls to the dr_with_seg_len
16389 constructor and to vect_compile_time_alias.
16390 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
16392 (vect_loop_versioning): Call it.
16393 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
16395 (vect_estimate_min_profitable_iters): Account for any bounds checks.
16397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16398 Alan Hayward <alan.hayward@arm.com>
16399 David Sherwood <david.sherwood@arm.com>
16401 * doc/sourcebuild.texi (vect_scatter_store): Document.
16402 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
16404 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
16406 * genopinit.c (main): Add supports_vec_scatter_store and
16407 supports_vec_scatter_store_cached to target_optabs.
16408 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
16409 IFN_MASK_SCATTER_STORE.
16410 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
16412 * internal-fn.h (internal_store_fn_p): Declare.
16413 (internal_fn_stored_value_index): Likewise.
16414 * internal-fn.c (scatter_store_direct): New macro.
16415 (expand_scatter_store_optab_fn): New function.
16416 (direct_scatter_store_optab_supported_p): New macro.
16417 (internal_store_fn_p): New function.
16418 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
16419 IFN_MASK_SCATTER_STORE.
16420 (internal_fn_mask_index): Likewise.
16421 (internal_fn_stored_value_index): New function.
16422 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
16423 for scatter stores.
16424 * optabs-query.h (supports_vec_scatter_store_p): Declare.
16425 * optabs-query.c (supports_vec_scatter_store_p): New function.
16426 * tree-vectorizer.h (vect_get_store_rhs): Declare.
16427 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16428 true for scatter stores.
16429 (vect_gather_scatter_fn_p): Handle scatter stores too.
16430 (vect_check_gather_scatter): Consider using scatter stores if
16431 supports_vec_scatter_store_p.
16432 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16433 scatter stores too.
16434 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16435 internal_fn_stored_value_index.
16436 (check_load_store_masking): Handle scatter stores too.
16437 (vect_get_store_rhs): Make public.
16438 (vectorizable_call): Use internal_store_fn_p.
16439 (vectorizable_store): Handle scatter store internal functions.
16440 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16441 when deciding whether the end of the group has been reached.
16442 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16443 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16444 (mask_scatter_store<mode>): New insns.
16446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16447 Alan Hayward <alan.hayward@arm.com>
16448 David Sherwood <david.sherwood@arm.com>
16450 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16451 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16452 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16454 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16455 Use vect_truncate_gather_scatter_offset if we can't treat the
16456 operation as a normal gather load or scatter store.
16457 (get_group_load_store_type): Take the gather_scatter_info
16458 as argument. Try using a gather load or scatter store for
16459 single-element groups.
16460 (get_load_store_type): Update calls to get_group_load_store_type
16461 and vect_use_strided_gather_scatters_p.
16463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16464 Alan Hayward <alan.hayward@arm.com>
16465 David Sherwood <david.sherwood@arm.com>
16467 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16468 optional tree argument.
16469 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16471 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16472 but continue to use the current value as a fallback.
16473 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16474 to compare the updates.
16475 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16476 (get_load_store_type): Use it when handling a strided access.
16477 (vect_get_strided_load_store_ops): New function.
16478 (vect_get_data_ptr_increment): Likewise.
16479 (vectorizable_load): Handle strided gather loads. Always pass
16480 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16483 Alan Hayward <alan.hayward@arm.com>
16484 David Sherwood <david.sherwood@arm.com>
16486 * doc/md.texi (gather_load@var{m}): Document.
16487 (mask_gather_load@var{m}): Likewise.
16488 * genopinit.c (main): Add supports_vec_gather_load and
16489 supports_vec_gather_load_cached to target_optabs.
16490 * optabs-tree.c (init_tree_optimization_optabs): Use
16491 ggc_cleared_alloc to allocate target_optabs.
16492 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16493 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16495 * internal-fn.h (internal_load_fn_p): Declare.
16496 (internal_gather_scatter_fn_p): Likewise.
16497 (internal_fn_mask_index): Likewise.
16498 (internal_gather_scatter_fn_supported_p): Likewise.
16499 * internal-fn.c (gather_load_direct): New macro.
16500 (expand_gather_load_optab_fn): New function.
16501 (direct_gather_load_optab_supported_p): New macro.
16502 (direct_internal_fn_optab): New function.
16503 (internal_load_fn_p): Likewise.
16504 (internal_gather_scatter_fn_p): Likewise.
16505 (internal_fn_mask_index): Likewise.
16506 (internal_gather_scatter_fn_supported_p): Likewise.
16507 * optabs-query.c (supports_at_least_one_mode_p): New function.
16508 (supports_vec_gather_load_p): Likewise.
16509 * optabs-query.h (supports_vec_gather_load_p): Declare.
16510 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16511 and memory_type field.
16512 (NUM_PATTERNS): Bump to 15.
16513 * tree-vect-data-refs.c: Include internal-fn.h.
16514 (vect_gather_scatter_fn_p): New function.
16515 (vect_describe_gather_scatter_call): Likewise.
16516 (vect_check_gather_scatter): Try using internal functions for
16517 gather loads. Recognize existing calls to a gather load function.
16518 (vect_analyze_data_refs): Consider using gather loads if
16519 supports_vec_gather_load_p.
16520 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16521 (vect_get_gather_scatter_offset_type): Likewise.
16522 (vect_convert_mask_for_vectype): Likewise.
16523 (vect_add_conversion_to_patterm): Likewise.
16524 (vect_try_gather_scatter_pattern): Likewise.
16525 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16526 (vect_vect_recog_func_ptrs): Add it.
16527 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16528 internal_fn_mask_index and internal_gather_scatter_fn_p.
16529 (check_load_store_masking): Take the gather_scatter_info as an
16530 argument and handle gather loads.
16531 (vect_get_gather_scatter_ops): New function.
16532 (vectorizable_call): Check internal_load_fn_p.
16533 (vectorizable_load): Likewise. Handle gather load internal
16535 (vectorizable_store): Update call to check_load_store_masking.
16536 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16537 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16538 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16539 (aarch64_gather_scale_operand_d): New predicates.
16540 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16541 (mask_gather_load<mode>): New insns.
16543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16544 Alan Hayward <alan.hayward@arm.com>
16545 David Sherwood <david.sherwood@arm.com>
16547 * optabs.def (fold_left_plus_optab): New optab.
16548 * doc/md.texi (fold_left_plus_@var{m}): Document.
16549 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16550 * internal-fn.c (fold_left_direct): Define.
16551 (expand_fold_left_optab_fn): Likewise.
16552 (direct_fold_left_optab_supported_p): Likewise.
16553 * fold-const-call.c (fold_const_fold_left): New function.
16554 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16555 * tree-parloops.c (valid_reduction_p): New function.
16556 (gather_scalar_reductions): Use it.
16557 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16558 (vect_finish_replace_stmt): Declare.
16559 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16560 (needs_fold_left_reduction_p): New function, split out from...
16561 (vect_is_simple_reduction): ...here. Accept reductions that
16562 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16563 (vect_force_simple_reduction): Also store the reduction type in
16564 the assignment's STMT_VINFO_REDUC_TYPE.
16565 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16566 (merge_with_identity): New function.
16567 (vect_expand_fold_left): Likewise.
16568 (vectorize_fold_left_reduction): Likewise.
16569 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16570 scalar phi in place for it. Check for target support and reject
16571 cases that would reassociate the operation. Defer the transform
16572 phase to vectorize_fold_left_reduction.
16573 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16574 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16575 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16579 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16580 call to ifc_temp_var.
16582 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16583 Alan Hayward <alan.hayward@arm.com>
16584 David Sherwood <david.sherwood@arm.com>
16586 * target.def (legitimize_address_displacement): Take the original
16587 offset as a poly_int.
16588 * targhooks.h (default_legitimize_address_displacement): Update
16590 * targhooks.c (default_legitimize_address_displacement): Likewise.
16591 * doc/tm.texi: Regenerate.
16592 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16593 as an argument, moving assert of ad->disp == ad->disp_term to...
16594 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16595 Try calling targetm.legitimize_address_displacement before expanding
16596 the address rather than afterwards, and adjust for the new interface.
16597 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16598 Match the new hook interface. Handle SVE addresses.
16599 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16600 new hook interface.
16602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16604 * Makefile.in (OBJS): Add early-remat.o.
16605 * target.def (select_early_remat_modes): New hook.
16606 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16607 * doc/tm.texi: Regenerate.
16608 * targhooks.h (default_select_early_remat_modes): Declare.
16609 * targhooks.c (default_select_early_remat_modes): New function.
16610 * timevar.def (TV_EARLY_REMAT): New timevar.
16611 * passes.def (pass_early_remat): New pass.
16612 * tree-pass.h (make_pass_early_remat): Declare.
16613 * early-remat.c: New file.
16614 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16616 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16619 Alan Hayward <alan.hayward@arm.com>
16620 David Sherwood <david.sherwood@arm.com>
16622 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16623 vfm1 with a bound_epilog parameter.
16624 (vect_do_peeling): Update calls accordingly, and move the prologue
16625 call earlier in the function. Treat the base bound_epilog as 0 for
16626 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16627 this base when peeling for gaps.
16628 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16629 with fully-masked loops.
16630 (vect_estimate_min_profitable_iters): Handle the single peeled
16631 iteration in that case.
16633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16634 Alan Hayward <alan.hayward@arm.com>
16635 David Sherwood <david.sherwood@arm.com>
16637 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16638 single-element interleaving even if the size is not a power of 2.
16639 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16640 accesses for single-element interleaving if the group size is
16643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16644 Alan Hayward <alan.hayward@arm.com>
16645 David Sherwood <david.sherwood@arm.com>
16647 * doc/md.texi (fold_extract_last_@var{m}): Document.
16648 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16649 * optabs.def (fold_extract_last_optab): New optab.
16650 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16651 * internal-fn.c (fold_extract_direct): New macro.
16652 (expand_fold_extract_optab_fn): Likewise.
16653 (direct_fold_extract_optab_supported_p): Likewise.
16654 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16655 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16656 EXTRACT_LAST_REDUCTION.
16657 (get_initial_def_for_reduction): Do not create an initial vector
16658 for EXTRACT_LAST_REDUCTION reductions.
16659 (vectorizable_reduction): Leave the scalar phi in place for
16660 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16661 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16662 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16663 transform phase to vectorizable_condition.
16664 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16666 (vect_finish_stmt_generation): ...here.
16667 (vect_finish_replace_stmt): New function.
16668 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16669 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16671 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16674 Alan Hayward <alan.hayward@arm.com>
16675 David Sherwood <david.sherwood@arm.com>
16677 * doc/md.texi (extract_last_@var{m}): Document.
16678 * optabs.def (extract_last_optab): New optab.
16679 * internal-fn.def (EXTRACT_LAST): New internal function.
16680 * internal-fn.c (cond_unary_direct): New macro.
16681 (expand_cond_unary_optab_fn): Likewise.
16682 (direct_cond_unary_optab_supported_p): Likewise.
16683 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16684 loops using EXTRACT_LAST.
16685 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16686 (extract_last_<mode>): ...this optab.
16687 (vec_extract<mode><Vel>): Update accordingly.
16689 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16690 Alan Hayward <alan.hayward@arm.com>
16691 David Sherwood <david.sherwood@arm.com>
16693 * target.def (empty_mask_is_expensive): New hook.
16694 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16695 * doc/tm.texi: Regenerate.
16696 * targhooks.h (default_empty_mask_is_expensive): Declare.
16697 * targhooks.c (default_empty_mask_is_expensive): New function.
16698 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16699 if the target says that empty masks are expensive.
16700 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16702 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16705 Alan Hayward <alan.hayward@arm.com>
16706 David Sherwood <david.sherwood@arm.com>
16708 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16709 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16710 (vect_use_loop_mask_for_alignment_p): New function.
16711 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16712 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16713 niters_skip argument. Make sure that the first niters_skip elements
16714 of the first iteration are inactive.
16715 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16716 Update call to vect_set_loop_masks_directly.
16717 (get_misalign_in_elems): New function, split out from...
16718 (vect_gen_prolog_loop_niters): ...here.
16719 (vect_update_init_of_dr): Take a code argument that specifies whether
16720 the adjustment should be added or subtracted.
16721 (vect_update_init_of_drs): Likewise.
16722 (vect_prepare_for_masked_peels): New function.
16723 (vect_do_peeling): Skip prologue peeling if we're using a mask
16724 instead. Update call to vect_update_inits_of_drs.
16725 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16727 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16728 alignment. Do not include the number of peeled iterations in
16729 the minimum threshold in that case.
16730 (vectorizable_induction): Adjust the start value down by
16731 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16732 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16733 Take the number of skipped iterations into account when calculating
16735 * tree-vect-stmts.c (vect_gen_while_not): New function.
16737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16738 Alan Hayward <alan.hayward@arm.com>
16739 David Sherwood <david.sherwood@arm.com>
16741 * doc/sourcebuild.texi (vect_fully_masked): Document.
16742 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16743 default value to 0.
16744 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16746 (vect_analyze_loop_2): ...here. Don't check the vectorization
16747 factor against the number of loop iterations if the loop is
16750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16751 Alan Hayward <alan.hayward@arm.com>
16752 David Sherwood <david.sherwood@arm.com>
16754 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16755 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16756 (dump_groups): Update accordingly.
16757 (iv_use::mem_type): New member variable.
16758 (address_p): New function.
16759 (record_use): Add a mem_type argument and initialize the new
16761 (record_group_use): Add a mem_type argument. Use address_p.
16762 Remove obsolete null checks of base_object. Update call to record_use.
16763 (find_interesting_uses_op): Update call to record_group_use.
16764 (find_interesting_uses_cond): Likewise.
16765 (find_interesting_uses_address): Likewise.
16766 (get_mem_type_for_internal_fn): New function.
16767 (find_address_like_use): Likewise.
16768 (find_interesting_uses_stmt): Try find_address_like_use before
16769 calling find_interesting_uses_op.
16770 (addr_offset_valid_p): Use the iv mem_type field as the type
16771 of the addressed memory.
16772 (add_autoinc_candidates): Likewise.
16773 (get_address_cost): Likewise.
16774 (split_small_address_groups_p): Use address_p.
16775 (split_address_groups): Likewise.
16776 (add_iv_candidate_for_use): Likewise.
16777 (autoinc_possible_for_pair): Likewise.
16778 (rewrite_groups): Likewise.
16779 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16780 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16781 (get_alias_ptr_type_for_ptr_address): New function.
16782 (rewrite_use_address): Rewrite address uses in calls that were
16783 identified by find_address_like_use.
16785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16786 Alan Hayward <alan.hayward@arm.com>
16787 David Sherwood <david.sherwood@arm.com>
16789 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16791 * gimple-expr.h (is_gimple_addressable: Likewise.
16792 * gimple-expr.c (is_gimple_address): Likewise.
16793 * internal-fn.c (expand_call_mem_ref): New function.
16794 (expand_mask_load_optab_fn): Use it.
16795 (expand_mask_store_optab_fn): Likewise.
16797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16798 Alan Hayward <alan.hayward@arm.com>
16799 David Sherwood <david.sherwood@arm.com>
16801 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16802 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16803 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16804 (cond_umax@var{mode}): Document.
16805 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16806 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16807 (cond_umin_optab, cond_umax_optab): New optabs.
16808 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16809 (COND_IOR, COND_XOR): New internal functions.
16810 * internal-fn.h (get_conditional_internal_fn): Declare.
16811 * internal-fn.c (cond_binary_direct): New macro.
16812 (expand_cond_binary_optab_fn): Likewise.
16813 (direct_cond_binary_optab_supported_p): Likewise.
16814 (get_conditional_internal_fn): New function.
16815 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16816 Cope with reduction statements that are vectorized as calls rather
16818 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16819 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16820 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16821 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16822 (UNSPEC_COND_EOR): New unspecs.
16823 (optab): Add mappings for them.
16824 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16825 (sve_int_op, sve_fp_op): New int attributes.
16827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16828 Alan Hayward <alan.hayward@arm.com>
16829 David Sherwood <david.sherwood@arm.com>
16831 * optabs.def (while_ult_optab): New optab.
16832 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16833 * internal-fn.def (WHILE_ULT): New internal function.
16834 * internal-fn.h (direct_internal_fn_supported_p): New override
16835 that takes two types as argument.
16836 * internal-fn.c (while_direct): New macro.
16837 (expand_while_optab_fn): New function.
16838 (convert_optab_supported_p): Likewise.
16839 (direct_while_optab_supported_p): New macro.
16840 * wide-int.h (wi::udiv_ceil): New function.
16841 * tree-vectorizer.h (rgroup_masks): New structure.
16842 (vec_loop_masks): New typedef.
16843 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16844 and fully_masked_p.
16845 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16846 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16847 (vect_max_vf): New function.
16848 (slpeel_make_loop_iterate_ntimes): Delete.
16849 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16850 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16851 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16852 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16853 internal-fn.h, stor-layout.h and optabs-query.h.
16854 (vect_set_loop_mask): New function.
16855 (add_preheader_seq): Likewise.
16856 (add_header_seq): Likewise.
16857 (interleave_supported_p): Likewise.
16858 (vect_maybe_permute_loop_masks): Likewise.
16859 (vect_set_loop_masks_directly): Likewise.
16860 (vect_set_loop_condition_masked): Likewise.
16861 (vect_set_loop_condition_unmasked): New function, split out from
16862 slpeel_make_loop_iterate_ntimes.
16863 (slpeel_make_loop_iterate_ntimes): Rename to..
16864 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16865 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16866 (vect_do_peeling): Update call accordingly.
16867 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16869 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16870 mask_compare_type, can_fully_mask_p and fully_masked_p.
16871 (release_vec_loop_masks): New function.
16872 (_loop_vec_info): Use it to free the loop masks.
16873 (can_produce_all_loop_masks_p): New function.
16874 (vect_get_max_nscalars_per_iter): Likewise.
16875 (vect_verify_full_masking): Likewise.
16876 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16877 retries, and free the mask rgroups before retrying. Check loop-wide
16878 reasons for disallowing fully-masked loops. Make the final decision
16879 about whether use a fully-masked loop or not.
16880 (vect_estimate_min_profitable_iters): Do not assume that peeling
16881 for the number of iterations will be needed for fully-masked loops.
16882 (vectorizable_reduction): Disable fully-masked loops.
16883 (vectorizable_live_operation): Likewise.
16884 (vect_halve_mask_nunits): New function.
16885 (vect_double_mask_nunits): Likewise.
16886 (vect_record_loop_mask): Likewise.
16887 (vect_get_loop_mask): Likewise.
16888 (vect_transform_loop): Handle the case in which the final loop
16889 iteration might handle a partial vector. Call vect_set_loop_condition
16890 instead of slpeel_make_loop_iterate_ntimes.
16891 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16892 (check_load_store_masking): New function.
16893 (prepare_load_store_mask): Likewise.
16894 (vectorizable_store): Handle fully-masked loops.
16895 (vectorizable_load): Likewise.
16896 (supportable_widening_operation): Use vect_halve_mask_nunits for
16898 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16899 (vect_gen_while): New function.
16900 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16901 (aarch64_uqdec<mode>): New insn.
16903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16904 Alan Hayward <alan.hayward@arm.com>
16905 David Sherwood <david.sherwood@arm.com>
16907 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16908 (reduc_xor_scal_optab): New optabs.
16909 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16910 (reduc_xor_scal_@var{m}): Document.
16911 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16912 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16913 internal functions.
16914 * fold-const-call.c (fold_const_call): Handle them.
16915 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16916 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16917 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16918 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16919 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16920 (UNSPEC_XORV): New unspecs.
16921 (optab): Add entries for them.
16922 (BITWISEV): New int iterator.
16923 (bit_reduc_op): New int attributes.
16925 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16926 Alan Hayward <alan.hayward@arm.com>
16927 David Sherwood <david.sherwood@arm.com>
16929 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16930 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16931 * optabs.def (vec_shl_insert_optab): New optab.
16932 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16933 (duplicate_and_interleave): Likewise.
16934 * tree-vect-loop.c: Include internal-fn.h.
16935 (neutral_op_for_slp_reduction): New function, split out from
16936 get_initial_defs_for_reduction.
16937 (get_initial_def_for_reduction): Handle option 2 for variable-length
16938 vectors by loading the neutral value into a vector and then shifting
16939 the initial value into element 0.
16940 (get_initial_defs_for_reduction): Replace the code argument with
16941 the neutral value calculated by neutral_op_for_slp_reduction.
16942 Use gimple_build_vector for constant-length vectors.
16943 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16944 but the first group_size elements have a neutral value.
16945 Use duplicate_and_interleave otherwise.
16946 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16947 Update call to get_initial_defs_for_reduction. Handle SLP
16948 reductions for variable-length vectors by creating one vector
16949 result for each scalar result, with the elements associated
16950 with other scalar results stubbed out with the neutral value.
16951 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16952 Require IFN_VEC_SHL_INSERT for double reductions on
16953 variable-length vectors, or SLP reductions that have
16954 a neutral value. Require can_duplicate_and_interleave_p
16955 support for variable-length unchained SLP reductions if there
16956 is no neutral value, such as for MIN/MAX reductions. Also require
16957 the number of vector elements to be a multiple of the number of
16958 SLP statements when doing variable-length unchained SLP reductions.
16959 Update call to vect_create_epilog_for_reduction.
16960 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16961 and remove initial values.
16962 (duplicate_and_interleave): Make public.
16963 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16964 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16966 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16967 Alan Hayward <alan.hayward@arm.com>
16968 David Sherwood <david.sherwood@arm.com>
16970 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16971 (can_duplicate_and_interleave_p): New function.
16972 (vect_get_and_check_slp_defs): Take the vector of statements
16973 rather than just the current one. Remove excess parentheses.
16974 Restriction rejectinon of vect_constant_def and vect_external_def
16975 for variable-length vectors to boolean types, or types for which
16976 can_duplicate_and_interleave_p is false.
16977 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16978 (duplicate_and_interleave): New function.
16979 (vect_get_constant_vectors): Use gimple_build_vector for
16980 constant-length vectors and suitable variable-length constant
16981 vectors. Use duplicate_and_interleave for other variable-length
16982 vectors. Don't defer the update when inserting new statements.
16984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16985 Alan Hayward <alan.hayward@arm.com>
16986 David Sherwood <david.sherwood@arm.com>
16988 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16989 min_profitable_iters doesn't go negative.
16991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16992 Alan Hayward <alan.hayward@arm.com>
16993 David Sherwood <david.sherwood@arm.com>
16995 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16996 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16997 * optabs.def (vec_mask_load_lanes_optab): New optab.
16998 (vec_mask_store_lanes_optab): Likewise.
16999 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17000 (MASK_STORE_LANES): Likewise.
17001 * internal-fn.c (mask_load_lanes_direct): New macro.
17002 (mask_store_lanes_direct): Likewise.
17003 (expand_mask_load_optab_fn): Handle masked operations.
17004 (expand_mask_load_lanes_optab_fn): New macro.
17005 (expand_mask_store_optab_fn): Handle masked operations.
17006 (expand_mask_store_lanes_optab_fn): New macro.
17007 (direct_mask_load_lanes_optab_supported_p): Likewise.
17008 (direct_mask_store_lanes_optab_supported_p): Likewise.
17009 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17011 (vect_load_lanes_supported): Likewise.
17012 * tree-vect-data-refs.c (strip_conversion): New function.
17013 (can_group_stmts_p): Likewise.
17014 (vect_analyze_data_ref_accesses): Use it instead of checking
17015 for a pair of assignments.
17016 (vect_store_lanes_supported): Take a masked_p parameter.
17017 (vect_load_lanes_supported): Likewise.
17018 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17019 vect_store_lanes_supported and vect_load_lanes_supported.
17020 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17021 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17022 parameter. Don't allow gaps for masked accesses.
17023 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17024 and vect_load_lanes_supported.
17025 (get_load_store_type): Take a masked_p parameter and update
17026 call to get_group_load_store_type.
17027 (vectorizable_store): Update call to get_load_store_type.
17028 Handle IFN_MASK_STORE_LANES.
17029 (vectorizable_load): Update call to get_load_store_type.
17030 Handle IFN_MASK_LOAD_LANES.
17032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17033 Alan Hayward <alan.hayward@arm.com>
17034 David Sherwood <david.sherwood@arm.com>
17036 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17038 * config/aarch64/aarch64-protos.h
17039 (aarch64_sve_struct_memory_operand_p): Declare.
17040 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17041 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17042 (VPRED, vpred): Handle SVE structure modes.
17043 * config/aarch64/constraints.md (Utx): New constraint.
17044 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17045 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17046 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17047 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17048 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17049 structure modes. Split into pieces after RA.
17050 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17051 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17053 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17054 SVE structure modes.
17055 (aarch64_classify_address): Likewise.
17056 (sizetochar): Move earlier in file.
17057 (aarch64_print_operand): Handle SVE register lists.
17058 (aarch64_array_mode): New function.
17059 (aarch64_sve_struct_memory_operand_p): Likewise.
17060 (TARGET_ARRAY_MODE): Redefine.
17062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17063 Alan Hayward <alan.hayward@arm.com>
17064 David Sherwood <david.sherwood@arm.com>
17066 * target.def (array_mode): New target hook.
17067 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17068 * doc/tm.texi: Regenerate.
17069 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17070 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17071 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17072 targetm.array_mode.
17073 * stor-layout.c (mode_for_array): Likewise. Support polynomial
17076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17077 Alan Hayward <alan.hayward@arm.com>
17078 David Sherwood <david.sherwood@arm.com>
17080 * fold-const.c (fold_binary_loc): Check the argument types
17081 rather than the result type when testing for a vector operation.
17083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17085 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17086 * doc/tm.texi: Regenerate.
17088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17089 Alan Hayward <alan.hayward@arm.com>
17090 David Sherwood <david.sherwood@arm.com>
17092 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17093 (sve): Document new AArch64 extension.
17094 * doc/md.texi (w): Extend the description of the AArch64
17095 constraint to include SVE vectors.
17096 (Upl, Upa): Document new AArch64 predicate constraints.
17097 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17099 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17100 (msve-vector-bits=): New option.
17101 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17102 SVE when these are disabled.
17103 (sve): New extension.
17104 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17105 modes. Adjust their number of units based on aarch64_sve_vg.
17106 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17107 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17108 aarch64_addr_query_type.
17109 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17110 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17111 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17112 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17113 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17114 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17115 (aarch64_simd_imm_zero_p): Delete.
17116 (aarch64_check_zero_based_sve_index_immediate): Declare.
17117 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17118 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17119 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17120 (aarch64_sve_float_mul_immediate_p): Likewise.
17121 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17122 rather than an rtx.
17123 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17124 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17125 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17126 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17127 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17128 (aarch64_regmode_natural_size): Likewise.
17129 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17130 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17132 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17133 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17134 for VG and the SVE predicate registers.
17135 (V_ALIASES): Add a "z"-prefixed alias.
17136 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17137 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17138 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17139 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17140 (REG_CLASS_NAMES): Add entries for them.
17141 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17142 and the predicate registers.
17143 (aarch64_sve_vg): Declare.
17144 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17145 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17146 (REGMODE_NATURAL_SIZE): Define.
17147 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17149 * config/aarch64/aarch64.c: Include cfgrtl.h.
17150 (simd_immediate_info): Add a constructor for series vectors,
17151 and an associated step field.
17152 (aarch64_sve_vg): New variable.
17153 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17154 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17155 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17156 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17157 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17158 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17159 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17160 (aarch64_get_mask_mode): New functions.
17161 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17162 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17163 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17164 predicate modes and predicate registers. Explicitly restrict
17165 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17166 to store a vector mode if it is recognized by
17167 aarch64_classify_vector_mode.
17168 (aarch64_regmode_natural_size): New function.
17169 (aarch64_hard_regno_caller_save_mode): Return the original mode
17171 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17172 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17173 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17174 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17176 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17177 does not overlap dest if the function is frame-related. Handle
17179 (aarch64_split_add_offset): New function.
17180 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17181 them aarch64_add_offset.
17182 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17183 and update call to aarch64_sub_sp.
17184 (aarch64_add_cfa_expression): New function.
17185 (aarch64_expand_prologue): Pass extra temporary registers to the
17186 functions above. Handle the case in which we need to emit new
17187 DW_CFA_expressions for registers that were originally saved
17188 relative to the stack pointer, but now have to be expressed
17189 relative to the frame pointer.
17190 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17192 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17193 IP0 and IP1 values for SVE frames.
17194 (aarch64_expand_vec_series): New function.
17195 (aarch64_expand_sve_widened_duplicate): Likewise.
17196 (aarch64_expand_sve_const_vector): Likewise.
17197 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17198 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17199 into the register, rather than emitting a SET directly.
17200 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17201 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17202 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17203 (offset_9bit_signed_scaled_p): New functions.
17204 (aarch64_replicate_bitmask_imm): New function.
17205 (aarch64_bitmask_imm): Use it.
17206 (aarch64_cannot_force_const_mem): Reject expressions involving
17207 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17208 (aarch64_classify_index): Handle SVE indices, by requiring
17209 a plain register index with a scale that matches the element size.
17210 (aarch64_classify_address): Handle SVE addresses. Assert that
17211 the mode of the address is VOIDmode or an integer mode.
17212 Update call to aarch64_classify_symbol.
17213 (aarch64_classify_symbolic_expression): Update call to
17214 aarch64_classify_symbol.
17215 (aarch64_const_vec_all_in_range_p): New function.
17216 (aarch64_print_vector_float_operand): Likewise.
17217 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17218 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17219 and the FP immediates 1.0 and 0.5.
17220 (aarch64_print_address_internal): Handle SVE addresses.
17221 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17222 (aarch64_regno_regclass): Handle predicate registers.
17223 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17225 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17226 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17227 (aarch64_convert_sve_vector_bits): New function.
17228 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17229 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17230 rather than an rtx.
17231 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17232 Handle SVE vector and predicate modes. Accept VL-based constants
17233 that need only one temporary register, and VL offsets that require
17234 no temporary registers.
17235 (aarch64_conditional_register_usage): Mark the predicate registers
17236 as fixed if SVE isn't available.
17237 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17238 Return true for SVE vector and predicate modes.
17239 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17240 rather than an unsigned int. Handle SVE modes.
17241 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17243 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17245 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17246 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17247 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17248 (aarch64_sve_float_mul_immediate_p): New functions.
17249 (aarch64_sve_valid_immediate): New function.
17250 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17251 Explicitly reject structure modes. Check for INDEX constants.
17252 Handle PTRUE and PFALSE constants.
17253 (aarch64_check_zero_based_sve_index_immediate): New function.
17254 (aarch64_simd_imm_zero_p): Delete.
17255 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17256 vector modes. Accept constants in the range of CNT[BHWD].
17257 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17258 ask for an Advanced SIMD mode.
17259 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17260 (aarch64_simd_vector_alignment): Handle SVE predicates.
17261 (aarch64_vectorize_preferred_vector_alignment): New function.
17262 (aarch64_simd_vector_alignment_reachable): Use it instead of
17264 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17265 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17267 (MAX_VECT_LEN): Delete.
17268 (expand_vec_perm_d): Add a vec_flags field.
17269 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17270 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17271 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17273 (aarch64_evpc_rev): Rename to...
17274 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17275 (aarch64_evpc_rev_global): New function.
17276 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17277 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17279 (aarch64_evpc_sve_tbl): New function.
17280 (aarch64_expand_vec_perm_const_1): Update after rename of
17281 aarch64_evpc_rev. Handle SVE permutes too, trying
17282 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17283 than aarch64_evpc_tbl.
17284 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17285 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17286 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17287 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17288 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17289 (aarch64_expand_sve_vcond): New functions.
17290 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17291 of aarch64_vector_mode_p.
17292 (aarch64_dwarf_poly_indeterminate_value): New function.
17293 (aarch64_compute_pressure_classes): Likewise.
17294 (aarch64_can_change_mode_class): Likewise.
17295 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17296 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17297 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17298 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17299 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17300 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17301 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17302 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17304 (Dn, Dl, Dr): Accept const as well as const_vector.
17305 (Dz): Likewise. Compare against CONST0_RTX.
17306 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17307 of "vector" where appropriate.
17308 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
17309 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
17310 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
17311 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
17312 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
17313 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
17314 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
17315 (v_int_equiv): Extend to SVE modes.
17316 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
17318 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
17319 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
17320 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
17321 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
17322 (SVE_COND_FP_CMP): New int iterators.
17323 (perm_hilo): Handle the new unpack unspecs.
17324 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
17326 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
17327 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
17328 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
17329 (aarch64_equality_operator, aarch64_constant_vector_operand)
17330 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
17331 (aarch64_sve_nonimmediate_operand): Likewise.
17332 (aarch64_sve_general_operand): Likewise.
17333 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
17334 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
17335 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
17336 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
17337 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
17338 (aarch64_sve_float_arith_immediate): Likewise.
17339 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
17340 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
17341 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
17342 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
17343 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
17344 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
17345 (aarch64_sve_float_arith_operand): Likewise.
17346 (aarch64_sve_float_arith_with_sub_operand): Likewise.
17347 (aarch64_sve_float_mul_operand): Likewise.
17348 (aarch64_sve_vec_perm_operand): Likewise.
17349 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
17350 (aarch64_mov_operand): Accept const_poly_int and const_vector.
17351 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
17352 as well as const_vector.
17353 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
17354 in file. Use CONST0_RTX and CONSTM1_RTX.
17355 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
17356 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
17357 Use aarch64_simd_imm_zero.
17358 * config/aarch64/aarch64-sve.md: New file.
17359 * config/aarch64/aarch64.md: Include it.
17360 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
17361 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
17362 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
17363 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
17364 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
17365 (sve): New attribute.
17366 (enabled): Disable instructions with the sve attribute unless
17368 (movqi, movhi): Pass CONST_POLY_INT operaneds through
17369 aarch64_expand_mov_immediate.
17370 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
17371 CNT[BHSD] immediates.
17372 (movti): Split CONST_POLY_INT moves into two halves.
17373 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
17374 Split additions that need a temporary here if the destination
17375 is the stack pointer.
17376 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
17377 (*add<mode>3_poly_1): New instruction.
17378 (set_clobber_cc): New expander.
17380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17382 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
17383 parameter and use it instead of GET_MODE_SIZE (innermode). Use
17384 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
17385 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
17386 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
17387 Change innermode from fixed_mode_size to machine_mode.
17388 (simplify_subreg): Update call accordingly. Handle a constant-sized
17389 subreg of a variable-length CONST_VECTOR.
17391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17392 Alan Hayward <alan.hayward@arm.com>
17393 David Sherwood <david.sherwood@arm.com>
17395 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
17396 (add_offset_to_base): New function, split out from...
17397 (create_mem_ref): ...here. When handling a scale other than 1,
17398 check first whether the address is valid without the offset.
17399 Add it into the base if so, leaving the index and scale as-is.
17401 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17404 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
17405 fold_for_warn before checking if arg2 is INTEGER_CST.
17407 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
17409 * config/rs6000/predicates.md (load_multiple_operation): Delete.
17410 (store_multiple_operation): Delete.
17411 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
17412 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
17413 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
17414 guarded by TARGET_STRING.
17415 (rs6000_output_load_multiple): Delete.
17416 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
17417 OPTION_MASK_STRING / TARGET_STRING handling.
17418 (print_operand) <'N', 'O'>: Add comment that these are unused now.
17419 (const rs6000_opt_masks) <"string">: Change mask to 0.
17420 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
17421 (MASK_STRING): Delete.
17422 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
17424 (load_multiple): Delete.
17431 (store_multiple): Delete.
17438 (movmemsi_8reg): Delete.
17439 (corresponding unnamed define_insn): Delete.
17440 (movmemsi_6reg): Delete.
17441 (corresponding unnamed define_insn): Delete.
17442 (movmemsi_4reg): Delete.
17443 (corresponding unnamed define_insn): Delete.
17444 (movmemsi_2reg): Delete.
17445 (corresponding unnamed define_insn): Delete.
17446 (movmemsi_1reg): Delete.
17447 (corresponding unnamed define_insn): Delete.
17448 * config/rs6000/rs6000.opt (mno-string): New.
17449 (mstring): Replace by deprecation warning stub.
17450 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17452 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17454 * regrename.c (regrename_do_replace): If replacing the same
17455 reg multiple times, try to reuse last created gen_raw_REG.
17458 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17459 main to workaround a bug in GDB.
17461 2018-01-12 Tom de Vries <tom@codesourcery.com>
17464 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17466 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17468 PR rtl-optimization/80481
17469 * ira-color.c (get_cap_member): New function.
17470 (allocnos_conflict_by_live_ranges_p): Use it.
17471 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17472 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17474 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17477 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17478 (*saddl_se_1): Ditto.
17479 (*ssubsi_1): Ditto.
17480 (*ssubl_se_1): Ditto.
17482 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17484 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17485 rather than wi::to_widest for DR_INITs.
17486 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17487 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17488 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17490 (vect_analyze_group_access_1): Note that here.
17492 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17494 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17495 polynomial type sizes.
17497 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17499 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17500 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17501 (gimple_add_tmp_var): Likewise.
17503 2018-01-12 Martin Liska <mliska@suse.cz>
17505 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17506 (gimple_alloc_sizes): Likewise.
17507 (dump_gimple_statistics): Use PRIu64 in printf format.
17508 * gimple.h: Change uint64_t to int.
17510 2018-01-12 Martin Liska <mliska@suse.cz>
17512 * tree-core.h: Use uint64_t instead of int.
17513 * tree.c (tree_node_counts): Likewise.
17514 (tree_node_sizes): Likewise.
17515 (dump_tree_statistics): Use PRIu64 in printf format.
17517 2018-01-12 Martin Liska <mliska@suse.cz>
17519 * Makefile.in: As qsort_chk is implemented in vec.c, add
17520 vec.o to linkage of gencfn-macros.
17521 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17522 passing the info to record_node_allocation_statistics.
17523 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17525 * ggc-common.c (struct ggc_usage): Add operator== and use
17526 it in operator< and compare function.
17527 * mem-stats.h (struct mem_usage): Likewise.
17528 * vec.c (struct vec_usage): Remove operator< and compare
17529 function. Can be simply inherited.
17531 2018-01-12 Martin Jambor <mjambor@suse.cz>
17534 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17535 * tree-ssa-math-opts.c: Include domwalk.h.
17536 (convert_mult_to_fma_1): New function.
17537 (fma_transformation_info): New type.
17538 (fma_deferring_state): Likewise.
17539 (cancel_fma_deferring): New function.
17540 (result_of_phi): Likewise.
17541 (last_fma_candidate_feeds_initial_phi): Likewise.
17542 (convert_mult_to_fma): Added deferring logic, split actual
17543 transformation to convert_mult_to_fma_1.
17544 (math_opts_dom_walker): New type.
17545 (math_opts_dom_walker::after_dom_children): New method, body moved
17546 here from pass_optimize_widening_mul::execute, added deferring logic
17548 (pass_optimize_widening_mul::execute): Moved most of code to
17549 math_opts_dom_walker::after_dom_children.
17550 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17551 * config/i386/i386.c (ix86_option_override_internal): Added
17552 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17554 2018-01-12 Richard Biener <rguenther@suse.de>
17557 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17558 inline instance vars.
17560 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17563 * config/rx/rx.c (rx_is_restricted_memory_address):
17564 Handle SUBREG case.
17566 2018-01-12 Richard Biener <rguenther@suse.de>
17568 PR tree-optimization/80846
17569 * target.def (split_reduction): New target hook.
17570 * targhooks.c (default_split_reduction): New function.
17571 * targhooks.h (default_split_reduction): Declare.
17572 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17573 target requests first reduce vectors by combining low and high
17575 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17576 (get_vectype_for_scalar_type_and_size): Export.
17577 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17578 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17579 * doc/tm.texi: Regenerate.
17580 * config/i386/i386.c (ix86_split_reduction): Implement
17581 TARGET_VECTORIZE_SPLIT_REDUCTION.
17583 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17586 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17587 in PIC mode except for TARGET_VXWORKS_RTP.
17588 * config/sparc/sparc.c: Include cfgrtl.h.
17589 (TARGET_INIT_PIC_REG): Define.
17590 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17591 (sparc_pic_register_p): New predicate.
17592 (sparc_legitimate_address_p): Use it.
17593 (sparc_legitimize_pic_address): Likewise.
17594 (sparc_delegitimize_address): Likewise.
17595 (sparc_mode_dependent_address_p): Likewise.
17596 (gen_load_pcrel_sym): Remove 4th parameter.
17597 (load_got_register): Adjust call to above. Remove obsolete stuff.
17598 (sparc_expand_prologue): Do not call load_got_register here.
17599 (sparc_flat_expand_prologue): Likewise.
17600 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17601 (sparc_use_pseudo_pic_reg): New function.
17602 (sparc_init_pic_reg): Likewise.
17603 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17604 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17606 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17608 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17609 Add item for branch_cost.
17611 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17613 PR rtl-optimization/83565
17614 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17615 not extend the result to a larger mode for rotate operations.
17616 (num_sign_bit_copies1): Likewise.
17618 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17621 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17623 Use values-Xc.o for -pedantic.
17624 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17626 2018-01-12 Martin Liska <mliska@suse.cz>
17629 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17631 (possible_polymorphic_call_targets): Use it.
17632 (ipa_devirt): Likewise.
17634 2018-01-12 Martin Liska <mliska@suse.cz>
17636 * profile-count.h (enum profile_quality): Use 0 as invalid
17637 enum value of profile_quality.
17639 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17641 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17642 -mext-string options.
17644 2018-01-12 Richard Biener <rguenther@suse.de>
17646 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17647 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17648 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17650 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17652 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17654 * configure.ac (--with-long-double-format): Add support for the
17655 configuration option to change the default long double format on
17657 * config.gcc (powerpc*-linux*-*): Likewise.
17658 * configure: Regenerate.
17659 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17660 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17661 used without modification.
17663 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17665 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17666 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17667 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17668 MISC_BUILTIN_SPEC_BARRIER.
17669 (rs6000_init_builtins): Likewise.
17670 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17672 (speculation_barrier): New define_insn.
17673 * doc/extend.texi: Document __builtin_speculation_barrier.
17675 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17678 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17679 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17680 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17682 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17683 integral modes instead of "ss" and "sd".
17684 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17685 vectors with 32-bit and 64-bit elements.
17686 (vecdupssescalarmodesuffix): New mode attribute.
17687 (vec_dup<mode>): Use it.
17689 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17692 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17693 frame if argument is passed on stack.
17695 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17698 * ree.c (combine_reaching_defs): Optimize also
17699 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17700 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17702 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17704 PR middle-end/83189
17705 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17707 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17709 PR middle-end/83718
17710 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17711 after they are computed.
17713 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17715 PR tree-optimization/83695
17716 * gimple-loop-linterchange.cc
17717 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17718 reset cached scev information after interchange.
17719 (pass_linterchange::execute): Remove call to scev_reset_htab.
17721 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17723 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17724 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17725 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17726 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17727 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17728 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17729 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17730 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17731 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17732 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17733 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17734 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17735 (V_lane_reg): Likewise.
17736 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17738 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17739 (vfmal_lane_low<mode>_intrinsic,
17740 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17741 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17742 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17743 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17744 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17745 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17747 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17749 * config/arm/arm-cpus.in (fp16fml): New feature.
17750 (ALL_SIMD): Add fp16fml.
17751 (armv8.2-a): Add fp16fml as an option.
17752 (armv8.3-a): Likewise.
17753 (armv8.4-a): Add fp16fml as part of fp16.
17754 * config/arm/arm.h (TARGET_FP16FML): Define.
17755 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17757 * config/arm/arm-modes.def (V2HF): Define.
17758 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17759 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17760 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17761 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17762 vfmsl_low, vfmsl_high): New set of builtins.
17763 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17764 (vfml_op): New code attribute.
17765 (VFMLHALVES): New int iterator.
17766 (VFML, VFMLSEL): New mode attributes.
17767 (V_reg): Define mapping for V2HF.
17768 (V_hi, V_lo): New mode attributes.
17769 (VF_constraint): Likewise.
17770 (vfml_half, vfml_half_selector): New int attributes.
17771 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17773 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17774 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17776 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17777 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17778 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17779 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17781 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17782 Document new effective target and option set.
17784 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17786 * config/arm/arm-cpus.in (armv8_4): New feature.
17787 (ARMv8_4a): New fgroup.
17788 (armv8.4-a): New arch.
17789 * config/arm/arm-tables.opt: Regenerate.
17790 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17791 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17792 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17793 Add matching rules for -march=armv8.4-a and extensions.
17794 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17796 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17799 * config/rx/rx.md (BW): New mode attribute.
17800 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17802 2018-01-11 Richard Biener <rguenther@suse.de>
17804 PR tree-optimization/83435
17805 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17806 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17807 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17809 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17810 Alan Hayward <alan.hayward@arm.com>
17811 David Sherwood <david.sherwood@arm.com>
17813 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17815 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17816 (aarch64_print_address_internal): Use it to check for a zero offset.
17818 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17819 Alan Hayward <alan.hayward@arm.com>
17820 David Sherwood <david.sherwood@arm.com>
17822 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17823 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17824 Return a poly_int64 rather than a HOST_WIDE_INT.
17825 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17826 rather than a HOST_WIDE_INT.
17827 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17828 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17829 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17830 final_offset from HOST_WIDE_INT to poly_int64.
17831 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17832 to_constant when getting the number of units in an Advanced SIMD
17834 (aarch64_builtin_vectorized_function): Check for a constant number
17836 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17838 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17839 attribute instead of GET_MODE_NUNITS.
17840 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17841 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17842 GET_MODE_SIZE for fixed-size registers.
17843 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17844 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17845 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17846 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17847 (aarch64_print_operand, aarch64_print_address_internal)
17848 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17849 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17850 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17851 Handle polynomial GET_MODE_SIZE.
17852 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17853 wider than SImode without modification.
17854 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17855 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17856 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17857 passing and returning SVE modes.
17858 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17859 rather than GEN_INT.
17860 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17861 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17862 (aarch64_allocate_and_probe_stack_space): Likewise.
17863 (aarch64_layout_frame): Cope with polynomial offsets.
17864 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17865 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17866 polynomial offsets.
17867 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17868 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17869 poly_int64 rather than a HOST_WIDE_INT.
17870 (aarch64_get_separate_components, aarch64_process_components)
17871 (aarch64_expand_prologue, aarch64_expand_epilogue)
17872 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17873 (aarch64_anchor_offset): New function, split out from...
17874 (aarch64_legitimize_address): ...here.
17875 (aarch64_builtin_vectorization_cost): Handle polynomial
17876 TYPE_VECTOR_SUBPARTS.
17877 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17879 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17880 number of elements from the PARALLEL rather than the mode.
17881 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17882 rather than GET_MODE_BITSIZE.
17883 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17884 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17885 (aarch64_expand_vec_perm_const_1): Handle polynomial
17886 d->perm.length () and d->perm elements.
17887 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17888 Apply to_constant to d->perm elements.
17889 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17890 polynomial CONST_VECTOR_NUNITS.
17891 (aarch64_move_pointer): Take amount as a poly_int64 rather
17893 (aarch64_progress_pointer): Avoid temporary variable.
17894 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17895 the mode attribute instead of GET_MODE.
17897 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17898 Alan Hayward <alan.hayward@arm.com>
17899 David Sherwood <david.sherwood@arm.com>
17901 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17902 x exists before using it.
17903 (aarch64_add_constant_internal): Rename to...
17904 (aarch64_add_offset_1): ...this. Replace regnum with separate
17905 src and dest rtxes. Handle the case in which they're different,
17906 including when the offset is zero. Replace scratchreg with an rtx.
17907 Use 2 additions if there is no spare register into which we can
17908 move a 16-bit constant.
17909 (aarch64_add_constant): Delete.
17910 (aarch64_add_offset): Replace reg with separate src and dest
17911 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17912 Use aarch64_add_offset_1.
17913 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17914 an rtx rather than an int. Take the delta as a poly_int64
17915 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17916 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17917 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17918 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17919 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17920 and aarch64_add_sp.
17921 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17922 aarch64_add_constant.
17924 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17926 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17927 Use scalar_float_mode.
17929 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17931 * config/aarch64/aarch64-simd.md
17932 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17933 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17934 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17935 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17936 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17937 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17938 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17939 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17940 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17941 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17943 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17946 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17947 targ_options->x_arm_arch_string is non NULL.
17949 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17951 * config/aarch64/aarch64.h
17952 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17954 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17957 * expmed.c (emit_store_flag_force): Swap if const op0
17958 and change VOIDmode to mode of op0.
17960 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17962 PR rtl-optimization/83761
17963 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17964 than bytes to mode_for_size.
17966 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17968 PR middle-end/83189
17969 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17970 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17973 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17975 PR middle-end/83575
17976 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17977 when in layout mode.
17978 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17979 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17982 2018-01-10 Michael Collison <michael.collison@arm.com>
17984 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17985 * config/aarch64/aarch64-option-extension.def: Add
17986 AARCH64_OPT_EXTENSION of 'fp16fml'.
17987 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17988 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17989 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17990 * config/aarch64/constraints.md (Ui7): New constraint.
17991 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17992 (VFMLA_SEL_W): Ditto.
17995 (VFMLA16_LOW): New int iterator.
17996 (VFMLA16_HIGH): Ditto.
17997 (UNSPEC_FMLAL): New unspec.
17998 (UNSPEC_FMLSL): Ditto.
17999 (UNSPEC_FMLAL2): Ditto.
18000 (UNSPEC_FMLSL2): Ditto.
18001 (f16mac): New code attribute.
18002 * config/aarch64/aarch64-simd-builtins.def
18003 (aarch64_fmlal_lowv2sf): Ditto.
18004 (aarch64_fmlsl_lowv2sf): Ditto.
18005 (aarch64_fmlalq_lowv4sf): Ditto.
18006 (aarch64_fmlslq_lowv4sf): Ditto.
18007 (aarch64_fmlal_highv2sf): Ditto.
18008 (aarch64_fmlsl_highv2sf): Ditto.
18009 (aarch64_fmlalq_highv4sf): Ditto.
18010 (aarch64_fmlslq_highv4sf): Ditto.
18011 (aarch64_fmlal_lane_lowv2sf): Ditto.
18012 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18013 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18014 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18015 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18016 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18017 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18018 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18019 (aarch64_fmlal_lane_highv2sf): Ditto.
18020 (aarch64_fmlsl_lane_highv2sf): Ditto.
18021 (aarch64_fmlal_laneq_highv2sf): Ditto.
18022 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18023 (aarch64_fmlalq_lane_highv4sf): Ditto.
18024 (aarch64_fmlsl_lane_highv4sf): Ditto.
18025 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18026 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18027 * config/aarch64/aarch64-simd.md:
18028 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18029 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18030 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18031 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18032 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18033 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18034 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18035 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18036 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18037 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18038 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18039 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18040 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18041 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18042 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18043 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18044 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18045 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18046 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18047 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18048 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18049 (vfmlsl_low_u32): Ditto.
18050 (vfmlalq_low_u32): Ditto.
18051 (vfmlslq_low_u32): Ditto.
18052 (vfmlal_high_u32): Ditto.
18053 (vfmlsl_high_u32): Ditto.
18054 (vfmlalq_high_u32): Ditto.
18055 (vfmlslq_high_u32): Ditto.
18056 (vfmlal_lane_low_u32): Ditto.
18057 (vfmlsl_lane_low_u32): Ditto.
18058 (vfmlal_laneq_low_u32): Ditto.
18059 (vfmlsl_laneq_low_u32): Ditto.
18060 (vfmlalq_lane_low_u32): Ditto.
18061 (vfmlslq_lane_low_u32): Ditto.
18062 (vfmlalq_laneq_low_u32): Ditto.
18063 (vfmlslq_laneq_low_u32): Ditto.
18064 (vfmlal_lane_high_u32): Ditto.
18065 (vfmlsl_lane_high_u32): Ditto.
18066 (vfmlal_laneq_high_u32): Ditto.
18067 (vfmlsl_laneq_high_u32): Ditto.
18068 (vfmlalq_lane_high_u32): Ditto.
18069 (vfmlslq_lane_high_u32): Ditto.
18070 (vfmlalq_laneq_high_u32): Ditto.
18071 (vfmlslq_laneq_high_u32): Ditto.
18072 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18073 (AARCH64_FL_FOR_ARCH8_4): New.
18074 (AARCH64_ISA_F16FML): New ISA flag.
18075 (TARGET_F16FML): New feature flag for fp16fml.
18076 (doc/invoke.texi): Document new fp16fml option.
18078 2018-01-10 Michael Collison <michael.collison@arm.com>
18080 * config/aarch64/aarch64-builtins.c:
18081 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18082 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18083 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18084 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18085 (AARCH64_ISA_SHA3): New ISA flag.
18086 (TARGET_SHA3): New feature flag for sha3.
18087 * config/aarch64/iterators.md (sha512_op): New int attribute.
18088 (CRYPTO_SHA512): New int iterator.
18089 (UNSPEC_SHA512H): New unspec.
18090 (UNSPEC_SHA512H2): Ditto.
18091 (UNSPEC_SHA512SU0): Ditto.
18092 (UNSPEC_SHA512SU1): Ditto.
18093 * config/aarch64/aarch64-simd-builtins.def
18094 (aarch64_crypto_sha512hqv2di): New builtin.
18095 (aarch64_crypto_sha512h2qv2di): Ditto.
18096 (aarch64_crypto_sha512su0qv2di): Ditto.
18097 (aarch64_crypto_sha512su1qv2di): Ditto.
18098 (aarch64_eor3qv8hi): Ditto.
18099 (aarch64_rax1qv2di): Ditto.
18100 (aarch64_xarqv2di): Ditto.
18101 (aarch64_bcaxqv8hi): Ditto.
18102 * config/aarch64/aarch64-simd.md:
18103 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18104 (aarch64_crypto_sha512su0qv2di): Ditto.
18105 (aarch64_crypto_sha512su1qv2di): Ditto.
18106 (aarch64_eor3qv8hi): Ditto.
18107 (aarch64_rax1qv2di): Ditto.
18108 (aarch64_xarqv2di): Ditto.
18109 (aarch64_bcaxqv8hi): Ditto.
18110 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18111 (vsha512h2q_u64): Ditto.
18112 (vsha512su0q_u64): Ditto.
18113 (vsha512su1q_u64): Ditto.
18114 (veor3q_u16): Ditto.
18115 (vrax1q_u64): Ditto.
18116 (vxarq_u64): Ditto.
18117 (vbcaxq_u16): Ditto.
18118 * config/arm/types.md (crypto_sha512): New type attribute.
18119 (crypto_sha3): Ditto.
18120 (doc/invoke.texi): Document new sha3 option.
18122 2018-01-10 Michael Collison <michael.collison@arm.com>
18124 * config/aarch64/aarch64-builtins.c:
18125 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18126 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18127 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18128 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18129 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18130 (AARCH64_ISA_SM4): New ISA flag.
18131 (TARGET_SM4): New feature flag for sm4.
18132 * config/aarch64/aarch64-simd-builtins.def
18133 (aarch64_sm3ss1qv4si): Ditto.
18134 (aarch64_sm3tt1aq4si): Ditto.
18135 (aarch64_sm3tt1bq4si): Ditto.
18136 (aarch64_sm3tt2aq4si): Ditto.
18137 (aarch64_sm3tt2bq4si): Ditto.
18138 (aarch64_sm3partw1qv4si): Ditto.
18139 (aarch64_sm3partw2qv4si): Ditto.
18140 (aarch64_sm4eqv4si): Ditto.
18141 (aarch64_sm4ekeyqv4si): Ditto.
18142 * config/aarch64/aarch64-simd.md:
18143 (aarch64_sm3ss1qv4si): Ditto.
18144 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18145 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18146 (aarch64_sm4eqv4si): Ditto.
18147 (aarch64_sm4ekeyqv4si): Ditto.
18148 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18149 (sm3part_op): Ditto.
18150 (CRYPTO_SM3TT): Ditto.
18151 (CRYPTO_SM3PART): Ditto.
18152 (UNSPEC_SM3SS1): New unspec.
18153 (UNSPEC_SM3TT1A): Ditto.
18154 (UNSPEC_SM3TT1B): Ditto.
18155 (UNSPEC_SM3TT2A): Ditto.
18156 (UNSPEC_SM3TT2B): Ditto.
18157 (UNSPEC_SM3PARTW1): Ditto.
18158 (UNSPEC_SM3PARTW2): Ditto.
18159 (UNSPEC_SM4E): Ditto.
18160 (UNSPEC_SM4EKEY): Ditto.
18161 * config/aarch64/constraints.md (Ui2): New constraint.
18162 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18163 * config/arm/types.md (crypto_sm3): New type attribute.
18164 (crypto_sm4): Ditto.
18165 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18166 (vsm3tt1aq_u32): Ditto.
18167 (vsm3tt1bq_u32): Ditto.
18168 (vsm3tt2aq_u32): Ditto.
18169 (vsm3tt2bq_u32): Ditto.
18170 (vsm3partw1q_u32): Ditto.
18171 (vsm3partw2q_u32): Ditto.
18172 (vsm4eq_u32): Ditto.
18173 (vsm4ekeyq_u32): Ditto.
18174 (doc/invoke.texi): Document new sm4 option.
18176 2018-01-10 Michael Collison <michael.collison@arm.com>
18178 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18179 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18180 (AARCH64_FL_FOR_ARCH8_4): New.
18181 (AARCH64_FL_V8_4): New flag.
18182 (doc/invoke.texi): Document new armv8.4-a option.
18184 2018-01-10 Michael Collison <michael.collison@arm.com>
18186 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18187 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18188 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18189 * config/aarch64/aarch64-option-extension.def: Add
18190 AARCH64_OPT_EXTENSION of 'sha2'.
18191 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18192 (crypto): Disable sha2 and aes if crypto disabled.
18193 (crypto): Enable aes and sha2 if enabled.
18194 (simd): Disable sha2 and aes if simd disabled.
18195 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18197 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18198 (TARGET_SHA2): New feature flag for sha2.
18199 (TARGET_AES): New feature flag for aes.
18200 * config/aarch64/aarch64-simd.md:
18201 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18202 conditional on TARGET_AES.
18203 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18204 (aarch64_crypto_sha1hsi): Make pattern conditional
18206 (aarch64_crypto_sha1hv4si): Ditto.
18207 (aarch64_be_crypto_sha1hv4si): Ditto.
18208 (aarch64_crypto_sha1su1v4si): Ditto.
18209 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18210 (aarch64_crypto_sha1su0v4si): Ditto.
18211 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18212 (aarch64_crypto_sha256su0v4si): Ditto.
18213 (aarch64_crypto_sha256su1v4si): Ditto.
18214 (doc/invoke.texi): Document new aes and sha2 options.
18216 2018-01-10 Martin Sebor <msebor@redhat.com>
18218 PR tree-optimization/83781
18219 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18222 2018-01-11 Martin Sebor <msebor@gmail.com>
18223 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18225 PR tree-optimization/83501
18226 PR tree-optimization/81703
18228 * tree-ssa-strlen.c (get_string_cst): Rename...
18229 (get_string_len): ...to this. Handle global constants.
18230 (handle_char_store): Adjust.
18232 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18233 Jim Wilson <jimw@sifive.com>
18235 * config/riscv/riscv-protos.h (riscv_output_return): New.
18236 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18237 (riscv_attribute_table, riscv_output_return),
18238 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18239 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18240 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18242 (riscv_expand_prologue): Add early return for naked function.
18243 (riscv_expand_epilogue): Likewise.
18244 (riscv_function_ok_for_sibcall): Return false for naked function.
18245 (riscv_set_current_function): New.
18246 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18247 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18248 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18249 * doc/extend.texi (RISC-V Function Attributes): New.
18251 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18253 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18254 check for 128-bit long double before checking TCmode.
18255 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18256 128-bit long doubles before checking TFmode or TCmode.
18257 (FLOAT128_IBM_P): Likewise.
18259 2018-01-10 Martin Sebor <msebor@redhat.com>
18261 PR tree-optimization/83671
18262 * builtins.c (c_strlen): Unconditionally return zero for the empty
18264 Use -Warray-bounds for warnings.
18265 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18266 for non-constant array indices with COMPONENT_REF, arrays of
18267 arrays, and pointers to arrays.
18268 (gimple_fold_builtin_strlen): Determine and set length range for
18269 non-constant character arrays.
18271 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18273 PR middle-end/81897
18274 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18277 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18279 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18281 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18284 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18285 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18286 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18287 indexed_or_indirect_operand predicate.
18288 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18289 (*vsx_le_perm_load_v8hi): Likewise.
18290 (*vsx_le_perm_load_v16qi): Likewise.
18291 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18292 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18293 (*vsx_le_perm_store_v8hi): Likewise.
18294 (*vsx_le_perm_store_v16qi): Likewise.
18295 (eight unnamed splitters): Likewise.
18297 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18299 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18300 * config/rs6000/emmintrin.h: Likewise.
18301 * config/rs6000/mmintrin.h: Likewise.
18302 * config/rs6000/xmmintrin.h: Likewise.
18304 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18307 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
18309 * tree.c (tree_nop_conversion): Return true for location wrapper
18311 (maybe_wrap_with_location): New function.
18312 (selftest::check_strip_nops): New function.
18313 (selftest::test_location_wrappers): New function.
18314 (selftest::tree_c_tests): Call it.
18315 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
18316 (maybe_wrap_with_location): New decl.
18317 (EXPR_LOCATION_WRAPPER_P): New macro.
18318 (location_wrapper_p): New inline function.
18319 (tree_strip_any_location_wrapper): New inline function.
18321 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
18324 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
18325 stack_realign_offset for the largest alignment of stack slot
18327 (ix86_find_max_used_stack_alignment): New function.
18328 (ix86_finalize_stack_frame_flags): Use it. Set
18329 max_used_stack_alignment if we don't realign stack.
18330 * config/i386/i386.h (machine_function): Add
18331 max_used_stack_alignment.
18333 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
18335 * config/arm/arm.opt (-mbranch-cost): New option.
18336 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
18339 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
18342 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
18343 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
18345 2018-01-10 Richard Biener <rguenther@suse.de>
18348 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
18349 early out so it also covers the case where we have a non-NULL
18352 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18354 PR tree-optimization/83753
18355 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
18356 for non-strided grouped accesses if the number of elements is 1.
18358 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18361 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
18362 * i386.h (TARGET_USE_GATHER): Define.
18363 * x86-tune.def (X86_TUNE_USE_GATHER): New.
18365 2018-01-10 Martin Liska <mliska@suse.cz>
18368 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
18369 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
18371 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
18372 CLEANUP_NO_PARTITIONING is not set.
18374 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18376 * doc/rtl.texi: Remove documentation of (const ...) wrappers
18377 for vectors, as a partial revert of r254296.
18378 * rtl.h (const_vec_p): Delete.
18379 (const_vec_duplicate_p): Don't test for vector CONSTs.
18380 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
18381 * expmed.c (make_tree): Likewise.
18384 * common.md (E, F): Use CONSTANT_P instead of checking for
18386 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
18387 checking for CONST_VECTOR.
18389 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18391 PR middle-end/83575
18392 * predict.c (force_edge_cold): Handle in more sane way edges
18393 with no prediction.
18395 2018-01-09 Carl Love <cel@us.ibm.com>
18397 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
18399 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
18400 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
18401 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
18402 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
18403 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
18404 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
18405 * config/rs6000/rs6000-protos.h: Add extern defition for
18406 rs6000_generate_float2_double_code.
18407 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
18409 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
18410 (float2_v2df): Add define_expand.
18412 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
18415 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
18416 op_mode in the force_to_mode call.
18418 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18420 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
18421 instead of checking each element individually.
18422 (aarch64_evpc_uzp): Likewise.
18423 (aarch64_evpc_zip): Likewise.
18424 (aarch64_evpc_ext): Likewise.
18425 (aarch64_evpc_rev): Likewise.
18426 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
18427 instead of checking each element individually. Return true without
18429 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18430 whether all selected elements come from the same input, instead of
18431 checking each element individually. Remove calls to gen_rtx_REG,
18432 start_sequence and end_sequence and instead assert that no rtl is
18435 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18437 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18438 order of HIGH and CONST checks.
18440 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18442 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18443 if the destination isn't an SSA_NAME.
18445 2018-01-09 Richard Biener <rguenther@suse.de>
18447 PR tree-optimization/83668
18448 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18450 (canonicalize_loop_form): ... here, renamed from ...
18451 (canonicalize_loop_closed_ssa_form): ... this and amended to
18452 swap successor edges for loop exit blocks to make us use
18453 the RPO order we need for initial schedule generation.
18455 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18457 PR tree-optimization/64811
18458 * match.pd: When optimizing comparisons with Inf, avoid
18459 introducing or losing exceptions from comparisons with NaN.
18461 2018-01-09 Martin Liska <mliska@suse.cz>
18464 * asan.c (shadow_mem_size): Add gcc_assert.
18466 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18468 Don't save registers in main().
18471 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18472 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18473 * config/avr/avr.c (avr_set_current_function): Don't error if
18474 naked, OS_task or OS_main are specified at the same time.
18475 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18477 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18479 * common/config/avr/avr-common.c (avr_option_optimization_table):
18480 Switch on -mmain-is-OS_task for optimizing compilations.
18482 2018-01-09 Richard Biener <rguenther@suse.de>
18484 PR tree-optimization/83572
18485 * graphite.c: Include cfganal.h.
18486 (graphite_transform_loops): Connect infinite loops to exit
18487 and remove fake edges at the end.
18489 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18491 * ipa-inline.c (edge_badness): Revert accidental checkin.
18493 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18496 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18497 symbols; not inline clones.
18499 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18502 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18503 hard registers. Formatting fixes.
18505 PR preprocessor/83722
18506 * gcc.c (try_generate_repro): Pass
18507 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18508 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18511 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18512 Kito Cheng <kito.cheng@gmail.com>
18514 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18515 (riscv_leaf_function_p): Delete.
18516 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18518 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18520 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18522 (do_ifelse): New function.
18523 (do_isel): New function.
18524 (do_sub3): New function.
18525 (do_add3): New function.
18526 (do_load_mask_compare): New function.
18527 (do_overlap_load_compare): New function.
18528 (expand_compare_loop): New function.
18529 (expand_block_compare): Call expand_compare_loop() when appropriate.
18530 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18531 option description.
18532 (-mblock-compare-inline-loop-limit): New option.
18534 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18537 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18538 Reverse order of second and third operands in first alternative.
18539 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18540 of first and second elements in UNSPEC_VPERMR vector.
18541 (altivec_expand_vec_perm_le): Likewise.
18543 2018-01-08 Jeff Law <law@redhat.com>
18545 PR rtl-optimizatin/81308
18546 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18547 (process_switch): If group_case_labels makes a change, then set
18549 (pass_convert_switch::execute): If a switch is converted, then
18550 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18552 PR rtl-optimization/81308
18553 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18556 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18558 PR target/83663 - Revert r255946
18559 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18560 generation for cases where splatting a value is not useful.
18561 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18562 across a vec_duplicate and a paradoxical subreg forming a vector
18563 mode to a vec_concat.
18565 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18567 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18568 -march=armv8.3-a variants.
18569 * config/arm/t-multilib: Likewise.
18570 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18572 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18574 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18576 (cceq_ior_compare_complement): Give it a name so I can use it, and
18577 change boolean_or_operator predicate to boolean_operator so it can
18578 be used to generate a crand.
18579 (eqne): New code iterator.
18580 (bd/bd_neg): New code_attrs.
18581 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18582 a single define_insn.
18583 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18584 decrement (bdnzt/bdnzf/bdzt/bdzf).
18585 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18586 with the new names of the branch decrement patterns, and added the
18587 names of the branch decrement conditional patterns.
18589 2018-01-08 Richard Biener <rguenther@suse.de>
18591 PR tree-optimization/83563
18592 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18595 2018-01-08 Richard Biener <rguenther@suse.de>
18597 PR middle-end/83713
18598 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18600 2018-01-08 Richard Biener <rguenther@suse.de>
18602 PR tree-optimization/83685
18603 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18604 references to abnormals.
18606 2018-01-08 Richard Biener <rguenther@suse.de>
18609 * dwarf2out.c (output_indirect_strings): Handle empty
18610 skeleton_debug_str_hash.
18611 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18613 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18615 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18616 (emit_store_direct): Likewise.
18617 (arc_trampoline_adjust_address): Likewise.
18618 (arc_asm_trampoline_template): New function.
18619 (arc_initialize_trampoline): Use asm_trampoline_template.
18620 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18621 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18622 * config/arc/arc.md (flush_icache): Delete pattern.
18624 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18626 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18627 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18630 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18633 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18634 by not USED_FOR_TARGET.
18635 (make_pass_resolve_sw_modes): Likewise.
18637 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18639 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18642 2018-01-08 Richard Biener <rguenther@suse.de>
18644 PR middle-end/83580
18645 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18647 2018-01-08 Richard Biener <rguenther@suse.de>
18649 PR middle-end/83517
18650 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18652 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18654 PR middle-end/81897
18655 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18656 basic blocks with a small number of successors.
18657 (convert_control_dep_chain_into_preds): Improve handling of
18659 (dump_predicates): Split apart into...
18660 (dump_pred_chain): ...here...
18661 (dump_pred_info): ...and here.
18662 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18663 (can_chain_union_be_invalidated_p): Improve check for invalidation
18665 (uninit_uses_cannot_happen): Avoid unnecessary if
18666 convert_control_dep_chain_into_preds yielded nothing.
18668 2018-01-06 Martin Sebor <msebor@redhat.com>
18670 PR tree-optimization/83640
18671 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18672 subtracting negative offset from size.
18673 (builtin_access::overlap): Adjust offset bounds of the access to fall
18674 within the size of the object if possible.
18676 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18678 PR rtl-optimization/83699
18679 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18680 extract_bit_field_as_subreg to cases in which the extracted
18681 value is also a vector.
18683 * lra-constraints.c (process_alt_operands): Test for the equivalence
18684 substitutions when detecting a possible reload cycle.
18686 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18689 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18690 by default if flag_selective_schedling{,2}. Formatting fixes.
18692 PR rtl-optimization/83682
18693 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18694 if it has non-VECTOR_MODE element mode.
18695 (vec_duplicate_p): Likewise.
18697 PR middle-end/83694
18698 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18699 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18701 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18704 * config/i386/i386-builtin.def
18705 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18706 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18707 Require also OPTION_MASK_ISA_AVX512F in addition to
18708 OPTION_MASK_ISA_GFNI.
18709 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18710 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18711 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18712 to OPTION_MASK_ISA_GFNI.
18713 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18714 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18715 OPTION_MASK_ISA_AVX512BW.
18716 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18717 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18718 addition to OPTION_MASK_ISA_GFNI.
18719 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18720 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18721 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18722 to OPTION_MASK_ISA_GFNI.
18723 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18724 a requirement for all ISAs rather than any of them with a few
18726 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18728 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18729 bitmasks to be enabled with 3 exceptions, instead of requiring any
18730 enabled ISA with lots of exceptions.
18731 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18732 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18733 Change avx512bw in isa attribute to avx512f.
18734 * config/i386/sgxintrin.h: Add license boilerplate.
18735 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18736 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18737 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18738 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18740 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18741 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18742 temporarily sse2 rather than sse if not enabled already.
18745 * config/i386/sse.md (VI248_VLBW): Rename to ...
18746 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18747 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18748 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18749 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18750 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18751 mode iterator instead of VI248_VLBW.
18753 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18755 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18756 (record_modified): Skip clobbers; add debug output.
18757 (param_change_prob): Use sreal frequencies.
18759 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18761 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18762 punt for user-aligned variables.
18764 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18766 * tree-chrec.c (chrec_contains_symbols): Return true for
18769 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18772 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18773 of (x|y) == x for BICS pattern.
18775 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18777 PR tree-optimization/83605
18778 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18779 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18782 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18784 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18785 * config/epiphany/rtems.h: New file.
18787 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18788 Uros Bizjak <ubizjak@gmail.com>
18791 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18792 QIreg_operand instead of register_operand predicate.
18793 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18794 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18795 comments instead of -fmitigate[-_]rop.
18797 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18800 * cgraphunit.c (symbol_table::compile): Switch to text_section
18801 before calling assembly_start debug hook.
18802 * run-rtl-passes.c (run_rtl_passes): Likewise.
18805 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18807 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18808 range_int_cst_p rather than !symbolic_range_p before calling
18809 extract_range_from_multiplicative_op_1.
18811 2018-01-04 Jeff Law <law@redhat.com>
18813 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18814 redundant test in assertion.
18816 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18818 * doc/rtl.texi: Document machine_mode wrapper classes.
18820 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18822 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18823 using tree_to_uhwi.
18825 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18827 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18828 the VEC_PERM_EXPR fold to fail.
18830 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18833 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18834 to switched_sections.
18836 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18839 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18840 test for d.testing.
18842 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18845 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18846 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18848 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18851 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18852 is BLKmode and bitpos not zero or mode change is needed.
18854 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18857 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18860 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18863 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18864 instead of MULT rtx. Update all corresponding splitters.
18865 (*saddl_se): Ditto.
18866 (*ssub<modesuffix>): Ditto.
18867 (*ssubl_se): Ditto.
18868 (*cmp_sadd_di): Update split patterns.
18869 (*cmp_sadd_si): Ditto.
18870 (*cmp_sadd_sidi): Ditto.
18871 (*cmp_ssub_di): Ditto.
18872 (*cmp_ssub_si): Ditto.
18873 (*cmp_ssub_sidi): Ditto.
18874 * config/alpha/predicates.md (const23_operand): New predicate.
18875 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18876 Look for ASHIFT, not MULT inner operand.
18877 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18879 2018-01-04 Martin Liska <mliska@suse.cz>
18881 PR gcov-profile/83669
18882 * gcov.c (output_intermediate_file): Add version to intermediate
18884 * doc/gcov.texi: Document new field 'version' in intermediate
18885 file format. Fix location of '-k' option of gcov command.
18887 2018-01-04 Martin Liska <mliska@suse.cz>
18890 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18892 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18894 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18896 2018-01-03 Martin Sebor <msebor@redhat.com>
18898 PR tree-optimization/83655
18899 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18900 checking calls with invalid arguments.
18902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18904 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18905 (vectorizable_mask_load_store): Delete.
18906 (vectorizable_call): Return false for masked loads and stores.
18907 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18908 instead of gimple_assign_rhs1.
18909 (vectorizable_load): Handle IFN_MASK_LOAD.
18910 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18914 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18916 (vectorizable_mask_load_store): ...here.
18917 (vectorizable_load): ...and here.
18919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18921 * tree-vect-stmts.c (vect_build_all_ones_mask)
18922 (vect_build_zero_merge_argument): New functions, split out from...
18923 (vectorizable_load): ...here.
18925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18927 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18929 (vectorizable_mask_load_store): ...here.
18930 (vectorizable_store): ...and here.
18932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18934 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18936 (vectorizable_mask_load_store): ...here.
18938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18940 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18941 (vect_model_store_cost): Take a vec_load_store_type instead of a
18943 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18944 (vect_model_store_cost): Take a vec_load_store_type instead of a
18946 (vectorizable_mask_load_store): Update accordingly.
18947 (vectorizable_store): Likewise.
18948 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18952 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18953 IFN_MASK_LOAD calls here rather than...
18954 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18957 Alan Hayward <alan.hayward@arm.com>
18958 David Sherwood <david.sherwood@arm.com>
18960 * expmed.c (extract_bit_field_1): For vector extracts,
18961 fall back to extract_bit_field_as_subreg if vec_extract
18964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18965 Alan Hayward <alan.hayward@arm.com>
18966 David Sherwood <david.sherwood@arm.com>
18968 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18969 they are variable or constant sized.
18970 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18971 slots for constant-sized data.
18973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18974 Alan Hayward <alan.hayward@arm.com>
18975 David Sherwood <david.sherwood@arm.com>
18977 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18978 handling COND_EXPRs with boolean comparisons, try to find a better
18979 basis for the mask type than the boolean itself.
18981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18983 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18984 is calculated and how it can be overridden.
18985 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18986 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18988 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18992 Alan Hayward <alan.hayward@arm.com>
18993 David Sherwood <david.sherwood@arm.com>
18995 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18996 Remove the mode argument.
18997 (aarch64_simd_valid_immediate): Remove the mode and inverse
18999 * config/aarch64/iterators.md (bitsize): New iterator.
19000 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19001 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19002 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19003 aarch64_simd_valid_immediate.
19004 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19005 (aarch64_reg_or_bic_imm): Likewise.
19006 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19007 with an insn_type enum and msl with a modifier_type enum.
19008 Replace element_width with a scalar_mode. Change the shift
19009 to unsigned int. Add constructors for scalar_float_mode and
19010 scalar_int_mode elements.
19011 (aarch64_vect_float_const_representable_p): Delete.
19012 (aarch64_can_const_movi_rtx_p)
19013 (aarch64_simd_scalar_immediate_valid_for_move)
19014 (aarch64_simd_make_constant): Update call to
19015 aarch64_simd_valid_immediate.
19016 (aarch64_advsimd_valid_immediate_hs): New function.
19017 (aarch64_advsimd_valid_immediate): Likewise.
19018 (aarch64_simd_valid_immediate): Remove mode and inverse
19019 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19020 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19021 and aarch64_float_const_representable_p on the result.
19022 (aarch64_output_simd_mov_immediate): Remove mode argument.
19023 Update call to aarch64_simd_valid_immediate and use of
19024 simd_immediate_info.
19025 (aarch64_output_scalar_simd_mov_immediate): Update call
19028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19029 Alan Hayward <alan.hayward@arm.com>
19030 David Sherwood <david.sherwood@arm.com>
19032 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19033 (mode_nunits): Likewise CONST_MODE_NUNITS.
19034 * machmode.def (ADJUST_NUNITS): Document.
19035 * genmodes.c (mode_data::need_nunits_adj): New field.
19036 (blank_mode): Update accordingly.
19037 (adj_nunits): New variable.
19038 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19040 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19041 listed in adj_nunits.
19042 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19043 listed in adj_nunits. Don't emit case statements for such modes.
19044 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19045 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19046 nothing if adj_nunits is nonnull.
19047 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19048 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19049 (emit_mode_fbit): Update use of print_maybe_const_decl.
19050 (emit_move_size): Likewise. Treat the array as non-const
19052 (emit_mode_adjustments): Handle adj_nunits.
19054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19056 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19057 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19058 (VECTOR_MODES): Use it.
19059 (make_vector_modes): Take the prefix as an argument.
19061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19062 Alan Hayward <alan.hayward@arm.com>
19063 David Sherwood <david.sherwood@arm.com>
19065 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19066 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19067 for MODE_VECTOR_BOOL.
19068 * machmode.def (VECTOR_BOOL_MODE): Document.
19069 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19070 (make_vector_bool_mode): New function.
19071 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19073 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19074 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19076 * stor-layout.c (int_mode_for_mode): Likewise.
19077 * tree.c (build_vector_type_for_mode): Likewise.
19078 * varasm.c (output_constant_pool_2): Likewise.
19079 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19080 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19081 for MODE_VECTOR_BOOL.
19082 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19083 of mode class checks.
19084 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19085 instead of a list of mode class checks.
19086 (expand_vector_scalar_condition): Likewise.
19087 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19090 Alan Hayward <alan.hayward@arm.com>
19091 David Sherwood <david.sherwood@arm.com>
19093 * machmode.h (mode_size): Change from unsigned short to
19095 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19096 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19097 or if measurement_type is not polynomial.
19098 (fixed_size_mode::includes_p): Check for constant-sized modes.
19099 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19100 return a poly_uint16 rather than an unsigned short.
19101 (emit_mode_size): Change the type of mode_size from unsigned short
19102 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19103 (emit_mode_adjustments): Cope with polynomial vector sizes.
19104 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19106 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19108 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19109 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19110 * caller-save.c (setup_save_areas): Likewise.
19111 (replace_reg_with_saved_mem): Likewise.
19112 * calls.c (emit_library_call_value_1): Likewise.
19113 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19114 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19115 (gen_lowpart_for_combine): Likewise.
19116 * convert.c (convert_to_integer_1): Likewise.
19117 * cse.c (equiv_constant, cse_insn): Likewise.
19118 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19119 (cselib_subst_to_values): Likewise.
19120 * dce.c (word_dce_process_block): Likewise.
19121 * df-problems.c (df_word_lr_mark_ref): Likewise.
19122 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19123 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19124 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19125 (rtl_for_decl_location): Likewise.
19126 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19127 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19128 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19129 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19130 (expand_expr_real_1): Likewise.
19131 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19132 (pad_below): Likewise.
19133 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19134 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19135 * ira.c (get_subreg_tracking_sizes): Likewise.
19136 * ira-build.c (ira_create_allocno_objects): Likewise.
19137 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19138 (ira_sort_regnos_for_alter_reg): Likewise.
19139 * ira-costs.c (record_operand_costs): Likewise.
19140 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19141 (resolve_simple_move): Likewise.
19142 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19143 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19144 (lra_constraints): Likewise.
19145 (CONST_POOL_OK_P): Reject variable-sized modes.
19146 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19147 (add_pseudo_to_slot, lra_spill): Likewise.
19148 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19149 * optabs-query.c (get_best_extraction_insn): Likewise.
19150 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19151 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19152 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19153 * recog.c (offsettable_address_addr_space_p): Likewise.
19154 * regcprop.c (maybe_mode_change): Likewise.
19155 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19156 * regrename.c (build_def_use): Likewise.
19157 * regstat.c (dump_reg_info): Likewise.
19158 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19159 (find_reloads, find_reloads_subreg_address): Likewise.
19160 * reload1.c (eliminate_regs_1): Likewise.
19161 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19162 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19163 (simplify_binary_operation_1, simplify_subreg): Likewise.
19164 * targhooks.c (default_function_arg_padding): Likewise.
19165 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19166 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19167 (verify_gimple_assign_ternary): Likewise.
19168 * tree-inline.c (estimate_move_cost): Likewise.
19169 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19170 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19171 (get_address_cost_ainc): Likewise.
19172 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19173 (vect_supportable_dr_alignment): Likewise.
19174 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19175 (vectorizable_reduction): Likewise.
19176 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19177 (vectorizable_operation, vectorizable_load): Likewise.
19178 * tree.c (build_same_sized_truth_vector_type): Likewise.
19179 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19180 * var-tracking.c (emit_note_insn_var_location): Likewise.
19181 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19182 (ADDR_VEC_ALIGN): Likewise.
19184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19185 Alan Hayward <alan.hayward@arm.com>
19186 David Sherwood <david.sherwood@arm.com>
19188 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19190 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19191 or if measurement_type is polynomial.
19192 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19193 * combine.c (make_extraction): Likewise.
19194 * dse.c (find_shift_sequence): Likewise.
19195 * dwarf2out.c (mem_loc_descriptor): Likewise.
19196 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19197 (extract_bit_field, extract_low_bits): Likewise.
19198 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19199 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19200 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19201 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19202 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19203 * reload.c (find_reloads): Likewise.
19204 * reload1.c (alter_reg): Likewise.
19205 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19206 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19207 * tree-if-conv.c (predicate_mem_writes): Likewise.
19208 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19209 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19210 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19211 * valtrack.c (dead_debug_insert_temp): Likewise.
19212 * varasm.c (mergeable_constant_section): Likewise.
19213 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19216 Alan Hayward <alan.hayward@arm.com>
19217 David Sherwood <david.sherwood@arm.com>
19219 * expr.c (expand_assignment): Cope with polynomial mode sizes
19220 when assigning to a CONCAT.
19222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19223 Alan Hayward <alan.hayward@arm.com>
19224 David Sherwood <david.sherwood@arm.com>
19226 * machmode.h (mode_precision): Change from unsigned short to
19228 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19230 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19231 or if measurement_type is not polynomial.
19232 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19233 in which the mode is already known to be a scalar_int_mode.
19234 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19235 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19237 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19238 for GET_MODE_PRECISION.
19239 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19240 for GET_MODE_PRECISION.
19241 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19243 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19244 (expand_field_assignment, make_extraction): Likewise.
19245 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19246 (get_last_value): Likewise.
19247 * convert.c (convert_to_integer_1): Likewise.
19248 * cse.c (cse_insn): Likewise.
19249 * expr.c (expand_expr_real_1): Likewise.
19250 * lra-constraints.c (simplify_operand_subreg): Likewise.
19251 * optabs-query.c (can_atomic_load_p): Likewise.
19252 * optabs.c (expand_atomic_load): Likewise.
19253 (expand_atomic_store): Likewise.
19254 * ree.c (combine_reaching_defs): Likewise.
19255 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19256 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19257 * tree.h (type_has_mode_precision_p): Likewise.
19258 * ubsan.c (instrument_si_overflow): Likewise.
19260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19261 Alan Hayward <alan.hayward@arm.com>
19262 David Sherwood <david.sherwood@arm.com>
19264 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19265 polynomial numbers of units.
19266 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19267 (valid_vector_subparts_p): New function.
19268 (build_vector_type): Remove temporary shim and take the number
19269 of units as a poly_uint64 rather than an int.
19270 (build_opaque_vector_type): Take the number of units as a
19271 poly_uint64 rather than an int.
19272 * tree.c (build_vector_from_ctor): Handle polynomial
19273 TYPE_VECTOR_SUBPARTS.
19274 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19275 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19276 (build_vector_from_val): If the number of units is variable,
19277 use build_vec_duplicate_cst for constant operands and
19278 VEC_DUPLICATE_EXPR otherwise.
19279 (make_vector_type): Remove temporary is_constant ().
19280 (build_vector_type, build_opaque_vector_type): Take the number of
19281 units as a poly_uint64 rather than an int.
19282 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19284 * cfgexpand.c (expand_debug_expr): Likewise.
19285 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19286 (store_constructor, expand_expr_real_1): Likewise.
19287 (const_scalar_mask_from_tree): Likewise.
19288 * fold-const-call.c (fold_const_reduction): Likewise.
19289 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19290 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19291 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19292 (fold_relational_const): Likewise.
19293 (native_interpret_vector): Likewise. Change the size from an
19294 int to an unsigned int.
19295 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19296 TYPE_VECTOR_SUBPARTS.
19297 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19298 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19299 duplicating a non-constant operand into a variable-length vector.
19300 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19301 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19302 * ipa-icf.c (sem_variable::equals): Likewise.
19303 * match.pd: Likewise.
19304 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19305 * print-tree.c (print_node): Likewise.
19306 * stor-layout.c (layout_type): Likewise.
19307 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19308 * tree-cfg.c (verify_gimple_comparison): Likewise.
19309 (verify_gimple_assign_binary): Likewise.
19310 (verify_gimple_assign_ternary): Likewise.
19311 (verify_gimple_assign_single): Likewise.
19312 * tree-pretty-print.c (dump_generic_node): Likewise.
19313 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19314 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
19315 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
19316 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
19317 (vect_shift_permute_load_chain): Likewise.
19318 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
19319 (expand_vector_condition, optimize_vector_constructor): Likewise.
19320 (lower_vec_perm, get_compute_type): Likewise.
19321 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19322 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
19323 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
19324 (vect_recog_mask_conversion_pattern): Likewise.
19325 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
19326 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
19327 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19328 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
19329 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
19330 (vectorizable_shift, vectorizable_operation, vectorizable_store)
19331 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
19332 (supportable_widening_operation): Likewise.
19333 (supportable_narrowing_operation): Likewise.
19334 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
19336 * varasm.c (output_constant): Likewise.
19338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19339 Alan Hayward <alan.hayward@arm.com>
19340 David Sherwood <david.sherwood@arm.com>
19342 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
19343 so that both the length == 3 and length != 3 cases set up their
19344 own permute vectors. Add comments explaining why we know the
19345 number of elements is constant.
19346 (vect_permute_load_chain): Likewise.
19348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19349 Alan Hayward <alan.hayward@arm.com>
19350 David Sherwood <david.sherwood@arm.com>
19352 * machmode.h (mode_nunits): Change from unsigned char to
19354 (ONLY_FIXED_SIZE_MODES): New macro.
19355 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
19356 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
19357 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
19359 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
19360 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
19361 or if measurement_type is not polynomial.
19362 * genmodes.c (ZERO_COEFFS): New macro.
19363 (emit_mode_nunits_inline): Make mode_nunits_inline return a
19365 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
19366 Use ZERO_COEFFS when emitting initializers.
19367 * data-streamer.h (bp_pack_poly_value): New function.
19368 (bp_unpack_poly_value): Likewise.
19369 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19370 for GET_MODE_NUNITS.
19371 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19372 for GET_MODE_NUNITS.
19373 * tree.c (make_vector_type): Remove temporary shim and make
19374 the real function take the number of units as a poly_uint64
19375 rather than an int.
19376 (build_vector_type_for_mode): Handle polynomial nunits.
19377 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
19378 * emit-rtl.c (const_vec_series_p_1): Likewise.
19379 (gen_rtx_CONST_VECTOR): Likewise.
19380 * fold-const.c (test_vec_duplicate_folding): Likewise.
19381 * genrecog.c (validate_pattern): Likewise.
19382 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
19383 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19384 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
19385 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
19386 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
19387 * rtlanal.c (subreg_get_info): Likewise.
19388 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19389 (vect_grouped_load_supported): Likewise.
19390 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19391 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
19392 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19393 (simplify_const_unary_operation, simplify_binary_operation_1)
19394 (simplify_const_binary_operation, simplify_ternary_operation)
19395 (test_vector_ops_duplicate, test_vector_ops): Likewise.
19396 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
19397 instead of CONST_VECTOR_NUNITS.
19398 * varasm.c (output_constant_pool_2): Likewise.
19399 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
19400 explicit-encoded elements in the XVEC for variable-length vectors.
19402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19404 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
19406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19407 Alan Hayward <alan.hayward@arm.com>
19408 David Sherwood <david.sherwood@arm.com>
19410 * coretypes.h (fixed_size_mode): Declare.
19411 (fixed_size_mode_pod): New typedef.
19412 * builtins.h (target_builtins::x_apply_args_mode)
19413 (target_builtins::x_apply_result_mode): Change type to
19414 fixed_size_mode_pod.
19415 * builtins.c (apply_args_size, apply_result_size, result_vector)
19416 (expand_builtin_apply_args_1, expand_builtin_apply)
19417 (expand_builtin_return): Update accordingly.
19419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19421 * cse.c (hash_rtx_cb): Hash only the encoded elements.
19422 * cselib.c (cselib_hash_rtx): Likewise.
19423 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
19424 CONST_VECTOR encoding.
19426 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19427 Jeff Law <law@redhat.com>
19430 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19431 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19432 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19433 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19436 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19437 explicitly probe *sp in a noreturn function if there were any callee
19438 register saves or frame pointer is needed.
19440 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19443 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19444 BLKmode for ternary, binary or unary expressions.
19447 * var-tracking.c (delete_vta_debug_insn): New inline function.
19448 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19449 insns from get_insns () to NULL instead of each bb separately.
19450 Use delete_vta_debug_insn. No longer static.
19451 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19452 delete_vta_debug_insns callers.
19453 * rtl.h (delete_vta_debug_insns): Declare.
19454 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19455 instead of variable_tracking_main.
19457 2018-01-03 Martin Sebor <msebor@redhat.com>
19459 PR tree-optimization/83603
19460 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19461 arguments past the endof the argument list in functions declared
19462 without a prototype.
19463 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19464 Avoid checking when arguments are null.
19466 2018-01-03 Martin Sebor <msebor@redhat.com>
19469 * doc/extend.texi (attribute const): Fix a typo.
19470 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19471 issuing -Wsuggest-attribute for void functions.
19473 2018-01-03 Martin Sebor <msebor@redhat.com>
19475 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19476 offset_int::from instead of wide_int::to_shwi.
19477 (maybe_diag_overlap): Remove assertion.
19478 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19479 * gimple-ssa-sprintf.c (format_directive): Same.
19480 (parse_directive): Same.
19481 (sprintf_dom_walker::compute_format_length): Same.
19482 (try_substitute_return_value): Same.
19484 2018-01-03 Jeff Law <law@redhat.com>
19486 PR middle-end/83654
19487 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19488 non-constant residual for zero at runtime and avoid probing in
19489 that case. Reorganize code for trailing problem to mirror handling
19492 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19494 PR tree-optimization/83501
19495 * tree-ssa-strlen.c (get_string_cst): New.
19496 (handle_char_store): Call get_string_cst.
19498 2018-01-03 Martin Liska <mliska@suse.cz>
19500 PR tree-optimization/83593
19501 * tree-ssa-strlen.c: Include tree-cfg.h.
19502 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19503 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19504 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19506 (strlen_dom_walker::before_dom_children): Call
19507 gimple_purge_dead_eh_edges. Dump tranformation with details
19509 (strlen_dom_walker::before_dom_children): Update call by adding
19510 new argument cleanup_eh.
19511 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19513 2018-01-03 Martin Liska <mliska@suse.cz>
19516 * cif-code.def (VARIADIC_THUNK): New enum value.
19517 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19520 2018-01-03 Jan Beulich <jbeulich@suse.com>
19522 * sse.md (mov<mode>_internal): Tighten condition for when to use
19523 vmovdqu<ssescalarsize> for TI and OI modes.
19525 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19527 Update copyright years.
19529 2018-01-03 Martin Liska <mliska@suse.cz>
19532 * ipa-visibility.c (function_and_variable_visibility): Skip
19533 functions with noipa attribure.
19535 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19537 * gcc.c (process_command): Update copyright notice dates.
19538 * gcov-dump.c (print_version): Ditto.
19539 * gcov.c (print_version): Ditto.
19540 * gcov-tool.c (print_version): Ditto.
19541 * gengtype.c (create_file): Ditto.
19542 * doc/cpp.texi: Bump @copying's copyright year.
19543 * doc/cppinternals.texi: Ditto.
19544 * doc/gcc.texi: Ditto.
19545 * doc/gccint.texi: Ditto.
19546 * doc/gcov.texi: Ditto.
19547 * doc/install.texi: Ditto.
19548 * doc/invoke.texi: Ditto.
19550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19552 * vector-builder.h (vector_builder::m_full_nelts): Change from
19553 unsigned int to poly_uint64.
19554 (vector_builder::full_nelts): Update prototype accordingly.
19555 (vector_builder::new_vector): Likewise.
19556 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19557 (vector_builder::operator ==): Likewise.
19558 (vector_builder::finalize): Likewise.
19559 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19560 Take the number of elements as a poly_uint64 rather than an
19562 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19563 from unsigned int to poly_uint64.
19564 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19565 (vec_perm_indices::new_vector): Likewise.
19566 (vec_perm_indices::length): Likewise.
19567 (vec_perm_indices::nelts_per_input): Likewise.
19568 (vec_perm_indices::input_nelts): Likewise.
19569 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19570 number of elements per input as a poly_uint64 rather than an
19571 unsigned int. Use the original encoding for variable-length
19572 vectors, rather than clamping each individual element.
19573 For the second and subsequent elements in each pattern,
19574 clamp the step and base before clamping their sum.
19575 (vec_perm_indices::series_p): Handle polynomial element counts.
19576 (vec_perm_indices::all_in_range_p): Likewise.
19577 (vec_perm_indices_to_tree): Likewise.
19578 (vec_perm_indices_to_rtx): Likewise.
19579 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19580 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19581 (tree_vector_builder::new_binary_operation): Handle polynomial
19582 element counts. Return false if we need to know the number
19583 of elements at compile time.
19584 * fold-const.c (fold_vec_perm): Punt if the number of elements
19585 isn't known at compile time.
19587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19589 * vec-perm-indices.h (vec_perm_builder): Change element type
19590 from HOST_WIDE_INT to poly_int64.
19591 (vec_perm_indices::element_type): Update accordingly.
19592 (vec_perm_indices::clamp): Handle polynomial element_types.
19593 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19594 (vec_perm_indices::all_in_range_p): Likewise.
19595 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19597 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19598 polynomial vec_perm_indices element types.
19599 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19600 * fold-const.c (fold_vec_perm): Likewise.
19601 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19602 * tree-vect-generic.c (lower_vec_perm): Likewise.
19603 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19604 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19605 element type to HOST_WIDE_INT.
19607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19608 Alan Hayward <alan.hayward@arm.com>
19609 David Sherwood <david.sherwood@arm.com>
19611 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19612 rather than an int. Use plus_constant.
19613 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19614 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19617 Alan Hayward <alan.hayward@arm.com>
19618 David Sherwood <david.sherwood@arm.com>
19620 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19621 a HOST_WIDE_INT to a poly_int64.
19623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19624 Alan Hayward <alan.hayward@arm.com>
19625 David Sherwood <david.sherwood@arm.com>
19627 * calls.c (load_register_parameters): Cope with polynomial
19628 mode sizes. Require a constant size for BLKmode parameters
19629 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19630 forces a parameter to be padded at the lsb end in order to
19631 fill a complete number of words, require the parameter size
19632 to be ordered wrt UNITS_PER_WORD.
19634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19635 Alan Hayward <alan.hayward@arm.com>
19636 David Sherwood <david.sherwood@arm.com>
19638 * reload1.c (spill_stack_slot_width): Change element type
19639 from unsigned int to poly_uint64_pod.
19640 (alter_reg): Treat mode sizes as polynomial.
19642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19643 Alan Hayward <alan.hayward@arm.com>
19644 David Sherwood <david.sherwood@arm.com>
19646 * reload.c (complex_word_subreg_p): New function.
19647 (reload_inner_reg_of_subreg, push_reload): Use it.
19649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19650 Alan Hayward <alan.hayward@arm.com>
19651 David Sherwood <david.sherwood@arm.com>
19653 * lra-constraints.c (process_alt_operands): Reject matched
19654 operands whose sizes aren't ordered.
19655 (match_reload): Refer to this check here.
19657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19658 Alan Hayward <alan.hayward@arm.com>
19659 David Sherwood <david.sherwood@arm.com>
19661 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19662 that the mode size is in the set {1, 2, 4, 8, 16}.
19664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19665 Alan Hayward <alan.hayward@arm.com>
19666 David Sherwood <david.sherwood@arm.com>
19668 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19669 Use plus_constant instead of gen_rtx_PLUS.
19671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19672 Alan Hayward <alan.hayward@arm.com>
19673 David Sherwood <david.sherwood@arm.com>
19675 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19676 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19677 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19678 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19679 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19680 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19681 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19682 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19683 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19684 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19686 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19687 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19688 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19689 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19690 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19691 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19692 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19693 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19694 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19695 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19697 * expr.c (emit_move_resolve_push): Treat the input and result
19698 of PUSH_ROUNDING as a poly_int64.
19699 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19700 (emit_push_insn): Likewise.
19701 * lra-eliminations.c (mark_not_eliminable): Likewise.
19702 * recog.c (push_operand): Likewise.
19703 * reload1.c (elimination_effects): Likewise.
19704 * rtlanal.c (nonzero_bits1): Likewise.
19705 * calls.c (store_one_arg): Likewise. Require the padding to be
19706 known at compile time.
19708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19709 Alan Hayward <alan.hayward@arm.com>
19710 David Sherwood <david.sherwood@arm.com>
19712 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19713 Use plus_constant instead of gen_rtx_PLUS.
19715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19716 Alan Hayward <alan.hayward@arm.com>
19717 David Sherwood <david.sherwood@arm.com>
19719 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19720 rather than an int.
19722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19723 Alan Hayward <alan.hayward@arm.com>
19724 David Sherwood <david.sherwood@arm.com>
19726 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19727 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19728 via stack temporaries. Treat the mode size as polynomial too.
19730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19731 Alan Hayward <alan.hayward@arm.com>
19732 David Sherwood <david.sherwood@arm.com>
19734 * expr.c (expand_expr_real_2): When handling conversions involving
19735 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19736 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19737 as a poly_uint64 too.
19739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19740 Alan Hayward <alan.hayward@arm.com>
19741 David Sherwood <david.sherwood@arm.com>
19743 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19746 Alan Hayward <alan.hayward@arm.com>
19747 David Sherwood <david.sherwood@arm.com>
19749 * combine.c (can_change_dest_mode): Handle polynomial
19750 REGMODE_NATURAL_SIZE.
19751 * expmed.c (store_bit_field_1): Likewise.
19752 * expr.c (store_constructor): Likewise.
19753 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19754 and polynomial REGMODE_NATURAL_SIZE.
19755 (gen_lowpart_common): Likewise.
19756 * reginfo.c (record_subregs_of_mode): Likewise.
19757 * rtlanal.c (read_modify_subreg_p): Likewise.
19759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19760 Alan Hayward <alan.hayward@arm.com>
19761 David Sherwood <david.sherwood@arm.com>
19763 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19764 numbers of elements.
19766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19767 Alan Hayward <alan.hayward@arm.com>
19768 David Sherwood <david.sherwood@arm.com>
19770 * match.pd: Cope with polynomial numbers of vector elements.
19772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19773 Alan Hayward <alan.hayward@arm.com>
19774 David Sherwood <david.sherwood@arm.com>
19776 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19777 in a POINTER_PLUS_EXPR.
19779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19780 Alan Hayward <alan.hayward@arm.com>
19781 David Sherwood <david.sherwood@arm.com>
19783 * omp-simd-clone.c (simd_clone_subparts): New function.
19784 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19785 (ipa_simd_modify_function_body): Likewise.
19787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19788 Alan Hayward <alan.hayward@arm.com>
19789 David Sherwood <david.sherwood@arm.com>
19791 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19792 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19793 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19794 (expand_vector_condition, vector_element): Likewise.
19795 (subparts_gt): New function.
19796 (get_compute_type): Use subparts_gt.
19797 (count_type_subparts): Delete.
19798 (expand_vector_operations_1): Use subparts_gt instead of
19799 count_type_subparts.
19801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19802 Alan Hayward <alan.hayward@arm.com>
19803 David Sherwood <david.sherwood@arm.com>
19805 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19806 (vect_compile_time_alias): ...this new function. Do the calculation
19807 on poly_ints rather than trees.
19808 (vect_prune_runtime_alias_test_list): Update call accordingly.
19810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19811 Alan Hayward <alan.hayward@arm.com>
19812 David Sherwood <david.sherwood@arm.com>
19814 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19816 (vect_schedule_slp_instance): Likewise.
19818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19819 Alan Hayward <alan.hayward@arm.com>
19820 David Sherwood <david.sherwood@arm.com>
19822 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19823 constant and extern definitions for variable-length vectors.
19824 (vect_get_constant_vectors): Note that the number of units
19825 is known to be constant.
19827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19828 Alan Hayward <alan.hayward@arm.com>
19829 David Sherwood <david.sherwood@arm.com>
19831 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19832 of units as polynomial. Choose between WIDE and NARROW based
19835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19836 Alan Hayward <alan.hayward@arm.com>
19837 David Sherwood <david.sherwood@arm.com>
19839 * tree-vect-stmts.c (simd_clone_subparts): New function.
19840 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19843 Alan Hayward <alan.hayward@arm.com>
19844 David Sherwood <david.sherwood@arm.com>
19846 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19847 vectors as polynomial. Use build_index_vector for
19848 IFN_GOMP_SIMD_LANE.
19850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19851 Alan Hayward <alan.hayward@arm.com>
19852 David Sherwood <david.sherwood@arm.com>
19854 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19855 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19856 for variable-length vectors.
19857 (vectorizable_mask_load_store): Treat the number of units as
19858 polynomial, asserting that it is constant if the condition has
19859 already been enforced.
19860 (vectorizable_store, vectorizable_load): Likewise.
19862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19863 Alan Hayward <alan.hayward@arm.com>
19864 David Sherwood <david.sherwood@arm.com>
19866 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19867 of units as polynomial. Punt if we can't tell at compile time
19868 which vector contains the final result.
19870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19871 Alan Hayward <alan.hayward@arm.com>
19872 David Sherwood <david.sherwood@arm.com>
19874 * tree-vect-loop.c (vectorizable_induction): Treat the number
19875 of units as polynomial. Punt on SLP inductions. Use an integer
19876 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19877 cast of such a series for variable-length floating-point
19880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19881 Alan Hayward <alan.hayward@arm.com>
19882 David Sherwood <david.sherwood@arm.com>
19884 * tree.h (build_index_vector): Declare.
19885 * tree.c (build_index_vector): New function.
19886 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19887 of units as polynomial, forcibly converting it to a constant if
19888 vectorizable_reduction has already enforced the condition.
19889 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19890 to create a {1,2,3,...} vector.
19891 (vectorizable_reduction): Treat the number of units as polynomial.
19892 Choose vectype_in based on the largest scalar element size rather
19893 than the smallest number of units. Enforce the restrictions
19896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19897 Alan Hayward <alan.hayward@arm.com>
19898 David Sherwood <david.sherwood@arm.com>
19900 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19901 number of units as polynomial.
19903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19904 Alan Hayward <alan.hayward@arm.com>
19905 David Sherwood <david.sherwood@arm.com>
19907 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19908 * target.def (autovectorize_vector_sizes): Return the vector sizes
19909 by pointer, using vector_sizes rather than a bitmask.
19910 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19911 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19912 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19914 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19915 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19916 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19917 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19918 * omp-general.c (omp_max_vf): Likewise.
19919 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19920 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19921 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19922 * tree-vect-slp.c (vect_slp_bb): Likewise.
19923 * doc/tm.texi: Regenerate.
19924 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19926 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19927 the vector size as a poly_uint64 rather than an unsigned int.
19928 (current_vector_size): Change from an unsigned int to a poly_uint64.
19929 (get_vectype_for_scalar_type): Update accordingly.
19930 * tree.h (build_truth_vector_type): Take the size and number of
19931 units as a poly_uint64 rather than an unsigned int.
19932 (build_vector_type): Add a temporary overload that takes
19933 the number of units as a poly_uint64 rather than an unsigned int.
19934 * tree.c (make_vector_type): Likewise.
19935 (build_truth_vector_type): Take the number of units as a poly_uint64
19936 rather than an unsigned int.
19938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19939 Alan Hayward <alan.hayward@arm.com>
19940 David Sherwood <david.sherwood@arm.com>
19942 * target.def (get_mask_mode): Take the number of units and length
19943 as poly_uint64s rather than unsigned ints.
19944 * targhooks.h (default_get_mask_mode): Update accordingly.
19945 * targhooks.c (default_get_mask_mode): Likewise.
19946 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19947 * doc/tm.texi: Regenerate.
19949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19950 Alan Hayward <alan.hayward@arm.com>
19951 David Sherwood <david.sherwood@arm.com>
19953 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19954 * omp-general.c (omp_max_vf): Likewise.
19955 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19956 (expand_omp_simd): Handle polynomial safelen.
19957 * omp-low.c (omplow_simd_context): Add a default constructor.
19958 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19959 (lower_rec_simd_input_clauses): Update accordingly.
19960 (lower_rec_input_clauses): Likewise.
19962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19963 Alan Hayward <alan.hayward@arm.com>
19964 David Sherwood <david.sherwood@arm.com>
19966 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19967 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19968 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19969 (vect_analyze_slp_cost): Likewise.
19970 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19971 (vect_model_load_cost): Likewise.
19973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19974 Alan Hayward <alan.hayward@arm.com>
19975 David Sherwood <david.sherwood@arm.com>
19977 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19978 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19979 from an unsigned int * to a poly_uint64_pod *.
19980 (calculate_unrolling_factor): New function.
19981 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19984 Alan Hayward <alan.hayward@arm.com>
19985 David Sherwood <david.sherwood@arm.com>
19987 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19988 from an unsigned int to a poly_uint64.
19989 (_loop_vec_info::slp_unrolling_factor): Likewise.
19990 (_loop_vec_info::vectorization_factor): Change from an int
19992 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19993 (vect_get_num_vectors): New function.
19994 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19995 (vect_get_num_copies): Use vect_get_num_vectors.
19996 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19997 to an unsigned int *.
19998 (vect_analyze_data_refs): Change min_vf from an int * to a
20000 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20001 than an unsigned HOST_WIDE_INT.
20002 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20003 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20004 to an unsigned int *.
20005 (vect_analyze_data_ref_dependences): Likewise.
20006 (vect_compute_data_ref_alignment): Handle polynomial vf.
20007 (vect_enhance_data_refs_alignment): Likewise.
20008 (vect_prune_runtime_alias_test_list): Likewise.
20009 (vect_shift_permute_load_chain): Likewise.
20010 (vect_supportable_dr_alignment): Likewise.
20011 (dependence_distance_ge_vf): Take the vectorization factor as a
20012 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20013 (vect_analyze_data_refs): Change min_vf from an int * to a
20015 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20016 vfm1 as a poly_uint64 rather than an int. Make the same change
20017 for the returned bound_scalar.
20018 (vect_gen_vector_loop_niters): Handle polynomial vf.
20019 (vect_do_peeling): Likewise. Update call to
20020 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20021 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20023 * tree-vect-loop.c (vect_determine_vectorization_factor)
20024 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20025 (vect_get_known_peeling_cost): Likewise.
20026 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20027 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20028 (vect_transform_loop): Likewise. Use the lowest possible VF when
20029 updating the upper bounds of the loop.
20030 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20031 rather than an int.
20032 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20033 polynomial unroll factors.
20034 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20035 (vect_make_slp_decision): Likewise.
20036 (vect_supported_load_permutation_p): Likewise, and polynomial
20038 (vect_analyze_slp_cost): Handle polynomial vf.
20039 (vect_slp_analyze_node_operations): Likewise.
20040 (vect_slp_analyze_bb_1): Likewise.
20041 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20042 than an unsigned HOST_WIDE_INT.
20043 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20044 (vectorizable_load): Handle polynomial vf.
20045 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20047 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20050 Alan Hayward <alan.hayward@arm.com>
20051 David Sherwood <david.sherwood@arm.com>
20053 * match.pd: Handle bit operations involving three constants
20054 and try to fold one pair.
20056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20058 * tree-vect-loop-manip.c: Include gimple-fold.h.
20059 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20060 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20061 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20062 Add a path that uses a step of VF instead of 1, but disable it
20064 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20065 and niters_no_overflow parameters. Update calls to
20066 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20067 Create a new SSA name if the latter choses to use a ste other
20068 than zero, and return it via niters_vector_mult_vf_var.
20069 * tree-vect-loop.c (vect_transform_loop): Update calls to
20070 vect_do_peeling, vect_gen_vector_loop_niters and
20071 slpeel_make_loop_iterate_ntimes.
20072 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20073 (vect_gen_vector_loop_niters): Update declarations after above changes.
20075 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20077 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20078 128-bit round to integer instructions.
20079 (ceil<mode>2): Likewise.
20080 (btrunc<mode>2): Likewise.
20081 (round<mode>2): Likewise.
20083 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20085 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20086 unaligned VSX load/store on P8/P9.
20087 (expand_block_clear): Allow the use of unaligned VSX
20088 load/store on P8/P9.
20090 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20092 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20094 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20095 swap associated with both a load and a store.
20097 2018-01-02 Andrew Waterman <andrew@sifive.com>
20099 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20100 * config/riscv/riscv.md (clear_cache): Use it.
20102 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20104 * web.c: Remove out-of-date comment.
20106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20108 * expr.c (fixup_args_size_notes): Check that any existing
20109 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20110 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20111 (emit_single_push_insn): ...here.
20113 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20115 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20116 (const_vector_encoded_nelts): New function.
20117 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20118 (const_vector_int_elt, const_vector_elt): Declare.
20119 * emit-rtl.c (const_vector_int_elt_1): New function.
20120 (const_vector_elt): Likewise.
20121 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20122 of CONST_VECTOR_ELT.
20124 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20126 * expr.c: Include rtx-vector-builder.h.
20127 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20128 directly on the tree encoding.
20129 (const_vector_from_tree): Likewise.
20130 * optabs.c: Include rtx-vector-builder.h.
20131 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20132 sequence of "u" values.
20133 * vec-perm-indices.c: Include rtx-vector-builder.h.
20134 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20135 directly on the vec_perm_indices encoding.
20137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20139 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20140 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20141 * rtx-vector-builder.h: New file.
20142 * rtx-vector-builder.c: Likewise.
20143 * rtl.h (rtx_def::u2): Add a const_vector field.
20144 (CONST_VECTOR_NPATTERNS): New macro.
20145 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20146 (CONST_VECTOR_DUPLICATE_P): Likewise.
20147 (CONST_VECTOR_STEPPED_P): Likewise.
20148 (CONST_VECTOR_ENCODED_ELT): Likewise.
20149 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20150 (unwrap_const_vec_duplicate): Likewise.
20151 (const_vec_series_p): Check for a non-duplicated vector encoding.
20152 Say that the function only returns true for integer vectors.
20153 * emit-rtl.c: Include rtx-vector-builder.h.
20154 (gen_const_vec_duplicate_1): Delete.
20155 (gen_const_vector): Call gen_const_vec_duplicate instead of
20156 gen_const_vec_duplicate_1.
20157 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20158 (gen_const_vec_duplicate): Use rtx_vector_builder.
20159 (gen_const_vec_series): Likewise.
20160 (gen_rtx_CONST_VECTOR): Likewise.
20161 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20162 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20163 Build a new vector rather than modifying a CONST_VECTOR in-place.
20164 (handle_special_swappables): Update call accordingly.
20165 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20166 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20167 Build a new vector rather than modifying a CONST_VECTOR in-place.
20168 (handle_special_swappables): Update call accordingly.
20170 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20172 * simplify-rtx.c (simplify_const_binary_operation): Use
20173 CONST_VECTOR_ELT instead of XVECEXP.
20175 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20177 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20178 the selector elements to be different from the data elements
20179 if the selector is a VECTOR_CST.
20180 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20181 ssizetype for the selector.
20183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20185 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20186 before testing each element individually.
20187 * tree-vect-generic.c (lower_vec_perm): Likewise.
20189 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20191 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20192 * selftest-run-tests.c (selftest::run_tests): Call it.
20193 * vector-builder.h (vector_builder::operator ==): New function.
20194 (vector_builder::operator !=): Likewise.
20195 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20196 (vec_perm_indices::all_from_input_p): New function.
20197 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20198 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20199 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20200 instead of reading the VECTOR_CST directly. Detect whether both
20201 vector inputs are the same before constructing the vec_perm_indices,
20202 and update the number of inputs argument accordingly. Use the
20203 utility functions added above. Only construct sel2 if we need to.
20205 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20207 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20208 the broadcast of the low byte.
20209 (expand_mult_highpart): Use an explicit encoding for the permutes.
20210 * optabs-query.c (can_mult_highpart_p): Likewise.
20211 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20212 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20213 (vectorizable_bswap): Likewise.
20214 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20215 explicit encoding for the power-of-2 permutes.
20216 (vect_permute_store_chain): Likewise.
20217 (vect_grouped_load_supported): Likewise.
20218 (vect_permute_load_chain): Likewise.
20220 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20222 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20223 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20224 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20225 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20226 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20227 (vect_gen_perm_mask_any): Likewise.
20229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20231 * int-vector-builder.h: New file.
20232 * vec-perm-indices.h: Include int-vector-builder.h.
20233 (vec_perm_indices): Redefine as an int_vector_builder.
20234 (auto_vec_perm_indices): Delete.
20235 (vec_perm_builder): Redefine as a stand-alone class.
20236 (vec_perm_indices::vec_perm_indices): New function.
20237 (vec_perm_indices::clamp): Likewise.
20238 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20239 (vec_perm_indices::new_vector): New function.
20240 (vec_perm_indices::new_expanded_vector): Update for new
20241 vec_perm_indices class.
20242 (vec_perm_indices::rotate_inputs): New function.
20243 (vec_perm_indices::all_in_range_p): Operate directly on the
20244 encoded form, without computing elided elements.
20245 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20246 encoding. Update for new vec_perm_indices class.
20247 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20248 the given vec_perm_builder.
20249 (expand_vec_perm_var): Update vec_perm_builder constructor.
20250 (expand_mult_highpart): Use vec_perm_builder instead of
20251 auto_vec_perm_indices.
20252 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20253 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20254 or double series encoding as appropriate.
20255 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20256 vec_perm_indices instead of auto_vec_perm_indices.
20257 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20258 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20259 (vect_permute_store_chain): Likewise.
20260 (vect_grouped_load_supported): Likewise.
20261 (vect_permute_load_chain): Likewise.
20262 (vect_shift_permute_load_chain): Likewise.
20263 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20264 (vect_transform_slp_perm_load): Likewise.
20265 (vect_schedule_slp_instance): Likewise.
20266 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20267 (vectorizable_mask_load_store): Likewise.
20268 (vectorizable_bswap): Likewise.
20269 (vectorizable_store): Likewise.
20270 (vectorizable_load): Likewise.
20271 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20272 vec_perm_indices instead of auto_vec_perm_indices. Use
20273 tree_to_vec_perm_builder to read the vector from a tree.
20274 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20275 vec_perm_builder instead of a vec_perm_indices.
20276 (have_whole_vector_shift): Use vec_perm_builder and
20277 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20278 truncation to calc_vec_perm_mask_for_shift.
20279 (vect_create_epilog_for_reduction): Likewise.
20280 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20281 from auto_vec_perm_indices to vec_perm_indices.
20282 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20283 instead of changing individual elements.
20284 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20285 the vector in d.perm.
20286 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20287 from auto_vec_perm_indices to vec_perm_indices.
20288 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20289 instead of changing individual elements.
20290 (arm_vectorize_vec_perm_const): Use new_vector to install
20291 the vector in d.perm.
20292 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20293 Update vec_perm_builder constructor.
20294 (rs6000_expand_interleave): Likewise.
20295 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20296 (rs6000_expand_interleave): Likewise.
20298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20300 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20301 to qimode could truncate the indices.
20302 * optabs.c (expand_vec_perm_var): Likewise.
20304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20306 * Makefile.in (OBJS): Add vec-perm-indices.o.
20307 * vec-perm-indices.h: New file.
20308 * vec-perm-indices.c: Likewise.
20309 * target.h (vec_perm_indices): Replace with a forward class
20311 (auto_vec_perm_indices): Move to vec-perm-indices.h.
20312 * optabs.h: Include vec-perm-indices.h.
20313 (expand_vec_perm): Delete.
20314 (selector_fits_mode_p, expand_vec_perm_var): Declare.
20315 (expand_vec_perm_const): Declare.
20316 * target.def (vec_perm_const_ok): Replace with...
20317 (vec_perm_const): ...this new hook.
20318 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
20319 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
20320 * doc/tm.texi: Regenerate.
20321 * optabs.def (vec_perm_const): Delete.
20322 * doc/md.texi (vec_perm_const): Likewise.
20323 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
20324 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
20325 expand_vec_perm for constant permutation vectors. Assert that
20326 the mode of variable permutation vectors is the integer equivalent
20327 of the mode that is being permuted.
20328 * optabs-query.h (selector_fits_mode_p): Declare.
20329 * optabs-query.c: Include vec-perm-indices.h.
20330 (selector_fits_mode_p): New function.
20331 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
20332 is defined, instead of checking whether the vec_perm_const_optab
20333 exists. Use targetm.vectorize.vec_perm_const instead of
20334 targetm.vectorize.vec_perm_const_ok. Check whether the indices
20335 fit in the vector mode before using a variable permute.
20336 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
20337 vec_perm_indices instead of an rtx.
20338 (expand_vec_perm): Replace with...
20339 (expand_vec_perm_const): ...this new function. Take the selector
20340 as a vec_perm_indices rather than an rtx. Also take the mode of
20341 the selector. Update call to shift_amt_for_vec_perm_mask.
20342 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
20343 Use vec_perm_indices::new_expanded_vector to expand the original
20344 selector into bytes. Check whether the indices fit in the vector
20345 mode before using a variable permute.
20346 (expand_vec_perm_var): Make global.
20347 (expand_mult_highpart): Use expand_vec_perm_const.
20348 * fold-const.c: Includes vec-perm-indices.h.
20349 * tree-ssa-forwprop.c: Likewise.
20350 * tree-vect-data-refs.c: Likewise.
20351 * tree-vect-generic.c: Likewise.
20352 * tree-vect-loop.c: Likewise.
20353 * tree-vect-slp.c: Likewise.
20354 * tree-vect-stmts.c: Likewise.
20355 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
20357 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
20358 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
20359 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
20360 (aarch64_vectorize_vec_perm_const): ...this new function.
20361 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20362 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20363 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
20364 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
20365 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20366 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20367 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
20369 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
20370 check for NEON modes.
20371 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
20372 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
20373 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
20374 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
20376 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
20377 the old VEC_PERM_CONST conditions.
20378 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
20379 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
20380 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
20381 (ia64_vectorize_vec_perm_const_ok): Merge into...
20382 (ia64_vectorize_vec_perm_const): ...this new function.
20383 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
20384 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
20385 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
20386 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
20387 * config/mips/mips.c (mips_expand_vec_perm_const)
20388 (mips_vectorize_vec_perm_const_ok): Merge into...
20389 (mips_vectorize_vec_perm_const): ...this new function.
20390 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
20391 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
20392 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
20393 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
20394 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
20395 (rs6000_expand_vec_perm_const): Delete.
20396 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
20398 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20399 (altivec_expand_vec_perm_const_le): Take each operand individually.
20400 Operate on constant selectors rather than rtxes.
20401 (altivec_expand_vec_perm_const): Likewise. Update call to
20402 altivec_expand_vec_perm_const_le.
20403 (rs6000_expand_vec_perm_const): Delete.
20404 (rs6000_vectorize_vec_perm_const_ok): Delete.
20405 (rs6000_vectorize_vec_perm_const): New function.
20406 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20407 an element count and rtx array.
20408 (rs6000_expand_extract_even): Update call accordingly.
20409 (rs6000_expand_interleave): Likewise.
20410 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
20411 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
20412 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
20413 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
20414 (rs6000_expand_vec_perm_const): Delete.
20415 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20416 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20417 (altivec_expand_vec_perm_const_le): Take each operand individually.
20418 Operate on constant selectors rather than rtxes.
20419 (altivec_expand_vec_perm_const): Likewise. Update call to
20420 altivec_expand_vec_perm_const_le.
20421 (rs6000_expand_vec_perm_const): Delete.
20422 (rs6000_vectorize_vec_perm_const_ok): Delete.
20423 (rs6000_vectorize_vec_perm_const): New function. Remove stray
20424 reference to the SPE evmerge intructions.
20425 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20426 an element count and rtx array.
20427 (rs6000_expand_extract_even): Update call accordingly.
20428 (rs6000_expand_interleave): Likewise.
20429 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20430 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20434 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20436 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20437 vector mode and that that mode matches the mode of the data
20439 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20440 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20441 directly using expand_vec_perm_1 when forcing selectors into
20443 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20445 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20447 * optabs-query.h (can_vec_perm_p): Delete.
20448 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20449 * optabs-query.c (can_vec_perm_p): Split into...
20450 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20451 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20452 particular selector is valid.
20453 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20454 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20455 (vect_grouped_load_supported): Likewise.
20456 (vect_shift_permute_load_chain): Likewise.
20457 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20458 (vect_transform_slp_perm_load): Likewise.
20459 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20460 (vectorizable_bswap): Likewise.
20461 (vect_gen_perm_mask_checked): Likewise.
20462 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20463 implementations of variable permutation vectors into account
20464 when deciding which selector to use.
20465 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20466 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20467 with a false third argument.
20468 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20469 to test whether the constant selector is valid and can_vec_perm_var_p
20470 to test whether a variable selector is valid.
20472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20474 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20475 * optabs-query.c (can_vec_perm_p): Likewise.
20476 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20477 instead of vec_perm_indices.
20478 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20479 (vect_gen_perm_mask_checked): Likewise,
20480 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20481 (vect_gen_perm_mask_checked): Likewise,
20483 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20485 * optabs-query.h (qimode_for_vec_perm): Declare.
20486 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20487 (qimode_for_vec_perm): ...this new function.
20488 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20490 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20492 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20493 does not have a conditional at the top.
20495 2018-01-02 Richard Biener <rguenther@suse.de>
20497 * ipa-inline.c (big_speedup_p): Fix expression.
20499 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20502 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20505 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20509 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20510 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20511 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20512 cond_taken_branch_cost 3->4.
20514 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20516 PR tree-optimization/83581
20517 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20518 TODO_cleanup_cfg if any changes have been made.
20520 PR middle-end/83608
20521 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20522 convert_modes if target mode has the right side, but different mode
20525 PR middle-end/83609
20526 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20527 last argument when extracting from CONCAT. If either from_real or
20528 from_imag is NULL, use expansion through memory. If result is not
20529 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20530 the parts directly to inner mode, if even that fails, use expansion
20533 PR middle-end/83623
20534 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20535 check for bswap in mode rather than HImode and use that in expand_unop
20538 Copyright (C) 2018 Free Software Foundation, Inc.
20540 Copying and distribution of this file, with or without modification,
20541 are permitted in any medium without royalty provided the copyright
20542 notice and this notice are preserved.